The role of Zic transcription factors in regulating hindbrain retinoic acid signaling

Background The reiterated architecture of cranial motor neurons aligns with the segmented structure of the embryonic vertebrate hindbrain. Anterior-posterior identity of cranial motor neurons depends, in part, on retinoic acid signaling levels. The early vertebrate embryo maintains a balance between retinoic acid synthetic and degradative zones on the basis of reciprocal expression domains of the retinoic acid synthesis gene aldhehyde dehydrogenase 1a2 (aldh1a2) posteriorly and the oxidative gene cytochrome p450 type 26a1 (cyp26a1) in the forebrain, midbrain, and anterior hindbrain. Results This manuscript investigates the role of zinc finger of the cerebellum (zic) transcription factors in regulating levels of retinoic acid and differentiation of cranial motor neurons. Depletion of zebrafish Zic2a and Zic2b results in a strong downregulation of aldh1a2 expression and a concomitant reduction in activity of a retinoid-dependent transgene. The vagal motor neuron phenotype caused by loss of Zic2a/2b mimics a depletion of Aldh1a2 and is rescued by exogenously supplied retinoic acid. Conclusion Zic transcription factors function in patterning hindbrain motor neurons through their regulation of embryonic retinoic acid signaling.


Background
During development, the vertebrate hindbrain is transiently divided into a series of lineage-restricted segments, termed rhombomeres, through the expression of distinct transcription factors. Notably, anterior-posterior patterning and segmentation of the hindbrain is critical in appropriately specifying neuronal cell types [1][2][3][4][5]. The identity of each hindbrain segment is regulated by the Hox family of homeobox transcription factors, the anterior expression limits of which correlate precisely with rhombomere boundaries [6][7][8][9][10][11][12][13][14][15][16][17][18]. The correct complement of hox genes expressed within each hindbrain segment specifies the identity of cells within that segment by activating regional expression of cell migration and axon guidance molecules. Blocking the functions of Hox proteins or their Pbx (Pre-B cell leukemia) and Meis (Myeloid ecotropic virus integration site) cofactors within the hindbrain leads to changes in rhombomere identity and corresponding defects in cranial motor neuron migration and axon guidance [6,17,[19][20][21].
Although the role and requirement of retinoic acid metabolism genes during embryogenesis has been extensively studied, the factors acting to initiate and maintain expression of RA metabolism genes remain largely unknown. Within vertebrates, transcription factors from the Zic (Zinc Finger of the Cerebellum) family of transcription factors are dynamically expressed in partially overlapping regions of the neural tube, indicative of a role in neural development. Recent evidence suggests a connection between Zic transcription factors and the retinoic acid signaling pathway: Maurus et al. demonstrated that loss of zebrafish Zic1 causes a decrease in presumptive forebrain expression of cyp26a1 and an increase in RA signaling as detected by RARE:eGFP transgenics [49]. Further, mutations in human ZIC2 result in holoprosencephaly (HPE), a forebrain defect where the cerebral hemispheres fail to separate during development [50,51] and HPE phenotypes have been connected to aberrant RA signaling, thus providing a plausible link between Zic2 and retinoic acid metabolism [29,36,52,53]. Based on these observations, we tested the hypothesis that Zic transcription factors play a key role in the initiation and maintenance of RA metabolism gene expression during zebrafish embryogenesis. The data presented here demonstrate that zebrafish zic2 genes act upstream of retinoic acid metabolism and suggest a novel regulatory interaction between Zic2a and Zic2b transcription factors and the RA-synthesizing gene aldh1a2. Further, we show that Zic2 signaling is necessary for proper hindbrain patterning.

Results and discussion
Zic transcription factors are expressed during the initiation of RA metabolism genes Retinoic acid levels are regulated by the precise action of synthesis and hydroxylation genes. Aldehyde dehydrogenase 1a2 (Aldh1a2, also known as Raldh2), the ratelimiting synthetic enzyme, catalyzes the conversion of retinal to retinoic acid [34,35,38]. The transcription of aldh1a2 is initiated early in development, with expression at the embryonic margin at 5 hours post fertilization (hpf) ( Figure 1A) and in lateral plate mesoderm by 8 hpf -10 hpf ( Figure 1B-C') [26,54,55]. Beginning at 18 hpf, aldh1a2 expression is restricted to the dorsal retina and anterior somites ( Figure 1D). Retinoic acid degradation occurs following its hydroxylation by the cytochrome p450 oxidase, Cyp26a1. cyp26a1 transcription is initiated early during zebrafish embryogenesis, starting with the presumptive anterior neural ectoderm at 5 hpf ( Figure 1E) [56,57]. By early somitogenesis (10 hpf), cyp26a1 mRNA is found in the presumptive forebrain, midbrain, anterior hindbrain, and part of the tailbud ( Figure 1G, G'). At the later somitogenesis stages of 16-21 hpf, its expression becomes restricted to parts of the retina, caudal notochord, and tailbud ( Figure 1H).
To ascertain whether Zic transcription factors could regulate embryonic initiation of RA metabolism, we first examined which of the seven zebrafish zics are present during the time when cyp26a1 and aldh1a2 mRNA expression are initiated. Using in situ hybridization in conjunction with known expression data, we determined that zic1, zic2a, zic2b, and zic3 are the best candidates for driving initiation of RA metabolism genes because they are expressed at the earliest stages of development [58][59][60]. zic2a, zic2b, and zic3 are all expressed broadly by 5 hpf, while zic1 expression is not detectable ( Figure 1I, M, Q, U). By 8-10 hpf, zic1 and zic2a are expressed in discrete regions consistent with cyp26a1 expression within the anterior neural ectoderm ( Figure 1J Figure 1L, P, S-T, W-X). Overall, these data show that zic2a/2b/3 transcription factor expression is consistent both temporally and spatially with the initiation of expression of cyp26a1 and aldh1a2. To ensure that zic gene expression corresponds to the same germ layers as cyp26a1 and aldh1a2, we sectioned 90% epiboly (9 hpf ) embryos following in situ hybridization. While cyp26a1 is expressed in the epiblast, and aldh1a2 is limited to the hypoblast, zic2a and zic2b are present diffusely in both germ layers and are therefore coexpressed with both RA metabolism genes (Figure 2A-D).
Zics act upstream of early retinoic acid metabolism genes, cyp26a1 and aldh1a2 Based on our observations of zic gene expression, the earliest expressed genes zic2a, zic2b, and zic3 were examined for possible roles in the initiation of RA metabolism. Due to the propensity for functional redundancy, we chose first to simultaneously knock down three transcription factors, Zic2a, Zic2b, and Zic3 with antisense morpholino oligonucleotides. Upon examining resulting phenotypes, we determined that Zic3 depletion is dispensable, and the Zic3 morpholino was removed from the injection mixture. Interestingly, embryos injected with splice-blocking morpholinos for Zic2a and Zic2b display a reduction in cyp26a1 ( Figure 3A, B) and aldh1a2 ( Figure 3C, D) expression by 7 hpf. While both the cyp26a1 expression level and domain are reduced, the aldh1a2 expression domain is reduced in size, but normal in intensity of staining. Results from quantitative real-time PCR showed a 29% reduction in aldh1a2 expression in Zic2a2b-depleted embryos as compared to control embryos (p-value ≤ 0.0001, unpaired t-test) ( Figure 3E). The downregulation of aldh1a2 expression in Zic2a2b-depleted embryos can be observed as late as 18 hpf and 24 hpf ( Figure 3F, G and Figure 3H, I respectively); however it is not as profound as the reduction seen in 7 hpf embryos.
Zic depletion causes mild alterations to retinoic acid-responsive genes and hindbrain patterning As Aldh1a2 and Cyp26a1 have opposing effects on RA levels, the knockdown of either gene alone would be expected to produce opposite phenotypes. However, since Zic2a2b depletion reduces both RA metabolism genes simultaneously, the outcome for RA signaling is difficult to predict. Thus, we performed in situ hybridization for retinoic acid-responsive genes to ascertain the overall level of retinoic acid signaling in Zic-depleted embryos. Previous work demonstrated that the expression levels of zebrafish meis3, hoxb1a, hoxd4a, and hnf1ba correlate with changing retinoic acid levels (Hernandez et al., 2004, Huang et al., 2002, Kudoh et al., 2002, Moroni et al., 1993, Zhang et al., 2000. Our results show a strong reduction in meis3 and a mild reduction in hoxd4a expression within the presumptive hindbrain following Zic2a2b knockdown ( Figure 4A, B, E, F). Two other retinoic acid-responsive genes, hoxb1a and hnf1ba, did not show significant changes in (See figure on previous page.) Figure 1 Temporal and spatial analysis of zic transcription factors and key RA synthesis and degradation genes. aldh1a2 expression is restricted to the embryonic margin at 5 hpf (A). By 8-10 hpf aldh1a2 is observed in the lateral plate mesoderm (B, C, C'). At 18 hpf, aldh1a2 is expressed in the dorsal retina and anterior somites (D). cyp26a1 is expressed in the embryonic margin and the presumptive anterior neural ectoderm at 5-8 hpf (E, F). By 10-18 hpf cyp26a1 is expressed in presumptive forebrain, midbrain, anterior hindbrain, retina, and part of the tailbud (G, G', H). zic1 is not detectable until 8-10 hpf (75% epiboly), when it is expressed within the presumptive anterior neural tissue (I, J, K, K'). By 18 hpf, zic1 is expressed strongly in the telencephalon, midbrain-hindbrain boundary, dorsal hindbrain and spinal cord (L). zic2a initiates earlier, with dorsally-restricted expression at 5 hpf (50% epiboly) (M). By 8-10 hpf zic2a becomes anteriorly restricted within presumptive anterior neural tissue with additional midline expression (N, O, O'). By 18 hpf, zic2a is within the anterior forebrain, ventral eye, dorsal hindbrain and spinal cord (P). Initially, zic2b is expressed in a broad domain encompassing the dorsal side of the embryo (Q; 5 hpf). This broad expression is maintained at 8 hpf (R) and 10-18 hpf where expression is strongest within the eye, midbrain-hindbrain boundary, and presumptive hindbrain (S, S', T). zic3 is also dorsally restricted at 5-8 hpf (U, V).   expression with Zic2a2b morpholino when compared to uninjected controls ( Figure 4C, D, G, H). To determine whether the reductions in meis3 and hoxd4a are caused by changes to retinoic acid signaling, we treated zic2a2b morphants with 5 nM retinoic acid. Indeed, 5 nM RA is able to rescue the expression of meis3 and hoxd4a to levels similar to those seen in our uninjected controls ( Figure 5A-H). In addition, morpholino knockdown of aldh1a2 causes a similar reduction in meis3 and hoxd4a to that seen in zic2a2b morphants ( Figure 5I-L). Thus, our data strongly argue that Zic2a2b promote the expression of RA-responsive genes through Aldh1a2.
We took advantage of manipulating RA levels to further examine the regulatory loops present in the early zebrafish embryo. Given that zic2a2b morpholinos cause a loss in cyp26a1 expression, it is plausible that this causes an indirect effect on the aldh1a2 domain. To test this, we examined expression of aldh1a2 in mutants lacking cyp26a1. Notably, aldh1a2 levels are unaffected in cyp26a1(giraffe) mutants, arguing against a scenario whereby the aldh1a2 reduction in zic2a2b morphants occurs because of reduced hydroxylation of RA by Cyp26a1 ( Figure 5M, N).
Based on the observation that retinoic acid signaling is reduced in zic2a2b morphants, we examined whether zic2a2b depletion also led to defects in hindbrain patterning. In situ hybridization for hindbrain markers (krox20, mafba, hoxb4a, hoxb1a, egfl6) was performed to examine hindbrain segmentation and rhombomere morphology ( Figure 6). We found that all rhombomeres are formed, but that rhombomeres 3 and 5 (egr2b/krox20 expression) are reduced in size ( Figure 6C, D) and there is reduced hoxa2b expression within r2-r6 ( Figure 6A, B). In addition, hoxb4a expression within the posterior hindbrain and spinal cord is also reduced ( Figure 6C, D). Other regions within the hindbrain appear normal as observed through appropriate marker expression patterns ( Figure 6E-L). This suggests that although there is a reduction in the posterior neural domain, the reduction of retinoic acid signaling in zic2a2b morphants does not lead to a complete loss of posterior regions.

Zic2a2b depletion reduces retinoic acid signaling
The transgenic zebrafish line Tg(12xRARE-ef1a:eGFP) sk71 contains twelve retinoic acid response elements (RAREs) upstream of a ubiquitous promoter linked to the gene for enhanced green fluorescent protein, eGFP [61,62]. Active retinoic acid signaling in Tg(12xRARE-ef1a:eGFP) sk71 fish can be detected either directly by observation of eGFP fluorescence, or by in situ hybridization for eGFP mRNA. The sensitivity of this transgenic line was examined by assaying its response to alterations in retinoic acid levels (  Figure 7D, G, J, J' , M, P). These results demonstrate that the Tg (12xRARE-ef1a:eGFP) sk71 zebrafish line is a suitable tool for studying retinoic acid signaling. Of note, we were unable to detect the transgene before 24 hpf, preventing the analysis of retinoic acid signaling levels at early embryonic stages, when retinoic acid metabolism genes are first transcribed.
As the Tg(12xRARE-ef1a:eGFP) transgenic line labels posterior retinoic acid signaling regions, we asked whether the reduction in retinoic acid metabolism genes following Zic depletion leads to an alteration in later embryonic retinoic acid signaling levels. As compared to control eGFP levels ( Figure 8A, B), there is a strong reduction in eGFP expression in 75% of the zic2a2b morphant embryos at 26 hpf ( Figure 8C, D, E). This suggests that there is a reduction in retinoic acid signaling within the Zic2a2bdepleted hindbrain at later stages. To test the hypothesis that the reduction in retinoic acid signaling in zic2a2b morphants is a result of reduced aldh1a2 expression, we compared retinoic acid signaling between Aldh1a2depleted and Cyp26a1-depleted embryos. eGFP levels appear only slightly affected in cyp26a1 MO-injected embryos as compared to control ( Figure 8A, F). How- ever, strikingly, there is a strong reduction in retinoic acid signaling levels across both Zic2a2b and Aldh1a2 knockdowns, as shown by strong reduction of eGFP transcript as compared to controls (Figure 8A, C, G).
Taken together, these data show that we have identified a novel regulatory mechanism whereby Zic2a2b regulates retinoic acid levels, most likely through the regulation of aldh1a2 transcription.  spinal cord [55]. Using the Tg(isl1:eGFP) transgenic line, we sought to analyze development of hindbrain branchiomotor neurons by confocal microscopy in response to Zic2a2b depletion [63]. In control embryos, trigeminal neuron cell bodies are located within rhombomeres 2 and 3, facial neurons within rhombomeres 5 and 6, and the vagal neuron domain extends from the posterior hindbrain down the spinal cord ( Figure 9A, D). To estimate the number of vagal neurons, we used ImageJ to quantify the area and length of the vagal neuron cluster ( Figure 9G, H). Upon treatment with retinoic acid, the vagal domain is greatly expanded and more neuronal cell bodies are visible as compared to control (Figure 9B, E, G, H). The average area in DMSO-treated control embryos is 5481 μm 2 (SD = 321 μm 2 ) and the anterior-posterior length is 165 μm (SD = 4.5 μm). Treatment with RA results in an average area of 10,353 μm 2 (SD = 1448 μm 2 ; p-value ≤ 0.01) and length of 214 μm (SD = 11 μm; p-value ≤ 0.01). Alternatively, treatment with DEAB, a pharmacological inhibitor of retinoic acid synthesis, significantly alters embryonic patterning such that all posterior rhombomeres are lost. The vagal domain is almost undetectable in these embryos (Figure 9C, F, G, H). DEAB treatment also results in mispatterning of anterior branchiomotor neuron classes such that it is difficult to identify trigeminal, facial, and possible remnants of vagal neuron populations. As a best estimate, the area of vagal neurons in DEAB-treated embryos is 1683 μm 2 (SD = 1264 μm 2 ; p-value ≤ 0.01) and length is 65 μm (SD = 20 μm; p-value ≤0.01).

Zic2a2b knockdown causes loss of vagal neurons
Zic2a2b-depleted embryos have reduced retinoic acid signaling levels within the posterior hindbrain and spinal cord. To ascertain whether this reduction in retinoic acid signaling causes a concomitant defect in vagal motor neurons, Zic2a2b knockdown was introduced in Tg(isl1:eGFP) transgenic zebrafish. Embryos were grown in media containing DMSO (control) or treated with exogenous retinoic acid (1 nM) or DEAB (1 μM) to determine if retinoic acid supplementation could rescue phenotypes. Zicdepleted embryos have significantly reduced vagal neuron area and length as compared to control embryos (Figure 10A, D, G, H). The average area of the vagal neuron domain in Zic-depleted embryos is 4979 μm 2 (SD = 1443 μm 2 ) as compared to 7382.5 μm 2 (SD = 377.2) in controls. The average length of the vagal neuron domain is 126.4 μm (SD = 13.8) as compared to 176.6 μm (SD = 10.5) in controls (n = 15, p-value < 0.01). This reduction in the vagal neuron domain would be expected with a mild reduction in retinoic acid levels. In many cases, the medial-lateral width of each vagal population was increased in morphants, suggesting a possible migrational error, as vagal neurons are typically formed medially and migrate laterally during development [1]. The physiological level of retinoic acid is estimated to be 3 nM [64,65]. Due to reduced retinoic levels in Zic-depleted embryos, we postulated that supplementation with near physiological levels (1 nM) of retinoic acid (a concentration that does not cause significant alteration to control embryos) may be sufficient to rescue the neural phenotype seen. There is a mild increase in vagal neuron area and length in control embryos in response to 1 nM retinoic acid treatment ( Figure 10B, G, H). Strikingly, with low dose treatment of retinoic acid, the vagal neuron area of Zic-depleted embryos is rescued to values similar to those for DMSO-treated morphants. While the domain length after RA treatment is similar to that of untreated control embryos, it is still significantly reduced compared to control embryos treated with retinoic acid. We asked whether zic2a2b morphants had residual RA by treating them with DEAB. By further reducing retinoic acid levels with DEAB treatment, there is a roughly equivalent shrinkage of the vagal neuron domain in both zic2a2b morphants and control embryos (Figure 10C, F, G, H).
Our data is consistent with a model whereby Zic2a and Zic2b act upstream of retinoic acid metabolism by activating transcription of aldh1a2. As shown, there is an early reduction in aldh1a2 transcript levels in Zic2a2b morphants, and the reduction in retinoic acid signaling levels within Tg(12xRARE-ef1a:eGFP) transgenics is similar in Zic2a2b and Aldh1a2-depleted embryos. Further, the reduction in vagal neurons in zic2a2b morphants is rescued by treatment with retinoic acid, the product of an Aldh1a2-catalyzed reaction. To determine if this relationship between Zic2a2b and aldh1a2 is Figure 9 Vagal neurons are sensitive to alterations in retinoic acid levels. Tg(isl1:eGFP) transgenic zebrafish are used to visualize alterations to branchiomotor neurons in response to 1% DMSO (A, D), 5 nM all-trans retinoic acid (B, E), or 5 μM DEAB (C, F) treatment. Images are dorsal views of 48 hpf embryo hindbrain using 20× or (A-C) or 40× (D-F) objective. Length (G) and area (H) were quantified using ImageJ and displayed graphically (error bar denotes standard deviation). The differences induced by either retinoic acid or DEAB treatment are statistically significant (ANOVA and Tukey's HSD post-hoc test, *p-value < 0.01). maintained in later neurons, vagal neurons were examined in aldh1a2 morphant Tg(isl1:eGFP) zebrafish. The Aldh1a2-depleted embryos also show a statistically significant reduction in length of the vagal domain, very similar to that seen in Zic2a2b-depleted embryos (compare Figures 10 and 11). The comparable reduction in vagal domain length supports the hypothesis that Zic2a2b act upstream of aldh1a2. Taken together, our data show that the most posterior class of branchiomotor neurons, the vagal neurons, is sensitive to alterations in retinoic acid levels. Increased retinoic acid causes an expansion in the area and length of the vagal neuronal population, while retinoic acid depletion reduces the population. Consistent with this observation, the reduction in retinoic acid signaling levels in Zic2a2b-depleted embryos elicits a reduction in the size of the vagal neuron domain. Importantly, the vagal neuron domain in Zic2a2b-depleted embryos can be rescued by supplementation of RA to physiological levels.

Conclusions
We have identified a novel regulatory mechanism between Zic2a2b transcription factors and early retinoic acid signaling levels. We propose a model whereby Zic2a and Zic2b act upstream of both retinoic acid synthesis (aldh1a2) and degradation (cyp26a1) genes in an early embryo. We see a persistent reduction in retinoic acid signaling in Zic2a2b morphants at 26 hpf that is similar (See figure on previous page.) Figure 10 Vagal neuron domain in Zic2a and Zic2b morpholino injected embryos is partially rescued with low doses of RA. Dorsal views of the vagal motor neurons in 48 hpf Tg(isl1:eGFP) embryos that were injected with p53 morpholino (A-C) or Zic2a;Zic2b and p53 morpholinos (D-F), and treated with 1% DMSO (A, D), 1 nM RA (B, E), or 1 μM DEAB (C, F). Quantification of vagal motor neuron domain area (G) demonstrates that Zic2a/2b depletion causes a statistically significant reduction in domain size, which is rescued by treatment with retinoic acid. Quantification of vagal neuron domain length (H) gives similar results, with a statistically significant reduction upon Zic2 depletion, which could again be counteracted by retinoic acid addition. Statistical significance calculated by ANOVA and Tukey's HSD post-hoc test (*p-value ≤ 0.01).

Figure 11
Knockdown of Aldh1a2 causes a reduction in the vagal neuron domain. Analysis of vagal motor neurons in Tg(isl1:eGFP) transgenic embryos injected with either a p53 control morpholino (1 ng, A) or an Aldh1a2 morpholino (2 ng, B). The length of the vagal domain was quantified using Image J (C). Significance calculated using an unpaired t-test (*p-value ≤0.01; n = 2, 6 for quantification of p53MO and Aldh1a2 MO, respectively).
to that observed in embryos depleted of the main retinoic acid synthesis enzyme, Aldh1a2. Consistent with reduced retinoic acid levels, Zic2a2b knockdown results in reduced vagal motor neuron formation within the posterior hindbrain and spinal cord. This phenotype is nearly identical to Aldh1a2-depleted embryos and is rescued by exogenous retinoic acid treatment.
While the role and requirement for retinoic acid metabolism genes has been well studied, less is known about the upstream regulators of these factors. During mouse development, Hoxa1-Pbx1 (Pre-B-cell leukemia homeobox 1) complexes directly regulate Raldh2 transcription within mesodermal tissue [66]. Additionally, deficiencies in zebrafish Tgif and Hmx4 cause defects associated with reduced retinoic acid levels and reduced aldh1a2 transcription [36,67]. Interestingly, Tgif-depletion also results in a concomitant reduction in cyp26a1 mRNA levels, similar to defects observed in Zic-depleted embryos. As expected, the Cyp26a1 murine mutant has defects associated with increased retinoic acid levels [41]. More notably, rescue of these defects is observed with heterozygous disruption of Aldh1a2 in a Cyp26a1 murine mutant, suggesting co-dependency between these genes, where less retinoic acid degradation is necessary when retinoic acid synthesis is reduced [38]. Thus, the fact that the transcription factors Zic2a and Zic2b regulate expression of both retinoic acid and synthesis genes strongly suggests that they play an important role in regulating retinoic acid levels during early embryogenesis.
Holoprosencephaly (HPE) is the most common forebrain birth defect, occurring when the cerebral hemispheres fail to separate. Mutations in human ZIC2 and TGIF are known to cause HPE. Our results from zebrafish studies provide evidence that both of these genes function to regulate RA metabolism [36]. Although both Zic and Tgif genes are implicated in regulating multiple signaling pathways, it is now plausible that altered levels of RA predispose embryos to HPE.

Zebrafish care
Danio rerio were maintained in accordance with published protocols [68]. Zebrafish adults and embryos were maintained at 28.5°C. Embryos were raised in embryo medium (EM) with 10 ml/l Penicillin/Streptomycin (Sigma) added to prevent bacterial growth, and with phenyl-thiourea (0.003%) used to prevent pigmentation after 24 hpf. The stage of developing embryos was assessed using reported morphological guidelines (Kimmel et al., 1995). Wild type AB, Tg(isl1:eGFP), and Tg(12xRARE-ef1μ: eGFP) strains were used as described [61][62][63]. Animal protocols were approved by the University of Alberta's Animal Care and Use Committee-Biosciences with protocol #427.

Examination of retinoic acid signaling
To manipulate retinoic acid levels in developing zebrafish embryos, we used pharmacological treatment (1 or 5 nM of all-trans retinoic acid (Sigma); 1 or 5 μM diethylaminobenzaldehyde (DEAB, Sigma)) or morpholinos (Cyp26a1 or Aldh1a2) that target enzymes known to regulate degradation and synthesis of endogenous RA. Embryos were treated at approximately 50% epiboly with DEAB or RA in Embryo Medium, with DMSO and ethanol used as vehicle controls. Embryos were manually dechorionated at 26 hpf and, maintaining treatment concentration, media was changed once per day. All treatments were carried out in 60 mm petri dishes each containing approximately 40 embryos grown at 28.5°C. Tg(12xRARE-ef1a:eGFP) transgenic zebrafish allow visualization of RA-signaling levels by assaying fluorescence (or by in situ hybridization for eGFP mRNA).

JB-4 sectioning
After in situ hybridization, embryos were post-fixed for 2 hours at room temperature in 4% paraformaldehyde. Embryos were then rinsed in PBS, dehydrated through a series of ethanol washes, and incubated over one hour in two rinses of JB-4 infiltration solution (made as per manufacturer's instructions, Polysciences Inc.). After transferring the embryos to molds, the infiltration solution was replaced with JB-4 embedding media and the blocks were left to harden at room temperature overnight. Seven μM sections were cut on a Leica RM2235 microtome and images were captured on a Zeiss Axio Imager.Z1 using Axovision SE64 Rel.4.8 software.

Quantitative real-time PCR
Quantitative Real-Time PCR was used to quantify in vivo mRNA levels to ascertain aldh1a2 gene expression in morpholino-injected embryos. Primers were designed using Roche Universal Probe Library for Zebrafish [Forward 5'-AACCACTGAACACGGACCTC-3' and Reverse 5'-ATGAGCTCCAGCACACGTC-3']. After RNA isolation (Ambion RNAqueous), DNA was removed by DNAseI digestion (19 μl DEPC-treated H 2 O, 10 μl 10× DNAseI buffer, 1 μl DNAseI (Ambion)) for 30 minutes at 37°C. RNA was further purified using Qiagen RNeasy columns and cDNA synthesized using AffinityScript qPCR cDNA Synthesis kit (Ambion) by the recommended protocol. Quantitative RT-PCR and primer validation was carried out as previously described [74,75]. For primer validation, control (AB) cDNA was isolated and used to create a dilution series: 1/8, 1/16, 1/32, 1/64, 1/128, 1/256. Ambion Brilliant II qPCR kit with thermocycler conditions 40× denaturation at 95°C for 30 seconds; annealing at 55°C for 1 minute; and extension at 72°C for 30 seconds. Finally, relative gene expression level was determined using the comparative Ct method (2 −ΔΔCt method) and unpaired t-test calculations for significance [74,75]. Reaction mixtures and thermocycler program were identical to primer validation protocol; a 1/16 dilution of experimental and control cDNA was used with each primer pair. Reactions were completed with at least 7 technical replicates from RNA isolated from 50 embryos, comparing to ef1a as endogenous control.

Imaging of cranial motor neurons
The Tg(isl1:eGFP) transgenic line allows visualization of a subset of motor neurons including those that are located within midbrain and hindbrain: oculomotor, trochlear, trigeminal, facial, glossopharyngeal, and vagal neurons [63]. For fluorescent analysis, embryos were fixed with 4% PFA for 4 hours at room temperature. Yolks were removed manually and embryos dehydrated in 50% and 70% glycerol before mounting and imaging using a Zeiss LSM 510 confocal microscope and Zeiss Zen software. Image preparation and analysis (quantification of vagal neuron fluorescent area and length of vagal domain) was done with ImageJ software. To calculate vagal area, Z-stacked images were converted to 8-bit color and scale was set to correspond with that of image (pixel length 1024 or 2048, length of field of view 319.93 μm). Image threshold was set capturing the greatest number of neurons while keeping noise low. Regions of fluorescence were selected and the area calculated with numerical information transferred to Microsoft excel, where totals and averages were calculated. Statistical significance was determined using ANOVA followed by Tukey's HSD post hoc test, or by an unpaired t-test.