Zic1 controls placode progenitor formation non-cell autonomously by regulating retinoic acid production and transport

All cranial placode progenitors arise from a common precursor field anterior to the neural plate, the pre-placodal region (PPR). We showed that transcription factor Zic1, expressed at the anterior neural plate, is necessary and sufficient to promote placode fate. Here we reveal a non-cell autonomous activity of Zic1 and implicate retinoic acid (RA) signaling as a key player in cranial placode progenitor specification. In a screen for genes activated by Zic1 we identify several factors involved in RA metabolism and function. Among them we show that retinaldehyde dehydrogenase 2 (RALDH2) and lipocalin-type prostaglandin D2 synthase (LPGDS), which respectively regulate the synthesis and transport of RA, directly participate in the establishment of the PPR. We propose that RALDH2 and LPGDS induction by Zic1 at the anterior neural plate allows for the localized production and transport of RA, which in turn activates a cranial placode developmental program in neighboring cells.


Introduction
Cranial sensory placodes are thickenings of the embryonic head ectoderm that give rise to the specialized paired sense organs and sensory cranial ganglia. While they produce very diverse cell types such as sensory neurons, lens fibers and hormone secreting cells [1][2][3] , all placode progenitors arise from a common precursor field that borders the anterior neural plate known as the pre-placodal region (PPR). Subsequently, in response to inductive interactions with surrounding tissues the PPR divides into territories with distinct identities Users may view, print, copy, and download text and data-mine the content in such documents, for the purposes of academic research, subject always to the full Conditions of use:http://www.nature.com/authors/editorial_policies/license.html#terms to generate the adenohypophyseal, olfactory, lens, trigeminal, otic, and epibranchial placodes.
Placode progenitors are induced by a combination of inductive signals primarily mediated by FGFs and attenuation of BMP and Wnt signals [4][5][6] . The zinc-finger transcription factor Zic1, is one of the earliest genes activated in response to these signaling events, and in Xenopus Zic1 is both necessary and sufficient to promote placodal fate by regulating the expression of the PPR-specific genes, Six1 and Eya1 7 . Interestingly, Zic1 is expressed at the anterior neural plate but does not overlap with the prospective PPR 8,9 , suggesting that Zic1 regulates placode formation in a non-cell autonomous manner. To gain insights into the mechanisms by which Zic1 regulates PPR formation, we performed a microarray analysis to identify genes activated by Zic1 in a Xenopus animal explant assay. Among the targets regulated by Zic1 we found a number of genes involved in the synthesis and metabolism of retinoic acid (RA) including lipocalin-type prostaglandin D2 synthase (LPGDS), retinaldehyde dehydrogenase 2 (RALDH2), two members of the cytochrome P450 enzyme family (Cyp26a and Cyp26c) and a cellular retinoic acid binding protein 2 (Crabp2) signifying the importance of this signaling pathway in placode formation. Here we present evidence that Zic1 regulates placode progenitor formation non-cell autonomously by controlling RA production and transport at the anterior neural plate.

Identification of downstream targets of Zic1
The transcription factor Zic1 is expressed at the anterior neural plate 7-10 and is required for the formation of sensory placode progenitors 7 . To gain insights into the mechanisms by which Zic1 regulates sensory placode formation we performed a microarray screen to identify targets of Zic1 11 . The screen was based on the observation that while expression of Zic1 favored placode fate in Xenopus animal cap explants, simultaneous expression of Pax3 repressed placode-specific genes to promote neural crest fate 7,11 (Fig. 1a). Among the genes that were both strongly upregulated by Zic1, as compared to Pax3 alone, and repressed by Pax3 co-injection, we found several well-characterized early placode-specific genes, including Six1, Eya1, Xanf1, Ebf2 and Sox11 (Fig 1b; Supplementary Table 1). The recovery of these genes was an important validation of our experimental design. We also found several novel potential regulators of placode formation (Supplementary Table 1). These genes were initially screened by whole-mount in situ hybridization to select factors expressed at the anterior neural plate, in a pattern similar to Zic1. One candidate that fulfilled this criterion was the lipocalin-type prostaglandin D2 synthase (LPGDS; Fig. 1c), also known as Cpl1 [12][13][14][15] . By in situ hybridization LPGDS is first expressed at stage 13 in the anterior region of the neural plate (Fig. 1d). This expression pattern is maintained throughout neurulation and then appears confined to the head region in tailbud stage embryos (Fig. 1d). Double in situ hybridization demonstrates that LPGDS completely overlaps with the anterior expression domain of Zic1 (Fig. 1e), but is excluded from the lateral expression domain of Zic1, which corresponds to the prospective neural crest region. In situ hybridization for the neural crest-specific gene Snail2 16 confirmed that there is no overlap between LPGDS expression domain and neural crest progenitors, LPGDS abuts the anterior expression domain of Snail2 (Fig 1e).
To further establish that LPGDS is a true target of Zic1 we analyzed LPGDS expression pattern in embryos injected with Zic1GR mRNA (a hormone-inducible version of Zic1 fused to human glucocorticoid receptor ligand-binding domain), or a morpholino antisense oligonucleotide that blocks Zic1 function (Zic1-MO) 7,9 . In embryos injected with Zic1GR mRNA and treated with dexamethasone, we observed a dramatic upregulation and expansion of the LPGDS expression domain (Fig. 1f, g). The same injection in the absence of dexamethasone had no effect on LPGDS expression (Fig. 1f, g). Conversely, injection of Zic1-MO completely inhibited LPGDS expression on the injected side (Fig. 1h, i). Interestingly, in both situations we observed a reduction of Foxi1c and Six1 expression, two early PPR-specific genes (Fig 1f-i). These observations confirm that LPGDS is a downstream target of Zic1 and indicate that placode formation is sensitive to Zic1 and LPGDS expression levels in the embryos.

LPGDS is required for placode formation
The expression pattern of LPGDS at the anterior neural plate and its regulation by Zic1 suggest a potential role in placode formation. To test this possibility, we used a translation blocking morpholino antisense oligonucleotide (LPGDS-MO) to interfere with LPGDS function. The activity of the morpholino was confirmed in an in vitro transcription/ translation assay in which LPGDS-MO blocked LPGDS protein production (Fig 2a;  Supplementary Fig. 1). Unilateral injection of LPGDS-MO (40 ng) in the animal pole region of 2-cell stage embryos resulted in a marked decrease expression of Six1 17,18 , Foxi1c 18,19 and Sox2 20 , three genes broadly expressed at the PPR (Fig 2b, d). We also analyzed the expression of genes restricted to individual placodal domains such as Dmrt4 for the adenohypophyseal and olfactory placodes 21 , Pax8 for the otic and lateral line placodes 22 and Tbx2 for otic and trigeminal/profundal placodes 23 . For each one of these genes a significant reduction of expression was observed on the injected side ( Fig. 2c) in more than 80% of the embryos (Fig 2d). Injection of a control morpholino (Cont-MO) had no effect on the expression of these genes ( Supplementary Fig. 2e). To further establish the specificity of LPGDS knockdown phenotype, we used a second morpholino (LPGDS-MO2) that specifically interfered with LPGDS pre-mRNA splicing by targeting intron1/exon2 junction ( Supplementary Fig. 2a), resulting in the production of a transcript of higher size, due to intron retention ( Supplementary Fig. 2b). The phenotype of LPGDS-MO2-injected embryos was indistinguishable from the phenotype generated by injection of the translation blocking morpholino, with inhibition of Foxi1c, Six1 and Dmrt4, though at a lower frequency (Supplementary Fig. 2c-d). These results indicate that LPGDS is critically required for the establishment of the PPR and for sensory placode formation.

LPGDS functions independently of its enzymatic activity
LPGDS has a dual function, as an enzyme involved in the biosynthesis of prostaglandin D2 (PGD2) from its precursor, and as a lipophilic ligand-binding protein when secreted 24 . The LPGDS molecule contains three cysteine residues, and site-directed mutagenesis has established that the cysteine residue at position 65 (Cyst65) is essential for the enzymatic activity of LPGDS 25 , a cysteine highly conserved from human to frogs (Fig 2e). To determine whether LPGDS regulates placode formation through PGD2 signaling, we performed a rescue experiment using mRNA encoding wild type (WT) mouse LPGDS or a mutated version in which Cyst65 has been replaced by an alanine (C65A), thereby completely abolishing the enzymatic activity of LPGDS 25 . Embryos at the 2-cell stage were sequentially injected with LPGDS-MO and either with WT or C65A mouse LPGDS mRNA. Both mRNA were equally efficient at rescuing Foxi1c and Six1 expression in morphant embryos ( Fig. 2f-g). Complementary to this rescue assay, we found that well-characterized pharmacological agents that specifically block PGD2 signaling pathway did not affect Foxi1c expression at the PPR ( Supplementary Fig. 3). Altogether, these results demonstrate that LPGDS regulates placode formation through a mechanism that does not involve its enzymatic activity and the production of PGD2.

LPGDS functions as a carrier for retinoic acid
Since LPGDS binds retinoids with high affinity 14,26 we wished to determine whether LPGDS regulated placode formation through this mechanism. As a preliminary evaluation we treated intact embryos at the gastrula stage (stage 11) with increasing doses of retinoic acid (RA; 0.01 μM to 10 μM) and analyze the consequence on PPR formation. We found that the lower doses of RA (0.01 μM and 0.1 μM) resulted in an expansion of Foxi1c and Six1 expression domains, while the higher doses of RA (1 μM and 10 μM) inhibited the expression of both genes (Fig. 3), indicating that PPR formation is very sensitive to RA levels in the embryo. Based on these observations, we performed a rescue experiment in which LPGDS-MO-injected embryos were exposed at stage 11 to 0.1 μM of RA (the dose of RA that promotes strong expansion of PPR genes; Fig. 3) or a RA receptor agonist, TTNPB 27 . Both treatments efficiently rescued Foxi1c and Six1 expression in morphant embryos and occasionally expanded Foxi1c and Six1 expression on the control side ( Fig.  4a-b). LPGDS can also bind retinaldehyde (retinal), the RA precursor, with high affinity 14,26 , however this compound was unable to rescue Foxi1c expression domain in LPGDS-depleted embryos ( Supplementary Fig. 4), indicating that in the context of PPR formation LPGDS is functioning primarily as a RA carrier, and that RA is the active compound promoting PPR formation.

Zic1 regulates RALDH2 expression
Our results suggest that LPGDS activity at the anterior neural plate is linked to RA activity. Consistent with this observation we found several components of RA metabolism and function strongly upregulated by Zic1 in the microarray screen (Supplementary Table 1). These include retinaldehyde dehydrogenase 2 (RALDH2) the enzyme responsible for the synthesis of RA from its precursor retinal, two RA-degrading enzymes (Cyp26a and Cyp26c) and a cellular retinoic acid binding protein 2 (Crabp2). In the microarray samples, RALDH2 induction levels followed the same pattern as LPDGS (Fig 4c). RALDH2 is expressed in the trunk mesoderm, as well as in a discrete U-shaped ectodermal domain around the anterior neural plate, similar to LPGDS (Fig 4d) 28 . Zic1 knockdown reduced RALDH2 expression in the embryo (Fig 4e), confirming that RALDH2 is functioning downstream of Zic1. Moreover, in the absence of RALDH2 function Foxi1c was severely reduced at the PPR (Fig 4f), further demonstrating the link between Zic1, RALDH2 activity and placode formation.

Zic1 induces placode fate non-cell autonomously
Our data indicate that Zic1 may control RA signaling at the anterior neural plate through the activation of RALDH2 to produce RA and LPGDS to transport RA extracellularly. Zic1, RALDH2 and LPGDS are all confined to the anterior neural plate, at a distance from the prospective PPR (Foxi1c-expressing cells; Fig 5a) suggesting that Zic1 regulates placode formation non-cell autonomously. To test this possibility, embryos were injected in one blastomere at the 2-cell stage with Zic1GR mRNA and mRNA encoding the lineage tracer GFP, animal explants isolated at the blastula stage, treated with dexamethasone and cultured for 8 hours (Fig 5b). Injection of GR mRNA was used as a negative control ( Supplementary  Fig. 5). Explants were then processed for in situ hybridization on sections. In these explants, Six1 and Foxi1c expression was exclusively confined to GFP-negative cells, derived from the uninjected blastomere (Fig 5c; upper panels). By contrast, LPGDS and RALDH2 expression was always closely associated with Zic1GR/GFP-positive cells (Fig 5c; lower panels). Altogether, these results demonstrate that Zic1 is inducing placode fate by a noncell autonomous mechanism involving activation of RA signaling.

Zic1-activated RA signaling uses non-canonical RA receptors
To further establish that the placode-inducing activity of Zic1 depends on active RA signaling, we analyzed by RT-PCR the expression Six1 and Eya1 in Zic1GR-injected animal explants treated with 100 μM of Disulfiram or Citral (Fig 6a), two general inhibitors of alcohol and aldehyde dehydrogenases 29,30 , the group of enzymes that sequentially catalyze the oxidation of retinol and retinaldehyde to produce RA. In these animal explants treated with the inhibitors, the induction of Six1 and Eya1 by Zic1GR was significantly reduced ( Fig. 6b; Supplementary Fig. 6a, b) consistent with the view that placode induction by Zic1 depends on RA production.
To determine whether the placode-inducing activity of Zic1 is mediated through canonical RA receptors (RARs), Zic1GR-injected animal explants were treated with a wellcharacterized pan-RAR antagonist, AGN193109 31,32 (Fig. 6a). Surprisingly, we found that AGN193109 was unable to block the expression of Six1 and Eya1 (Fig 6c; Supplementary  Fig. 6c), while the same concentration of antagonist (10 μM) was extremely efficient at repressing Hnf1b expression in the whole embryo (Fig 6d), a gene directly regulated by RAR signaling 33 . These results suggest that Zic1-activated RA signaling mediates its activity through a mechanism that does not involve canonical RARs, and may signal through other nuclear receptors that also bind RA, such as peroxisome proliferator-activated receptor β (PPARβ) 34 , and members of the RAR-related orphan receptors (RORs) 35,36 .

Discussion
Our findings uncover a non-cell autonomous activity of Zic1 in the control of placode fate, and implicate RA signaling as a major player in cranial placode progenitor formation. We have identified several genes activated by Zic1 that are involved in RA metabolism and function. Among them we demonstrate that RALDH2 and LPGDS, respectively, are responsible for the synthesis and the transport of RA, and directly participate in the establishment of the PPR. We propose that the activation of RALDH2 and LPGDS at the anterior neural plate by Zic1 allows for the localized production and transport of RA, which in turn activates a placode developmental program in neighboring cells (Fig 7).
Xenopus LPGDS, also known as cpl1 (choroid plexus lipocalin 1), was first isolated in a screen for genes activated at gastrulation and specifically expressed in the nervous system 15 , and has been primarily used as a forebrain marker 12,13,37 . Its specific role at the anterior neural plate has not been studied. Cpl1/LPGDS was also isolated in another screen designed to identify direct targets of Zic1 38 , however further studies are needed to confirm that Zic1 directly activates LPGDS expression. LPGDS is a bifunctional protein that acts as a prostaglandin D2-producing enzyme and a lipophilic ligand-binding protein 39 . More specifically LPGDS can bind retinaldehyde and RA with high affinities 13,14,26 . In a rescue assay, using WT and mutant versions of mouse LPGDS we completely ruled out the involvement of its enzymatic activity, and demonstrated that LPGDS functions during placode formation through a mechanism implicating RA signaling. By exposing the entire embryo to RA, the expression of PPR genes was efficiently rescued in LPGDS-depleted embryos (Fig 4a, b). In these experiments, the exogenous source of RA bypasses the requirement for an LPGDS-mediated transport of RA to the PPR. We posit that LPGDS is essential to transport RA outside the producing cells. LPGDS knockdown in the embryo leads to a loss of genes expressed at the PPR, as well as genes that are more specific for individual cranial placodes, suggesting a broad requirement for LPGDS and RA signaling in placode progenitor formation.
A recent study has proposed that signaling through RA receptors (RARα2), which are expressed in the caudal half of the embryo 40 , was critical to set up the posterior boundary of the PPR 41 . Our experiments using the pan-RAR antagonist (AGN193109) suggest that the placode-inducing activity of Zic1 may use an RAR-independent pathway, pointing to the intriguing possibility that another set of receptors may mediate this activity. A similar activity of RA independent of canonical RARs has been described in the context of the developing mouse neural retina 42 . Several classes of nuclear receptors have been reported to bind RA, including peroxisome proliferator-activated receptor β (PPARβ) 34 and members of the RAR-related orphan receptors (RORs) 35,36 , and are therefore potential candidate to mediate the placode-inducing activity of Zic1. Future work will test this hypothesis.
During embryonic development, RA signaling is essential for the regionalization of the embryo along the anteroposterior axis. Studies in several organisms have shown that RA is an important posteriorizing signal in all three germ layers, acting in concert with molecules of the Wnt and FGF families [43][44][45] . However, RA signaling may have a more direct role in PPR specification, independent of its posteriorizing activity, in a similar manner as the posteriorizing activity of FGF and canonical Wnt can be uncoupled from their role in neural crest induction 10,46 . Consistent with this possibility, RALDH2 has two major expression domains in the early embryo, the paraxial and lateral plate mesoderm in the trunk region, as well as a discrete U-shaped ectodermal domain around the anterior neural plate 28 (Fig 4d). While the former has been linked to the posteriorizing activity of RA, the later represents an independent source of RA, and we propose that this anterior domain is essential to pattern the cranial region of the embryo. In a recent study using novel transgenic RA sensor lines, RA activity has been visualized at the very anterior region of the zebrafish embryo and at the polster, and a candidate source of this activity is RALDH3 47 . In the mouse, RALDH2 is responsible for RA signaling in the early embryonic head, and knockout embryos die around E9.5-10.5 with severe forebrain and craniofacial defects 48 . In birds, RALDH3 is expressed anteriorly at early stages 49 , and vitamin A-deficient quail embryos have defective otic vesicles and lack the pituitary gland, two derivatives of cranial placodes 50,51 .
Zic1 gain-of-function resulted in a dramatic upregulation of LPGDS in the embryo, which was surprisingly associated with a loss of early placode markers (Six1 and Foxi1c; Fig. 1f). However, in light of our results linking Zic1 to RA signaling, this result is not completely unexpected and indicates that PPR formation is very sensitive to RA levels. As in other RAregulated developmental processes, excess or insufficient RA signaling often result in similar outcomes 45 . Excess RA can shut down its own production through the downregulation of RALDH2 52 . Indeed, in embryos treated with RA, RALDH2 is preferentially lost anteriorly, leading to severe head truncation 28 . Furthermore, exposure of intact embryos to varying doses of RA affects PPR genes expression in a concentration dependent manner (Fig 3). Altogether these observations support the view that RA levels at the anterior neural plate are critical for placode formation.
The non-cell autonomous activity of Zic1 in the control of placode fate was especially evident in the animal explant system (Fig 5). Interestingly, we never observed the activation of Six1 or Foxi1c in Zic1GR-expressing cells, rather the expression Six1 and Foxi1c was always confined to adjacent cells, derived from the uninjected blastomere. A possible explanation is related to the activity of Cyp26c and Cyp26a, the RA metabolizing enzymes, which were strongly upregulated by Zic1GR in the microarray. Typically, RA is released by RA-producing cells and can enter adjacent cells where it faces two scenarios: (i) in cells expressing Cyp26, RA is degraded preventing RA signaling; and (ii) in cells that do not express Cyp26, RA enters the nucleus and initiates the transcription of target genes. It is possible that Cyp26 expression renders Zic1-expressing cells refractory to RA signaling, thereby allowing for the unidirectional propagation of the RA signal, and the establishment of sharp boundaries of gene expression. Consistent with this interpretation, in Xenopus Cyp26c is expressed in the neural plate but is excluded from the most anterior region of the embryo 53 .
While the placode inducing activity of Zic1 requires active RA signaling, RA is likely to act in concert with other signals to impart placode fate. Work in zebrafish suggests that Zic1 is functioning not only upstream of RA, but also by regulating Nodal and Hedgehog signaling to control midline tissue development in the head 54,55 . These factors are therefore good candidates to synergize with RA to promote cranial placode fate.

Embryos, injections and explants culture
Xenopus laevis embryos were staged according to Nieuwkoop 60 . Since GFP can no longer be detected following this procedure, sections of GFP-labeled animal explants were first individually photographed and then processed for in situ hybridization.

Microarray analysis
RNAs were extracted from animal explants expressing GR, Pax3GR, Zic1GR or a combination of Pax3GR and Zic1GR using the RNeasy microRNA isolation kit. During the extraction procedure, the samples were treated with DNase I to eliminate possible genomic DNA contamination. The amount of isolated RNA was quantified using a spectrophotometer. Extracted RNAs (five replicate for each injection set) were used for the microarray experiment. Probe labelling, hybridization and initial data analysis were performed at the University of Pennsylvania Microarray Facility. Around 0.1-0.2 μg total RNA was converted into first-strand cDNA using Superscript II reverse transcriptase primed by a poly(T) oligomer that incorporates the T7 promoter. Second-strand cDNA synthesis is followed by in vitro transcription for linear amplification of each transcript and incorporation of biotinylated nucleotides. The cRNA products were fragmented to 200 nucleotides or less, heated at 99°C for 5 min and hybridized for 16 hr at 45°C to a GeneChip Xenopus laevis Genome 2.0 Array (Affymetrix, Cleveland, OH). Hybridized arrays were processed by the GeneChip Fluidics system, and scanned in the GeneChip Scanner. Probes intensities were exported in cel files using GCOS software (Affymetrix, Clevland, OH). Subsequent quantification of expression levels and statistical identification of differentially regulated genes was performed using Partek Genomics Suite (Partek, Inc., St Louis, MO). Normalized, log2-transformed probeset intensities were calculated using GC-RMA. ANOVA with multiple testing corrections was used to determine p-values for likelihood of differential expression, based on reproducibility in five replicates to identify cDNAs both activated by Zic1GR and repressed in Zic1GR/Pax3GR samples. Pairwise comparisons yielding fold changes were performed between these conditions. The microarray data have been deposited in the GEO databa se under the accession number GSE68546.

PGD2 signaling inhibitors and treatment
To interfere with PGD2 signaling pathway we used well-characterized pharmacological inhibitors: AT-56 (Tocris Bioscience, Minneapolis, MN) to block LPGDS enzymatic activity, and two antagonists of PGD2 receptors, BWA868C (Cayman Chemical, Ann Arbor, MI) specific for DP1 and ramatroban (Sigma-Aldrich, St Louis, MO) specific for DP2. For treatment, embryos at stage 11 devoid of the vitelline membrane were incubated in the presence of the inhibitors (1 μM, 10 μM and 100 μM in NAM 0.1X), collected at stage 15, and processed for in situ hybridization. DMSO-treated embryos were used as control.

Imaging
Images were captured using an Olympus SZX9 microscope and a QImaging Micro Publisher 3.3 RTV camera. Images from sections of animal explants were obtained on a Nikon SMZ1500 microscope equipped with a Nikon Ds-U3 camera. Composite images were assembled using Adobe Photoshop. In some experiments RFP mRNA was injected as a lineage tracer, in place of GFP mRNA, for consistency the images were then digitally colored (green) using ImageJ, this applies specifically to Fig 5c (RALDH2; lower panels).