Phosphorylation states change Otx2 activity for cell proliferation and patterning in the Xenopus embryo

Gene regulation by transcription factors (TFs) is fundamental to cellular and developmental regulatory processes, and some TFs are likely to be regulated by various modifications, including phosphorylation, forming regulatory axes or loops to drive a tissue-specific program of gene expression (Spitz and Furlong, 2012). However, it is not fully understood how modifications of TFs orchestrate regulation in cellular or developmental processes.

The homeobox gene orthodenticle (otd), which encodes the bicoid-type homeodomain TF, is required for the formation of the anterior portion of the head in bilaterians (Holland et al., 2013). otx2 is a vertebrate homolog of otd, and is required for head and eye development (Blitz and Cho, 1995; Matsuo et al., 1995; Suda et al., 2009; Andreazzoli et al., 1997; Isaacs et al., 1999; Pannese et al., 1995). The expression of otx2 begins in the anterior neuroectoderm, which further develops into the forebrain and midbrain, and the gene has an essential role in anteroposterior patterning to prevent caudalization by Meis3 and Gbx2 (Agoston and Schulte, 2009; Katahira et al., 2000). In eye development, Otx2 upregulates an eye field-specific gene, rax, in the center of the otx2-expressing domain at the gastrula stage (Danno et al., 2008); in turn, Rax downregulates otx2 to discriminate the eye field from the diencephalon (Andreazzoli et al., 1999; Zuber, 2010). During later eye development, expression of otx2 restarts in the developing optic vesicle (Viczian et al., 2003; Martinez-Morales et al., 2001), and is involved in the specification of presumptive neural retina (pax6-positive cells) from the optic stalk (pax2-positive cells) in mice (Martinez-Morales et al., 2001). After retinal specification, otx2 expression becomes restricted to postmitotic retinal progenitors, and determines bipolar cell fate (Brzezinski and Reh, 2015). Thus, we are gradually beginning to understand the role of Otx2 in head and eye formation.

Otx2 has both transactivation and transrepression activities in the regulation of downstream target genes. It has a transactivation domain in the C-terminal region (Gammill and Sive, 2001) and directly upregulates a cement gland marker, xcg1 (muc2) in Xenopus (Gammill and Sive, 1997). Otx2 also has a repression domain with a motif of SIWSPAS, which belongs to engrailed homology region 1 (eh1) (Heimbucher et al., 2007). This domain is conserved in vertebrate Otx family proteins and interacts with the co-repressor TLE/Groucho (Heimbucher et al., 2007). With this domain, Otx2 functions as a repressor that inhibits the posterior gene gbx2 to form the midbrain–hindbrain boundary (MHB) (Nakamura et al., 2005). Furthermore, genome-wide analysis of chromatin immunoprecipitation (ChIP) sequencing using Xenopus tropicalis gastrula embryos showed that Otx2 cooperates with the transcriptional activator Lim1 (Lhx1) to bind specific cis-regulatory modules (CRMs) to activate anterior genes in the head organizer, whereas Otx2 cooperates with the repressor Goosecoid (Gsc) to inhibit nonhead-organizer genes (Yasuoka et al., 2014). Thus, both the transactivation and transrepression activities of Otx2 are important for regionalization and tissue patterning and function during embryogenesis.

Recently, several reports have shown that Otx2 directly influences both cell proliferation and differentiation. In mice, aberrant expression of Otx2 in the hindbrain induces the ectopic proliferation of neural progenitor cells (Wortham et al., 2012), and overexpression of Otx2 causes proliferation of dopaminergic progenitors in the ventral mesencephalon (Omodei et al., 2008; Vernay et al., 2005). In addition, in a medulloblastoma cell line, Otx2 upregulates cell cycle-positive regulator genes, such as ccnd3 (cyclin D3), and downregulates cell cycle-negative regulator genes, including p27, and differentiation-specific genes, such as neuroD (Bunt et al., 2012). Thus, Otx2 is likely to be involved in both proliferation and differentiation. However, the molecular mechanism of how Otx2 coordinates cell proliferation and differentiation by negatively and positively regulating direct target genes remains unknown.

Here, we show that Otx2 is phosphorylated, which confers transrepression activity; if Otx2 is not phosphorylated then it can only act as a transactivator. Functional analyses using phosphomimetic or nonphosphorylatable Otx2 mutants suggest that phosphorylated Otx2 represses the cyclin-dependent kinase (Cdk) inhibitor gene p27xic1 (cdknx) and the posterior gene gbx2 to promote cell proliferation and anteriorize the neuroectoderm, respectively, whereas unphosphorylated Otx2 enhances the formation of the eye field and later the retina in the optic stalk. Thus, our data provide a new model in which the post-translational regulation of Otx2 results in versatile functions of this protein, including cell proliferation- and differentiation-orientated activities.

During western blot analyses of exogenous Otx2 in the Xenopus laevis embryo (Sudou et al., 2012), we noticed several bands of Otx2 that were migrating more slowly than an in vitro translated product. Similarly, several previous studies using western blots showed multiple bands of exogenous mouse Otx2 in transfected cultured cells (Mallamaci et al., 1996) and endogenous Otx2 in the mouse embryo (Boyl et al., 2001), although these have yet to be characterized. To evaluate the modification of Otx2, we constructed N-terminally Myc-tagged full-length Otx2 (Myc-Otx2FL) and its deletion constructs (Myc-Otx2HD, ΔAD, AD and ΔRD) (Fig. 1A), and expressed them by mRNA injection into X. laevis embryos for western blot analysis. The Myc-Otx2FL and ΔAD but not ΔRD, HD or AD constructs clearly showed several slowly migrating bands (magenta arrowheads in Fig. 1B and Fig. S1A), compared with the in vitro translated product (blue arrowheads in Fig. 1B and Fig. S1A). These data suggest that modification mainly occurs in the region of amino acid positions 96-184 (Fig. 1A).

The amino acid sequence from position 96 to 184 indicated possible phosphorylation sites for Cdk (S*/T*-P or S*/T*-P-X-R/K; asterisk indicates phosphorylatable residues), mitogen-activated protein kinase (MAPK), casein kinase (S*/T*-X_0-2-D/E/S^p_1-3; S^p, primary phosphorylatable serine residue), and Akt kinase (R/K-X₂-S*/T*) (Fig. 1C). Mouse Otx2 reportedly has an Akt phosphorylation site at S115 (the corresponding site of X. laevis Otx2 is T115; Fig. S2), according to the database for the prediction of phosphorylation sites (see Materials and Methods). To identify modification sites, we first examined seven serine residues (S116, S122, S123, S132, S153, S158 and S161) and one threonine (T115) by introducing point mutations in Myc-Otx2ΔAD. Replacement of a single serine with alanine revealed that S116A, S132A and S158A constructs eliminated the uppermost band and increased the lower bands compared with the wild type (WT) and other mutants (S122A, S123A, S153A and S161A) (Fig. S1B,C). Furthermore, a triple-alanine mutant of Myc-Otx2ΔAD for S116, S132 and S158 (named ΔAD-3A) showed no slowly migrating bands (Fig. 1D, lane 3). These data suggest that S116, S132 and S158, which have the S-P consensus motif for Cdk and MAPK, are responsible for multiple shifted bands (up to seven bands by combinations of three sites) of Myc-Otx2. To examine this suggestion, embryonic lysates expressing Myc-Otx2 constructs were treated with λ-protein phosphatase (λ-PP). λ-PP treatment removed all of the modified bands of Myc-Otx2ΔAD (Fig. 1D, lanes 1 and 2) and did not change the band of ΔAD-3A itself (Fig. 1D, lanes 3 and 4), suggesting that slowly migrating bands of Myc-Otx2ΔAD were caused by phosphorylation at S116, S132 and S158. The same experiment was performed with Myc-Otx2FL, with similar results, whereby all modified bands of Myc-Otx2FL were removed by λ-PP treatment except for a weak band that is shifted (Fig. S1D). Consistently, Myc-Otx2FL-3A had only a single weak shifted band, which was not removed with λ-PP (Fig. S1D, orange arrowhead), suggesting another distinct modification site around the activation domain.

We next asked whether T115 of Xenopus Otx2 is an Akt phosphorylation site, although we did not detect a shifted band in the Otx2 mutant 3A (ΔAD-3A) using western blotting (Fig. 1D, lanes 3 and 4). In addition, there was no difference in band positions between ΔAD-3A (remaining T115) and ΔAD-4A (alanine mutation at T115 in addition to 3A) regardless of λ-PP (Fig. 1D, lanes 3-6). However, it was still possible that phosphorylation at T115 either did not cause a band shift in ΔAD-3A in western blotting or was inhibited by alanine mutations at the three sites including S116 next to T115. Therefore, we performed immunoprecipitation (IP)-western blot assays using anti-Phospho-Akt Substrate antibody, which recognizes the R-X-X-S^p/T^p motif (S^p/T^p, phosphorylated residues). Myc-Otx2ΔAD in embryonic lysates was immunoprecipitated with anti-Myc antibody and then subjected to western blotting with anti-Phospho-Akt Substrate antibody. A band was detected with mutant ΔAD-3A but not with ΔAD-4A (Fig. 1E, upper panel, lanes 2 and 3), suggesting T115 as an Akt site. Thus, we identified four possible phosphorylation sites for Akt (T115) and Cdk/MAPK (S116, S132 and S158) in Otx2, and these sites are well conserved in vertebrate Otx2 as well as an Otx2 paralog/ohnolog, Otx5/Crx (Fig. S2).

Although it is important to determine whether endogenous Otx2 is phosphorylated, no antibodies against phosphorylated sites of Otx2 are available. As far as we know, there is no report on detecting the phosphorylation of TFs expressed region specifically and endogenously in Xenopus embryos, such as Otx2, as evidenced by band shift in western blotting, possibly because of the limited amounts of such TFs in the whole embryo. Therefore, we first estimated the sensitivity of detection using a diluted series of otx2 mRNA injected into Xenopus embryos. Modified bands of full-length Otx2 without a Myc-tag (Otx2FL) were clearly detected when >200 pg otx2FL mRNA was exogenously injected per embryo (Fig. S3). Based on RNA-seq data from X. tropicalis (Owens et al., 2016), the amount of otx2 mRNA was estimated to be in the order of 1 pg per embryo at the neurula stage. Therefore, we predicted that endogenous Otx2 can be detected by western blotting of samples from more than 200 X. tropicalis embryos. As expected, we successfully detected multiple bands of endogenous Otx2 protein immunoprecipitated from lysates of approximately 800 X. laevis neurula embryos in total, and the upper bands were reduced by λ-PP treatment (Fig. 1F, left). Similarly, shifted bands of exogenous Otx2FL were detected, and were abolished by λ-PP treatment (Fig. 1F, right). Thus, endogenous Otx2 is phosphorylated, although the intensities of phosphorylated bands of endogenous Otx2 appeared to be lower than those of exogenous Otx2 (see Discussion).

Given that S116, S132 and S158 sites have the S-P motif (Fig. 1C), we speculated that MAPK or cyclin/Cdk complexes phosphorylate Otx2. We first observed that western blot bands of Otx2 were not affected by either MAPKK (Kosako et al., 1993) or a MAPKK inhibitor (Fig. S4A,B). We then tried to hyperactivate endogenous Cdk activity in Xenopus embryos by overexpressing stable mutants of cyclin B1 (cyclin B1* for Cdk1) and cyclin A1 (cyclin A1* for Cdk1 and Cdk2), both of which lack the destruction box (Geley et al., 2001; Yamano et al., 2004). Hyperactivation of Cdks by mRNA injection of HA-cyclin B1* or HA-cyclin A1* was verified by cell-cycle arrest of mRNA-injected blastomeres (Fig. 2A, white asterisks). Under these conditions, modifications of Myc-Otx2ΔAD were dramatically enhanced (Fig. 2B,C), and these modified bands were almost removed by λ-PP treatment (Fig. 2B,C). These data suggest that Cdk is involved in the phosphorylation of Otx2.

If Otx2 is phosphorylated by cyclin/Cdks, then the S116, S132 and S158 sites might prefer either cyclin B/Cdk or cyclin A/Cdks. The data suggest that S116 and S132 preferred both cyclin B/Cdk and cyclin A/Cdks, whereas S158 preferred cyclin A/Cdks more than cyclin B/Cdk (Fig. S5), implying that cyclin B/Cdk and cyclin A/Cdks are differentially involved in the phosphorylation of Otx2.

To further examine Cdk-dependent phosphorylation of Otx2, we overexpressed the Cdk inhibitor p27xic1 in Xenopus embryos (Vernon, 2003). We first verified cleavage arrest by p27xic1 mRNA injection (Fig. 2D, left, white asterisks) as reported elsewhere (Ohnuma et al., 1999). When a mixture of p27xic1 and Myc-Otx2ΔAD mRNAs were injected at the two-cell stage, modified bands of Myc-Otx2ΔAD were reduced, but a weak band remained, which might result from phosphorylation occurring prior to the accumulation of p27xic1 protein translated from the injected mRNA (Fig. 2D, lanes 1, 2). Therefore, we first injected p27xic1 mRNA alone at the two-cell stage, followed by injection of Myc-Otx2ΔAD mRNA into the cleavage-arrested blastomeres at the 32-cell-stage equivalent. p27xic1 completely abolished modified bands of Myc-Otx2ΔAD (Fig. 2D, lanes 3 and 4), suggesting that phosphorylation of Otx2 depends on Cdk activity. These data also imply that the reduction of modified Otx2 by p27xic1 expression was not caused by dephosphorylation, because, once Otx2 was phosphorylated at the 2- to 32-cell stages (lane 2), it appeared to be stable until the late blastula stage when western blotting was performed (compare lanes 2 and 4 in Fig. 2D). Given that phosphorylation of Otx2 appears to start early in development, we examined developmental changes in phosphorylation levels using Myc-Otx2ΔAD and Myc-Otx2FL (Fig. S5D,E). The data showed that modification of Otx2ΔAD was already detectable at the 16-cell stage (stage 5), gradually increased until the early gastrula stage (stage 10.5) (Fig. S5D; see boxed areas for WT), and then reached almost a plateau at stage 10.5 onwards (Fig. S5E). The increase in phosphorylation levels until stage 9 correlates well with the cleavage cycle, supporting again the possibility that Otx2 is a substrate of cyclin/Cdks.

Taken together, Otx2 has four possible phosphorylation sites, and we refer to T115 as A-site (putative Akt site) and S116, S132 and S158 as C-sites (putative Cdk sites) in the following experiments.

To determine whether known Otx2 activities are affected by phosphorylation states, we overexpressed phosphomimetic or nonphosphorylatable Otx2 in Xenopus embryos and analyzed the resulting eye phenotypes, which have been well documented for Otx2 functions (Martinez-Morales et al., 2001; Nishihara et al., 2012). A phosphomimetic mutant, named Otx2-4E, was made by replacing A-site (T115) and three C-sites (S116, S132 and S158) with glutamate in Otx2FL. Nonphosphorylatable mutants, named Otx2-T115A, -3A and -4A, were similarly made by replacing A-site, C-sites or all four sites with alanine, respectively. Each Otx2 construct was expressed unilaterally in the anterior neuroectoderm, which includes the eye field, and the phenotypes were scored at stages 38-42. We categorized abnormal eye phenotypes into three types compared with the normal-looking phenotype (Fig. 3A-D). Expression of Otx2-WT, -4E -T115A, or -3A caused small/trace eye and ventral defect phenotypes in a dose-dependent manner (Fig. 3E), in which stronger phenotypes appeared to be obtained by expression of Otx2-WT, -4E, -3A and -T115A in order. By contrast, only Otx2-4A caused enlarged eye phenotypes, which have not previously been reported following the expression of any other Otx2 mutants in Xenopus (Andreazzoli et al., 1997; Isaacs et al., 1999). Notably, both Otx2-T115A and -3A reduced rather than enlarged the eyes (Fig. 3E), indicating that both mutations are necessary for the activity of Otx2-4A. Quantitative analysis of the area of the eye vesicle showed that eye size in Otx2-4A-expressing embryos was enlarged by approximately 1.3 times, whereas that in Otx2-4E-expressing embryos was reduced by approximately 0.5 times, compared with control (Fig. 3F). Section examination of the enlarged eye phenotype by Otx2-4A showed that the retinal tissue was expanded ventrally with a reduction in the retinal pigment epithelium (Fig. 3H,H′, arrowheads) compared with the control (Fig. 3G,G′). These results clearly demonstrated that Otx2-4A has different activities in eye formation compared with the other mutants, and that exogenous Otx2-WT functions in a phosphorylated form, given that WT exhibits similar activities to 4E (Fig. 3E).

To further examine whether all C-site mutations are required for the activity of Otx2-4A, we constructed triple-alanine mutants of Otx2 (named A2A constructs), in which only one of the C-sites (S116, S132 or S158) was intact, referred to as A2A-S116, -S132 or -S158, respectively. A2A-S116 and A2A-S132 both caused small/trace and ventral defect eye phenotypes (Fig. 3I). By contrast, A2A-S158 caused an enlarged eye phenotype in a dose-dependent manner, although the fraction of enlarged eye phenotypes by A2A-S158 was lower than that of Otx2-4A. These data indicate that mutations at T115, S116 and S132 are essential for enlarged eye phenotypes.

We next examined the early effect of Otx2-4A and -4E by focusing on the eye field marker gene rax. Overexpression of Otx2-4A increased the expression of rax compared with the uninjected side (Fig. 4D), whereas Otx2-WT and -4E as well as the globin control did not affect rax expression (Fig. 4A-C). The expansion of rax expression by Otx2-4A could explain the enlarged-eye phenotypes seen in these embryos. The enlarged eye phenotypes with aberrant ventral eye structures (Fig. 3H,H′) also suggested that Otx2-4A converts the presumptive optic stalk to the retina. Therefore, we examined the expression of pax2 and pax6 in the late neurula, in which pax2 is expressed in the optic stalk (OS) and MHB (Fig. 4E), whereas pax6 is expressed in the optic vesicle (OV) and diencephalon (Di) side by side with different whole-mount in situ hybridization (WISH) intensities (Fig. 4I). Overexpression of Otx2-WT, -4E and -4A all inhibited pax2 expression in the OS (Fig. 4E-H), whereas Otx2-WT and -4E, but not -4A caused a posterior shift in pax2 expression in the MHB (open arrowheads in Fig. 4E-H). The reduction in pax2 in the OS by Otx2-WT, -4E and -4A (Fig. 4F-H) is consistent with the aberrant eye morphology at the tailbud stage on the ventral side where the OS is formed (Fig. 3C,E). Otx2-WT, -4E and -4A differently affected the expression of pax6. WT reduced OV but expanded Di pax6 expression (Fig. 4J,J′; dashed black lines, Di; dashed white lines, OV); and Otx2-4E slightly reduced pax6 OV expression similar to WT (Fig. 4K,K′), consistent with the small/trace eye phenotypes of these embryos (Fig. 3B,E). By contrast, Otx2-4A expanded OV pax6 expression (Fig. 4L,L′), consistent with its enlarged eye phenotypes (Fig. 3D,E). These data demonstrated that Otx2-4E and -4A have different activities in eye formation. This raised the next question of whether Otx2-4E and -4A have different activities in cell proliferation and anteroposterior patterning, which have been reported for Otx2 (Pilo et al., 2001; Acampora et al., 2000).

To investigate the effect of Otx2 on cell proliferation, Otx2-4E or -4A was expressed in the dorsoanterior region of the embryo, and immunostaining was then performed for phospho-histone H3 (PH3) as a mitosis marker. The numbers of PH3-positive nuclei were counted in EGFP-positive and negative areas for quantitative analysis (Fig. 5A-C). The data showed that the ratio of the number of PH3-positive nuclei in injected versus uninjected areas was 1.27±0.07 (mean±s.e.m.) and 1.38±0.05 in Otx2-WT and -4E-expressing embryos, respectively, whereas it was 0.727±0.062 in Otx2-4A-expressing embryos. These ratios significantly differed from the control (0.944±0.014) (Fig. 5D), suggesting that Otx2-WT and -4E both stimulate cell proliferation, whereas Otx2-4A reduces it. To confirm this, we counted the number of nuclei stained by DAPI (Fig. 5E-G). Quantitative analysis showed that the ratio of cell number in injected versus uninjected areas was 1.08±0.061 and 1.13±0.02 in Otx2-WT- and -4E-expressing embryos, respectively, whereas it was 0.894±0.019 in Otx2-4A-expressing embryos. The ratios in 4E and 4A embryos differed significantly from the control (1.02±0.04; P<0.05) (Fig. 5H). Thus, these data suggest that Otx2-4E and -4A have opposite effects on cell proliferation: Otx2-4E stimulates whereas -4A reduces proliferation.

To examine the in vivo relevance of Otx2 function in cell proliferation, we next performed loss-of-function experiments using antisense morpholino oligos (MOs) for otx2 and its functionally redundant paralog otx5, according to our previous report (Yasuoka et al., 2014). The number of PH3-positive nuclei was almost unchanged in the anterior neuroectoderm injected with standard control MO (Fig. S6A-A″,C), whereas it significantly decreased in the region injected with otx2/otx5-MOs (Fig. S6B-B″,C). Consistent with this, the number of nuclei decreased in the otx2/otx5-MO-injected area (Fig. S6E-E″,F), but not affected in the control MO-injected area (Fig. S6D-D″,F). These data indicate that Otx2 and Otx5 are required for cell proliferation in the anterior neuroectoderm, and these in vivo data are consistent with the enhancement of cell proliferation by exogenous Otx2-WT and -4E (Fig. 5). This also suggests that the reduction in cell proliferation caused by Otx2-4A results from the dominant-negative action of Otx2-4A against endogenous phosphorylated Otx2, which stimulates cell proliferation, and the unphosphorylated form of Otx2 is unable to take its place.

To determine how Otx2 stimulates proliferation, we focused on p27xic1 because Otx2 reportedly inhibits the p27 gene (Bunt et al., 2012). Analysis of our ChIP-seq data using the X. tropicalis gastrula (Yasuoka et al., 2014) showed the co-occupancy of Otx2 and Tle1 of several potential CRMs near p27xic1 (Fig. 6A), suggesting that p27xic1 is a direct target gene of Otx2. Using mRNA injection, Otx2-WT, -4E or -4A with nuclear β-galactosidase (nβ-gal) as a tracer were unilaterally expressed, and the expression of p27xic1 was examined by WISH. Expression of Otx2-WT and -4E, but not of -4A, downregulated p27xic1, compared with globin-injected negative control and on the uninjected side (Fig. 6B-E′), consistent with the stimulation of cell proliferation by Otx2-4E (Fig. 5D,H).

To assess the gain-of- and loss-of-function data discussed above, the expression domains of otx2 and p27xic1 and the distribution of PH3-positive cells in the early neurula were examined using sagittal hemisections. The data showed that the otx2-expressing region did not overlap with that of p27xic1 (Fig. 6F-G′) and had more mitotic cells than the posterior region (Fig. 6H,H′), consistent with the involvement of otx2 in cell proliferation in the anterior neuroectoderm. These data suggest the existence of a positive feedback loop involving Otx2, Cdk and p27xic in the regulation of cell proliferation (Fig. 6I).

In the neural plate, Otx2 is involved in anteroposterior patterning by directly inhibiting posterior genes, which are suggested to include gbx2 and meis3 (Yasuoka et al., 2014), and by directly activating anterior genes, such as xcg1 (Gammill and Sive, 1997) and rax (Danno et al., 2008). We first examined gbx2 by ectopically expressing Otx2-WT, -4E or -4A, in the posterior neuroectoderm. Both Otx2-WT and -4E downregulated the expression of gbx2 (Fig. 7B-C′, arrowheads) compared with the uninjected side and globin-injected control (Fig. 7A,A′). This activity explains the posterior shift in pax2 expression in the MHB by Otx2-WT and -4E (Fig. 4F,G). Conversely, Otx2-4A did not downregulate gbx2 expression (Fig. 7D,D′), suggesting that only phosphorylated Otx2 has the ability to inhibit posteriorization.

xcg1 is expressed in the cement gland, which is the anterior-most organ (Fig. 7E). Ectopic expression of Otx2-WT, -4E and -4A all upregulated xcg1 (Fig. 7F-H), and this activation was observed in the ventrolateral ectoderm, but not in the neuroectoderm, as reported elsewhere (Gammill and Sive, 1997). To assess whether the ectopic upregulation of xcg1 by the Otx2 mutants required their transactivation activity, we developed transactivation domain-deleted constructs of 4E (ΔAD-4E) and 4A (ΔAD-4A). Neither ΔAD-4E- nor ΔAD-4A upregulated xcg1 (Fig. 7I,J), indicating that the transactivation domain of Otx2-4E and -4A is required for their activation of xcg1. These data also suggest that Otx2 has transactivation activity regardless of its phosphorylation state.

The above data suggest that the difference in function between Otx2-4E and -4A is repression activity. To further examine whether phosphorylation states affect the repression and activation activity of Otx2, we performed luciferase reporter assays. We first examined the response of the silencer of meis3, a posterior gene, which is expected to be repressed by Otx2. The meis3-D2 silencer (Yasuoka et al., 2014) was inserted upstream of the SV40 promoter in the reporter construct, and this reporter alone showed a high level of luciferase activity in the animal pole region. Under this condition, reporter activity was significantly repressed by Otx2-WT and -4E, but not by -4A (Fig. 7K). We next tested the combinatorial repressive activity of Otx2 with the interacting partner Gsc and the co-repressor Tle1, which form the repression complex for negative target genes of Otx2 (Yasuoka et al., 2014). Repression of the meis3-D2 reporter by Gsc and Tle1 was enhanced by co-expression of Otx2-WT, and this enhancement was also observed with Otx2-4A (Fig. 7L), suggesting that phosphorylation of Otx2 is not required for, and neither does unphosphorylated Otx2 interfere with, the formation of a ternary repression complex. Given that Otx2-4A and Tle1 did not repress the reporter (Fig. 7L), repression activity by a combination of Otx2-4A, Gsc and Tle1 probably results from recruitment of Tle1 to Gsc. Although Otx2-4A does not interfere with complex formation, Otx2-4A might weaken the repression activity of the complex, as implied in Fig. S7 (compare Otx2-4A with Otx2-4E).

We next examined an activated target gene of Otx2, rax, using the reporter gene SOP-FLASH, which has eight tandemly repeated rax enhancer elements binding for Otx2 and Sox2 (Danno et al., 2008). The data showed that both Otx2-4E and -4A exerted transactivation activity (Fig. 7M), similarly to activation of the xcg1 gene by Otx2-4E and -4A (Fig. 7G,H). Given that Sox2 is expressed in the neuroectoderm (Mizuseki et al., 1998) and has a transactivation domain (Nowling et al., 2000), a combination of Sox2 and Otx2 could function as an activator regardless of the phosphorylation states of Otx2. In addition, because Otx2-4E did not upregulate the endogenous rax gene (Fig. 4C), activation of the rax reporter by Otx2-4E might result from the CRM boosted by eight tandem repeats. Taken together, it is suggested that phosphorylation of Otx2 is required for its repression activity but does not affect the transactivation activity of this protein.

The eh1 repression domain of Otx2 reportedly interacts with TLE/Groucho (Heimbucher et al., 2007). Therefore, we tested the physical interaction between Otx2 mutants and HA-tagged Tle1 by co-immunoprecipitation (Co-IP) assays (Fig. 7N). A co-immunoprecipitated band of HA-Tle1 was detected with Myc-Otx2-4E as well as with Myc-Otx2-WT (magenta arrowheads in Fig. 7N), but barely with Myc-Otx2-4A, consistent with the data that neither Otx2-4A alone nor co-expression with Tle1 exerted repressive activity on the meis3-D2 luciferase reporter (Fig. 7K,L). The reason why the intensity of the co-immunoprecipitated Tle1 band with WT was weaker than that of 4E (Fig. 7N) might be because of the partial phosphorylation of WT. These data suggest that phosphorylation of Otx2 is required for the interaction with Tle1, which enables Otx2 to then exert its own repression activity. The phosphorylation-dependent interaction of eh1 domains with TLE has not previously been reported.

In this study, we showed that Otx2 exists in two distinct states: phosphorylated and unphosphorylated. Our data suggest that phosphorylated Otx2 functions as both a repressor and activator, whereas unphosphorylated Otx2 functions only as an activator. To date, only a few TFs have been reported to be converted from being an activator and a repressor by post-translational modification. The SoxE group of TFs, including Sox9, are converted by SUMOylation from an activator to a repressor by displacing CBP/p300 and recruiting Tle4/Grg4 (Lee et al., 2012). Another case is Pax2, transactivation of which is enhanced through JNK kinase-mediated phosphorylation in the activation domain by blocking TLE/Groucho (Cai et al., 2003). Thus, we have shown a third case, in which phosphorylation of Otx2 enables it to function as a repressor through its interaction with Tle1 (Fig. 7N), opposite to the case of Pax2.

Otx2 has a novel eh1 domain with a Cdk site (SIWSPASISP; underlined is the possible Cdk site), which is only present in the Otx family proteins (Otx1, Otx2 and Otx5/Crx) as well as a few other TFs (Copley, 2005). The eh1 domain is present in many metazoan TFs (Copley, 2005) and interacts with TLE family proteins through its conserved C-terminal WD40 repeat domain (Heimbucher et al., 2007; Pickles et al., 2002). Given that some WD40 domains of ubiquitin ligases and phosphatases, such as β-Trcp and Cdc4, interact with a phosphorylated serine residue in their substrates (Yaffe and Elia, 2001), it is possible that one of the WD40 repeats of TLE interacts with the eh1 domain with a phosphorylated site in Otx2.

We propose a positive feedback loop model for Otx2 functions in cell proliferation (Fig. 6I). In this model, in a proliferative state, Otx2 is phosphorylated to repress target genes, including p27xic1 as well as gbx2 and meis3, whereas, when the cell cycle is arrested, unphosphorylated Otx2 increases to activate the target genes, rax and pax6. Thus, the phosphorylated and unphosphorylated states of Otx2 can be toggled by the cell cycle. However, it is not known how the phosphorylated state of Otx2 is changed to the unphosphorylated state. There are two possibilities: (1) dephosphorylation of Otx2 by phosphatase; and (2) degradation of phosphorylated Otx2. We did not observe any differences in the instability of the Otx2-4E mutant compared with -4A by western blotting (not shown), which does not support the second possibility, and denies phosphorylation-dependent degradation. A putative PEST sequence, an indicator of rapidly degraded protein, is present in the Xenopus Otx2 protein (Mori et al., 1994; Williams and Holland, 1998), and Otx2 might have a short half-life. Therefore, unphosphorylated Otx2 could be generated through the rapid turnover of Otx2 regardless of its phosphorylation state.

We showed that endogenous Otx2 is phosphorylated, but that its phosphorylation levels were lower than those of exogenous Otx2 (Fig. 1F). This difference might be caused by the timing of protein production and the number of proliferating cells where phosphorylation of Otx2 supposedly occurs. For exogenous Otx2, mRNA was injected into the animal pole region at two- or four-cell stages, and translated in the presumptive ectoderm from early cleavage stages; the translation product was phosphorylated until the blastula stage (Fig. S5D,E) by Cdk1 activity, which oscillates with a period of ∼30 min during the cleavage cycle (Hörmanseder et al., 2013). By contrast, in the in vivo situation, the transcription of otx2 starts in the dorsal endoderm and mesoderm at the late blastula stage and in the dorsal ectoderm at the early gastrula stage (Sudou et al., 2012); endogenous Otx2 protein then accumulates in the head organizer and the anterior neuroectoderm, and the proliferation rates in these regions dramatically decrease, especially in the head organizer (Fig. 6H). Therefore, even if phosphorylation levels of endogenous Otx2 in proliferating cells are high, the average phosphorylation level of Otx2 in major nonproliferating and minor proliferating cell populations in a gastrula or neurula embryo is lower than those of exogenous Otx2.

We have proposed a positive feedback loop involving Cdk-dependent phosphorylation of Otx2 for cell proliferation (Fig. 6I). Similar feedback loops involving other TFs are also reported, but the regulatory mode is different from that of Otx2. That is, TFs are switched ‘on’ or ‘off’ by Cdk-dependent phosphorylation to promote cell cycle progression. An example of regulation by an ‘on’ switch is Foxm1. Foxm1 is activated by the phosphorylation of the N-terminal inhibitory domain by cyclin/Cdk to upregulate cyclin B and cdc25B, thereby forming a positive feedback loop for cell proliferation (Park et al., 2008; Major et al., 2004). An example for an ‘off’ switch is Foxo. Foxo upregulates the Cdk inhibitor genes p27kip1 and p21WAF1, but, upon phosphorylation by cyclin/Cdk, undergoes nuclear export, which almost switches off the function of Foxo, leading to cell proliferation (Schuff et al., 2010; Liu et al., 2008; Stahl et al., 2002). In the case of Otx2, Cdk-dependent phosphorylation confers repressor activity to Otx2 for repressing p27xic; otherwise, it functions as an activator on its own, as mentioned above, which differs from simple switch ‘on’ or ‘off’ regulation.

Regarding tissue specification and differentiation, Cdk-dependent phosphorylation generally negatively regulates the activity of TFs that directly activate differentiation genes. For example, the phosphorylation of the myogenic TF MyoD by cyclin/Cdk leads to its degradation to prevent muscle differentiation (Kitzmann et al., 1999). During neurogenesis, the Cdk-dependent phosphorylation of neurogenin 2 (Ngn2, or Neurog2) inhibits its neural differentiation activity (Ali et al., 2011; Hindley et al., 2012). In these cases, Cdk-dependent phosphorylation of the TFs switches off their functions. Taken together, phosphorylation of TFs by cyclin/Cdks either promotes cell cycle progression (e.g. Foxm1 and Foxo) or inhibits tissue differentiation (e.g. MyoD and Ngn2). By contrast, Cdk-dependent phosphorylation not only inhibits the function of Otx2 in differentiation, but also converts it to be able to function in proliferation. Thus, this study has revealed a new category of TFs, in which post-translational modification of a single TF converts it from having a differentiation- to a proliferation-orientated role.

How Otx2 properly functions as an activator or a repressor for its target genes, and coordinates proliferation, tissue patterning and differentiation during embryogenesis, has been unclear. One answer to this question was reported in our previous paper (Yasuoka et al., 2014), showing the combinatorial regulation of target genes with Lim1 (activator) or Gsc (repressor). In the current study, we showed the post-translational regulation of Otx2 activity to be another answer, in which phosphorylation of Otx2 confers repressor activity at least partly through its interaction with Tle1. However, how the combinatorial and post-translational regulations work together remains elusive.

During eye formation, it was reported that expression of a repressor form of Otx2 (Otx2-EnR) by mRNA injection in Xenopus embryos results in a small eye or eyeless phenotype (Isaacs et al., 1999), and that expression of another type of repressor form, EnR-Otx2, in chick eyes by electroporation caused pigmentation defects of the retinal pigment epithelium and a reduction of pax6 (Nishihara et al., 2012). These activities of repressor forms of Otx2 are similar to that of Otx2-4E in Xenopus embryos (Fig. 3E, Fig. 4J,K), consistent with the repressor activity of Otx2-4E (Fig. 7). In contrast with Otx2-4E, Otx2-4A expanded the expression of rax at early neurula stages (Fig. 4D) without stimulating cell proliferation (Fig. 5D,H); at later stages, Otx2-4A expanded the expression of pax6 (retina) but inhibited pax2 (OS) on the ventral side, leading to ventral expansion of the retina. These data suggest that Otx2-4A expands the eye field, and subsequently alters patterning of the neural retina and OS. In other words, enlarged eye phenotypes caused by Otx2-4A result from changes in patterning, not proliferation. Our data are reminiscent of a previous report on the function of Otx2 for Xenopus retina formation, in which an activator type of Otx2 (Otx2-VP16) promotes the bipolar cell fate without stimulating retinal proliferation (Viczian et al., 2003), supporting the possibility that unphosphorylated Otx2 acts mainly as an activator in retina formation.

Several heterozygous OTX2 mutations in human were reportedly linked with severe ocular malformation. For example, point mutations of P133T, P134A and P134R, which occur near one of the four putative phosphorylation sites, S132, found in our study (Fig. S2), are associated with microphthalmia, anophthalmia, sclerocornea and retinal detachment (Beby and Lamonerie, 2013). It is not certain whether these mutations affect the phosphorylation of Otx2 or the interaction with putative binding partners of Otx2, but it is possible that ocular defects in humans are caused by the alteration of phosphorylation states of Otx2.

In summary, we demonstrated that Otx2 undergoes phosphorylation in vivo, and that phosphomimetic and nonphosphorylatable mutants of Otx2 exhibit distinct activities in cell proliferation, patterning and differentiation in Xenopus embryos. Post-translational regulation of Otx2 repressor activity (shown here) in addition to combinatorial regulations with partner TFs (Yasuoka et al., 2014) could enable Otx2 to have versatile roles during development. In vivo analysis of the phosphorylation sites, such as a mutated gene knock-in approach, is now awaited to explore the role of phosphorylated Otx2 in embryonic development.

Point mutants and deletion constructs of Otx2 were made by using PCR methods. PCR fragments of Otx2 mutants were cloned into the pCSf107mT (Mii and Taira, 2009), pCSf107_MTmT (for Myc tagging) and pCSf107_4HAmT (for HA tagging) vectors (Shibano et al., 2015), which contain SP6 terminator sequences. pGEX-GST-ΔN106cyclin B1 (Iwabuchi et al., 2002) was recloned into pCSf107_HAmT. pGEM-Δcyclin A (deletion of 55 amino acids from the N terminus of the protein) was gifted by Dr N. Furuno (Hiroshima University) before publication, and recloned into pCSf107_HAmT. All constructs made in this study are listed in Table S1.

Xenopus laevis and X. tropicalis embryos were obtained by artificial fertilization, dejellied, incubated in 0.1×Steinberg's solution (Peng, 1991), and staged according to Nieuwkoop and Faber (1967). Microinjection experiments were conducted as previously described (Yasuoka et al., 2014). For mRNA synthesis, pCSf107mT constructs with 4×SP6 terminators were transcribed with SP6 polymerase (mMESSAGE mMACHINE SP6 kit, Ambion). mRNAs were injected into one blastomere at the two- or four-cell stage. Nuclear β-galactosidase (nbgal) mRNA (50-100 pg/embryo) or EGFP mRNA (250 pg/embryo) was co-injected for lineage tracing. Antisense or standard control morpholino oligos (MOs) (Gene Tools) were injected into a dorsal blastomere at the four-cell stage. FITC-dextran (5 ng/embryo) was co-injected for lineage tracing. Sequences and specificities of antisense MOs for X. tropicalis otx2 and otx5 were as described previously (Yasuoka et al., 2014).

Lysates were prepared from embryos (stages 10-10.5 unless mentioned otherwise) that had been injected with mRNA (500 pg/embryo) into both of blastomeres in the animal pole region at the two-cell stage. In vitro translation was performed using the TNT SP6 Quick-Coupled Transcription/Translation System (Promega) as previously described (Yamamoto et al., 2003). λ-PP treatment of lysates was carried out as previously described (Shibano et al., 2015). SDS-PAGE was carried out with 12-12.5% polyacrylamide gels. Western blotting was performed essentially as described elsewhere (Shibata et al., 2005), using anti-Myc (9E10), anti-Otx2 (Sudou et al., 2012), anti-p44/42 MAP kinase and anti-phospho-p44/42 MAPK (CST) and anti-HA (12CA5) antibodies. Apparent molecular masses (kDa) of bands were estimated based on the standard curve of size markers (BioDynamics Laboratory).

Possible Akt sites of Otx2 were searched for using the database PhosphoSitePlus, (www.phosphosite.org/). IP-western assays were performed essentially as described elsewhere (Shibano et al., 2015). Immunoprecipitates with anti-Myc antibody were boiled in 2×SDS sample buffer and analyzed by western blotting with anti-Phospho-Akt Substrate antibody (CST). To detect the endogenous Otx2, we modified the preparation of the embryonic lysates. Xenopus neurula embryos were homogenized with lysis buffer A (50 mM Tris-Cl, pH 8.0, 10% glycerol, 0.1%NP-40, 40 μg/ml aprotinin, 20 μg/ml leupeptin, 1 mM PMSF) plus 400 mM NaCl, and homogenates were centrifuged at 4°C for 1 h at 13,200 rpm (16,000 g). Supernatants were diluted four times with buffer A and centrifuged at 4°C for 1 h at 13,200 rpm. Equivalent amounts of lysate were incubated with the anti-Otx2 antibody (Sudou et al., 2012) and 25 μl protein A-agarose beads (Roche). The following wash steps of beads for IP-western assays were performed essentially as described elsewhere (Shibano et al., 2015). Immunoprecipitates with anti-Otx2 antibody were boiled in 2×SDS sample buffer and analyzed by SDS-PAGE with a 15% polyacrylamide gel and western blotting with anti-Otx2 antibody (Abcam, ab21990).

Embryos injected for phenotypic analysis were reared until stages 38-42 and fixed with MEMFA for 1 h. Experiments were carried out with at least three clutches. The lengths of semi-major and semi-minor axes of the eye vesicle were measured, and the eye size were calculated by using the formula for the area of an ellipse.

Immunostaining with anti-phospho-Histone H3 (Ser10) antibody (anti-PH3, 06-570, Merck), DAPI staining of nuclei, and counting of PH3-positive and DAPI-stained nuclei were performed as described elsewhere (Shibano et al., 2015). mRNA or MO with EGFP mRNA or FITC dextran, respectively, as a lineage tracer was injected into one blastomere at the four-cell stage. Embryos with a lineage tracer in the anterior neuroectoderm were selected for counting PH3- or DAPI-positive nuclei in an area (0.44 mm²) in the anterior neuroectoderm on the injected or uninjected side. The effect was scored by dividing the number of nuclei on the injected side with that on the uninjected side. The statistical significance (P value) was calculated using Student's t-test after a one-way analysis of variance (ANOVA). Experiments were carried out with either four or three clutches for the mRNA or MO injection experiments, respectively. WISH was performed according to Harland (1991) with BM Purple (Roche) as a substrate. Antisense RNA probes were transcribed from linearized plasmids as described in Table S2. Experiments for WISH were carried out with two clutches for rax, pax2 and p27xic1, and with three clutches for pax6, gbx2 and xcg1. Hemisection of fixed embryos was performed as described elsewhere (Sudou et al., 2012).

Embryos were co-injected with pGL4p-meis3-D2-luc reporter DNA and mRNA into two animal blastomeres at the two-cell stage, or co-injected with SOP-FLASH reporter (Danno et al., 2008) and mRNA into two dorsal animal blastomeres at the four-cell stage. Five pools of three injected embryos were assayed for luciferase activity at stage 10.5 for pGL4p-meis3-D2-luc reporter, and at stage 12 for SOP-FLASH reporter using the luciferase assay system (Promega) according to the manufacturer's protocol and as described elsewhere (Mii and Taira, 2009). Experiments were carried out at least three times; when similar results were obtained, a representative result is presented.

mRNA for HA-Tle1 (700 pg/embryo) was co-injected with or without mRNA for Myc-Otx2-WT, Myc-Otx2-4A or Myc-Otx2-4E (500 pg/embryo) into the animal pole region of both blastomeres at the two-cell stage. Embryonic lysates were prepared at stage 10.5, and subjected to Co-IP as described elsewhere (Shibano et al., 2015). Myc-Otx2 protein was co-immunoprecipitated with anti-Myc antibody and subjected to SDS-PAGE with a 10% polyacrylamide gel, followed by western blotting with anti-HA and anti-Myc antibodies.

We thank Keita Ohsumi for pGEX-GST-ΔN106cyclin B1 (a stable mutant of cyclin B1), Nobuaki Furuno for pGEM-Δcyclin A (a stable mutant of cyclin A1), Yukiko Gotoh for pSP64T-MAPKK*, and Hidetoshi Saiga and Mariko Kondo for critical reading of the manuscript. Male and female frogs of X. tropicalis were obtained from the National Bioresource Project (Amphibian Research Center, Hiroshima University).