In the Absence of Sonic Hedgehog, p53 Induces Apoptosis and Inhibits Retinal Cell Proliferation, Cell-Cycle Exit and Differentiation in Zebrafish

Background Sonic hedgehog (Shh) signaling regulates cell proliferation during vertebrate development via induction of cell-cycle regulator gene expression or activation of other signalling pathways, prevents cell death by an as yet unclear mechanism and is required for differentiation of retinal cell types. Thus, an unsolved question is how the same signalling molecule can regulate such distinct cell processes as proliferation, cell survival and differentiation. Methodology/Principal Findings Analysis of the zebrafish shh −/− mutant revealed that in this context p53 mediates elevated apoptosis during nervous system and retina development and interferes with retinal proliferation and differentiation. While in shh −/− mutants there is activation of p53 target genes and p53-mediated apoptosis, an increase in Hedgehog (Hh) signalling by over-expression of dominant-negative Protein Kinase A strongly decreased p53 target gene expression and apoptosis levels in shh −/− mutants. Using a novel p53 reporter transgene, I confirm that p53 is active in tissues that require Shh for cell survival. Proliferation assays revealed that loss of p53 can rescue normal cell-cycle exit and the mitotic indices in the shh −/− mutant retina at 24, 36 and 48 hpf. Moreover, generation of amacrine cells and photoreceptors was strongly enhanced in the double p53 −/− shh −/− mutant retina suggesting the effect of p53 on retinal differentiation. Conclusions Loss of Shh signalling leads to the p53-dependent apoptosis in the developing nervous system and retina. Moreover, Shh-mediated control of p53 activity is required for proliferation and cell cycle exit of retinal cells as well as differentiation of amacrine cells and photoreceptors.


Introduction
During organogenesis, cell numbers are tightly controlled by balancing proliferation and cell death. Hh signaling pathway regulates cell proliferation and survival during development in addition to its pattern formation and differentiation functions in many different biological contexts (reviewed in [1,2,3]). Work from many labs has shown that secreted Hh proteins bind to the transmembrane protein Patched (Ptc) and release the inhibition of the transmembrane protein Smoothened (Smo) by Ptc. The signal transduction downstream of Smo inhibits proteolytic processing of Gli2 and Gli3 to repressive forms, which is mediated by Protein Kinase A (PKA) and a few other factors [4]. The resulting fulllength Gli factors can activate their target genes.
Interestingly, in the neural tube Shh signalling promotes cell survival before its requirement for tissue patterning and differentiation [13] arguing against the hypothesis that Shh prevents cell death by providing patterning cues.
Nevertheless, there is still some controversy about the underlying mechanism of the anti-apoptotic role of Shh. Activation of anti-apoptotic genes by Shh signaling is the simplest scenario, and indeed, Gli1 can activate BCL2 gene expression in human cell lines and in the mouse keratinocytes by binding to several nearconsensus sites in the BCL2 promoter [14]. Likewise, overexpression of the Gli3 dominant-active form in developing chick embryos leads to induction of Bcl2 gene [13]. Moreover, another study has proposed that in the absence of Shh, Patched1 mediates apoptosis in the neural tube through caspase activation independent of the canonical Shh signaling pathway [15]. Patched1 could function in this context as a dependency receptor, i.e. a receptor inducing cell death in the absence of cognate ligand binding [16,17]. Recently, Abe and colleagues (2008) have shown that activation of Hh signaling can suppress p53 pathway in human cell lines [18]. The authors have proposed that Hh signaling induces an unknown Mdm2-activating factor leading to phosphorylation of serines 166 and 186 on Hdm2 (human Mdm2), which becomes activated and catalyses ubiquitination of p53. Degradation of ubiquitinated p53 disrupts p53-mediated tumour suppression under conditions of DNA damage and oncogenic stress. Furthermore, Stecca and Ruiz i Altaba (2009) have described a negative feedback loop between p53 and Gli1 in transgenic mouse models [19]. The authors found that p53 inhibits activity of Gli1 by interfering with its normal nuclear localization and directing it to proteosomal degradation. Closing the loop, Gli1 overexpression leads to a striking increase of Mdm2 expression and downregulation of p53 in the mouse brain and several cell lines.
Beside this anti-apoptotic function, Shh is also known to stimulate proliferation by direct regulation of cell-cycle promoting genes or through activation of secondary signaling pathways.
In particular, Hh ligands stimulate proliferation in the mouse skin [20], in the vertebrate central nervous system [13,21,22,23,24,25], and in Drosophila wing [26,27] and eye imaginal discs [28]. In the vertebrate retina, Shh stimulates proliferation of retinal progenitors in vitro [29,30] and in the ciliary marginal zone [31]. Moreover, during mouse retina development Shh-mediated activation of cyclinD1 expression is essential for proliferation and cell-cycle exit of retinal ganglion cell progenitors [32]. Consistent with this, in the zebrafish shh 2/2 mutant retina, proliferation is reduced during neurogenesis [12]. In Xenopus, Shh signalling has been shown to promote retinal progenitor proliferation by activating expression of cell cycle regulators, which shortens G1and G2-phases of the cell cycle, and thereby drives cells towards the final mitosis and cell-cycle exit [33]. Interestingly, in several other contexts Hh signalling has been shown to negatively regulate cell cycle progression [34]. Indeed, in zebrafish shh 2/2 mutant retina and after Hh signaling inhibition by forskolin, most retinal cells fail to express p57Kip2 cyclin-dependent kinase (CDK) inhibitor and do not exit the cell cycle [35,36].
This study shows that loss of shh in zebrafish leads to p53mediated apoptosis in the developing nervous system and retina. Consistent with this, activation of Hh signalling suppresses p53 activity and apoptosis in the shh 2/2 mutant. To monitor p53 activation, I created a p53 reporter transgenic line expressing nuclear green fluorescent protein under control of p53-sensitive promoter, verified by p53 knock-down and being induced by roscovitine treatment. By using this line, I have determined the spatio-temporal activation of p53 and apoptosis in the shh 2/2 mutant. Further, comparisons of double p53 2/2 shh 2/2 mutants with wild-type, and shh 2/2 mutant embryos identified p53mediated inhibition of retinal progenitor proliferation, cell-cycle exit and differentiation of amacrine cells and photoreceptors. Taken together, these data provide novel genetic evidence that Shh regulates cell survival during nervous system and retina development as well as retinal proliferation and differentiation through inhibition of p53 pathway.

Results
p53 mediates apoptosis in shh 2/2 mutant via the intrinsic apoptotic pathway To determine the mechanism of cell death due to loss of shh, I analysed the expression of anti-apoptotic Bcl-2 family genes bcl2, bcl2l, mcl1a and mcl1b in shh 2/2 mutant and wild-type embryos, their expression being comparable in embryos of both genotypes at 24 hours post fertilization (hpf) (Fig. S1). I next looked at the expression of p53-dependent genes by in situ hybridisations which revealed consistently stronger expression of p53 target genes p53, cyclinG1, bax1, puma and p21 at 24 hpf in the neural tube, retina and somites of shh 2/2 mutant relative to wild-type embryos (Fig. 1A) and elevated expression of p53 at 56 hpf in the shh 2/2 mutant retina, midbrain and branchial arches (Fig. 1B). The observed elevated p53 expression was due to induction of the short D113 p53 transcript isoform [37,38] in the shh 2/2 mutant, whereas the level of the full-length p53 transcript was comparable between shh 2/2 mutant and wild-type embryos (Fig. 1C). To better measure p53 target gene expression levels, I performed quantitative polymerase chain reaction (qPCR) analyses on shh 2/2 mutant versus wild-type embryos at 24 hpf and 56 hpf confirming elevated expression of p53, p21 mdm2, cycG1, bax1 and puma in shh 2/2 mutants at both stages (Fig. 1D, E). The pro-apoptotic genes bax1 and puma, but not noxa, showed higher expression in shh 2/2 mutants compared to wild-type embryos at both 24 hpf (Fig. 1D) and 56 hpf (Fig. 1E), the relative expression of puma, the most potent inducer of mitochondrial apoptosis [39,40], being higher in shh 2/2 mutant embryos at 24 hpf (Fig. 1D) than at 56 hpf (Fig. 1E) correlating with a higher level of apoptosis at 24 hpf (Fig. 2).
To confirm that p53 is essential for elevated apoptosis in the shh 2/2 mutant, I used morpholino oligonucleotides (MO) [41] knock-down against p53 previously shown to effectively block p53 translation during zebrafish development [42]. Injection of p53 MO but not of 4-mismatch control MO into the shh 2/2 mutant embryos suppressed elevated apoptosis in the neural tube at 24 hpf ( Fig. 2A) as confirmed using image quantification of apoptosis stainings ( Fig. 2A') and in the retina at 56 hpf (Fig. 2C). Apoptosis can occur through either extrinsic pathway when cells receive external signals activating death receptors or intrinsic (mitochondrial) pathway when pro-apoptotic factors are released from mitochondria after their permeabilization, p53 activating primarily the intrinsic rather than the extrinsic apoptotic pathway [43]. To test whether p53 activates the mitochondrial pathway of apoptosis in the shh 2/2 mutant, I injected mRNA of EGFP-bcl2, an inhibitor of the mitochondrial apoptotic pathway, or EGFP into shh 2/2 mutant and wild-type embryos. Expression of EGFP-bcl2 significantly reduced the level of apoptosis in shh 2/2 mutants at 24 hpf, whereas EGFP did not have any effect (Fig. 2B, B') suggesting that the p53-mediated apoptosis in the shh 2/2 mutant occurs via the intrinsic pathway.
p53-dependent apoptosis in the shh 2/2 mutant retina To determine if genetic loss of p53 can rescue excessive apoptosis in the shh 2/2 mutant, p53 2/2 shh 2/2 double mutant animals were generated using the previously published tp53 M214K mutant zebrafish [44]. Indeed, high level of apoptosis in the shh 2/2 mutants at 24 hpf ( Fig. 3A-b) and in the retina at 72 hpf ( Fig. 3Bb) compared to wild-type ( Fig. 3A-a, 3B-a) was reduced in the p53 2/2 shh 2/2 mutant (Fig. 3A-c) to the wild-type level (Fig. 3A-c, 3B-c). I further examined differentiation status of apoptotic cells in the wild-type and shh 2/2 mutant retinas at 72 hpf by staining cells for neuronal marker HuC and apoptotic cells using TdT-mediated dUTP-biotin nick end labelling (TUNEL). In the wild-type retina, both HU-positive and negative apoptotic were present ( Fig. 3C-a). By contrast, most apoptotic cells in the shh 2/2 mutant retina were undifferentiated (HuC-negative) ( Fig. 3C-b), but some were HUpositive ( Fig. 3C-b) suggesting that differentiating cells also die by apoptosis in the shh 2/2 mutant retina.
PG13p21::nlsEGFP, a p53 reporter in the living zebrafish embryos To study the behaviour and characteristics of cells which have undergone activation of p53, I constructed a transgenic reporter to p53 Activation in shh Mutant Figure 1. Induction of p53 target gene expression in the shh 2/2 mutant. (A) 24 hpf wild-type and shh 2/2 mutant embryos were in situ hybridised with probes against p53, cyclinG1, bax1, puma and p21 mRNAs. These p53 target genes are expressed at a higher level in the shh 2/2 mutant in neural tube, retina and somites than in wild-type embryos. (B) 56 hpf wild-type and shh 2/2 mutant embryos were hybridized with the probe against p53 mRNA. The hybridizations show increased expression of p53 in shh 2/2 mutant compared to wild-type embryos in the brain, retina and branchial arches. The following structures are labelled: retina, midbrain (mb) and branchial arches (ba). (C) p53 transcript isoforms in shh 2/2 mutant and wild-type at 24 and 56 hpf were analysed by semi-quantitative PCR assays for full-length p53, D113 p53 transcripts and b-actin transcript as a loading control. At both stages, full-length p53 and b-actin transcripts were expressed at similar levels in wild-type and mutant samples, whereas D113 p53 was strongly upregulated in shh 2/2 mutant samples. (D, E) qPCR assays for p53, p21, mdm2, cyclinG1, bax1, puma, noxa and gapdh were performed at 24 hpf (D) and 56 hpf (E) on wild-type and shh 2/2 mutant embryos. cDNA from whole embryos was used for 24 hpf samples and cDNA from embryo heads was used for 56 hpf samples. Gene expression was normalized using gapdh expression as reference. The panels show p53 target gene expression in wild-type and shh 2/2 mutant embryos (D, E). Error bars represent standard deviation and an asterisk indicates statistical significance. There are significant expression differences between wild-type and shh 2/2 mutant samples (t-test, * -p,0.05) for all genes except noxa. qPCR assays were performed in triplicates on three different clutches of embryos and standard deviations are shown. doi:10.1371/journal.pone.0013549.g001 p53 Activation in shh Mutant PLoS ONE | www.plosone.org monitor p53 activity using a strategy similar to the one used to construct a p53 reporter in human cell lines [45]. A 2,4 kb promoter fragment of human CDKN1A encoding p21 CDK inhibitor and 13 copies of the optimal p53 binding site (PG13) [46] were combined into a p53-sensitive promoter. Tol2kit transgenic system [47] was then used to assemble the final construct containing PG13p21 promoter, nlsEGFP coding sequence and a polyA signal (Fig. 4A). (A) Apoptosis labelling using ApopTag staining in wild-type and shh 2/2 mutant embryos at 24 hpf after injection of p53 morpholino or control morpholino. The whole-mount images of the head (head) and neural tube (nt) are shown. (A') Quantification of ApopTag signal area in wild-type and shh 2/2 mutant embryo head images from (A) shows that p53MO successfully decreased apoptosis in shh 2/2 mutant to the wild-type level. Asterisks (***) on top of the shh 2/2 mutant bar indicate its significant statistical differences from other samples (t-test, P-value,0,001). (B) Apoptosis labelling using ApopTag staining in wild-type and shh 2/2 mutant embryos at 24 hpf after injection of of EGFP-bcl2 or EGFP mRNA. The whole-mount images of the head (head) and neural tube (nt) are shown. (B') Quantification of ApopTag signal area in wild-type and shh 2/2 mutant embryo head images from (B) shows that EGFP-bcl2 successfully decreased apoptosis in shh 2/2 mutant to the wild-type level. Asterisks (***) on top of the shh 2/2 mutant bar indicate its significant statistical differences from other samples (t-test, P-value,0,001). (C) Retinal apoptosis observed in shh 2/2 mutant at 56 hpf was suppressed by injection of p53 morpholino but not by injection of control morpholino (n = 10 and 40 sections for shh 2/2 mutant injected with p53 morpholino; n = 10 and 34 sections for shh 2/2 mutant injected with control morpholino). *** -t-test, P-value,0,001. The p53 reporter zebrafish transgenic line is a general tool to study p53 activation and regulation. Studies of p53 in zebrafish employed camptothecin, a genotoxic drug, or roscovitine, a CDK inhibitor, to induce p53-dependent apoptosis [42,48]. To verify that p53 reporter line can be used as a bona fide p53 reporter in vivo, I have treated p53 reporter transgenic fish with roscovitin. Wild-type embryos carrying the p53 reporter transgene were incubated either with 50 mM roscovitine or 0.4% DMSO as a control for 6 hours starting at 24 hpf. p53 reporter expression was strongly up-regulated in the brain and retina of roscovitine-treated but not in control DMSO-treated embryos (Fig. 4B). Furthermore, knocking down p53 with morpholino in shh 2/2 mutant strongly inhibited expression of p53 reporter transgene at 24 hpf (Fig. 4C). Moreover, in the shh 2/2 mutant retina at 56 hpf, most apoptotic cells showed co-localisation of active Caspase3 labeling with p53 reporter nlsEGFP in the cytoplasm (Fig. 4D) suggesting that apoptosis is induced in cells with active p53. Co-localised pixels made 20,9 +/2 3,8% of total active Caspase 3 pixels, but only 11,7 +/2 4,5% of total p53 reporter nlsEGFP pixels, the difference being significant (t-test, P-value,0,001), (Fig. 4E) suggesting that apoptotic cells expressing p53 reporter are more frequent than p53 reporter-positive cells undergoing apoptosis, consistent with p53 activating caspases as a downstream result of its activity.
Expression of p53 reporter in tissues requiring Shh for survival I have analysed p53 reporter expression at 24 hpf and 56 hpf in the shh 2/2 mutant and wild-type embryos (Fig. 5). At 24 hpf, shh 2/2 mutants expressed much higher level of nlsEGFP and contained many more TUNEL-positive cells in midbrain, hindbrain and spinal cord than wild-type embryos (Fig. 5B, C, D). I also detected that at 56 hpf shh 2/2 mutant hindbrain expressed the p53 reporter at a higher level than wild-type embryos (Fig. 5F). By contrast, shh 2/2 mutant retina showed strong expression of the p53 reporter but no elevated apoptosis at 24 hpf (Fig. 5A). This result is interesting and unexpected as shh expression starts in the retina at 28 hpf [49]. However, at 56 hpf in the shh 2/2 mutant retina, both p53 reporter expression and number of TUNEL-positive cells are high compared to wild-type (Fig. 5E). Activation of p53 in shh 2/2 mutant retinal cells before 28 hpf indicates that they require Shh signalling at an earlier stage to control p53 activity. In the absence of Shh, neural crest cells undergo apoptosis [10] and indeed some p53 reporter positive cells present in the midbrain are located in a region populated by neural crest cells (Fig. 5B). p53 reporter also labeled somite cells (Fig. 5D), which are likely sclerotome cells requiring Shh for their survival [7].
Hh signalling activation by dnPKA-GFP decreases cyclinG1 expression and apoptosis in shh 2/2 mutant embryos at 12-somite stage In tissue culture experiments, p53 activation has been shown to be inhibited by Gli-mediated activation of an unknown p53 pathway inhibitor [18]. This leads to the question whether activation of Hh signalling inhibits p53 pathway in vivo in the absence of Shh. To activate the canonical Hh/Gli signaling pathway, I injected mRNA of dominant-negative PKA fused with GFP (dnPKA-GFP) [50,51] or EGFP mRNA as a control into progeny of fish heterozygous for the shh deletion. Given the limited mRNA stability, I analysed the effects of RNA injections at the 12 somites stage, when shh 2/2 mutant embryos already show a higher level of p53 target gene expression and apoptosis than wild-type embryos (Fig. S2). While all the EGFP-injected embryos showed unaltered wild-type or shh 2/2 mutant patched1 expression pattern ( Fig. 6A, D, ''low''), over-expression of dnPKA-GFP led to ectopic spreading of patched1 expression in 87,7 +/2 3,8% of embryos ( Fig. 6A, D, ''high''). I next assessed expression of cyclinG1, p53 target gene, after mRNA injections by visually scoring expression strength as ''high'' or ''low'' based on staining of uninjected embryos (Fig. 6B), where embryos with ''high'' cyclinG1 expression were inferred to be shh 2/2 mutant and those with ''low'' expression were wild-type (control stainings of wild-type embryos, not shown). In this experiment, the effect of dnPKA-GFP on p53 target gene expression is inferred from the change in proportion of embryos with stronger cyclinG1 expression. While after EGFP mRNA injection the proportion of embryos with ''high'' expression of cyclinG1 was 25,3 +/2 1,2% (Fig. 6D), only 6 +/2 2% (Fig. 6D) of dnPKA-GFP -injected embryos had perceptibly stronger cyclinG1 expression than the other embryos in the samples (Fig. 6D). Overexpression of dnPKA-GFP also reduced the proportion of embryos with higher apoptosis levels (  (Fig. 6D). Taken together, these data indicate that activation of Hh signalling using dnPKA-GFP can suppress p53 target expression and apoptosis.
I then addressed the hypothesis that p53 loss in the shh 2/2 retina induces expression of p57kip2, a cyclin-dependent kinase inhibitor essential for cell-cycle exit in the zebrafish retina. Indeed, at 38 hpf, p57kip2 was expressed in large areas of wild-type and p53 2/2 shh 2/2 mutant retinas, but absent from the shh 2/2 mutant retina (Fig. 8C), whereas p53 was much more strongly expressed in the shh 2/2 mutant retina than in the wild-type and p53 2/2 shh 2/2 mutant retinas (Fig. 8C). Altogether, these results suggest that the delay of cell-cycle exit in the shh 2/2 mutant retina is dependent on the p53 activity inhibiting expression of the CDK inhibitor p57kip2.
p53 loss rescues amacrine and photoreceptor cells in the shh 2/2 mutant retina To address the question whether p53 affects differentiation in the shh 2/2 mutant, retinal sections of wild-type, shh 2/2 and p53 2/2 shh 2/2 mutant embryos were stained for markers of retinal cell types. Staining for HuC neuronal marker strongly labelled both ganglion and amacrine cells in the wild-type (Fig. 9A-a), but only ganglion cells in the shh 2/2 mutant retina ( Fig. 9A-b), whereas in the p53 2/2 shh 2/2 mutant retina, both ganglion and amacrine cells were labelled (Fig. 9A-c). Consistently, retinal ganglion cells (RGCs) identified by anti-zn5 staining were present in retinas of all three genotypes (Fig. 9A-d, e, f). However, amacrine cells labelled by antiparvalbumin antibodies in the wild-type retina (Fig. 9A-g), were absent in the shh 2/2 retina (Fig. 9A-h) and partially rescued in p53 2/2 shh 2/2 embryos (Fig. 9A-i, C). The ratio of area covered by parvalbumine signal to the total retinal section area in the wild-type retina (5,19 +/2 1,6%) was significantly different from this ratio in the p53 2/2 shh 2/2 retina (2,03 +/2 0,32%), the efficiency of rescue being 39% (Fig. 9C). Investigating whether p53 affects differentiation of photoreceptors, it was found that red-green cone and rod photoreceptors present in the wild-type retina ( Fig. 9A-j, m) are absent in the shh 2/2 retina (Fig. 9A-k, n) and rescued in the p53 2/2 shh 2/2 mutant retina ( Fig. 9A-l, o), the rescue being complete since the relative areas of photoreceptor labeling were not different between retinas of wild-type and p53 2/2 shh 2/2 embryos (Fig. 9C). Müller glia, a cell type required to maintain the retinal environment, was present in the wild-type retina at 72 hpf ( Fig. 9A-p) but nearly absent in both shh 2/2 (Fig. 9A-q) and p53 2/2 shh 2/2 mutant retinas (Fig. 9A-r). Bipolar cells, a retinal cell type involved in signal transduction during light perception, were labelled by staining for PKCa in the wild-type retina at 72 hpf (Fig. 9A-s), but were absent in both shh 2/2 and p53 2/2 shh 2/2 mutant retinas (Fig. 9A-t, u). However, at 5 days post fertilization (dpf), Müller glia and bipolar cells appeared in both shh 2/2 ( Fig. 9Bb, e) and p53 2/2 shh 2/2 (Fig. 9B-c, f) mutant retinas at similar levels indicating that their differentiation is strongly delayed in the absence of Shh and is not affected by p53 loss. As control, I checked whether loss of p53 alone affects retinal differentiation and found that differentiation of retinal cell types in wild-type and p53 2/2 mutant embryos was comparable (Fig. S3). These results suggest that p53 inhibits differentiation of amacrine cells and photoreceptors in the shh 2/2 mutant, without affecting differentiation of other cell types.

p53 is essential for elevated apoptosis in the absence of Shh
In vertebrates, Shh signalling is crucial to prevent cell death in many different contexts, but the mechanism of Shh-mediated survival regulation is still unclear. To clarify this mechanism, I focused on apoptosis in the neural tube and retina of the shh 2/2 mutant.
I first examined Bcl2 family gene expression as a potential cause of apoptosis in the shh 2/2 mutant but haven't detected any differences between wild-type and shh 2/2 mutant embryos (Fig.  S1). However, activation of Hh signalling has been shown to induce expression of bcl2 gene in mouse keratinocytes [14], chick spinal cord [13] and to maintain Bcl2 gene expression in the mouse lymphoma cells [53]. Thus, Hh signalling activates Bcl2 expression in certain contexts, but my results suggest that Shh signaling is not required for global regulation of Bcl2 family genes during zebrafish development.

PG13p21::nlsEGFP transgenic line faithfully reports regions with p53 activation
To detect cells that have undergone p53 activation, I constructed the PG13p21::nlsEGFP transgenic p53 reporter. Expression of this reporter in shh 2/2 mutant was dependent on p53, inducible by roscovitine and co-localised with activeCaspase3 labeling in the shh 2/2 mutant at 56 hpf (Fig. 4) suggesting that the PG13p21::nlsEGFP line is a bona fide p53 reporter.
Expression of p53 reporter in the shh 2/2 mutant retina at 24 hpf was much higher than in the wild-type retina but few apoptotic cells were detected in both cases (Fig. 5), suggesting that extraretinal Shh controls p53 activity during early retinal development before the onset of Shh production in the retina at 28 hpf [49]. However, at 56 hpf, apoptotic cells and p53 reporterpositive cells are both present in the retina indicating a delay between initial p53 activation and apoptosis induction in the shh 2/2 mutant retina probably due to a high level of pro-survival factors in the early retina. Extraretinal Hh signalling is known to be required for early retinal patterning, for example, Stenkamp and Frey (2003) showed that inhibition of Hh signalling from 10 hpf reduces ath5 retinal patterning gene expression stronger than the same treatments during retinal neurogenesis [56], and Kay and co-authors (2005) demonstrated that Hh signalling inhibition from 13 hpf but not from 25 hpf can delay expression of ath5 [57]. Moreover, Masai et al. (2005) found using forskolin, an activator of PKA, that Hh signalling during 22-26 hpf period is essential for ath5 expression wave, also highlighting the role of extraretinal Hh signalling for retinal patterning [36]. My findings are consistent with these studies, and also implicate extraretinal Hh signalling in regulation of p53 activity in the retina.
In addition, I observed p53 reporter activation and apoptosis in the midbrain, hindbrain and spinal cord regions of the neural tube (Fig. 5), consistent with the previous studies on the survival function of Shh in these tissues [6,9]. Although p53 reporter expression but no elevated apoptosis were detected in somites of the shh 2/2 mutant (Fig. 5D), p53 may mediate apoptosis in the somites of the shh 2/2 mutant at earlier stages, which would make the elevated p53 reporter expression fully consistent with the role of Shh in sclerotome cell survival and myogenic differentiation [7]. Locations of p53 reporter-positive cells in the lateral midbrain are suggestive of their neural crest origin (Fig. 5), agreeing with the requirement of Shh for their survival [10]. Taken together, these results show high expression of the p53 reporter transgenic reporter in the regions requiring Shh for cell survival and the advantages of p53 reporter over in situ hybridizations for p53 target genes or antibody stainings against p53 [48] thanks to the possibility to do live imaging, label specific cell parts and change the fluorescent protein used for labeling.
Activation of Hh signalling suppresses p53 pathway in the shh 2/2 mutant Activation of p53 in the absence of Shh raises the question whether this elevated p53 activity can be suppressed by Hh signaling activation. Indeed, Hh signaling activation using dnPKA-GFP effectively suppressed p53 target gene expression and apoptosis in the shh 2/2 mutant at 12somite stage (Fig. 6). While this study was in progress, Abe et al. (2008) have suggested that Hh signaling regulates stability of p53 by activating Mdm2 in human cell lines by inducing an unknown Mdm2-activating factor [18]. Stecca and Altaba (2009) characterized the interaction between p53 and Gli1, a downstream Hh signalling effector, in wild-type mouse neural stem cells [19]. These authors found that Gli1 can suppress p53 activity by up-regulating Mdm2 level. Since p53 activity can be inhibited by Hh signaling activation in human cell lines, mouse neural stem cells and zebrafish embryos, the same mechanism is likely involved.
Concerning p53 regulation, the question is whether Hh signaling is required for establishment or maintenance of p53 activity control. Given the localized nature of Shh expression, cells perceiving Shh signal are unlikely to require persistent Shh signaling for control of p53 activity, but rather only during a certain period of development. Shh is clearly not required to establish control of p53 activity since p53 can already be activated in the late gastrulation [40], which is much earlier than the 10somite stage. However, upon the onset of Shh expression, control of p53 pathway appears to become dependent on Shh signaling.
Regulation of retinal cell proliferation by p53 in the shh 2/2 mutant p53 loss in the shh 2/2 mutant rescues two aspects of retinal proliferation: cell-cycle exit of retinal cell progenitors (Fig. 8) and  retina (A, b). In the p53 2/2 shh 2/2 retina, however, labelling of both ganglion cells and amacrine cells was partially restored (A, c). Ganglion cells labelled by an anti-zn5 antibody were present in comparable numbers in wild-type (A, d), shh 2/2 mutant (A, e) and p53 2/2 shh 2/2 mutant retinas (A, f). Amacrine cells labelled by anti-parvalbumine antibody were present in the wild-type retina (A, g), absent in the shh 2/2 mutant (A, h) and partially rescued in the p53 2/2 shh 2/2 mutant retina (A, i). Red-green double cones (zpr1 antibody labelling) and rod photoreceptors (zpr3 antibody labeling) were present in the wild-type retina (A, j, m), absent in the shh 2/2 mutant retina (A, k, n) and rescued in the p53 2/2 shh 2/2 mutant retina (A, l, o). Staining for glutamine synthetase (GS) labelled Mü ller glia cells in wild-type embryos (A, p). In both shh 2/2 and p53 2/2 shh 2/2 retinas, very few Mü ller glia were present (A, q, r). Bipolar cells were labelled in the wild-type retina by anti-PKCa antibody staining (A, s), but absent in shh 2/2 and p53 2/2 shh 2/2 retinas (A, t, u). At 5 dpf, Mü ller glia and bipolar cells were detected in shh 2/2 and p53 2/2 shh 2/2 retinas (B, b,c,e,f) but in lower numbers than in the wild-type retina (B, a,d). All stainings were performed on at least 6 embryos for each genotype and representative images are shown. (C) Quantitation of rescue of amacrine cells (parvalbumine labelling), red-green double cone (zpr1) and rod photoreceptors (zpr3) in the p53 2/2 shh 2/2 mutant retina by plotting ratios of segmented area for each staining to the total retina area. Amacrine cells occupy a significantly larger relative area in the wild-type than in the p53 2/2 shh 2/2 mutant retina (* on top of the bars, t-test, Pvalue,0,05), whereas relative areas of both types of photoreceptors are not significantly different in wild-type and p53 2/2 shh 2/2 mutant retinas. doi:10.1371/journal.pone.0013549.g009 the mitotic indices before and during neurogenesis (Fig. 7). Moreover, expression of the cyclin-dependent kinase inhibitor p57kip2 is induced in the p53 2/2 shh 2/2 mutant retina (Fig. 8C) arguing for normal cell-cycle exit. A normal mitotic index in the p53 2/2 shh 2/2 mutant retina before neurogenesis (24 hpf) is likely to result from lack of p53-activated cell cycle inhibitory gene expression, while the rescue of cell-cycle exit and normal mitotic indices in the retina during neurogenesis may be due to both lack of cell-cycle inhibitor expression and more mitoses of differentiating cells upon cell-cycle exit. Previous studies of cell-cycle exit in the zebrafish retina [35,36] showed that Hh signalling is essential for p57kip2 expression in the retina. I support this conclusion and propose that Shh signalling regulates retinal expression of p57kip2 by controlling activity of p53.
p53-mediated Notch1 expression may be responsible for failure of shh 2/2 mutant embryos to express p57kip2. Notch1 is a p53 target in human keratinocytes [58] and myeloid and lymphoid cells [59]. Moreover, bioinformatic analyses suggest that Gli2 can prevent p53 from activating Notch1 gene expression, which could be a connection between p53, Hh and Notch pathways [60]. In zebrafish, Notch1 homologs are also likely to be p53 targets since they contain optimal p53 binding sites (unpublished results). Notch signaling inhibits p57kip2 expression during zebrafish neural tube development [61], during mouse pancreas development [62], in the adult mouse intestinal crypts [63] and during eye lense development in the mouse [64]. Whether p53 blocks retinal cellcycle exit in the absence of shh by activating Notch signalling will undoubtedly be uncovered by future studies.

Coordinate regulation of cell cycle, cell survival and differentiation in the retina
This study shows that p53 inhibits generation of amacrine and photoreceptor cells in the shh 2/2 mutant retina, raising a question whether p53 loss is essential only for their survival or also for differentiation. My data suggest that in the shh 2/2 mutant, p53 loss is essential both for survival and cell-cycle exit of retinal progenitors, which must occur before cells can differentiate. By contrast, differentiation of ganglion cells, bipolar and Müller glia cells the shh 2/2 mutant retina is not affected by p53 suggesting that Shh signalling promotes retinal differentiation not only by controlling p53 activity but also by other p53-independent mechanisms.
In the retinal progenitor cells, differentiation decisions are coupled to regulation of cell cycle. This study has shown that p53 affects both retinal cell cycle and differentiation phenotypes in the shh 2/2 mutant embryos implicating p53 pathway in coupling of cell cycle regulation and differentiation in the retina. Other studies identified some of the mechanisms coordinating cell cycle and differentiation decisions in the retina. Retinoblastoma 1 (Rb1) promotes retinal differentiation through inhibition of both E2f1 and E2f3a and its loss in the developing mouse retina leads to ectopic proliferation and elevated apoptosis rescued by E2f1 loss [65]. There is a functional analogy between Rb1-E2f and Shh-p53 interactions, Rb1 and Shh being required for proper cell cycle regulation, differentiation and control of E2f factors and p53, respectively. By contrast, in the postnatal mouse retina, Rb has been proposed to promote rod photoreceptor differentiation through activation of Nrl transcription factor [66]. CDK inhibitors (CDKI) are essential for cell-cycle exit of retinal progenitors, but are known regulate their differentiation through cell-cycle independent mechanisms, the importance of CDKI-mediated cell cycle inhibition still being unclear (reviewed in [67]). Studies by Ohnuma and colleagues showed that p27Xic1, a Xenopus CDK inhibitor, promotes generation of Müller glia at the expense of bipolar cells [68]. In this context, p27Xic1 interacts functionally with Notch/Delta signalling and proneural genes such as Xath5 to precisely regulate neurogenesis sequence [69,70]. Consistent with the lack of amacrine cells in the shh 2/2 mutant, mouse p57Kip2 is involved in development of a subset of amacrine cells [71].
Regulation of cell death is another powerful mechanism known to shape organs. Several studies of zebrafish mutants addressed whether suppression of p53-mediated apoptosis affects retinal differentiation. In mutants of DNA polymerase delta 1 subunit (pold1), suppression of apoptosis in the retina by p53 knockdown rescues eye size and morphology [72] and in the primase1 mutant, p53 knockdown led to the rescue of amacrine cell and photoreceptor differentiation [73]. By contrast, p53 knockdown in chromatin assembly factor 1b mutant didn't improve retinal or cartilage differentiation [74], since this gene is required for both cell survival and differentiation of most cell types. Rescue of amacrine cells and photoreceptors by p53 loss in the shh 2/2 mutant retina is similar to p53 knockdown results in pold1 and primase1 mutants, but Shh likely affects differentiation by regulating p53 activity and through other mechanisms, whereas mutations of pold1 or primase1 were proposed to induce genotoxic damage. Other studies identified context-dependence of p53 effects on differentiation, for example, p53 promotes differentiation in mouse embryonic stem cells after DNA damage by decreasing Nanog expression [75] and reduces spontaneous differentiation of human stem cells [76].

Conclusions
These results identify p53 pathway as the essential mediator of the elevated apoptosis in the developing nervous system and retina in the absence of Shh in zebrafish. I also find that Shh signalling promotes cell survival in this context by inhibiting p53 pathway. In the retina, Shh controls p53 to regulate both cell survival and proliferation, since in the shh mutant loss of p53 suppresses elevated apoptosis, leads to normal mitotic indices in the retina and rescues cell-cycle exit. Control of p53 by Shh is also essential for differentiation of some amacrine cells and photoreceptors.

Fish stocks, maintenance, identification and care
Wild-type WIK, sonic-you t4 [77] and p53 zdf1 [44] heterozygous fish were used. Fish were maintained according to standard protocols. Embryos were grown in E3 embryo medium at 28uC with or without the addition of 0.003% 1-phenyl-2-thiourea (Sigma) to inhibit pigmentation. Staging was performed according to hours post-fertilization (hpf) [78]. Identification of p53 genotype was performed according to a published protocol [44]. Fin clips were obtained by anesthesizing adult fish in 0.02% Tricaine and then cutting small pieces of the caudal fin. All DNA samples for identification were prepared as described previously [79]. sonicyou t4 mutants were identified by their curled body shape and Ushaped somites.

BrdU labelling
Embryos were anesthesized with 0.02% Tricaine and injected with a small drop (5-10 nl) of 10 mM 5-Bromodeoxyuridine (BrdU) into the yolk sac. After injection, embryos were allowed to develop for a required period of time and fixed in PFA for analysis. BrdU-labeled embryo were then cryosectioned and the sections were dried, washed in PBST (PBS + 0.1% Tween20 (Sigma, cat# P9416)), incubated in 2N HCl for 10 minutes at 23uC and then transferred to 37uC for another 30 minutes and washed again in PBST. Anti-BrdU staining was done as described in the next section of Materials and Methods, the sections finally being stained with 20 mg/ml of DAPI. Quantification of BrdU-positive nuclei was done using ''Colocalization Threshold'' analysis of BrdU and DAPI signals and ''Cell Counter'' plugins of Fiji software (http:// pacific.mpi-cbg.de) as described below.

Immunohistochemical methods
Antibody stainings were done on cryosections as described before [35]. The following antibodies were used: anti-zn5, anti-  p53 Activation in shh Mutant blocking buffer and added to secondary antibody solutions. Images were taken using Leica SP2 confocal microscope and processed using Adobe Creative Suite CS4.

Co-localisation analysis
Co-localisation was analysed using ''Colocalisation Threshold'' plugin of Fiji software. The plugin takes two images and outputs several statistics including percent of co-localised pixels relative to total pixels in each of the co-localising channels and an image, where pixels present in both images are white.
qPCR and semi-quantitative PCR assays RNA from whole embryos or embryo heads was purified using RNeasy kit (Qiagen, cat# 74104) and cDNA samples were prepared using Quantitect Reverse Transcription Kit (Qiagen, cat# 205311) and used for quantitative real-time PCR after 20fold dilution. SYBRH Green PCR Master Mix (Applied Biosystems, cat# 4309155) was used to prepare qPCR reactions, which were run on the ABI7500 qPCR machine. Primers used for qPCR assays and appropriate references are listed in Table 2. Semiquantitative PCR for full-length p53, its D113 isoform and b-actin transcripts was performed essentially as described before [38].
Counting procedure TUNEL-positive, DAPI-stained nuclei and pH 3-labelled cells were counted manually using the Cell Counter plugin of ImageJ, which allows labelling already counted objects and automatically records their number.

Image quantification
Image quantification in Figs. 2A-B and Fig.9 was done using Fiji software. To quantify the area of particular signal in an image, raw RGB images were split into channels, and single channel images were processed to remove irrelevant parts of the image (select and Edit Clear Outside). The processed images were thresholded, the threshold being the same for all images in an experiment and manually chosen for optimal segmentation of the signal. Thresholded images were then segmented using the Analyze Particles plugin (Analyze Analyze Particles), which outputs the area of the segmented signal.

Statistical analyses
The differences of cell proportions and counts were statistically evaluated using Student's t-test in Microsoft Excel 2007. qPCR data were initially processed using the SDS1.2 software from Applied Biosystems to obtain Ct values, that is the amplification cycle number at which a reaction reaches a certain threshold of expression. Relative gene expression was calculated using the DDCt method ( [82]) assuming equal amplification efficiencies. Ct values of all samples were normalised by reference gene expression and by average wild-type gene expression before being converted to fold expression differences. Student's t-test was then applied to fold expression differences (p,0.05 was set as a significance level). Statistical significance of embryo proportions in the Hedgehog signalling activation experiment (Fig. 6) was determined using Fisher's exact test using fisher.test function from R language environment (R version 2.9.0). Figure S1 Expression of anti-apoptotic Bcl2 family genes in shh2/2 mutant versus wild-type zebrafish embryos at 24 hpf. In situ whole-mount analysis of bcl2, bcl2l, mcl1a and mcl1b expression in shh2/2 mutant and wild-type embryos at 24 hpf.

Author Contributions
Conceived and designed the experiments: SVP. Performed the experiments: SVP. Analyzed the data: SVP. Contributed reagents/materials/ analysis tools: SVP. Wrote the paper: SVP.