Stage-dependent modes of Pax6-Sox2 epistasis regulate lens development and eye morphogenesis

The Pax6 and Sox2 genes are both known to have major roles in eye development (Callaerts et al.,1997; Chow and Lang,2001; Treisman,2004; Kondoh,2008). Pax6 is a paired domain and homeodomain-containing transcription factor that is essential for eye development in both invertebrates and vertebrates (Quiring et al., 1994). It also has the remarkable ability to induce ectopic eyes, both in flies and frogs, when misexpressed(Halder et al., 1995; Chow et al., 1999). Pax6 is expressed in the presumptive lens in a region that includes the lens placode and adjacent ectoderm (Grindley et al.,1997; Furuta and Hogan,1998; Ashery-Padan et al.,2000; Dimanlig et al.,2001). According to tissue recombination experiments(Fujiwara et al., 1994), the generation of Pax6 mutant chimeric mice(Collinson et al., 2000),lineage-traced ectopic lenses (Chow et al.,1999) and Pax6 conditional deletion(Ashery-Padan et al., 2000),Pax6 has an autonomous role in lens development. When Pax6 is misexpressed in the frog, it has been observed that ectopic lenses can form in isolation,whereas ectopic retina development is always accompanied by adjacent ectopic lens tissue (Chow et al.,1999). This has suggested that the developing lens may provide important signals for formation of the retina.

Sox2 is one member of the larger family of high mobility group (HMG) domain transcription factors (Kamachi et al.,2000). Members of the Sox family have diverse tissue-specific expression patterns throughout early development and have been implicated in cell fate decisions in numerous processes(Uwanogho et al., 1995). Sox1,Sox2 and Sox3 are all expressed in the lens(Kamachi et al., 1998). Sox2 and Sox3 expression in the early lens placode is dependent on the optic vesicle and this implies that they are responsive to inductive signals(Kamachi et al., 1998). Sox1 is first expressed later during invagination of the lens placode, and also during lens morphogenesis in lens fiber cells(Kamachi et al., 1998). Sox2 has been implicated in lens development through its regulation of theδ 1-crystallin gene in the chick(Kamachi et al., 2001) and,more recently, of N-cadherin(Matsumata et al., 2005), an adhesion molecule known to be required for normal lens morphogenesis and the differentiation of lens fiber cells(Pontoriero et al., 2009). It has also been proposed that the Sox2 gene is regulated by the Sox2 protein product in combination with Pax6 through the N-3 enhancer that is active in the presumptive lens (Inoue et al., 2007). Consistent with an important role for Sox2 in eye development, it was recently shown that, in humans, heterozygous Sox2mutation can result in anophthalmia-esophageal-genital (AEG) syndrome(Fantes et al., 2003; Taranova et al., 2006; Bakrania et al., 2007).

It has been recognized, based on the analysis of cis-acting regulatory elements, that Pax6 and Sox2 are likely to be cross-regulated during development (Kondoh et al., 2004; Hever et al.,2006; Inoue et al.,2007). Despite this, to date, an assessment of the genetic and functional interactions between Pax6 and Sox2 has not been performed. In the current study, we generated a Sox2 conditional allele in the mouse and, in combination with the existing Pax6 conditional allele(Ashery-Padan et al., 2000),formally tested the genetic and functional relationships between Pax6and Sox2 in the lens. This showed that in pre-placodal ectoderm, Pax6 and Sox2 expression is not inter-dependent, but that the two proteins cooperate functionally in the very early steps of eye development. Unexpectedly, we show that various combinations Pax6 and Sox2 deletion in the pre-placodal ectoderm have profound effects on the morphogenesis of the eye. This suggests that Pax6 and Sox2 are required for the production of signals that initiate eye morphogenesis. We also show that, in pre-placodal ectoderm,N-cadherin is dependent on Sox2, but not Pax6. After the lens placode has formed, Pax6 and Sox2 change to a different genetic relationship, whereby Sox2 expression is dependent on Pax6. N-cadherin expression at this later stage is dependent on Pax6. These data support a model in which the mode of Pax6-Sox2 inter-regulation is stage dependent, and point to an underlying mechanism in which DNA binding site availability is regulated.

Animals were housed in a pathogen-free vivarium in accordance with institutional policies. Gestational age was determined through detection of a vaginal plug. At specific gestational ages, fetuses were removed by hysterectomy after the dams had been anesthetized with isoflurane. In this analysis, eight to 20 embryonic eyes were analyzed for each genotype and stage of development.

The Sox2^flox allele was generated using conventional gene-targeting methods. A 5′ loxP site was placed in the 5′ untranslated region of Sox2 and a 3′ loxP site downstream of the single Sox2 exon(Fig. 1A). The Frt site-flanked neo gene of the targeting vector was excised in vivo using a flippase-expressing mouse line (Rodriguez et al., 2000). A similar allele of mouse Sox2 has been generated by others (Miyagi et al.,2008).

The following transgenic and gene-targeted mice were used in this study: Le-cre (Ashery-Padan et al.,2000), AP2α-cre(Macatee et al., 2003), Pax6^flox (Ashery-Padan et al., 2000), Pax6^sey(Hill et al., 1991), N-cadherin^Lacz (Radice et al., 1997), N-cadherin^flox(Kostetskii et al., 2005) andα MHC-Ecad (Luo et al.,2001). All mouse lines used in this study were genotyped by PCR using primers and protocols described previously. Yolk sacs from staged embryos or tail tips were digested overnight at 55°C in lysis buffer and genomic DNA extracted using a Kingfisher 96 Magnetic Particle Processor. Primers for genotyping of the Sox2^flox allele were as follows: VS635, TGGAATCAGGCTGCCGAGAATCC; VS636, TCGTTCTGGCAACAAGTGCTAAAGC; and VS369, CTGCCATAGCCACTCGAGAAG. The PCR protocol used was 95°C for 4 minutes followed by 25 cycles of 94°C for 30 seconds, 58°C for 30 seconds and 72°C for 30 seconds, with a final extension period of 72°C for 7 minutes. These primers produce bands of 421 bp for the wild-type allele and 546 bp for the targeted allele.

Immunofluorescence labeling for cryosections was performed as previously described (Smith et al.,2005). All sections were permeabilized with freshly prepared ice-cold 0.5% Triton X-100, 1% sodium citrate in PBS for 5 minutes and then washed three times in 0.1% Tween in PBS prior to blocking. Primary antibody dilutions were as follows: rabbit polyclonal anti-N-cadherin antibody (ABCAM,ab18203), 1:300; sheep polyclonal CHX10 (Exalpha, X1180P), 1:1000; rabbit polyclonal anti-Pax6 (Covance, PRB-278P), 1:2000; rabbit polyclonal anti-Sox2(Chemicon, AB5603), 1:1000; goat polyclonal anti-P-cadherin (R&D Systems,AF761), 1:100; polyclonal rabbit anti-β-crystallin (generated in our laboratory), 1:5000; and rabbit polyclonal Mitf-1 (a gift from H. Arnheiter,National Institute of Neurological Disorders and Stroke, NIH, USA), 1:2500. Alexa Fluor secondary antibodies and Alexa phalloidins were obtained from Invitrogen or Molecular Probes and used at a dilution of 1:5000 (A-11072,A-11070, A-11016, A11058, A11015, A11055, A-12379). All sections were counterstained with Hoechst 33342 (Sigma, B-2261) for the visualization of nuclei.

In order to effectively study the cooperative roles of Pax6 and Sox2 in the developing mouse lens, we generated a conditional Sox2^floxallele (see also Miyagi et al.,2008) using conventional gene targeting in ES cells. The mouse Sox2 gene has a single exon and so loxP sites were placed in the 5′ untranslated region and downstream of the 3′ UTR(Fig. 1A). A Pax6^flox allele was generated previously(Ashery-Padan et al.,2000).

To assess the functions of Pax6 and Sox2 in the developing lens, we took advantage of two Cre-expressing drivers. The Lecre transgenic mouse line (Ashery-Padan et al.,2000) uses the Pax6 ectoderm enhancer (EE)(Williams et al., 1998; Kammandel et al., 1999; Xu et al., 1999) to drive Cre and GFP expression in the developing lens from the placode stage(approximately E9.0) onwards (Fig. 1B). At later stages of development, Le-cre is expressed in the periocular surface ectoderm that includes the presumptive conjunctiva,the corneal ectoderm and the periocular gland epithelia(Williams et al., 1998; Kammandel et al., 1999; Xu et al., 1999; Ashery-Padan et al., 2000; Smith et al., 2005). The AP2α-cre mouse line was generated by inserting the Cre recombinase-coding region into the 3′ untranslated region of the AP2α gene (Tcfap2a - Mouse Genome Informatics) using gene-targeting methods (Macatee et al.,2003). This results in a Cre expression domain that includes the dorsal neural tube, neural crest-derived periocular mesenchyme and the head surface ectoderm that includes the presumptive lens at a pre-placodal stage(Fig. 1B). A comparison of the consequences of AP2α-cre and Le-cre mediated gene deletion can be informative (Song et al., 2007), as they represent genetically distinct phases of lens development (Grindley et al.,1995; Lang,2004).

To understand the roles of Pax6 and Sox2 at the placodal stage of lens development, we first performed deletions of each gene separately using Le-cre. Le-cre-mediated deletion of Pax6 confirmed(Ashery-Padan et al., 2000)that Pax6 immunoreactivity in the lens ectoderm was reduced at E9.5(Fig. 2B) and absent at E10.5(Fig. 2E). Pax6 labeling in the presumptive retina of Le-cre; Pax6^FL/FL embryos was unchanged (Fig. 2B,E) and served to illustrate the specificity of the Le-cre driver. Phenotypically, Pax6 deletion in the presumptive lens resulted in a failure of lens development from the placode stage onwards(Fig. 2B,E,H,K), as would be anticipated (Ashery-Padan et al.,2000). In addition, the optic vesicle failed to undergo normal morphogenesis; although the expected thickness differential was sometimes observed, the presumptive retinal and retinal pigmented epithelia were not apposed or cupped (Fig. 2E,K). When we labeled for Sox2 in Lecre; Pax6^FL/FL embryos,there was little change apparent at E9.5(Fig. 2H) but dramatically reduced immunoreactivity by E10.5 (Fig. 2K). This indicated that at the pit stage of lens development,Sox2 expression is dependent on Pax6.

Conditional deletion of Sox2^FL/FL with Le-cregave reliably diminished Sox2 immunoreactivity in the presumptive lens of E9.5 embryos (Fig. 2I) and an absence at E10.5 (Fig. 2L). The phenotypic consequences of Sox2 deletion were milder than in the Pax6 mutant: some placodal thickening was apparent (e.g. Fig. 2C) and a lens pit, albeit shallow compared with that of wild type(Fig. 2J), was observed(Fig. 2L). Furthermore, the optic cup of Le-cre; Sox2^FL/FL embryos was formed fairly normally with apposed presumptive retina and retinal pigment epithelial layers of appropriate thickness. Immunolabeling for Pax6 in Le-cre;Sox2^FL/FL embryos did not detect any changes in Pax6 levels either at E9.5 (Fig. 2C) or at E10.5 (Fig. 2F), indicating that Pax6 expression is not dependent on Sox2. When combined with the above data showing the dependence of Sox2 expression on Pax6, this suggests that after the placodal stage of lens development there is a simple linear genetic pathway with Pax6 upstream of Sox2.

AP2α-cre(Macatee et al., 2003) is expressed earlier than Le-cre and is active at E8.5 in the head surface ectoderm that encompasses the pre-placodal presumptive lens. Because Pax6 and Sox2 are not expressed in the crest-derived periocular mesenchyme where AP2α-cre is also active, this component of driver activity is not of concern. To determine the function of Pax6and Sox2 and to examine their epistatic relationship at pre-placodal stages, we performed conditional deletion of each individual gene with AP2α-cre.

In AP2α-cre; Pax6^FL/FL embryos, typical nuclear Pax6 immunoreactivity was largely lost at E8.5 in the surface ectoderm(Fig. 3B), but was somewhat patchy, presumably because some late-deleting cells retained residual Pax6. By E9.5 (Fig. 3E) and E10.5(Fig. 3H) the presumptive lens ectoderm of AP2α-cre; Pax6^FL/FL embryos showed no Pax6 immunoreactivity. We were not able to discern any change in Pax6 immunoreactivity in the optic vesicle at any stage in any (n=20)of the AP2α-cre; Pax6^Fl/Fl embryos examined. A phenotypic response to the absence of pre-placodal Pax6 did not become apparent until E9.5 and beyond but manifested as an absence of placodal thickening (Fig. 3E) and a complete failure of eye morphogenesis (Fig. 3H) that was more severe than when Lecre was used to delete Pax6^FL (Fig. 2). Indeed, the phenotype of AP2α-cre;Pax6^FL/FL embryos most closely resembles that of Pax6^Sey homozygotes in which the lens placode does not thicken, there is no invagination of either lens pit or optic cup and the optic stalk fails to constrict in the proximal eye region(Grindley et al., 1995). In AP2α-cre; Sox2^FL/FL embryos (see Fig. 4 for confirmation of deletion) there was no impact on Pax6 immunoreactivity in the presumptive lens from E8.5 to E10.5 (Fig. 3C,F,I). This indicates that during these stages of lens development, Sox2 is not upstream of Pax6.

As expected, AP2α-cre; Sox2^FL/FLresulted in the absence of Sox2 immunoreactivity in the surface ectoderm and presumptive lens from E8.5-E10.5 (Fig. 4C,F,I). The phenotypic consequence of this pre-placodal Sox2 deletion was similar to the consequence of placodal deletion with Le-cre, in that the presumptive lens and retina underwent a modest invagination that arrested at the equivalent of E10.0(Fig. 4I). However, there were also distinctions between AP2α-cre and Le-cremediated Sox2 deletion. Unlike Le-cre mediated Sox2deletion where the optic cup layers were well formed, AP2α-cre-mediated Sox2 deletion resulted in a failure of the presumptive retina and the RPE to form nested cups. Instead,the RPE was widely separated from the presumptive retina(Fig. 4I) and transitioned to an optic stalk region that, as in AP2α-cre;Pax6^FL/FL embryos, showed no proximal constriction. This suggests that the early phase of Sox2 expression in the surface ectoderm is required for the production of signals that regulate some aspects of optic cup morphogenesis.

Sox2 labeling of AP2α-cre; Pax6^FL/FLembryos revealed that immunoreactivity was retained from E8.5-E10.5(Fig. 4B,E,H). At first glance,this might appear surprising given the loss of Sox2 in the presumptive lens of Le-cre; Pax6^FL/FL embryos at E10.5(Fig. 2K), but is likely explained by arrested lens and eye development in this genotype. In other words, the eye of the E10.5 AP2α-cre;Pax6^FL/FL embryo is developmentally equivalent to an E9.0 eye,and at this pre-placodal stage, Sox2 expression is independent of Pax6. The notion that this represents a developmental arrest is reinforced by the pattern of Chx10 labeling in presumptive retina of AP2α-cre; Pax6^FL/FL embryos. Normally, at E9.0, Chx10 is expressed at a low level in a small domain of the central presumptive retina. This region expands to encompass the entire presumptive retina by E9.5 (Fig. 6K) and E10.5 (Burmeister et al., 1996)(Fig. 6C). By contrast, in E10.5 AP2α-cre; Pax6^FL/FL embryos, Chx10 was observed in a central domain of presumptive retina at the low expression levels (Fig. 4L) characteristic of the E9.0 eye.

The stage dependence of the severity of the Sox2 mutant phenotype is nicely illustrated by the expression of β-crystallin in both conditional mutants (Fig. 5A-C). In control embryos (Fig. 5A), the lower half of the lens vesicle expressedβ-crystallin at E11.5. With AP2α-cre-mediated deletion, only a few differentiated cells could be detected(Fig. 5C), whereas Le-cre-mediated deletion gave an intermediate-sizedβ-crystallin-expressing region (Fig. 5B). Thus, the phenotype is more severe when Sox2 is deleted earlier.

In some AP2α-cre; Sox2^FL/FLembryos, lenses of a reasonable size but abnormal morphology were observed later in development. At E17.5, control eyes showed robust β-crystallin and Prox1 labeling (Fig. 5D,G). By contrast, most AP2α-cre; Sox2^FL/FL eyes (n=6/10) had no morphologically recognizable lenses but did have an occasional ectopicβ-crystallin-positive cell (Fig. 5E,H). The remaining embryos of this genotype (n=4/10)had lenses, albeit of abnormal morphology, that expressed bothβ-crystallin and Prox1 (Fig. 5F,I). In Pax6 conditional mutants generated using either Cre driver, neither morphologically recognizable lenses nor β-crystallin immunoreactivity was ever detected (data not shown)(Ashery-Padan et al., 2000). This is consistent with the idea that Pax6 has the more upstream role in lens development.

Unchanged Sox2 immunoreactivity in the head surface ectoderm of AP2α-cre; Pax6^FL/FL embryos(Fig. 4B) is perhaps in contrast with earlier data examining Pax6^Sey/Sey mutants(Furuta and Hogan, 1998),suggesting that Sox2 expression in this location is dependent on Pax6. This information emerged from Sox2 in situ hybridization studies performed on Pax6^{Sey-Neu/Sey-Neu} embryos at 27 somites (approximately E10). To examine this issue further, we performed immunolabeling on Pax6^Sey/Sey embryos. We confirmed that,at E9.5, Pax6^Sey/Sey embryos showed no nuclear Pax6 immunoreactivity (Fig. 5K;cytoplasmic immunoreactivity was frequently detected but was not greater than background levels). Furthermore, Pax6^Sey/Sey embryos showed Sox2 immunoreactivity at apparently normal levels despite obvious morphological defects (Fig. 5M). Thus, both germline (Pax6^Sey/Sey) and conditional (AP2α-cre; Pax6^FL/FL) Pax6 deletion suggests that prior to E9.5, Sox2 expression in the surface ectoderm and presumptive lens is independent of Pax6.

The possibility that Pax6 and Sox2 cooperate developmentally is raised by their co-expression, by their ability to form a transcription regulation complex (Kamachi et al., 2001)and by the identification of potential binding sites in cis-elements that might mediate cross-regulation (Kondoh et al., 2004; Hever et al.,2006; Inoue et al.,2007). To assess the possibility of Pax6 and Sox2 cooperation in lens development, we generated conditional compound heterozygotes using Le-cre and assessed the phenotypic consequences. In E10.5 Le-cre;Pax6^+/Fl; Sox2^+/Fl embryos, we observed phenotypic variation that ranged from a small but otherwise normal eye (n=10/20;data not shown) to an eye that showed arrested development at an early stage(n=10/20; Fig. 6E-H). Compared with normal E10.5 eyes that showed a lens pit or lens vesicle(Fig. 4A-D), conditional compound heterozygotes that were severely affected showed no placodal thickening and no lens pit or optic cup invagination. The morphology of these mutant eyes most closely resembled that of the Pax6^Sey/Sey(Grindley et al., 1995) or the AP2α-cre; Pax6^Fl/Fl mutants (Figs 3, 4).

Immunolabeling of severely affected Le-cre; Pax6^+/Fl;Sox2^+/Fl embryos revealed that at E10.5, both Pax6 and Sox2 were absent from the surface ectoderm (Fig. 6E,F). Because these embryos are conditional mutants in which Pax6 and Sox2 heterozygote deletion took place approximately one day earlier, the complete loss of Pax6 and Sox2 immunoreactivity probably represents a secondary consequence of loss of the entire lens development program. Chx10 and Mitf are markers for presumptive retina and RPE, respectively, that reveal whether the optic cup has been patterned(Nguyen and Arnheiter, 2000; Horsford et al., 2005). Despite the absence of eye morphogenesis, Chx10 was expressed in central presumptive retina (Fig. 6G,compare with E9.5 control, Fig. 6K), and Mitf in presumptive RPE(Fig. 6H, compare with E9.5 control, Fig. 6L). This indicates that the first steps of optic cup patterning have occurred in Le-cre; Pax6^+/Fl; Sox2^+/Fl embryos, and suggests that a major cooperative function of Pax6 and Sox2 is to signal optic cup morphogenesis.

The adhesion molecule N-cadherin has a complex pattern of expression in the epithelia of the developing eye. At E8.5, N-cadherin was expressed in both the optic pit and the surface ectoderm of the head fold(Fig. 7A). N-cadherin expression was maintained in the optic vesicle at E9.5 but was downregulated in the lens placode region of the surface ectoderm(Fig. 7B). N-cadherin expression was further retained in the presumptive retina and the RPE of E10.5 embryos and was upregulated in the lens pit during invagination(Fig. 7C). Because it has been suggested that N-cadherin is regulated by Sox family members(Matsumata et al., 2005) or Pax6 (van Raamsdonk and Tilghman,2000), we assessed this possibility using the conditional mutants.

AP2α-cre; Pax6^Fl/Fl embryos at E8.5 showed a wild-type distribution of N-cadherin immunoreactivity(Fig. 7E). By contrast, AP2α-cre; Sox2^Fl/Fl embryos at E8.5 had lost N-cadherin expression from the surface ectoderm(Fig. 7F). Even though the N-cadherin signal was not high at this stage of development, this change was consistently detected in an analysis of six AP2α-cre;Sox2^Fl/Fl embryos. Previously, it has been shown that Pax6^Sey heterozygotes have reduced N-cadherintranscript levels in the lens pit (van Raamsdonk and Tilghman, 2000). Because neither the Pax6nor the Sox2 homozygous Le-cre conditional mutants developed a lens pit (Fig. 2), it was not possible to assess N-cadherin expression at this stage in these genotypes. However, we have confirmed that in Lecre and AP2α-cre conditional Pax6 heterozygotes(Fig. 7H,K), N-cadherin levels were reduced in the lens pit. In the case of AP2α-cre;Pax6^Fl/Fl embryos, we quantified this by measuring N-cadherin immunoreactivity and expressing the data as the ratio of presumptive retina to lens pit signal intensity (Fig. 7M); the N-cadherin signal was significantly reduced in the Pax6 mutant (control, n=4; AP2α-cre;Pax6^+/Fl, n=9; P=0.00004). A similar analysis for Sox2 conditional heterozygotes suggested a reduced N-cadherin signal in the lens pit according to observation of micrographs(Fig. 7I,L), but quantification(Fig. 7M) produced only a trend of reduced N-cadherin signal (control, n=4; AP2α-cre; Sox2^+/Fl, n=8; P=0.10). These findings confirm that in the lens pit N-cadherin expression is dependent on Pax6, and leave open the possibility that part of this regulation might be mediated by Sox2 (Fig. 8). The latter suggestion would be consistent with the identification of Sox2-binding sites in N-cadherin enhancers(Matsumata et al., 2005).

To determine whether N-cadherin was an important component of Sox2-dependent lens development, we deleted N-cadherin at the pre-placodal stage by using an existing Ncad^floxconditional allele (Kostetskii et al.,2005; Li et al.,2005). AP2α-cre; Ncad^Fl/Flembryos are difficult to produce at lens development stages owing to an early developmental lethality that probably results from neural tube and heart development defects (Radice et al.,1997). However, in embryos that were viable at E12.5, we could confirm that the normally robust level of N-cadherin in the lens vesicle(Fig. 7N) was absent in the conditional mutants (Fig. 7O,P), even though Pax6 expression(Fig. 7R,S) was retained. Furthermore, in AP2α-cre; Ncad^Fl/Flembryos, as expected (Pontoriero et al.,2009), we observed a failure of the lens vesicle to separate from the surface ectoderm and a persistence of P-cadherin expression(Fig. 7O,P). Even though this phenotype was striking, it was milder that that observed when Sox2was deleted with AP2α-cre (Figs 3, 4) and suggests that N-cadherin is just one of several downstream genes that Sox2 regulates during lens development. Furthermore, because this early deletion of N-cadherin resulted in a phenotype that manifested only at E11.5-E12.5, this suggests that N-cadherin does not have a crucial function in pre-placodal lens ectoderm.

In this analysis, we have assessed the genetic relationship and developmental functions of Pax6 and Sox2 in the early stages of lens development in the mouse. We show that there is an epistatic relationship between Pax6 and Sox2, but that this exists only in a defined developmental window. We also show that, when deleted only in presumptive lens, Pax6 and Sox2 have a cooperative action that regulates lens development but that also initiates morphogenesis in the adjacent optic cup. Finally, we show that N-cadherin is regulated by Sox2 and that, during the placodal phase of lens development when Pax6 regulates Sox2, N-cadherin is also dependent on Pax6. These data raise a number of questions.

Our data show that during lens development there are dynamic changes in the genetic relationship between Pax6 and Sox2. In pre-placodal lens ectoderm, even though there is evidence for a functional cooperation of the gene products, Pax6 and Sox2 transcription is regulated independently. After lens placode formation, the mode of interaction changes to one in which Sox2 expression is dependent on Pax6. This changing relationship is also illustrated by an assessment of N-cadherin expression. In pre-placodal presumptive lens, N-cadherin expression is dependent on Sox2, but not Pax6. After lens placode formation, N-cadherin expression is dependent on Pax6. These data suggest a model in which Pax6 becomes a regulator of Sox2 after lens induction signaling has been initiated(Fig. 8). The N-3 enhancer of Sox2 (Inoue et al.,2007) is a good candidate for mediating Pax6 regulation of Sox2 after placode formation.

In earlier experiments in which Pax6 was conditionally deleted in the presumptive lens with Le-cre(Ashery-Padan et al., 2000),Sox2 immunoreactivity was retained and this contrasts with the current data in which Le-cre deletion of Pax6 results in the loss of Sox2. An explanation for this difference might lie in the genotype of the experimental animals. Previously, Pax6^flox was conditionally deleted on a Pax6 heterozygous[Pax6^lacZ (St-Onge et al., 1997)] background, whereas here we used the homozygous conditional allele. It might be that the Pax6 heterozygous background produces an earlier developmental defect and that Pax6^+/lacZ; Le-cre embryos more closely resemble AP2α-cre; Pax6^flox/flox embryos,in which there is an early developmental arrest and in which Sox2 expression is retained. It will be very interesting to compare the eye transcriptomes of these mutants to understand whether these differences define early steps in lens induction.

The observation that the Pax6-Sox2 genetic relationship changes with developmental stage suggests that an additional level of transcriptional regulation is at play. Specifically, these data indicate that, regardless of whether regulation is direct or indirect, the Sox2 transcriptional control element that mediates Pax6-dependent regulation in the lens pit is inactive at earlier stages. Clearly there are many mechanisms that could explain this switching. Given the stage of development at which this regulatory switching occurs, it might be that optic vesicle-dependent inductive signaling can throw the switch. Switching might be mediated by co-regulator availability or perhaps by chromatin remodeling (Li et al., 2007). Further investigation will be required to gain an understanding of this mechanism.

Pax6 and Sox2 function in lens induction in the context of a larger set of transcription factors and several signaling pathways(Lang, 2004; Medina-Martinez and Jamrich,2007; Kondoh,2008). For example, the Six3 transcription factor is known to be important for the early stages of lens development(Liu et al., 2006) and is likely to function in a positive-feedback loop with Pax6 that would result in the enhanced expression of both (Liu et al., 2006). It has been suggested(Liu et al., 2006) that the mechanism of Six3 regulation of Pax6 is direct binding to the ectoderm enhancer (Williams et al.,1998; Kammandel et al.,1999; Xu et al.,1999). It has also been suggested that Six3 is upstream of Sox2 (Liu et al.,2006) and that, here too, the mechanism is direct transcriptional regulation, in this case, via the N4 enhancer(Uchikawa et al., 2003). The positive regulation of Sox2 by a positive-feedback loop provides a strong rationale for its upregulation during early lens development.

An unexpected finding from these studies was the absence of optic cup morphogenesis when various combinations of Pax6 and Sox2 were deleted in the surface ectoderm. A mild form of this morphogenesis failure is seen in Le-cre; Pax6^flox/flox embryos in which the retina becomes convoluted (Ashery-Padan et al.,2000). When Pax6 is deleted earlier in pre-placodal ectoderm with AP2α-cre, eye morphogenesis fails completely and the phenotype most closely resembles the changes observed in the Pax6^Sey/Sey mice, which are Pax6 germline null(Grindley et al., 1995). This implies that Pax6 expression in the surface ectoderm is required for the production of signals that initiate optic cup morphogenesis, including the epithelial bending that leads to the formation of nested cups of retina and RPE. The distances from the surface ectoderm to the presumptive RPE at the relevant stage of E9.5 are quite large and so, presumably, ectoderm and Pax6-dependent morphogenesis signaling must use a mechanism that can be transmitted or relayed over distance.

The transcription factors Chx10 and Mitf are expressed in the presumptive retina and the RPE, respectively. It has been shown that Chx10 represses Mitf expression and that this is an important element of defining the retinal and RPE territories in the optic cup (Nguyen and Arnheiter, 2000; Horsford et al., 2005). The failure of Chx10 expression to propagate throughout the presumptive retina in AP2α-cre;Pax6^flox/flox embryos might provide an explanation for the failure of optic cup morphogenesis. Specifically, Pax6 in the presumptive lens appears to be required for the lens-to-retina signaling that establishes Chx10 expression. In turn, retinal and RPE territories might remain undefined and this might lead to a failure of region-specific morphogenesis. Further analysis will be required to identify the morphogenesis mechanisms involved.

Our findings were similar when Sox2 was deleted in pre-placodal ectoderm,except that the morphogenesis defects were milder. In AP2α-cre; Sox2^flox/flox embryos, we typically observed E10.5 eyes with optic stalk regions that had not constricted, although in others there were nested cups of retina and RPE. The difference in the severity of eye morphogenesis defects following pre-placodal deletion of Pax6 and Sox2 presumably reflects the degree to which ectodermal morphogenesis signals are dependent on each transcription factor. Clearly, the dramatic eye morphogenesis failure apparent in some Pax6, Sox2conditional heterozygotes nicely illustrates the functional cooperation of the two transcription factors.

An earlier study (Donner et al.,2007) generated mice that were double heterozygotes for Sox2 and Pax6 by using the germline alleles Sox2^βgeo2(Avilion et al., 2003) and Pax6^Sey-Neu (Hill et al., 1991). In contrast to the current data, the analysis of different stages of eye development in Sox2^βgeo2/+; Pax6^Sey-Neu/+ embryos revealed no exacerbation of the Pax6^Sey-Neu/+ small eye phenotype by Sox2 heterozygosity. Although this might seem difficult to reconcile with the current analysis, there may be explanations. One possibility is that with Pax6 and Sox2 germline mutations producing a defect very early in development, the embryo might have the developmental plasticity to accommodate the change without major consequences. The rapid deletion of Pax6 and Sox2 conditional alleles at a later stage of development, as in this analysis, is unlikely to allow compensation due to developmental plasticity. There is also the possibility that the germline and conditional alleles for Sox2 and Pax6 do not produce mutations that are functionally equivalent, especially given the complex gene structure and multiple isoforms of Pax6. Further work will be required to better define these issues.

In humans, Sox2 heterozygosity leads to anophthalmia-esophageal-genital(AEG) syndrome (Taranova et al.,2006; Bakrania et al.,2007). This contrasts with findings in the mouse where Sox2 heterozygosity does not lead to this syndrome or any obvious eye defects(Avilion et al., 2003). We have also observed that in AP2α-cre or Le-cre;Sox2 conditional heterozygotes there are no apparent eye defects. However, the dramatic consequences of conditional Pax6, Sox2 heterozygosity do indicate that, on a sensitized background in which there is only half the normal level of Pax6, Sox2 heterozygosity can lead to anophthalmia. Perhaps individuals with AEG syndrome arise when the genetic variability of the human population provides a sensitized background in which Sox2 heterozygosity can have serious consequences.