Stage-dependent requirement of neuroretinal Pax6 for lens and retina development

Proper eye development is dependent on the coordinated formation of two main tissues in the eye: the retina and the lens. Vertebrate eye development begins with invagination of the optic vesicle (OV) toward the lens-competent head surface ectoderm (SE). As OV contacts SE, a series of reciprocal inductive signals elicit formation of the lens placode (LP) and subsequent invagination of both LP and OV to form a two-layered optic cup (OC), with retinal pigmented epithelium (RPE) surrounding the retina (reviewed by Fuhrmann, 2010; Chow and Lang, 2001; Ogino et al., 2012). Genetic studies have identified multiple transcription factors and signaling pathways interacting in a complex network orchestrating early eye development (reviewed by Fuhrmann, 2010; Chow and Lang, 2001; Ogino et al., 2012; Xie and Cvekl, 2011). Among the signaling pathways, BMP (Furuta and Hogan, 1998; Rajagopal et al., 2009; Sjödal et al., 2007; Wawersik et al., 1999) and FGF (Faber et al., 2001; Garcia et al., 2011; Gotoh et al., 2004; Pan et al., 2006) were found to be essential for lens induction and coordinated OV-to-OC transition, as severe eye defects are associated with their inactivation.

At the time the LP is formed, the dorsal region of the OV becomes specified to the retina and is populated with mitotically active retinal progenitor cells (RPCs) (Fuhrmann, 2010; Levine and Green, 2004). Lineage-tracing studies have shown that RPCs are multipotent, with a single progenitor cell competent to give rise to all retinal neuron and glia cell types (Holt et al., 1988; Turner and Cepko, 1987; Turner et al., 1990). The defining feature of RPCs is co-expression of the transcription factors Rx (Rax), Pax6, Lhx2, Six3/Six6, Chx10 (Vsx2) and Hes1, which are expressed prior to activation of the neurogenic program and contribute to the proliferative and retinogenic potential of RPCs (Burmeister et al., 1996; Grindley et al., 1995; Jean et al., 1999; Li et al., 2002; Liu et al., 2010; Marquardt et al., 2001; Mathers et al., 1997; Oliver et al., 1995; Porter et al., 1997; Tomita et al., 1996; Walther and Gruss, 1991). In a defined birth order, RPCs differentiate into seven retinal cell types: retinal ganglion cells, horizontal cells and cone photoreceptors differentiate first, followed by amacrine cells and rod photoreceptors, bipolar cells, and finally Muller glia cells (Young, 1985). As retinogenesis proceeds, RPCs are exposed to the changing environment of extrinsic cues (Cepko, 1999). These, in cooperation with intrinsic factors represented by transcription factors, most prominently of the basic helix-loop-helix (bHLH) and homeodomain class, regulate progenitor proliferation and operate to direct the bias towards particular cell types (Brown et al., 1998; Cepko, 1999; Hatakeyama and Kageyama, 2004; Inoue et al., 2002; Lillien, 1995; Morrow et al., 1999; Perron and Harris, 2000; Tomita et al., 1996).

At the time of neuronal differentiation, a subpopulation of progenitors undergoes a transition from the proliferative stage toward the lineage-restricted neurogenic stage, upon which they withdraw from the cell cycle and take up a neuronal or glial fate. Accordingly, a proper balance between cell cycle exit and re-entry is required to ensure the temporal generation of all retinal cell types (reviewed by Agathocleous and Harris, 2009). It is in cell cycle phase G1 that growth-promoting and growth-inhibiting signals determine whether a progenitor cell will exit or re-enter the cell cycle. In mammalian retina, the KIP proteins p57^Kip2 (Cdkn1c) and p27^Kip1 (Cdkn1b) and cyclin D1 have been implicated in direct regulation of progenitor proliferative potential (Das et al., 2009; Dyer and Cepko, 2000; Dyer and Cepko, 2001; Geng et al., 2001; Levine et al., 2000; Levine and Green, 2004), promoting either cell cycle exit or progression. The observation that some neurogenic factors promote both neuronal fate determination and cell cycle withdrawal implies that the processes of cell type specification and cell cycle exit are tightly coupled (Farah et al., 2000; Ochocinska and Hitchcock, 2009). However, the mechanism that orchestrates these complex events remains largely elusive.

The paired and homeodomain transcription factor Pax6 plays a pivotal role in both vertebrate and invertebrate eye development (Kozmik, 2005). Since Pax6-deficient (Pax6^-/-) mice are anophthalmic (Hill et al., 1991; Hogan et al., 1986), with eye development arrested at the OV stage, much attention has been paid to elucidation of Pax6 function in the development of individual ocular structures (reviewed by Shaham et al., 2012). Pax6 is expressed from very early stages of eye formation, in SE and OV, and later in developing lens, RPE and all mitotically active RPCs of the retina (Walther and Gruss, 1991). Conditional ablation of Pax6 revealed its autonomous requirement for lens development (Ashery-Padan et al., 2000; Shaham et al., 2009) as well as for later retinal neurogenesis (Marquardt et al., 2001; Oron-Karni et al., 2008). Nevertheless, the autonomous role of Pax6 in the progenitors of the OV and newly formed OC remains unresolved. Here we employed the Cre-loxP system to conditionally inactivate Pax6 specifically in retina-committed eye progenitors.

Pax6 is expressed in both the OV and SE of the developing eye (Walther and Gruss, 1991). In Pax6^-/- (Sey/Sey) mouse embryos, eye development is arrested at the OV stage and no eyes are formed (Hill et al., 1991; Hogan et al., 1986). Although Pax6 expression in SE was found to be essential for lens induction (Ashery-Padan et al., 2000), the role of Pax6 in the OV is less well understood. We used the mRx-Cre transgenic mouse line (Klimova et al., 2013) to conditionally inactivate Pax6 in retina-committed progenitor cells selectively. Expression of Cre recombinase in the mRx-Cre line is controlled by regulatory sequences of the mouse Rx gene (supplementary material Fig. S1B). To monitor mRx-Cre, the ROSA26R reporter line was employed in which Cre-expressing cells can be traced by X-gal staining for lacZ expression (β-galactosidase) after Cre-mediated recombination (Soriano, 1999). Consistent with the expression pattern of the endogenous Rx gene (Mathers et al., 1997), strong Cre activity was observed in the OVs (Fig. 1A-C) of mRx-Cre/ROSA26R embryos at E9.0 and later in the neuroretina, RPE and optic stalk (Fig. 1D-F). Next, mRx-Cre mice were crossed to Pax6^fl/fl mice to inactivate Pax6 in the OV. Pax6^fl/fl/mRx-Cre mice were viable and fertile. The eyes of Pax6^fl/fl/mRx-Cre (i.e. Pax6 loss-of-function) mutants were analyzed for the presence of Pax6 protein by immunohistochemistry (Fig. 1G-J). At E9.5, we observed decreased Pax6 levels in OV neuroepithelium of Pax6^fl/fl/mRx-Cre embryos, whereas Pax6 protein expression was very high in the SE and OV of Pax6^fl/fl control embryos (compare Fig. 1G with 1H). One day later, very few Pax6⁺ cells were detected in the neuroretina and optic stalk of Pax6^fl/fl/mRx-Cre (Fig. 1J). Importantly, the Pax6 protein levels remained unchanged in the SE and lens pit upon OV-specific Pax6 elimination (Fig. 1H,J).

The consequences of Pax6 inactivation in retina-committed progenitor cells of Pax6^fl/fl/mRx-Cre embryos were first investigated at the histological level. The first manifestation of abnormal retina development was observed at E10.5. Already at this stage, the retina was thinner than in wild type (compare Fig. 2A with 2B), suggesting a decreased number of RPCs. We counted DAPI-stained (DAPI⁺) cells per retinal section and found that the number of RPCs was decreased by 44±4.8% (± s.d.) in Pax6-deficient retinae (Fig. 2I). Hypocellularity became more obvious at E14.5, when the number of retinal cells reached only 19.7±3.1% of wild-type levels (Fig. 2C,D,I). At E16.5, Pax6-deficient retinae became progressively smaller (Fig. 2E-F′) and, whereas the retina in wild-type newborns was properly laminated (Fig. 2G), Pax6-deficient retinae reproducibly formed a thin layer around the lens with no sign of lamination (Fig. 2H,H′, red arrowheads). At postnatal stages, the eye was generally smaller with hardly distinguishable retinal cells (data not shown).

Tissue hypocellularity could be due to a decreased proliferation rate, premature cell cycle exit or cell death. To address this issue in Pax6-deficient retina, we first analyzed the proliferation potential of RPCs. The proportion of actively proliferating cells (BrdU⁺/DAPI⁺ per retinal section) was counted at several stages of embryonic development using incorporation of BrdU applied 1 hour prior to analysis. Despite the fact that Pax6-deficient retinae appeared strongly hypocellular, we did not observe a dramatic difference in the proportion of BrdU⁺ cells relative to all retinal cells between E10.5 and E13.0 (Fig. 3A-D,I). However, at E14.5 a rapid decrease of BrdU⁺ retinal cells was observed (Fig. 3E,F,I). Only 9.1±2% of RPCs were BrdU⁺, compared with 34±6.2% BrdU⁺ cells in wild-type retinae (Fig. 3I). Although BrdU⁺ cells were localized in the neuroblastic layer (NBL) throughout the whole retina of the wild-type OC (Fig. 3E), in Pax6-deficient retinae cycling cells were localized mostly in the central part of the retina with almost no BrdU⁺ cells localized peripherally (Fig. 3F). From E16.5 onwards, almost no dividing BrdU⁺ cells were found in Pax6-deficient retinae (Fig. 3H,I).

Cell death, as a possible cause of hypocellularity, was analyzed on retinal sections between E10.5 and E15.5 using an antibody against cleaved caspase 3 (supplementary material Fig. S2A-J). Although we revealed no significant difference between Pax6-deficient and wild-type retinae at E10.5-E12.5, increased cell death was observed at E13.5 and E14.5 (supplementary material Fig. S2K).

Because the BrdU incorporation assay between E10.5 and E13.0 did not indicate perturbed S-phase re-entry and since no increase in cell death was observed in Pax6-deficient RPCs of these stages, we next analyzed a potential M-phase arrest and the length of RPC cell cycle as potential contributors to the phenotype. For M-phase arrest, staining for phosphorylated histone H3 (PH3) was performed at E10.5, E11.5 and E14.5 (supplementary material Fig. S3A). No difference in the proportion of PH3⁺ cells was observed at E10.5 and E11.5. However, a decreased proportion of PH3⁺ cells was observed in Pax6-deficient retinae at E14.5 (supplementary material Fig. S3B). This decrease corresponds to the decreased proliferation rate at E14.5 observed in the BrdU incorporation assay (Fig. 3I). To measure the length of the cell cycle, we used window labeling based on two thymidine analogs that can be differentially detected (Burns and Kuan, 2005; Das et al., 2009). Retinal sections of E11.5 and E13.0 embryos were co-stained for BrdU, EdU and Pcna (Fig. 3J; data not shown) and the lengths of the whole cell cycle (T_c), S phase (T_s) and G1+G2+M phase (T_c-T_s) were determined as previously described (Das et al., 2009). At both stages analyzed, T_c of Pax6-deficient RPCs was significantly prolonged compared with wild-type littermates (Fig. 3K). The prolonged T_c was not caused by a lengthened S phase, as T_s was unchanged, but instead T_c-T_s was increased (Fig. 3K). Furthermore, quantification of EdU⁺ cells relative to Pcna⁺ proliferating progenitors showed that overall progenitor proliferation was also affected at E11.5 and E13.0 (Fig. 3K′), at stages when no significant difference in the proportion of BrdU⁺ cells relative to all retinal cells was detected in Pax6-deficient retinae (Fig. 3I). This difference can be attributed to neurogenesis in wild-type retinae, which decreases the fraction of BrdU⁺ cells relative to all (DAPI⁺) retinal cells. Taken together, these data indicate that Pax6 positively regulates progression through the RPC cell cycle.

During development, decreased proliferation usually coincides with cell cycle exit and subsequent differentiation. To address whether premature cell cycle exit might contribute to the phenotype observed in Pax6-deficient retina, we analyzed the expression of cyclin D1 and of the cyclin-dependent kinase inhibitors p27^Kip1 and p57^Kip2, which are known regulators of RPC proliferation. At E14.5, the stage when the decrease in BrdU⁺ cells was noted (Fig. 3F), we observed a decreased level of cyclin D1 and p27^Kip1 and elevated expression of p57^Kip2 (Fig. 3L-S). For cyclin D1 and p57^Kip2, expression changes were obvious mostly in the distal parts of retinae (Fig. 3M,Q,S), in a pattern complementary to that of BrdU staining (Fig. 3F). This suggested that peripherally located cells have just left the cell cycle, as they were p57^Kip2+, cyclin D1^- and BrdU^-. To address the possible gradual peripheral-to-central progression of this phenomenon, we analyzed the retinae of E16.5 embryos. Whereas no p57^Kip2+ cells were detected in any wild-type retina, p57^Kip2 was detected in the majority of Pax6-deficient retinal cells (Fig. 3V,W). At the same time, cyclin D1 was downregulated in Pax6-deficient retinae (compare Fig. 3T with 3U). These results indicate that Pax6 depletion from early retinal progenitors dramatically restricted their proliferation potential and shifted RPCs to forced cell cycle exit.

As Pax6-deficient retinae exhibit severe proliferation defects, we next analyzed whether retinal progenitor characteristics were maintained in mutants. We assessed the expression of known markers such as Rx, Lhx2, Chx10, Sox2, Six3, Hes1 and cyclin D1 at E10.5, when the cell number was already decreased. However, the expression of none of these factors was significantly changed (Fig. 4A-G′).

It has been reported that conditional inactivation of Pax6 in the distal parts of the OC at later stages (E12) of retinogenesis using α-Cre leads to exclusive generation of amacrine cells (Marquardt et al., 2001; Oron-Karni et al., 2008). In accordance, generation of amacrine cells was observed in Pax6^fl/fl/α-Cre mice, as documented by staining for the amacrine cell marker syntaxin and Vc1.1 (HNK-1) immunoreactivity in the distal part of Pax6^fl/fl/α-Cre retinae (supplementary material Fig. S1C). We therefore tested the differentiation potential of RPCs in which Pax6 is absent from E10.5 in the whole retina. We used RNA in situ hybridization to analyze the expression of the pro-neural bHLH transcription factors Atoh7, Ngn2 (Neurog2), Neurod1, Mash1 (Ascl1) and Math3 (Neurod4), which that have been shown to initiate the differentiation program and exert bias towards particular cell fates. The expression of these bHLH factors was not initiated at E14.5 (Fig. 4H′-L′), suggesting that not only proliferation but also the retina-specific differentiation program was severely affected in the absence of Pax6.

To test whether general neuronal differentiation took place in the mutants, retinae were stained with antibody against the pan-neuronal marker acetylated beta III tubulin (Tuj1, also known as Tubb3), which marks differentiating cells and reveals formation of the differentiated cell layer (DCL) (Sharma and Netland, 2007; Sigulinsky et al., 2008). Although strong Tuj1 staining was observed in the DCL of wild-type retinae at E14.5 and E16.5 (Fig. 4M,N), only a very few Tuj1⁺ cells were detected in Pax6-deficient retinae, as DCL was not established at all (Fig. 4M′,N′, arrowheads). In addition, it should be noted that staining for the early amacrine cell-specific factor bHLHb5 (Bhlhe22) and Vc1.1 immunoreactivity at E15.5 revealed no appearance of amacrine cells in the Pax6^fl/fl/mRx-Cre retina (supplementary material Fig. S4A-D). Taken together, these results indicate that the overall differentiation potential of Pax6-deficient early progenitors is severely compromised.

Previous studies have shown that Pax6 inactivation in RPCs located in the most peripheral region of the OC leads to premature activation of the photoreceptor differentiation program (Oron-Karni et al., 2008). This process is accompanied by upregulation of cone-rod homeobox protein (Crx), which is the earliest expressed photoreceptor determinant (Furukawa et al., 1997; Chen et al., 1997). We therefore analyzed early RPCs in Pax6^fl/fl/mRx-Cre for the presence of Crx transcripts. Already at E10.5, following Pax6 protein elimination, Crx mRNA was detected throughout the invaginating Pax6-deficient retina (Fig. 5B), whereas in wild-type controls Crx expression was not detectable (Fig. 5A). In E13.5 control retina, Crx protein was immunohistochemically detected in a few photoreceptor-committed cells of the OC (Fig. 5C). Strikingly, in Pax6-deficient retina of the same stage, Crx protein was produced by virtually all RPCs (Fig. 5D). At the protein level, elevated Crx expression was reproducibly detected between E11.5 and E14.5 in Pax6-deficient RPCs (supplementary material Fig. S4F,H).

Crx protein is known to enhance the expression of photoreceptor-specific genes (Hennig et al., 2008; Chen et al., 1997; Mitton et al., 2000; Peng and Chen, 2005); however, Crx alone does not determine the specific photoreceptor cell fate and is supposed to activate transcription in cooperation with other transcription factors (Akagi et al., 2005; Furukawa et al., 1999; Hennig et al., 2008). To further test the ability of Crx to induce photoreceptor differentiation, we analyzed Pax6-deficient RPCs for the expression of Otx2, a key regulator of the photoreceptor lineage (Nishida et al., 2003). Otx2 expression failed to be activated, with the exception of a few cells in the most distal part of the OC (Fig. 5F). To rule out a possible delay of Otx2 expression in Pax6-deficient retinae, the expression of Otx2 and its photoreceptor-specific target Blimp1 (Prdm1) was analyzed at E15.5. Although the expression of both Otx2 and Blimp1 was apparent in the outer layer, where differentiating photoreceptors reside in wild-type retinae (Fig. 5I,K), their expression was not initiated in Pax6^fl/fl/mRx-Cre (Fig. 5J,L). We observed elevated levels of the cone-specifying nuclear receptor retinoid X receptor-γ (Rxrγ) (Fig. 5N); however, other factors essential for cone lineage specification, such as thyroid hormone nuclear receptor TRβ2 (Thrb), or for rod specification, such as Nr2e3, were not expressed upon Pax6 inactivation (Fig. 5H,P). In summary, although Crx expression was efficiently induced by all Pax6-deficient RPCs, these cells were not able to fully accomplish the photoreceptor differentiation program, as they failed to express other factors indispensable for this process.

Despite the fact that Pax6^fl/fl/mRx-Cre animals did not exhibit severe lens defects in general (Fig. 2), we occasionally observed morphological abnormalities interfering with lens pit/OC formation in these mutants (supplementary material Fig. S5A,B′). Since SE and OV continuous interaction is required for lens and OC morphogenesis, our observation led us to hypothesize that, at the OV stage, OV-expressed Pax6 might also play a role in lens formation. To test whether earlier removal of Pax6 protein in the OV enhances the lens phenotype, we introduced a single Sey allele (Pax6^Sey/+) into the mRx-Cre/Pax6^fl background. Under these conditions, only one allele of Pax6 has to be recombined since the second allele is genetically inactive in Sey. Although there are several lens phenotypes associated with the inactivation of one Pax6 allele, including small lens size or its incomplete separation from the overlying ectoderm, the lens is always formed (Hogan et al., 1986). This genetic combination resulted in downregulation of Pax6 protein levels in the OV neuroepithelium of Pax6^Sey/fl/mRx-Cre embryos before its transition to the OC (Fig. 6G). Note that this effect can be partially attributed to slightly delayed OC/lens pit morphogenesis in Pax6^Sey/fl embryos (Fig. 6F). At E11.0, Pax6^Sey/fl/mRx-Cre embryos reproducibly exhibited defective lens/OC formation (Fig. 6B,B′). In all embryos analyzed, the lens was completely missing, and the OV either remained arrested or occasionally showed invagination into a small OC-like structure (Fig. 6D; supplementary material Fig. S5D,D′). The arrest of eye development can also be observed when Pax6 protein is eliminated specifically in the OV neuroepithelium already at E9.5 using an earlier deleting founder of mRx-Cre (EmRx-Cre) (supplementary material Fig. S5E-H′). To further test whether the lens fate was established in Pax6^Sey/fl/mRx-Cre embryos, we analyzed expression of the LP-specific transcription factors Pax6, Six3 and Sox2. Although their expression in the SE was maintained (Fig. 6G-G′), indicating that the LP was initially formed, expression of the lens differentiation genes Foxe3 and Prox1 was not initiated and the LP did not invaginate to form the lens vesicle (Fig. 6J,J′).

Since mRx-Cre-mediated gene manipulation was performed in the OV and not in the SE, a non-cell-autonomous process is likely to regulate lens development in a fashion dependent on OV-expressed Pax6. The BMP and FGF signaling pathways are known to play an important role in the lens-inductive ability of OV. Using antibody staining we examined the intracellular mediators of these pathways: phosphorylated Erk proteins (pErk1/2; also known as pMapk3/1) for FGF signaling and phosphorylated Smad proteins (pSmad1/5) for BMP signaling in Pax6^Sey/fl/mRx-Cre LP. In both wild type and the Pax6^Sey/fl/mRx-Cre mutant, strong pErk1/2 staining was observed in the LP/lens (Fig. 7A), indicating that FGF signaling was unaffected even when lens formation was disrupted. Similarly, we did not observe any significant difference in pSmad1/5 levels in the LP/lens between wild-type and Pax6^Sey/fl/mRx-Cre mutant eyes (Fig. 7B). Furthermore, the expression of BMP ligands essential for eye development, Bmp4 and Bmp7, was not abolished, indicating that BMP signaling was not grossly affected upon OV-specific Pax6 inactivation (Fig. 7C).

Unlike BMP and FGF, activation of the Wnt/β-catenin pathway has been shown to inhibit lens fate since stabilization of β-catenin in lens primordium prevents lens formation (Kreslova et al., 2007; Machon et al., 2010; Smith et al., 2005). Activation of the Wnt/β-catenin pathway results in the accumulation of β-catenin in the nucleus, which allows the TCF/Lef family of transcription factors to activate downstream target genes. We therefore assayed the activity of the canonical Wnt/β-catenin pathway using the BAT-gal reporter mouse line carrying lacZ driven by multimerized TCF/Lef binding sites (Maretto et al., 2003). Although we observed decreased mRNA expression of known Wnt/β-catenin inhibitors Sfrp1 and Sfrp2 in Pax6^Sey/fl/mRx-Cre E10.5 eyes (supplementary material Fig. S6A-F), expression of another Wnt inhibitor, Dkk1, remained unchanged (supplementary material Fig. S6G-I), and BAT-gal reporter mice did not exhibit overall aberrant Wnt activation in the LP (Fig. 7D, arrowheads). Since the BAT-gal reporter does not always display sufficient sensitivity (Barolo, 2006), we used staining for Lef1 as a target of Wnt signaling (Planutiene et al., 2011; Wu et al., 2012). As with BAT-gal reporter activity, Lef1 was not aberrantly expressed in the LP of Pax6^Sey/fl/mRx-Cre mutants (Fig. 7D), indicating no activation of Wnt signaling. By contrast, aberrant Wnt signaling activity was observed in the OV. However, this is unlikely to cause the lens development arrest, since intentional activation of the Wnt/β-catenin pathway in OV neuroepithelium in Catnb^lox(ex3) mice (Harada et al., 1999) did not interfere with lens formation (supplementary material Fig. S6K). In summary, our findings suggest that, although OV-expressed Pax6 is essential for lens formation, this process is independent of FGF, BMP and Wnt/β-catenin signaling.

In our study, we focused on Pax6 function in retina-committed eye progenitors of the OV and early OC. The current model assumes that prospective retina-expressed Pax6 is dispensable for lens and OC development (Fujiwara et al., 1994). In addition it is known that, at the stage of neurogenesis, Pax6 loss leads to the exclusive generation of amacrine interneurons, indicating Pax6 requirement for progenitor cell multipotency (Marquardt et al., 2001). Here we show that, in prospective retina, Pax6 is required for three crucial processes: lens induction, propagation of early retinal progenitors, and initiation of the retinal differentiation program.

Although Pax6 has been found to be involved in neural progenitor proliferation, the response to Pax6 loss seems to be dependent on the developmental context. Pax6^-/- mutants display an increased number of early cortical progenitors in S phase (Estivill-Torrus et al., 2002; Götz et al., 1998; Warren et al., 1999), but a reduction in proliferation was observed in the diencephalon and optic rudiment (Philips et al., 2005; Warren and Price, 1997). Conditional inactivation of Pax6 in RPCs of the peripheral OC at E12 results in hypocellularity accompanied by a decreased proportion of cells in S phase (Marquardt et al., 2001; Oron-Karni et al., 2008), indicating a pro-proliferative effect of Pax6 in RPCs. The molecular mechanism of how Pax6 regulates cell proliferation remains elusive. One possibility includes the regulated expression or function of general components of the cell cycle machinery either directly by Pax6 or indirectly by some of its targets (Cvekl et al., 1999; Estivill-Torrus et al., 2002; Farah et al., 2000; Holm et al., 2007; Ochocinska and Hitchcock, 2009). Here, we show that after Pax6 inactivation the cyclin-dependent kinase inhibitor p57^Kip2 exhibits aberrant accumulation. This process was accompanied by RPC incompetence to re-enter S phase and by downregulation of cyclin D1. Although cyclin D1 is normally expressed by cycling RPCs, promoting progression through the cell cycle, its expression is rapidly downregulated in emerging postmitotic cells (Barton and Levine, 2008; Das et al., 2009; Dyer and Cepko, 2001). By contrast, the expression of p57^Kip2 is upregulated in a small subset of RPCs between E14.5 and E17.5 as they exit the cell cycle (Dyer and Cepko, 2000). Loss-of-function and overexpression studies performed in the mouse retina demonstrated that p57^Kip2 is both necessary and sufficient to induce cell cycle exit (Dyer and Cepko, 2000). Thus, the pro-proliferative effect of Pax6 in the retina might be mediated, at least in part, by the inhibition of premature cell cycle exit through regulation of p57^Kip2 protein levels. As we also observed upregulation of p57^Kip2 mRNA, the negative control appears to occur at the transcriptional level. The mechanism by which p57^Kip2 mediates cell cycle exit in Pax6-deficient RPCs might include blocking of phosphorylation of the retinoblastoma protein (reviewed by Sherr and Roberts, 1995).

It is worth noting that, before Pax6-deficient RPCs exit the cell cycle, a p57^Kip2/cyclin D1-independent mechanism regulates the proliferation rate. The cell cycle length of the Pax6-deficient RPC population is significantly increased relative to that of the wild-type RPC population at E11.5 and E13. In contrast to Pax6-deficient cortical progenitors manifesting prolonged S phase (Estivill-Torrus et al., 2002), the cumulative time spent in the G1, G2 and M phases was increased in Pax6-deficient RPCs, indicating Pax6 function in these phases of the cell cycle.

Once the OV starts to invaginate to form the OC, the population of RPCs is established. Previous studies have indicated that some retinal progenitor characteristics are maintained in the arrested OV rudiment of germline Pax6^-/- embryos (Bäumer et al., 2003; Bernier et al., 2001). The mRx-Cre line allows inactivation of Pax6 precisely at the time when the RPC population is being established and before the differentiation program has been initiated (at E10). Our analysis shows that Pax6 is absolutely essential for the generation of all retinal cell types, since no sign of general neuronal differentiation was observed upon Pax6 inactivation, pointing out a specific Pax6 role in the maintenance of RPC multipotency. This can be explained by the ability of Pax6 to activate the expression of proneurogenic bHLH factors, including Atoh7, Mash1, Math3, Ngn2 and Neurod1 (Hatakeyama and Kageyama, 2004; Marquardt et al., 2001; Oron-Karni et al., 2008; Riesenberg et al., 2009) (this study).

Our observation that Pax6 is indispensable for neuronal differentiation in the retina is seemingly inconsistent with previous studies. Marquardt and colleagues (Marquardt et al., 2001) showed that inactivation of Pax6 at the OC stage using α-Cre leads to the exclusive generation of amacrine interneurons. Further detailed analysis revealed two populations of RPCs that differentially responded to Pax6 loss: whereas progenitors located more centrally in the OC adopted amacrine cell fate, those located peripherally activated expression of Crx (Oron-Karni et al., 2008). Nevertheless, our data show that Pax6 is also indispensable for amacrine cell genesis, as Neurod1, Math3, Atoh7 and other amacrine cell-specific factors are not expressed in the absence of Pax6. This difference can be attributed to the timing of Pax6 inactivation. When using mRx-Cre, Pax6 is completely eliminated before the differentiation program is initiated (E10) (this study); for α-Cre (Marquardt et al., 2001; Oron-Karni et al., 2008), Pax6 is eliminated 2 days later (E12) (Riesenberg et al., 2009) (our observation). At E12 the differentiation program has already been initiated, as some proneurogenic factors, including Neurod1 and Atoh7, are expressed (reviewed by Hatakeyama and Kageyama, 2004). The amacrine cell genesis is likely to be the result of biphasic inactivation of Pax6 by α-Cre with respect to the onset of neurogenesis. Since progenitors located in the central OC differentiate earlier, the presence of two populations of RPCs in the OC of α-Cre/Pax6^fl/fl conditional mutants, with the amacrine cell population located more centrally, then apparently reflects the different degree of neuronal differentiation along the central-to-peripheral axis (Oron-Karni et al., 2008).

In Pax6^-/- embryos eye development is arrested at the OV stage and neither lens nor OC is formed (Grindley et al., 1995; Hill et al., 1991; Hogan et al., 1986). Since Pax6 is expressed in both SE and OV (Walther and Gruss, 1991), it was not clear which component is the source of the defect. Several studies indicated that SE-expressed Pax6 might be responsible (Collinson et al., 2000; Fujiwara et al., 1994; Grindley et al., 1995; Quinn et al., 1996), leading to a general acceptance of the notion that Pax6 activity in the OV is, by and large, not required for lens formation (reviewed by Ashery-Padan and Gruss, 2001; Lang, 2004; Mathers and Jamrich, 2000; Ogino and Yasuda, 2000). However, such a conclusion has not been tested genetically. Experiments in which anti-Pax6 morpholinos were electroporated into chick embryo OV indicated that OV-expressed Pax6 might play an essential role in retina and lens formation as well (Canto-Soler and Adler, 2006). Conditional inactivation of Pax6 in the SE revealed that SE-expressed Pax6 is autonomously required for LP/lens but not retina formation (Ashery-Padan et al., 2000). In this study we present evidence that early expression of Pax6 in the OV is indispensable for the development of both tissue components: cell-autonomously for OC/retina development and non-cell-autonomously for lens formation. In OV Pax6 mutants, eye development was arrested at the OV stage in a manner morphologically reminiscent of the Pax6^-/- (Sey) phenotype. Thus, in Pax6^-/- embryos, the defect in eye formation is apparently attributable to Pax6 function in both OV and SE (this study) (Canto-Soler and Adler, 2006), in sharp contrast to the current, prevailing view (Ashery-Padan and Gruss, 2001; Ogino and Yasuda, 2000). Interestingly, Pax6 is required for lens formation only before the OV-to-OC transition. Once the lens pit starts to emerge from the LP, lens development is no longer dependent on OC-expressed Pax6. This accords with the idea that lens development becomes independent of OV/OC when the lens has reached a certain developmental stage (Adler and Canto-Soler, 2007; Lang, 2004).

How Pax6 regulates the ability of OV to induce lens formation remains elusive. It has been demonstrated that lens formation is dependent on the deposition of molecules of the extracellular matrix between the LP and OV, and that this process is dependent on Pax6 expression (Huang et al., 2011). There is good evidence that signaling from the OV is essential for the activation of lens-specific expression in the SE and subsequent lens formation (Faber et al., 2001; Furuta and Hogan, 1998; Kamachi et al., 1998; Wawersik et al., 1999; Yun et al., 2009). We have not been able to detect conspicuous changes in components of the BMP, FGF and Wnt/β-catenin pathways. It remains possible that the causal changes in Pax6^Sey/fl/mRx-Cre are transient or too subtle to be detected in our analysis. Alternatively, additional, as yet ill-defined signals might be involved in lens induction. The identification of further molecules acting downstream of Pax6 in the OV might help uncover the molecular details of this process.

A mouse with floxed Pax6 alleles (Pax6^fl/fl) was generated by homologous recombination in embryonic stem cells, with loxP sites flanking exons 3 and 6 (supplementary material Fig. S1A). ROSA26R (Soriano, 1999), ROSA26R-EYFP [R-EYFP (Soriano, 1999; Srinivas et al., 2001)], Sey^1Neu (Hill et al., 1991), BAT-gal reporter line (Maretto et al., 2003), Catnb^lox(ex3) (Harada et al., 1999) and mRx-Cre (Klimova et al., 2013) (supplementary material Fig. S1B) mice have been described previously.

Mouse embryos were harvested from timed pregnant females. The morning of vaginal plug was considered embryonic day (E) 0.5. Embryos were fixed in 4% (w/v) paraformaldehyde (PFA), washed with PBS, cryopreserved in 30% (w/w) sucrose, frozen in OCT (Tissue Tek, Sekura Finetek) and sectioned.

Cryosections for immunohistochemistry were permeabilized with PBT (PBS with 0.1% Tween 20), blocked with 10% BSA in PBT and incubated with primary antibody overnight at 4°C. Primary antibodies are listed in supplementary material Table S1. Sections were washed with PBT, incubated with secondary antibody (Molecular Probes) coupled to Alexa fluorophore, counterstained with DAPI and mounted in Mowiol (Sigma). To detect phosphorylated Erk (pErk), the TSA Indirect Tyramide Signal Amplification Kit (Perkin Elmer) was used. To detect phosphorylated Smad1/5 (pSmad1/5), the Vectastain ABC Kit (Vector Labs) was used. Antisense mRNA probes for in situ hybridization were synthesized using RNA polymerase and digoxigenin-labeled nucleotides (Roche) following the manufacturer’s instructions (for a list of probes see supplementary material Table S1). RNA in situ hybridization and X-gal staining were carried out as previously described (Fujimura et al., 2009).

For analysis of all proliferation markers, a minimum of three animals from two individual litters were used. Wild-type littermates were used as control. To determine the proportion of actively proliferating RPCs, timed pregnant females were injected 1 hour prior to dissection with BrdU (0.1 mg/g body weight). Embryos were fixed with 4% PFA, cryopreserved in 30% sucrose, embedded in OCT and sectioned. Antigen retrieval was performed by microwave heating in 10 mM sodium citrate (pH 6.5) followed by incubation in 2 M HCl and neutralization (0.1 M borate buffer pH 8.3). Sections were blocked in 10% BSA and incubated overnight at 4°C with anti-BrdU antibody (Abcam, ab6326; 1:100). The cell proliferation rate was always calculated from at least two central sections per individual eye as the ratio of BrdU⁺ cells versus DAPI⁺ cells and statistical significance analyzed by Student’s t-test.

To determine the cell cycle rate at E11.5 and E13, two thymidine analogs were used as previously described (Das et al., 2009). Pregnant females were injected 2.5 hours and 0.5 hours prior to dissection with BrdU and EdU (5-ethynyl-2′-deoxyuridine), respectively. BrdU was detected using anti-BrdU antibody, EdU using the Click-it Reaction (Molecular Probes), and Pcna staining was used to identify all cycling RPCs. The length of the cell cycle (T_c) and of S phase (T_s) in hours (h) was determined by: T_c=2h(Pcna⁺ cells/BrdU⁺ only cells); T_s=2h(EdU⁺ cells/BrdU⁺ only cells). The combined length of G1, G2 and M phase was calculated as: T_c-T_s. Cell counts were performed from a single field of central sections of the eye (as depicted in Fig. 3J). Three fields per individual eye were counted. T_c and T_s for individual fields were determined and analyzed by Student’s t-test.

We thank J. Lachova, A. Zitova and V. Noskova for technical assistance; Drs N. Brown, P. Bovolenta, B. Hogan, S. Chen, C. Craft, P. Carlsson, J. Rubenstein, S. Pleasure, M. Taketo, J. Favor and H. Edlund for reagents and mice; and O. Machon for reading the manuscript.