The homeobox gene Six3 regulates forebrain development. Here we show that Six3 is also crucial for lens formation. Conditional deletion of mouse Six3 in the presumptive lens ectoderm (PLE) disrupted lens formation. In the most severe cases, lens induction and specification were defective, and the lens placode and lens were absent. In Six3‐mutant embryos, Pax6 was downregulated, and Sox2 was absent in the lens preplacodal ectoderm. Using ChIP, electrophoretic mobility shift assay, and luciferase reporter assays, we determined that Six3 activates Pax6 and Sox2 expression. Misexpression of mouse Six3 into chick embryos promoted the ectopic expansion of the ectodermal Pax6 expression domain. Our results position Six3 at the top of the regulatory pathway leading to lens formation. We conclude that Six3 directly activates Pax6 and probably also Sox2 in the PLE and regulates cell autonomously the earliest stages of mammalian lens induction.
In vertebrates, the lens originates from the presumptive lens ectoderm (PLE). The lens placode, a thickened region of the head surface ectoderm (SE), is the first morphologic sign of lens development. Interactions between the lens placode and the optic vesicle lead to the formation of the lens pits and neuroretina. Lens induction begins in the head SE before contact between the PLE and optic vesicle is established (Grainger et al, 1992, 1997; Sullivan et al, 2004).
The homeobox gene Pax6 is an evolutionarily conserved key regulator of metazoan eye development (Hill et al, 1991; Jordan et al, 1992; Glaser et al, 1992; Quiring et al, 1994; Ashery‐Padan et al, 2000), and Pax6 heterozygosity results in ocular abnormalities such as aniridia in humans (Glaser et al, 1994) and small eye in rodents (Hogan et al, 1986; Hill et al, 1991; Matsuo et al, 1993). Furthermore, in Pax6−/− mice, the lens and other eye structures do not form (Hill et al, 1991; Grindley et al, 1995). Conditional deletion of Pax6 in the mouse lens prevents lens placode formation (Ashery‐Padan et al, 2000).
Lang (2004) has proposed that the preplacodal phase of Pax6 expression regulates the later placodal phase of expression. In this context, BMP and FGF signaling act upstream of the Pax6 pathway controlling lens placode formation (Lang, 2004). Approximately 40 conserved, noncoding sequences potentially involved in Pax6 regulation have been identified in the Pax6 locus (see trafac.cchmc.org), and two Pax6 enhancers, ectoderm enhancer (EE) (Williams et al, 1998; Kammandel et al, 1999; Xu et al, 1999; Dimanlig et al, 2001) and SIMO enhancer (Kleinjan et al, 2001), are important for lens lineage expression. However, the upstream regulators of the preplacodal phase of Pax6 have remained elusive. Meis family proteins regulate Pax6 activity in the PLE (Zhang et al, 2002); however, Meis1‐deficient embryos have partially duplicated retinae and small lenses, which suggests that either Meis2 compensates for Meis1 or other unidentified factors regulate Pax6 in the PLE (Hisa et al, 2004).
Heterozygous mutations in the transcription factor Sox2 that produce hypomorphic conditions occur in about 10% of humans with anophthalmia or severe microphthalmia (Fantes et al, 2003; Hagstrom et al, 2005; Ragge et al, 2005; Zenteno et al, 2005). Sox2 in known to regulate δ‐crystallin expression in the chick in a complex with Pax6 (Kamachi et al, 2001). Sox2‐null mice die at preimplantation (Avilion et al, 2003), thereby preventing analysis of this gene's role(s) in eye development. Sox2 expression is absent in mutant mouse strains with severe lens defects such as Bmp4, Pax6sey/sey and Bmp7 (Furuta and Hogan, 1998; Wawersik et al, 1999).
Sox2 expression precedes lens placode formation (Furuta and Hogan, 1998) and its upregulation in the PLE is one of the earliest indicators of lens specification and is dependent on Pax6 (Furuta and Hogan, 1998; Zygar et al, 1998). Upregulation of Sox2 was not detected in the SE of Pax6Sey‐1Neu‐homozygous embryos in which lens induction is defective (Furuta and Hogan, 1998). However, the early unaffected phase of Pax6 expression in the SE of Pax6‐Le conditional mutant embryos was sufficient to upregulate Sox2 expression in this tissue (Ashery‐Padan et al, 2000), a finding that indicates that upregulation of Sox2 in the PLE is dependent on early, but not late Pax6 activity.
Six3 protein expression was reported to start at around E9.5 in the vertebrate lens placode (Oliver et al, 1995), and forced Six3 expression in fish and mouse embryos promotes ectopic lens (Oliver et al, 1996) and neural retina formation (Loosli et al, 1999; Lagutin et al, 2001). Until recently, investigating the in vivo role of Six3 in eye development was not possible, because the prosencephalon, the brain region from which eyes form, is severely truncated in Six3−/− mice (Lagutin et al, 2003). However, by using a Cre/loxP approach to inactivate Six3 in the developing eye, we can now address this question.
Cre/loxP‐mediated removal of Six3 activity from the PLE
To address the in vivo role of Six3 during lens morphogenesis, we generated the floxed Six3 mouse line Six3f/w (nomenclature and details in Supplementary Figure S1). During the initial crosses of Six3f/w mice with the Le‐Cre mouse strain (Ashery‐Padan et al, 2000), the transgenic Cre strain was occasionally germline active. This allowed us to generate the Six3Δ allele in which the floxed sequence encoding the Six domain and homeodomain of Six3 was permanently deleted (Supplementary Figure S1). The generation of the Six3Δ/Δ mice confirmed that the Six3Δ allele was a true null allele; the prosencephalon truncation in Six3Δ/Δ mice was indistinguishable from that previously reported in Six3−/− mice (Lagutin et al, 2003). Furthermore, antibody staining of the generated Six3‐mutant lens failed to detect any signal (see below) when we used an anti‐Six3 antibody (our unpublished data) that recognizes a Six3 epitope downstream of the floxed DNA sequence.
Six3Δ/w;Le‐Cre mice were generated by crossing Six3Δ/w mice with Le‐Cre mice in which Cre activity in the PLE was reported to start at around E9.0 (Kammandel et al, 1999; Ashery‐Padan et al, 2000). Six3f/f mice were generated by intercrossing Six3f/w mice. Finally, Six3f/Δ;Le‐Cre mice were obtained by crossing Six3f/f females with Six3Δ/w;Le‐Cre males. Six3f/w embryos, which were indistinguishable from wild‐type embryos, were used as controls.
Six3 mediates lens morphogenesis
Gross morphological analysis of the generated Six3f/Δ;Le‐Cre mice revealed several lens phenotypes, including drastically reduced lens size, cataracts, or absence of the lens (Figure 1A–D, A′–D′). These defects were even more obvious in the analysis of lens sections. Three major phenotypes were observed at any particular stage: mild (type I), moderate (type II), and severe (type III). A summary of this analysis is included in Supplementary Table 1.
Defects in lens development were evident as early as E10.5. Reduction in the size of the invaginating lens pit was observed in mildly (not shown) and moderately affected mutant embryos (Figure 1F and G). Embryos with the severe (type III) lens phenotype showed minimal thickening and no invagination of the PLE (Figure 1H). The shape of the optic vesicle was also defective (Figure 1H), although Six3 was not removed from it (Figure 3B′ and E′). Conditional deletion of Pax6 from the PLE also resulted in defective invagination of the optic vesicle (Ashery‐Padan et al, 2000). Variability in the severity of lens defects was also observed at later stages. At E12.5, the lens vesicle was small but relatively normal in mildly affected mutant embryos (Figure 1J), smaller and abnormal in others (Figure 1K), and totally absent in type III embryos (neuroretina was also deformed; Figure 1L). At E14.5, some embryos had an abnormally persistent lens stalk and disorganized lens fibers (Figure 1N), and others had a very small and abnormal lens (Figure 1O and P). Thus, removal of Six3 from the PLE during lens induction disrupts lens formation.
The time and extent of Le‐Cre‐mediated Six3 excision from the PLE is variable
As mentioned before, Le‐Cre activity in the PLE was reported to start around E9.0. (Kammandel et al, 1999; Ashery‐Padan et al, 2000). To determine whether the variability in the severity of the lens phenotype reflects differences in the timing and extent of Six3 excision from the PLE, we characterized the temporal and spatial activity of Le‐Cre by using the R26R strain (Soriano, 1999). We found that the time and extent of Le‐Cre activity was variable (Supplementary Figure S2). At around the 22‐somite stage, the most common and rather specific Cre activity was detected in the lens SE (Supplementary Figure S2A and B). In addition, a less common and more variable phase of Le‐Cre activity was also identified in the putative lens SE as early as the 7‐ to 11‐somite stage (Supplementary Figure S2C–I); variations in the time and extent of Le‐Cre were observed between the left and right eyes of individual embryos (Supplementary Figure S2C and D). X‐gal+ cells were also found in mesenchymal and neural cells (Supplementary Figure S2G–I), and in a few cases, X‐gal‐expressing cells were ubiquitously distributed along most of the embryo (Supplementary Figure S2F). Consistent with these results, we also found variability in the time and extent of Cre protein expression in the PLE; extensive Cre expression was seen in the PLE of a 16‐somite‐stage embryo (Supplementary Figure S2J); in contrast, fewer positive cells were seen in the same region of a somite‐matched littermate (Supplementary Figure S2K).
According to these results, the variability in the time and extent of Six3 removal from the PLE (compare Figure 3B′ with Supplementary Figure S3D and G) and, therefore, the severity of the Six3 lens phenotype are probably direct consequences of the variable onset of Le‐Cre activity. In some E9.5 Six3‐mutant embryos, high levels of Six3 remained in the PLE (Supplementary Figure S3D and G); therefore, the levels of Pax6 and Sox2 expressed in the PLE were normal or only mildly reduced (Supplementary Figure S3E and F). These differences could account for the milder (types I and II) phenotypes observed in some Six3‐mutant embryos at later stages.
In this experimental setting, Six3 deletion from the PLE probably ranged from about the 7‐somite stage to E9.5; from E9.0 to E9.5 for most of the characterized less severe lens mutant embryos; and starting from the 7‐somite stage onward for the reduced percentage of embryos that exhibited the most severe lens phenotype.
Lens differentiation is not affected in mildly or moderately affected Six3‐mutant lenses
Six3 mutants with mild lens phenotypes (types I and II) have lens pits and lens vesicles, indicating that lens placode formation was not severely affected. This conclusion was supported by the normal expression of the early lens placodal markers Pax6, Sox2, Mab21l1, sFrp2, and FoxE3 at E10.5 (data not shown), and FoxE3 (Figure 2A′) and Prox1 (Figure 2B′) at E12.5. Lens differentiation also appeared unaffected, as shown by the normal expression of β‐crystallin (Figure 2C′). However, an increase in cell death was observed in the lens pit at E10.5 (see Figure 2D′), and although the percentage of BrdU+ cells was unaffected at this stage (Figure 2E′), at E12.5 proliferating cells were abnormally located in the posterior lens compartment (Figure 2F′). Therefore, in moderately affected Six3‐mutant lenses, cells in the posterior margin do not exit the cell cycle, and lens polarity is probably not established, as also indicated by the ectopic expression of FoxE3 in that region (Figure 2A′).
Lens induction/specification is defective in the severely affected Six3‐mutant lens
Our studies focused on the severe lens phenotype (type III), which most likely resulted from early (7‐somite stage onwards), extensive removal of Six3 activity from the PLE. Le‐Cre activity was already abundant in the PLE of Six3‐mutant embryos at E9.5 (Figure 3A) and started decreasing at E10.5 (Figure 3A′). Lens specification normally occurs at around E9.0–E9.5. In severely affected embryos, Six3 expression was already removed from the mutant PLE at E9.5 (Figure 3B′); Six3 expression in the neural retina remained unaffected. Pax6 expression was drastically reduced (Figure 3C′), and Sox2 expression was absent in the mutant PLE (Figure 3D′). The few Pax6‐expressing cells remaining in some mutant PLE were most likely remnants of Pax6 that was expressed before Cre‐mediated deletion of Six3.
Consistent with these results, Six3 expression was not detected in the mutant PLE at E10.5 (Figure 3E′) but was normal in the neural retina, and no lens‐like structures were present, as indicated by the absence of thickening and invagination of the SE. In adjacent sections, Pax6 expression was barely detectable (Figure 3F′) in the mutant SE, and Sox2 was absent (Figure 3G′). Consequently, expression of downstream markers of the invaginating lens pit such as Prox1 (Figure 3H′), sFrp2 (Figure 3I′), and FoxE3 (Figure 3J′) was also undetected in the Six3‐mutant lens. Expression of Meis1 (Figure 3K′) and Meis2 (data not shown), which directly regulate Pax6 expression in the lens placode (Zhang et al, 2002), appeared unaffected.
Six3 is a direct transcriptional regulator of Wnt1 in the developing mouse forebrain (Lagutin et al, 2003), and the Wnt/β‐catenin signaling pathway mediates mouse lens development (Smith et al, 2005). To determine whether alterations in the Wnt/β‐catenin pathway influenced lens formation in Six3‐mutant embryos, we used the available Bat‐Gal reporter mouse that expresses lacZ in response to activation of the canonical Wnt pathway (Maretto et al, 2003). Control and Six3‐mutant E9.5 and E10.5 PLE isolated from crosses of Bat‐Gal and Six3f/Δ;Le‐Cre mutant mice were both negative for X‐gal staining (data not shown). In addition, no nuclear localization of β‐catenin was observed in the SE of mutant embryos (data not shown).
These results indicate that the lens phenotypes resulting from the specific removal of Six3 are probably not related to defects in the Wnt/β‐catenin pathway. However, the drastically reduced expression of Pax6 and the absence of Sox2 in the type III Six3‐mutant PLE indicate that lens induction/specification are defective in these severely affected embryos. Thus, Six3 activity in the PLE is essential during lens induction.
The time of Six3 removal determines the severity of the lens phenotype
To remove Six3 activity at different developmental stages as efficiently as possible but without compromising its role in forebrain formation (Lagutin et al, 2003), we used the CAGG‐CRE‐ER mouse strain in which ubiquitous Cre activity is induced by tamoxifen (Hayashi and McMahon, 2002). Floxed Six3 (Six3f/f) females were bred with Six3Δ/w;CAGG‐CRE‐ER males, and two injections of tamoxifen were given to the pregnant dams at approximately E7.8 and early E8.5 for early Six3 deletion, and at E9.5 and E10.0 for late deletion; embryos were then isolated at E10.5. Both protocols deleted Six3 activity efficiently (Figure 4A′ and G′), although defective forebrain patterning was observed in some of the early‐injected embryos, which were then no longer considered for the analysis of the lens.
No thickening or invagination of the PLE was seen at least in one side of six of the seven E10.5 mutant embryos in which tamoxifen was administered early (E7.8 and E8.5; Figure 4A′ and B′). Pax6 and Sox2 expression were not detected in the PLE but were observed at lower levels in the optic cup (Figure 4C′ and D′). The expression of FoxE3, sFrp2, and Prox1 was not detected in the mutant PLE (data not shown), but that of Meis1 (Figure 4E′) and Meis2 (data not shown) was unaffected. Thus, lens specification was defective, a phenotype similar to that of the severely affected type III embryos described above when using Le‐Cre. Interestingly, although the shape of the developing optic cup was abnormal, typical retinal progenitor markers (Pax6, Sox2, Chx10) were still detected (Figure 4C′, D′ and F′).
Lens specification was not obviously affected (i.e., the lens pit was normal or slightly smaller) in E10.5 mutant embryos (n=7/7) in which tamoxifen was administered later (E9.5 and E10.0; Figure 4G′). Pax6 expression appeared normal in the mutant lens pit (Figure 4H′), and that of Sox2 was normal or slightly reduced (Figure 4I′), characteristics reminiscent of those described for the types I and II Six3‐mutant lenses when using Le‐Cre. A graphic representation of these phenotypes is presented in Figure 4J.
To complement these studies, we also examined early deleted (tamoxifen injection at approximately E7.8 and early E8.5) Six3‐mutant embryos at around E9.0 (18‐somite stage, prior to the upregulation of Sox2 in the PLE). In two independent Six3‐mutant embryos (Figure 5A′ and D′), only a few Six3‐expressing cells remained in the PLE, and Six3 expression was efficiently deleted in the evaginating optic cup (bilateral differences in the rate of Six3 deletion and severity of the lens phenotype were occasionally observed). In adjacent sections, Pax6 expression was consistently reduced in the mutant PLE (Figure 5B′ and E′), although at different levels; it was almost undetectable in some embryos (Figure 5E′) and reduced in others (Figure 5B′). Normally, low levels of Sox2 expression are detected in the PLE at this early stage (Figure 5C and F); however, Sox2 expression was further reduced in the PLE of one of the analyzed mutant littermates (Figure 5C′) and not detected in the PLE of the second one (Figure 5F′). Our results suggest that the downregulation of Pax6 and consequent lack of upregulation of Sox2 in the PLE results from the removal of Six3 at early stages.
Together, these results argue that the time of Six3 deletion is critical for the lens phenotype, and early Six3 deficiency is responsible for the failure in lens specification. Early Six3 deletion drastically downregulates Pax6 and Sox2 expression in the PLE, and late Six3 deletion minimally affects their expression.
Six3 expression in the PLE precedes that of Pax6
We have now demonstrated that early conditional removal of Six3 from the PLE downregulates Pax6 preplacodal expression and arrests lens formation; the resulting lens phenotype is similar to that of Pax6−/− embryos (Grindley et al, 1995). Therefore, the Six3‐promoted lens phenotype may be caused by Six3's direct regulation of preplacodal Pax6 in the head SE. To support this proposal, Six3 expression in the presumptive lens SE must start as early as (if not before) that of Pax6. To test this, we characterized the expression profiles of the two proteins starting at around the 5‐somite stage (E8.0). Besides strong expression in the anterior neuroectoderm, Six3 was also detected in the head SE that will become lens ectoderm (Figure 6A); barely detectable levels of Pax6 were seen in the adjacent sections (Figure 6A′). Around the 7‐ to 8‐somite stage, Pax6 become visible in the PLE (Figure 6B′); at around E8.5–9.5, Six3 and Pax6 were both localized in the PLE (Figure 6C and C′), lens placode, and optic vesicle (Figure 6D and D′).
Six3 PLE expression at around early E9.5 is initially independent of Pax6; later, it becomes Pax6‐dependent (Purcell et al, 2005). The finding that Six3 expression in the head SE precedes that of Pax6 suggests that Six3‐mediated upregulation of Pax6 is a necessary early step in the commitment toward a lens fate. Supporting this argument, Six3 expression in the PLE was not overtly affected in 6‐ or 8‐somite‐stage Pax6Sey‐1Neu/Sey‐1Neu‐mutant embryos (Figure 6E′ and F′), but was downregulated at later stages (Figure 6G′) and was absent at around E9.5–E10.5 (Figure 6H′). In contrast, Six3 expression in the neural retina was unaffected.
We conclude that Six3 activity in the SE precedes that of Pax6, and that Pax6 activity, although not necessary to induce Six3 expression, is later required for Six3 maintenance and upregulation in the PLE.
Six3 binds to Pax6 and Sox2 lens enhancers
Pax6 deficiency causes a lens phenotype (Grindley et al, 1995) similar to that of the severe type III Six3 mutants. This similarity led us to hypothesize that Six3 directly activates Pax6 expression during early lens morphogenesis (i.e., during lens induction). In addition, it has been suggested that Sox2 is downstream of the Pax6 preplacodal phase (Furuta and Hogan, 1998). Sox2 expression was absent from the PLE of type III Six3‐mutant lenses. In this context, Six3 could also directly or indirectly activate Sox2 expression in the PLE.
During lens formation, Pax6 is regulated by the EE (Williams et al, 1998; Kammandel et al, 1999; Xu et al, 1999) and SIMO enhancers (Kleinjan et al, 2001). In the case of Sox2, three lens enhancers (N3, N4, L) have been identified in the chick (Uchikawa et al, 2003); however, only N3 and N4 function during lens specification. We have identified a consensus core ATTA motif as the typical Six3 DNA‐binding motif (Zhu et al, 2002). Visual inspection of the available DNA sequences of EE, SIMO, N3 and N4 enhancer regions identified multiple ATTA motifs (Supplementary Figure S4B–E).
To determine whether Six3 binds Pax6 and Sox2 lens enhancers in vitro, we performed an electrophoretic mobility shift assay (EMSA) using DNA fragments representing EE, SIMO, N3, and N4. Full‐length Six3 protein specifically retarded the four different enhancer elements (Figure 7A). The binding specificity was reflected by the reduced amount of complex that formed when nonradioactive probes were added, and a super‐shift of the GST‐Six3 probe complex was seen when using an anti‐Six3 antibody (Figure 7A). The presence of more than one retarded band for each enhancer suggests that more than one Six3‐binding site is present in each element.
To determine whether Six3 directly binds to Pax6 and Sox2 regulatory elements in vivo, we initially performed ChIP assays on the prospective head (Six3‐expressing) and trunk (Six3‐free) regions of E8.5 (11‐ to 17‐somite stage) wild‐type mice. TaqMan‐based real‐time PCR was used for quantification (details in Supplementary data). For EE, SIMO, and N4, the ratio for the head was significantly higher (>5) than the ratio for the trunk (1.20–2.97), indicating that Six3 protein was bound to these enhancers when using chromatin derived from the Six3‐expressing head region (Figure 7B). For the N3 enhancer, there were no considerable differences in the ratio between heads and trunks, indicating that Six3 did not bind to it. This result could be due to the fact that in the chick, N3 is normally active around the lens placode stage (equivalent to approximately 25‐somite stage in mice) (Uchikawa et al, 2003), and the chromatin sample used for the ChIP assay was from 11‐ to 17‐somite mouse embryos.
Mouse Six3 is a transcriptional repressor (Kobayashi et al, 2001; Zhu et al, 2002; Lopez‐Rios et al, 2003); however, transcription factors often behave as activators or repressors, depending on the context. To determine whether Six3 functions as a transcriptional activator in the PLE, we generated a Pax6 lens lineage luciferase reporter by inserting a luciferase‐coding fragment under the control of Pax6 EE. Cotransfection with a pCAB‐Six3 expression plasmid activated the Pax6 lens lineage reporter (Pax6‐EE‐luc) (Figure 7C); pCAB‐Six3VP16 expression plasmid enhanced the activation more than two‐fold, and the repressor form of Six3 (pCAB‐Six3EnR) attenuated it. Furthermore, reporter activation was attenuated when the Pax6 EE (Pax6EE) was missing (Figure 7D). Similar experiments using DNA fragments from the SIMO enhancer did not produce consistent results (data not shown). Therefore, at least in the context of Pax6 EE, Six3 activates Pax6 lens lineage expression.
A similar approach was used with N3 and N4 enhancer elements, except that the pRL‐TK reporter plasmid was selected, as carried out in the original Sox2 enhancer study (Uchikawa et al, 2003). Cotransfection with the pCAB‐Six3 expression plasmid slightly increased N4 lens lineage reporter and slightly decreased N3. The pCAB‐Six3VP16 expression plasmid slightly enhanced the activation of N4, and pCAB‐Six3EnR did not have any obvious effect (data not shown). These results suggest that Six3 activates Sox2 expression, although the two lens enhancers could respond differently.
Electroporation of mouse Six3 promotes ectopic Pax6 expression in the chick SE
Our data suggest that Six3 specifically activates preplacodal Pax6 expression in the PLE during the early‐somite stage. To test this hypothesis further, we electroporated a full‐length Six3 expression construct (pCAB‐Six3‐IRES‐GFP) or GFP (pCAB‐IRES‐GFP) into one side of the ectoderm, posterior and lateral to the PLE of Hamburger–Hamilton (HH) stage 8 chick embryos (4‐ to 5‐somite stage). After overnight culture, we assessed the expression of Pax6 and Sox2 and performed GFP immunostaining to visualize the electroporated cells. Misexpression of Six3, but not that of GFP, led to an enlarged Pax6 domain on the electroporated side of the SE (n=7/10 embryos; arrow in Figure 7F). Ectopic Pax6 expression was not induced in the trunk, and ectopic expression of Sox2 was never observed (n=0/10; data not shown). Similar expansion of Pax6 expression and no change in Sox2 expression were observed when an activated form of Six3 was electroporated (pCAB‐Six3VP16) (data not shown, n=10/20). Misexpression of Six3‐EnR did not overtly affect Pax6 expression (data not shown, n=21).
Six3‐mediated activation of Pax6 lens lineage expression was highly temporally specific. When HH9–HH10 embryos were electroporated and analyzed 24 h later (when the lens vesicle had formed), we never observed ectopic Pax6 expression in the Six3‐electroporated cells of the SE. Thus, Six3 directly activated Pax6 preplacodal expression in the PLE in vivo most likely only during a very brief period of early lens development.
In this paper, we identified Six3 as a critical early regulator of Pax6 and Sox2 expression during lens induction and specification. Conditional removal of Six3 activity in the PLE promoted a range of eye defects; in the most severe cases, no lenses were formed. We also showed that Six3 expression in the head SE precedes that of Pax6 and that Six3 expression is initially not affected in the SE of Pax6−/− small eye‐mutant embryos.
Six3 activity in PLE is essential for lens specification
A key consequence of removing Six3 from the PLE at the early somite stage was the lack of thickening and invagination of the PLE, the efficient downregulation of Pax6 expression, and the complete removal of Sox2 expression from the PLE of the most severely affected embryos. This finding suggests that, similar to the Pax6 small eye−/− phenotype, the identified defective upregulation of Pax6 and Sox2 in the PLE of the Six3‐mutant embryos affects the acquisition of the critical expression threshold of genes necessary to regulate the onset of lens placode formation. Ultimately, this defective regulation results in defective lens induction/specification (i.e., type III lenses).
As mentioned above, Pax6 maintains its own expression in the PLE (Grindley et al, 1995), and Pax6 activity is necessary to upregulate that of Sox2. In the generated Six3‐mutant embryos, the reduction in the levels of Sox2 expression was more sensitive to the removal of Six3 activity than that of Pax6. This result suggests that in addition to the Pax6‐dependent regulation of Sox2, upregulation of Sox2 expression in the PLE also requires activation by Six3 in a Pax6‐independent manner.
Phenotypic variability in Six3‐mutant lenses
The generated Six3‐mutant embryos display a range of lens phenotypes: lens specification failed in the most severely affected mutants, but was normal in the milder ones. This variability most likely reflects differences in the time and extent of Six3 deletion, which ultimately affects the rate and time of Pax6 and Sox2 downregulation in the PLE: early deletion of Six3 causes the failure in lens specification, but deletion at later stages results in milder lens phenotypes. In support of this argument, we report that in the genetic background used for our analysis, there is variability in the timing and extent of Le‐Cre activity and in the rate and efficiency of Six3 deletion (as a consequence also in the extent of Pax6 and Sox2 downregulation).
Spatial and temporal variation in the extent of Sox2 downregulation is seen when comparing standard Pax6‐null mice (Grindley et al, 1995) with conditional Pax6‐null animals (Ashery‐Padan et al, 2000); however, no variability in the lens phenotype has been reported in the generated Pax6‐conditional mutants in which Le‐Cre was also used (Ashery‐Padan et al, 2000). It could be argued that as indicated by the authors, the reported phenotype (no lens placode formation but normal Sox2 expression; Ashery‐Padan et al, 2000) is the result of late Pax6 deletion by Le‐Cre (E9.0–E9.5). However, based on our analysis of the Le‐Cre activity profile using the R26R reporter strain, early removal of Pax6 activity and therefore complete downregulation of Sox2 expression could have also taken place in a reduced but undetected percentage of the generated Pax6f/f;Le‐Cre embryos (assuming their genetic backgrounds were similar to the ones we used for these studies).
A requirement of a timely critical threshold of Pax6 expression for lens placode formation has already been suggested (van Raamsdonk and Tilghman, 2000; Davis‐Silberman et al, 2005), and specific conditional deletion of a single copy of Pax6 from the developing lens results in a smaller lens (Davis‐Silberman et al, 2005). This dosage sensitivity may be more relevant during early stages of lens placode formation, because the proposed Pax6 autoregulatory loop maintains Pax6 expression starting at this stage (Grindley et al, 1995). Therefore, the fewer the number of Pax6‐expressing cells that remain at the preplacodal stage, the more severe the resulting lens phenotype will be.
According to this, the mildly affected Six3 mutant lens (type I and type II) are most likely the consequence of less extensive and/or later deletion of Six3 in the PLE, which in turn lead to milder downregulation of Pax6 (similar to that of Pax6 small eye+/– mice that have smaller lenses) and Sox2 levels. It should be mentioned that we did not find any lens phenotype in the Six3‐heterozygous lenses (data not shown), a result indicating that one functional allele of Six3 is sufficient for normal lens specification and that more than 50% of Six3 should be depleted to promote Pax6 or Sox2 downregulation.
Six3 is a direct activator of Pax6 expression
Results from the analysis of the generated Six3‐mutant embryos suggested that Six3 directly regulates Pax6 and Sox2 expression in the PLE. Support for this proposal was provided by a combination of in vitro and in vivo experimental assays.
Pax6 EE is located about 3.9‐kb upstream of the Pax6 P0 promoter (Kammandel et al, 1999) and can direct reporter expression to the PLE from E9.0 onward (Kammandel et al, 1999; Ashery‐Padan et al, 2000). However, lens development is not completely blocked in EE‐deficient mice (Dimanlig et al, 2001); this finding supports the proposal that Pax6 expression in the lens is mediated by multiple regulatory elements. Pax6 expression in the lens placode of EE‐deficient embryos is reduced, and these embryos exhibit a relatively mild lens phenotype (i.e., reduced placodal cell proliferation and a small lens pit and lens vesicle) (Dimanlig et al, 2001). The Pax6 SIMO enhancer was identified because patients with aniridia have mutations in this regulatory element (Kleinjan et al, 2001). The SIMO enhancer also directs reporter expression in the lens lineage starting at around E9.5 (Kleinjan et al, 2001). However, given that Pax6 expression in the PLE can be detected as early as the 7‐ to 8‐somite stage (E8.0), additional regulatory elements, by themselves or in cooperation with the EE or SIMO enhancers, are probably responsible for the early preplacodal phase of Pax6 expression.
The Sox2 N4, N3, and L enhancers were identified by extensive sequence analyses and reporter assays in chick embryos (Uchikawa et al, 2003).
We have now demonstrated in vitro that Six3 directly binds Pax6 EE and SIMO enhancers and Sox2 N3 and N4 enhancers. Furthermore, we demonstrated in vivo that Six3 binds the EE, SIMO and N4 enhancers during early lens specification. The lack of enrichment of Six3 in Sox2 N3 could be due to the fact that the embryonic stage used for the ChIP analysis was earlier than the stage at which Sox2 N3 was reported to be active in the chick (Uchikawa et al, 2003). Those data demonstrate that Six3 directly regulates Pax6 and Sox2 in vivo, although activation of Sox2 by Six3 might be in cooperation with additional factors.
Six3 can act as a transcriptional repressor (Kobayashi et al, 1998; Zhu et al, 2002; Lopez‐Rios et al, 2003). Yet, injection of Six3 RNA causes the ectopic expansion of Pax6 expression into the prospective midbrain and cerebellum of medaka fish (Loosli et al, 1999), which suggests that Six3 also activates Pax6 expression in some contexts. Results from our luciferase reporter assay have now determined that Six3 activates the Pax6 lens lineage expression. Furthermore, our chick electroporation studies demonstrated that the Pax6 expression domain expanded in the electroporated SE of HH8 chick embryos; no expansion was seen when HH9–HH10 embryos were used. Together, these findings support the proposal that Six3 directly regulates the expression of Pax6 (and most likely also of Sox2) during a brief, crucial period of early lens development, that is, during the preplacodal stage. Later on, Pax6 expression appears to be independent of Six3.
A working model of lens induction
Six3 expression in the PLE was first detected at around the 5‐somite stage. We propose that during lens development two phases of Six3 activity occur, and it is the initial phase of Six3 expression in the PLE that is crucial for the regulation of lens induction and specification. Based on our results, Six3 can act as either an activator or repressor, depending on the cellular context and its specific protein partners; the early phase of Six3 activity that precedes that of Pax6 is necessary to activate Pax6 and Sox2 expression in the PLE (Figure 7G).
In addition, there are evidences to suggest that these three genes regulate each other in a complex regulatory network (Loosli et al, 1999; Goudreau et al, 2002; Aota et al, 2003). Our analysis of Pax6Sey‐1Neu/Sey‐1Neu‐mutant embryos (Figure 6) provides further support to this proposal. In this case, although initial Six3 expression in the PLE is unaffected, the lack of Pax6 activity in the PLE fails to maintain Six3 expression in that structure during later stages (it is not yet know whether maintenance of Six3 expression requires direct activation by Pax6). This results in the absence of the lens placode and failure in lens formation.
In summary, we propose that as lens development progresses, the initial downregulation of preplacodal Pax6 expression promoted by the lack of Six3 activity leads to the eventual failure of Pax6 own activation and its maintenance in the PLE. Absence of Six3 and Pax6 activity in the mutant PLE results in the defective upregulation of Sox2 expression in this tissue. Altogether, these defects lead to the failure in lens induction and specification observed in the most severe Six3‐mutant embryos.
Materials and methods
Descriptions of the generation of the Six3‐targeting construct and mouse strains used are presented as Supplementary data.
Immunohistochemistry and in situ hybridization
Standard procedures were used. The list of used primary antibodies and in situ probes is provided in the Supplementary data.
BrdU and TUNEL assays
Time‐mated female mice were injected with BrdU (0.1 mg/g. body weight, intraperitoneally), and embryos were harvested 1.5 h later. Anti‐BrdU monoclonal antibody (1:10; BD Biosciences) was used. The TUNEL assay was performed using the ApopTag Kit (Chemicon), according to the manufacturer's instructions. For counting of BrdU+ cells and TUNEL+ cells, sections through the middle of the lens pits were used. The results represent the average percentage of three embryos.
The ChIP Kit (Upstate) protocol was followed, except that the nuclear fraction was used. Briefly, 13 E8.5 wild‐type embryos were harvested. Heads (Six3‐expressing) and trunks (Six3‐negative) were dissected, and nuclear fractions were subjected to separate ChIP assays. Half of the chromatin was immunoprecipitated with the Six3 antibody (4 μl). Chromatin was purified by QIAquick PCR Purification Kit (Qiagen) and suspended in 50 μl EB buffer (Qiagen). A 2‐μl sample of DNA from each of the four samples (head and trunk, with and without Six3 immunoprecipitation) was used for TaqMan real‐time PCR (ABI7900HT) analysis (Supplementary data).
DNA sequence analyses
To generate the EMSA templates, PCR‐amplified DNA fragments containing selected regions from Pax6 EE, Pax6 SIMO, Sox2 N3, and Sox2 N4 were first cloned into the pGEM‐T easy plasmid (Promega) and verified by DNA sequencing. DNA sequences of the primers used are provided in the Supplementary data and in Supplementary Figure S4.
For the luciferase assay, the mouse Pax6 lens lineage reporter plasmids Pax6‐EE‐luc and Pax6‐ΔEE‐luc were generated by replacing the lacZ‐coding sequence from the P0‐3.9‐lacZ parental plasmid (Williams et al, 1998) with the luciferase reporter gene. For Sox2, a conserved 1380‐bp DNA region that included Sox2 N3 enhancer was generated by PCR using the following primers:
5′‐AGATCTCTGAGCTTACCCCTGCACTC‐3′ and cloned into the BglII site of the pRL‐TK (Promega) reporter plasmid. For Sox2 N4, the same amplified DNA fragment (502 bp) used for the EMSA assay was cloned into the BglII site of pRL‐TK.
Effectors for the luciferase assay were made as follows: the plasmids pCAB‐Six3‐IRES‐GFP, pCAB‐Six3VP16‐IRES‐GFP, and pCAB‐Six3EnR‐IRES‐GFP were obtained by inserting Six3 cDNA, Six3VP16, and Six3EnR, respectively, into blunted BamHI–ClaI sites of the pCAB‐IRES‐GFP vector (McLarren et al, 2003).
NotI digestion was used to release the four DNA templates from their parental plasmids. Klenow enzyme was used to end‐label these double‐stranded DNA fragments with [α‐32P]dCTP. The EMSA was then performed as described in the Supplementary data.
Cos7 cells were maintained in culture at 37°C in DMEM with 10% FBS, 50 μg/ml penicillin–streptomycin, and 2 mM glutamine. Cos7 cells were seeded into 24‐well plates and transfected with the proper reporter/effector mixtures using Fugene 6 (Roche) according to the manufacture's instructions. Details of the Pax6 and Sox2 reporter luciferase assays are presented in the Supplementary data.
Fertilized chicken eggs (Henry Stewart, UK) were incubated at 38°C in a humidified incubator for 26–36 h, until they reached HH8–HH10 stage. Overexpression of pCAB‐Six3‐IRES‐GFP in HH8 embryos was performed in modified New culture as described previously (Stern and Ireland, 1981) using 1–2 mg/ml DNA and four 50‐ms pulses of 7 V (McLarren et al, 2003). Embryos were grown for 12–14 h, and the expression of Pax6 and Sox2 was analyzed by whole‐mount in situ hybridization followed by immunostaining with polyclonal anti‐GFP antibodies (Molecular Probes). HH9–HH10 embryos were electroporated in ovo as described previously, with the exception that the electrical stimulus was increased to 20 V (Bhattacharyya et al, 2004). After 24 h, the expression of β‐crystallin and Pax6 was analyzed using the same methods described for the younger chick embryos.
Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).
Supplementary Figure S1
Supplementary Figure S2
Supplementary Figure S3
Supplementary Figure S4
Supplementary Table 1
Supplementary Figure Legends
We thank P Gruss and R Ashery‐Padan for the Le‐Cre mice, A McMahon for the CAGG‐Cre‐ER mice, B Hogan for Pax6Sey‐1Neu/+ mice, and S Piccolo for BAT‐gal mice; for antibodies, M Torres (Meis1 and Meis2), J Zigler (β‐crystallin), and T Jessell (Chx10); and for probes and plasmids, N Takahashi (Mab21l1), M Jamrich (FoxE3), R Maas (P0‐3.9‐LacZ). We also thank Y Lee for help with real‐time PCR, M Dillard and M Self for excellent technical assistance, and C Zhu for reagents and advice. We thank B Sosa‐Pineda, G Grosveld, and members of G Oliver's lab for helpful discussions and A McArthur for editing the manuscript. This work was supported in part by the NIH Grant Y12162 to GO, Cancer Center Support Grant CA‐21765, the American Lebanese Syrian Associated Charities (ALSAC), and the BBSRC (G20323 and D010659/1) to AS.
- Copyright © 2006 European Molecular Biology Organization