Depletion of Retinoic Acid Receptors Initiates a Novel Positive Feedback Mechanism that Promotes Teratogenic Increases in Retinoic Acid

Normal embryonic development and tissue homeostasis require precise levels of retinoic acid (RA) signaling. Despite the importance of appropriate embryonic RA signaling levels, the mechanisms underlying congenital defects due to perturbations of RA signaling are not completely understood. Here, we report that zebrafish embryos deficient for RA receptor αb1 (RARαb1), a conserved RAR splice variant, have enlarged hearts with increased cardiomyocyte (CM) specification, which are surprisingly the consequence of increased RA signaling. Importantly, depletion of RARαb2 or concurrent depletion of RARαb1 and RARαb2 also results in increased RA signaling, suggesting this effect is a broader consequence of RAR depletion. Concurrent depletion of RARαb1 and Cyp26a1, an enzyme that facilitates degradation of RA, and employment of a novel transgenic RA sensor line support the hypothesis that the increases in RA signaling in RAR deficient embryos are the result of increased embryonic RA coupled with compensatory RAR expression. Our results support an intriguing novel mechanism by which depletion of RARs elicits a previously unrecognized positive feedback loop that can result in developmental defects due to teratogenic increases in embryonic RA.


Introduction
Improper retinoic acid (RA) signaling during development can cause congenital malformations that affect the forelimbs, ocular, cardiovascular, respiratory, urogenital and nervous systems [1][2][3][4]. Despite almost a century of investigation, the mechanisms underlying many congenital defects due to fluctuations in RA signaling are still not understood. RA acts as a ligand for RA receptors (RARs), members of the nuclear hormone family of transcription factors [5]. Work using disparate embryonic models has provided critical insight into the molecular mechanisms and developmental requirements of RAR function in vertebrate embryos [6][7][8][9][10][11][12]. In addition, RAR deficiency and inappropriate RA signaling are associated with numerous types of cancers [13]. In the majority of cases, the mechanism by which loss of RARs promote tumorigenesis is not understood. Therefore, understanding the roles of RARs during development will help elucidate the mechanisms underlying congenital defects, and possibly cancers, caused by inappropriate RA signaling [3,4].
RA signaling employs a number of feedback mechanisms in order to maintain appropriate levels in the embryo and tissues. The best characterized feedback mechanism is through regulation of the RA producing [retinol dehydrogenases (RDHs) and retinaldehyde dehydrogenases (Aldh1a)] and degrading (Cyp26) enzymes. Specifically, increased RA signaling inhibits the expression of the RA producing enzymes, while promot-ing Cyp26a1 expression. Conversely, decreased RA signaling promotes expression of the RA producing enzymes, while inhibiting Cyp26a1 expression [14][15][16][17][18]. While modulation of RA signaling also affects the expression of other factors that control RA signaling [5,19], less well understood are feedback mechanisms that may influence RAR expression. RA response elements (RAREs) have been found in murine RARa2 and RARb2 promoters and RARb2 has been shown to be RA responsive [20][21][22]. However, if decreases in RA signaling, in particular due to loss of RAR expression, lead to compensatory expression of other RARs is less clear. While initial studies of mouse RAR KO mice suggested that there was not compensatory RAR expression in RAR deficient mice [11,12], more recent studies using siRNA to deplete RARa have challenged this model and suggested that there may be compensatory RAR expression in RARa deficient embryos [23]. Therefore, if there are RA feedback mechanisms that influence RAR expression and how the employment of these feedback mechanisms impact embryonic development are not well understood.
Here, we find that depletion of RARab1, a previously unrecognized yet conserved zebrafish RARa splice variant, causes an increase in CM specification and heart size, which is due to the triggering of a feedback mechanism that surprisingly promotes increased RA signaling from surplus embryonic RA and compensatory RAR expression. Our results provide insight into a newly recognized positive feedback mechanism that we posit resists fluctuations in RA signaling due to perturbation in RAR expression. However, if improperly maintained, the positive feedback can result in RA induced congenital defects. Altogether, the results from this study significantly enhance our understanding of the feedback mechanisms that are used to maintain appropriate RA signaling levels and previously unexplored mechanisms that potentially underlie congenital defects.

RARab1 deficient embryos have enlarged hearts and increased CM specification
In contrast to the studies of RARs in mice [9][10][11][12], depletion of RARs has not been able to recapitulate all of the consequences of loss of RA signaling in zebrafish [8], which prompted us to determine if additional conserved RAR variants exist in zebrafish beyond what has already been reported [24]. We cloned a previously unrecognized RARa splice variant that is orthologous to human, mouse and Xenopus RARa1 termed RARab1 ( Figure 1A-1C). The previously cloned zebrafish RARa homologs RARaa and RARab are teleost specific paralogs and both are orthologous to the splice variant 2 found in tetrapods ( Figure 1B, 1D) [24]. Both rarab1 and rarab2 are expressed maternally and zygotically ( Figure 1E), with ubiquitous expression until the tailbud stage (Figure S1A-S1I). After the tailbud stage, their expression patterns deviate ( Figure 1F-1H and Figure S1J-S1O).
We used a translation blocking morpholino (MO) to examine the function of RARab1 ( Figure 1B). By 48 hours post-fertilization (hpf), RARab1 deficient embryos had enlarged hearts with increased CM number and expression of CM marker genes myl7, vmhc and amhc (Figure 2A, 2B, 2M, 2N and Figure S2A-S2D). Similar increases in CM number were also found at 55 hpf ( Figure S3A-S3C), suggesting the major addition of surplus CMs occurs during earlier stages of development. Consistent with this idea, we observed an expansion of CM differentiation (myl7, vmhc, and amhc) and progenitor (nkx2.5 and hand2) marker expression in RARab1 deficient embryos at earlier stages via in situ hybridization (ISH) and quantitative real-time PCR (qPCR; Figure 2C-2L, 2O-2Q). Injecting the RARab1 MO along with rarab1 mRNA that lacks the 59UTR MO binding sequence is able to rescue the increased heart size, supporting the specificity of the phenotype ( Figure S4A-S4D). Together, these results suggest that RARab1 deficient embryos have increased CM specification, number and heart size.

Depletion of zebrafish RARab paralogs promotes RA signaling
The increased atrial and ventricular CM number in RARab1 deficient embryos are reminiscent of RA signaling deficient embryos [25,26]. Therefore, we examined hoxb5b expression, which functions downstream of RA signaling to restrict atrial CM number [26] and is likely a direct target of RARs ( Figure S5A-S5D). Unexpectedly, we found that hoxb5b expression was increased in RARab1 deficient embryos ( Figure 3A-3C). While this was initially perplexing, our recent studies showed that Hoxb5b overexpression is able to mimic many of the teratogenic effects of RA treatment [27]. Therefore, we asked if the increases in hoxb5b expression in RARab1 deficient embryos could be a cause of the enlarged hearts. While depletion of hoxb5b alone using a low concentration of hoxb5b MO does not affect CM number ( Figure S6A-S6C), we found that concurrent depletion of RARab1 and Hoxb5b largely restored heart morphology, CM differentiation marker expression, and CM number relative to the RARab1 deficient embryos ( Figure 3F-3N), suggesting that the increased CM number in RARab1 deficient embryos is in part a consequence of the increased hoxb5b expression.
We next examined the expression of additional RA signaling responsive genes. Similar to hoxb5b, we found that the expression of additional RA signaling responsive genes, including cyp26a1, dhrs3a, hoxb6b and hoxb5a, was increased in RARab1 deficient embryos ( Figure 3A). Comparing RA responsive gene expression in RA treated and RARab1 deficient embryos, we found that the trends were similar, but that RA treatment typically induced a greater increase in expression ( Figure 3A). Conversely, treatment with DEAB, an antagonist of the RA producing enzyme Aldh1a, inhibited RA responsive gene expression ( Figure 3A). These findings indicate that RARab1 depletion paradoxically results in increased expression of RA signaling responsive genes.
We next wanted to determine if increases in RA signaling responsive genes were specific to RARab1 depletion, so we examined RA responsive gene expression in RARab2 deficient embryos. Previous studies found that RARab2 deficient embryos lack forelimbs (pectoral fins) and tbx5a expression [8,28], which we confirmed ( Figure S7A, S7C, S7D, S7F, S7H, S7I). However, similar to RARab1 depletion ( Figure 3A and Figure 4A), RARab2 deficient embryos had increased expression of RA signaling responsive genes ( Figure 4A). While the previous studies found a loss of forelimbs, defects in heart development were not reported. Despite the loss of forelimbs and increase in RA signaling responsive genes, we did not observe an increase in heart size, CM number or CM gene expression ( Figure S8A-S8D). Therefore, although eliciting similar increases in RA signaling responsive gene expression, individual depletion of RARab1 and RARab2 results in distinct defects.
To determine the functional consequences of concurrent RARab1 and RARab2 depletion, we co-injected a suboptimal dose of each MO. Unfortunately, co-injection of an optimal dose of each MO resulted in significant non-specific toxicity even when injected along with p53 MO. However, concurrent depletion of the RARabs using suboptimal MO doses resulted in a dramatic increase in RA signaling responsive genes, above what was seen with depletion of RARab1 and RARab2 alone using the optimal

Author Summary
Retinoic acid (RA) is the most active metabolic product of Vitamin A. Appropriate levels of RA are required for proper embryonic development and tissue maintenance in all vertebrates. Inappropriate levels of RA in human embryos can cause congenital defects that affect many organs, including the heart and limbs, and lead to numerous types of cancers. Understanding how animals maintain appropriate RA levels and the consequences of inappropriate RA signaling will therefore provide insight into human congenital defects and diseases. RA signaling is mediated by RA receptors (RARs), which are transcription factors that are activated when binding RA. We have found that depletion of RARs in zebrafish results in defects that are surprisingly due to increases in embryonic RA and not a deficiency of RA signaling. Our results are the first to demonstrate that RAR depletion elicits a positive feedback mechanism that promotes RA signaling through complementary increases in both embryonic RA and RAR expression. Therefore, our analysis provides novel insight into the molecular mechanisms that are required to maintain appropriate RA signaling and will positively impact our understanding of the mechanisms underlying congenital defects.
MO doses ( Figure 4A). Additionally, there was an anterior shift of hoxb5a expression in the spinal cord of RARab1+2 deficient embryos, suggesting the spinal cords are posteriorized ( Figure  S9A-S9E). Increased RA signaling inhibits aldh1a2 expression through a negative feedback mechanism [16][17][18]. Although aldh1a2 expression in individual RARab1 and RARab2 deficient embryos was not suppressed ( Figure 4B), aldh1a2 expression was decreased in embryos depleted for both RARab variants ( Figure 4B). To corroborate the increases in endogenous RA signaling responsive genes, we used the RA signaling reporter line Tg(12XRARE-ef1a:EGFP) sk72 [29]. Again, co-depletion of both RARabs resulted in a greater expansion of egfp expression, compared to the individual depletion of each RARab ( Figure 4D-4H). Therefore, these experiments support the hypothesis that the RARab1+2 deficient embryos are sensing more significant increases in RA signaling than embryos deficient for either RARab variant alone.
We next examined the consequences of this functional interaction on heart development. We found that the hearts of RARab1+2 deficient embryos had increased atrial size, CM number, and a dramatic increase in amhc expression ( Figure 4I, 4L-4N and Figure S10A-S10D). Significant effects on CM number or heart size were not found when using a suboptimal dose of either RARab1 or RARab2 MO alone ( Figure 4I-4K, 4M), though we did find a modest increase in CM marker gene expression in the RARab1 deficient embryos ( Figure 4N). Interestingly, in RARab1+2 deficient embryos we found more significant increases in atrial CM number and amhc expression ( Figure 4M, 4N), which were remarkably similar to the consequences of modest increases in RA signaling due to RA treatment [27]. Increased RA signaling can also inhibit forelimb development [17] and RARab1 deficient embryos also have smaller forelimbs and a modest reduction of tbx5a expression ( Figure S7A, S7B, S7D, S7F, S7G, S7I). A functional interaction with the RARab variants that resulted in loss of forelimbs was also observed ( Figure S7D, S7E). Therefore, concurrent depletion of RARab variants elicits increases in RA signaling with heart and forelimb phenotypes that are strikingly similar to increases in RA signaling caused from RA treatment.

RARab1 deficient embryos have increased embryonic RA
We sought to understand the mechanism underlying the increase in RA signaling in RARab deficient embryos. In the absence of RA, RARs are thought to interact with transcriptional co-repressors, while binding of RA converts the RARs to transcriptional activators [1,5]. A previous study in Xenopus suggested that RARs are required as transcriptional repressors in some developmental contexts [6]. However, our gain-of-function analysis did not support that these zebrafish RARs function as transcriptional repressors (Figure S11A-S11L), consistent with what we have reported previously [29]. However, Manshouri et al. [23] found a compensatory increase in the expression of other RARs when using siRNA to deplete RARa in mice. Similarly, we found that the expression of other zebrafish RARs [24] was increased in RARab deficient embryos ( Figure 4C and Figure  S12A-S12L), suggesting that compensatory RAR expression is a conserved response to depletion of RARa homologs in vertebrates. Although Manshouri et al. [23] proposed the compensatory RAR expression was RA signaling dependent, our results suggest that the expression of most RARs is potentially regulated independent of RA signaling ( Figure 4C), because the effects on RAR expression did not parallel modulation of RA signaling using RA and DEAB. While we observed compensatory expression of other RARs in RARab deficient embryos, it is difficult to conclude that increased RAR expression is the sole cause of the increase in RA signaling since overexpression of RARs in zebrafish embryos does not produce significant positive or negative effects on RA responsive gene expression (Figure S11A-S11J) [29]. Nevertheless, our results suggest that when depleting RARabs in zebrafish embryos compensatory RARs are present that can mediate RA signaling.
Because we did not have evidence that RARs act as transcriptional repressors or that the increased expression of RARs alone contributes to the increases in RA signaling in RARab deficient embryos, we hypothesized that the depletion of RARs may trigger an increase in embryonic RA. Although aldh1a2 expression was suppressed in RARab1+2 deficient embryos similar to when embryos sense increases in RA signaling ( Figure 4B) [16][17][18], the expression of rdh10a and rdh10b, which control a limiting step in RA production in vertebrates by generating retinal from retinol [14,15], was increased in RARab1 and RARab1+2 depleted embryos ( Figure 4B and Fig. S13A-S13C). Interestingly, rdh10b expression, which was not sensitive to modulation of RA signaling, was increased in RARab deficient embryos ( Figure 4B). Therefore, our results suggest that depletion of RARabs triggers an increase in RA through promoting rdh10 expression. In addition to inhibiting aldh1a2 expression, increased RA signaling promotes a negative feedback mechanism that limits RA levels by positively regulating Cyp26a1 expression [16][17][18]. Since we observe an increase in cyp26a1 expression in RARab1 deficient embryos ( Figure 3A, 3D, 3E and Figure 4A), which was also consistent with the hypothesis that there is increased embryonic RA, we postulated that the increased Cyp26a1 may be protecting the RARab1 deficient embryos from teratogenic increases in embryonic RA. Therefore, we concurrently depleted RARab1 and Cyp26a1 to determine if there was a functional interaction indicative of increased embryonic RA. For these experiments, a suboptimal dose of cyp26a1 MOs (Figure S14A-S14E) was used to more easily discern a functional interaction. In either the RARab1 or Cyp26a1 deficient embryos alone, we never observed absence of the MHB or defects in tail elongation ( Figure 5A-5C, 5E-5G). However, co-depletion of RARab1 and Cyp26a1 resulted in a loss of the MHB and truncated tails ( Figure 5D, 5H), similar to increases in RA signaling [17,19,29,30]. Furthermore, we found that RARab1+Cyp26a1 deficient embryos had dismorphic hearts with a specific reduction in ventricular CM number compared to controls embryos hearts ( Figure 5I-5L, 5Q), which interestingly resembles the trend we previously found in embryos with intermediate increases in RA signaling [27]. Although one interpretation of the functional interaction of RARab1 and Cyp26a1 depletion is that there is increased embryonic RA levels in these embryos, we wanted to further test this hypothesis using additional assays. First, we sought to use a distinct readout of embryonic RA, so we made a novel stable transgenic RA sensor line which incorporated the RARab ligand binding domain (RLBD) fused to the Gal4 DNA binding domain (GDBD) expressed under the b-actin promoter (Figure S15A-S15G) [31]. Previous studies have found that similar GDBD fusions with nuclear hormone receptor LBDs create an effective reporter of nuclear hormone activity [6,32,33]. We observed a dramatic increase in reporter expression when RARab1 and Cyp26a1 were depleted together in Tg(b-actin:GDBD-RLBD); Tg(UAS:EGFP) embryos ( Figure 5M-5P, 5R) [34]. Second, our hypothesis predicted that reducing embryonic RA levels should be able to rescue teratogenic phenotypes found in RARab1+Cyp26a1 and RARab1 deficient embryos. Consistent with this hypothesis, DEAB treatment of RARab1+Cyp26a1 deficient embryos was able to rescue the loss of MHB ( Figure 6A-6J). Additionally, treatment of RARab1 deficient embryos with DEAB partially rescue the enlarged heart phenotype and restored atrial CM number ( Figure 6K-6O). Lastly, our hypothesis predicts that exogenous treatment with a concentration of RA that causes a minor increase in RA signaling should result in aberrant heart phenotypes that are similar to RARab1 deficient embryos. Indeed, embryos treated with low concentrations of exogenous RA (lower than we had reported using previously [27]) had enlarged hearts with an increase in both atrial and ventricular CM number at 48 hpf ( Figure 6P-6R). Altogether, our results suggest that increases in embryonic RA, coupled with compensatory RAR expression, contribute to the developmental defects found in RARab1 deficient embryos.

Discussion
Together, our study supports a novel paradigm whereby RARab depletion elicits a positive feedback mechanism that can result in teratogenic increases in RA signaling. Importantly, our work highlights that loss and gain of RA signaling can cause similar developmental defects. RA signaling is required to restrict CM specification [25,26], while high increases in RA signaling can eliminate CM specification ( Figure 7A) [27]. However, our present findings suggest that low increases in RA signaling, achieved when treating embryos with mM concentrations of RA or through RARab depletion, can also promote increases in both atrial and ventricular CM specification ( Figure 7A). As we found previously, modest, but slightly higher increases of RA signaling can promote atrial CM specification without significantly affecting ventricular CM specification [27], which is strikingly similar to what we found with concurrent depletion of the RARab variants here ( Figure 7A). Moreover, intermediate increases in RA signaling can inhibit ventricular CM specification, which is similar what we observed when concurrently depleting RARab1 and Cyp26a1 ( Figure 7A). It also appears that modulation of Hox activity downstream of both gain and loss RA signaling is at least partially responsible for the increases in CM specification, suggesting the hypothesis that the similar effects on CM number are actually due to opposite perturbations of anterior-posterior patterning within the ALPM. Therefore, our analysis corroborates and extends previous observations that there are differential effects on atrial and ventricular CM populations as there is a progressive increase from low to intermediate levels of RA signaling in the early embryo.
It is interesting that depletion of RARa homologs using MOs in zebrafish, presented in this study, and Xenopus [6] elicit similar phenotypic responses. In Xenopus embryos, RARa depletion alone results in loss of the MHB [6]. While depletion of RARab1 alone does not result in MHB defects in zebrafish embryos, we have found that RARab1+Cyp26a1 deficient embryos completely lack the MHB. Taken together, these results suggest that the underlying consequences of increased RA signaling due to depletion of RARa homologs are likely conserved at least in Xenopus and zebrafish embryos, but that in Xenopus perhaps the role of Cyp26 enzymes in protecting the brain has been lost. Despite similarities in the phenotypes that both point to an increase in RA signaling in RARa and RARab deficient Xenopus and zebrafish embryos, our results contrast with the model proposed by Koide et al. [6], which concluded that RARs are required to function as transcriptional repressors. Importantly, the tools used in the previous study, including dominant-negative RARs, transcriptional co-repressors, and inverse agonists, would not have allowed them to distinguish between a transcriptional de-repressive model and the positive feedback mechanism involving the production of excess RA supported here.
In addition to the phenotypic similarities when depleting RARa homologs in Xenopus and zebrafish, depletion of zebrafish RARabs results in compensatory RAR expression similar to RARa depletion in mice [23], supporting the hypothesis that this feedback response to RARa deficiency is conserved in vertebrates. Importantly, the response to RAR depletion is likely different than complete ablation of RARs. RAR KO mice have not been reported to have compensatory increases in other RARs [11,12], suggesting that a complete loss of RAR expression may cause a breakdown of this feedback loop. However, when considering the probability that RAR expression would be completely lost vs. depleted, we postulate that insults resulting in depletion of RAR expression would be much more likely. Consistent with this idea, variable levels of RAR expression deficiency, which in the case of RARb can be due to epigenetic silencing, is commonly observed in a variety of cancers [13].
Given the conserved feedback mechanisms already recognized that limit fluctuations in RA signaling in vertebrates [16,17,19,23], it seems logical that a conserved mechanism that senses RAR deficiency would also exist to prevent loss of RA signaling. We propose that this newly recognized positive feedback mechanism would be more suitable to prevent transient deficiency in RARs. As demonstrated here, persistent RARab depletion can result in a hypervigilant response of RA signaling and RA-induced teratogenic defects. Overall, these data provide insight into a previously unappreciated RAR-dependent positive feedback mechanism ( Figure 7B), which is active during development. Further elucidation of this RA signaling feedback mechanism may illuminate the etiology of poorly understood RA-insensitive cancers [13,23] and congenital defects [1,3].

Ethics Statement
All zebrafish husbandry and experiments were performed in accordance with protocols approved by the Institutional Animal Care and Use Committee (IACUC) of Cincinnati Children's Hospital Medical Center.

ISH
Whole-mount ISH was carried out using standard procedures [40]. All probes except rarab1 (accession number: KF030797) and rarab2 were reported previously.

MO and mRNA injections
The rarab1 MO (59-TGCAGGTCATCCGTAATGCCC-GATC) was designed to the 59 UTR of rarab1. Additional MOs targeting another region of the 59 UTR and the donor splice junction, which saturated the available MO target sites, were also tried. However, injection of these MOs resulted in significant toxicity and were not able to be used for analysis. Sequences to the rarab2 and hoxb5b MOs were reported previously [8,26]. The total amount of rarab1 MO injected was 16 ng. The total amount of rarab2 MO injected was 7 ng. The suboptimal doses used to test genetic interactions were half these concentrations. The amount of hoxb5b MO used was 0.25 ng. A cocktail of 4 ng cyp26a1 MO1 (59-TCTTATCATCCTTACCTTTTTCTTG) and 2 ng cyp26a1 MO2 (59-TAAAAATAATACACTACCTGCAAAC) produced a phenotype similar to gir mutants [17]. Suboptimal doses used in experiments were 0.9 ng (cyp26a1 MO1) and 0.45 ng of (cyp26a1 MO2). For all injection experiments, 3 ng of p53 MO were used to help suppress non-specific MO-induced cell death [41]. For experiments, the total amount of MO injected was always kept constant by equilibrating the concentrations with Standard Control MO (Gene Tools).
Capped mRNA was made using a Message Machine Kit (Ambion). 150 pg of mRNA was used for over-expression of all mRNAs in all experiments.

Cell culture and luciferase assay
Luciferase reporter assays were performed in HEK 293 cells as previously described [29].

Western blot analysis and ChIP
Western blots were performed as previously described [29]. Mouse monoclonal anti-myc antibody (Covance) was used for both Western blot analysis and ChIP experiments. The dynabeads (Invitrogen) ChIP protocol was adapted from the Dorsky Lab (University of Utah) ZFIN Protocol. qPCR was used to quantify the enrichment of the fragment containing the RARE (DR5) in embryos injected with the myc-rarab1 mRNA with respect to control uninjected embryos.

Comparison of genomic sequences
The genomic sequence flanking zebrafish hoxb5b (28 to +8 kb) was compared with the corresponding region for Hoxb5 in mouse using mVista. NHR SCAN was used to identify binding sites for nuclear receptor.

Identification of rarab1 and RT-PCR
Rarab1 was identified by using BLAST against the zebrafish genome (Ensemble_V7) with the human and mouse RARa1 A domains. MacVector was used for sequences alignments. For RT-PCR, primer pairs were designed so that they specifically recognized rarab1 and rarab2 ( Figure 1B). Primer sequences are available upon request.

Cloning
The full-length coding sequence for rarab1 was cloned into pCS2p+. The rarab2-pCS2p+ construct used for overexpression was reported previously [29]. The myc tagged RARab1 was made using the pCS2+MT vector. For rarab1 and rarab2 probes, 536 base pairs (bps) of rarab1 and 443 bps of rarab2, which include the 59 untranslated region (UTR) and the specific A domains with no overlap, were cloned ( Figure 1B). These fragments were cloned into pGEM-T easy (Promega). qPCR Total RNA was isolated from 25 embryos, homogenized in TRIzol (Ambion) and collected using Pure link RNA Micro Kit (In Vitrogen). 1 mg or 0.5 mg RNA was used for cDNA synthesis using the ThermoScript Reverse Transcriptase kit (Invitrogen). Quantitative real time PCR (qPCR) for myl7, amhc, vmhc, nkx2.5, hand2, hoxb5b, hoxb5a, hoxb6b, dhrs3a, cyp26a1, aldhh1a2, rdh10a, rdh10b, raraa, rarab1, rarab2, rarca and rarcb, egfp and mcherry was performed using standard PCR conditions in a Bio-Rad CFX PCR machine with Power SYBR Green PCR Master Mix (Applied Biosystems). Expression levels were standardized to ef1a expression and all the data were analyzed using the 2 2DDCT Livak Method. All experiments were performed in a biological triplicate. Primer sequences are available upon request.

Area and length measurements
Areas of myl7, vmhc and amhc expressing cells were measured using ImageJ and statistical analysis was performed as previously described [26]. Length of egfp expression and distance between hoxb5b and egr2b were measured also using ImageJ and statistical analysis was performed as previously described.

Imaging of zebrafish heart and cell counting
Immunohistochemistry, cell counting and statistical analysis were done as previously described [26].

RA and DEAB treatment
RA and DEAB, treatment of embryos was done as previously described [26,27]. Embryos that have been used for gene expression analysis at 8 somites were treated with 1 mM DEAB, an Aldh1a2 inhibitor, beginning at 40% epiboly or with 1 mM RA for 1 hr beginning at 40% epiboly. For analysis of the effects of low concentrations of RA on heart development, embryos were treated with 0.05 mM RA for 1 hr beginning at 40% epiboly and harvested at 48 hpf. For rescue experiments related to the heart phenotype of RARab1 deficient embryos, embryos were treated with 0.025 mM DEAB beginning at 40% epiboly until 24 hpf. For rescue experiments related to the MHB in RARab1+Cyp26a1 deficient embryos, embryos were treated with 0.25 mM DEAB.

Statistical analysis
To assess whether the means of two groups are statistically different from each other, we applied the Student's t-test. A p value of ,0.05 was considered statistically significant. (C) In the presence of RA (red triangles), the GDBD-RLBD is able to promoted egfp (UAS responsive gene) transcription. (D-G) Tg(b-actin:GDBD-RLB);Tg(UAS:nfsB-mcherry) embryos are responsive to RA treatment. ISH for mcherry. Equivalent results were found when the Tg(b-actin:GDBD-RLB) line was crossed to Tg(UAS:EGFP) fish (data not shown) as were used for experiments in Figure 5. More detailed characterization of the stable transgenic RA sensor lines is reported in [31]. (D, E) Lateral views with dorsal right. (F, G) Dorsal views. In images D-G anterior is up.