The appearance of gill slits (endodermal pouches) and pharyngeal arches constitutes a crucial event of chordate evolution. Serially reiterated gill slits first appeared in hemichordates (e.g., acorn worms) and are still present as endodermal pouches during the development of vertebrates (such as ourselves). Pharyngeal arches are separated by endodermal pouches, and their development is a complex process that involves tissues from all three germ layers: neural crest cells and epidermis (ectoderm), muscles and blood vessels (mesoderm), and the endodermal pouches (Fig. 1; Graham and Smith, 2001). It was believed that the pharyngeal arches obtain their identity from neural crest cells that migrate in segmental streams from the hindbrain into the arches (Hunt et al., 1991). However, several recent studies have shown that the pharyngeal endoderm plays a much more important role in pharyngeal arch development than previously anticipated (Veitch et al., 1999; Piotrowski and Nusslein-Volhard, 2000) and patterns neural crest cells along the anterior-posterior (a/p) axis (Couly et al., 2002; Ruhin et al., 2003). Studies of zebrafish mutants, Xenopus, and knock-out mice have significantly contributed to our understanding of how the endodermal germ layer is initially formed (Alexander and Stainier, 1999; Gritsman et al., 1999; Yasuo and Lemaire, 1999). However, less is known about how the endoderm becomes patterned along the a/p axis (reviewed in Kiefer, 2003; Grapin-Botton, 2005; Stainier, 2005).
The pharyngeal arches consist of several tissues and express a large number of molecules belonging to multiple signaling pathways. We are only beginning to understand how these molecules interact with each other to generate these evolutionary conserved structures. In this study, we focus on how retinoic acid (RA) signaling influences endodermal pouch development at different embryonic stages. It is understood that RA affects pharyngeal arch development because of the phenotypes observed in infants whose mothers were exposed to high levels of Vitamin A during pregnancy (Lammer and Opitz, 1986; de Die-Smulders et al., 1995). In addition, these defects can be phenocopied by treating pregnant experimental animals with vitamin A (a precursor of RA) or synthetic RA (Kalter and Warkany, 1961; Shenefelt, 1972; Fantel et al., 1977; Davis and Sadler, 1981; Scambler, 2000). Also, genetic disruption of the RA signaling pathway leads to defects in structures derived from pharyngeal arches 3–4 in mice and zebrafish (Lohnes et al., 1994; Mendelsohn et al., 1994; Begemann et al., 2001; Grandel et al., 2002; Matt et al., 2003; Niederreither et al., 2003). While extensive evidence exists implicating RA signaling in pharyngeal arch development, it is not as clear on which cells or tissues of the developing arches RA may act upon.
Traditionally, the pharyngeal arch defects had been attributed to the deleterious effects of RA on neural crest cells until results from a study by Holland and Holland (1996) aimed to elucidate the effects of RA on amphioxus development suggested that RA affects the pharynx directly. Amphioxus larvae have a set of gill slits, which express genes homologous to genes expressed in vertebrate pharynx. In contrast to vertebrates, however, they lack neural crest cells. Nevertheless, after exposure to RA the mouth and gill slits fail to form (Holland and Holland, 1996).
The first indications that vertebrate endodermal pouch defects arise independently from neural crest defects in the absence of RA signaling was provided by targeted inactivation of retinoic acid receptors alpha and beta in mice (Dupe et al., 1999). In these mice, neural crest cells are specified and migrate normally, even though pharyngeal pouches 3 and 4 develop aberrantly. This study was followed by publications specifically focusing on the effects of RA deficiency on endodermal pouch development. These investigations were performed in mice in which RA signaling was blocked pharmacologically (Wendling et al., 2000; Matt et al., 2003) in vitamin A–deficient quails (Quinlan et al., 2002) and in zebrafish and mice with a mutation in raldh2 (Begemann et al., 2001; Grandel et al., 2002; Niederreither et al., 2003).
Over the past few years, numerous studies have demonstrated key roles for RA during multiple stages of development. For example, RA is required during gastrulation for proper hindbrain segmentation and also postembryonically for neuron specification (Begemann et al., 2001; Guidato et al., 2003; Linville et al., 2004; Maves and Kimmel, 2005). Similarly, during heart development RA is important for myocardial cell specification but also later for tube morphogenesis, ventricular maturation, and growth (Noden, 1984; Stuckmann et al., 2003; Keegan et al., 2005).
RA has been shown to be required for pancreas specification during gastrulation (Stafford and Prince, 2002) and members of the RA pathway (raldh2 and the RA-catabolizing enzyme cyp26a1) are expressed during gastrulation and in the developing endodermal pouches (Kudoh et al., 2002; Dobbs-McAuliffe et al., 2004, unpublished data). This lead us to hypothesize that RA also might play multiple roles in pharyngeal arch development, e.g., endoderm specification and endodermal pouch morphogenesis. In this study, we describe the defects caused by inhibition of RA during gastrulation and during pharyngeal arch morphogenesis in the hindbrain, neural crest cells, and endodermal pouches. Our data demonstrate that RA is not required for pouch endoderm specification, but rather for proper morphogenesis and segmentation of the endodermal pouches.
Since exogenous RA probably acts through a different pathway than endogenous RA (Mark et al., 2004), we chose to block endogenous RA activities at precise developmental time points to examine the function(s) of this molecule in the development of the hindbrain, neural crest, and endodermal pouches. The first set of blocking experiments was performed during gastrulation at a time when the endoderm becomes specified (Warga and Nusslein-Volhard, 1999; Stafford and Prince, 2002; Warga and Stainier, 2002). A second set of blocking experiments was performed during somitogenesis stages to test whether RA is also crucial for proper morphogenesis of the pharyngeal pouches. In order to abrogate RA signaling, we chose DEAB, a competitive, reversible inhibitor of retinaldehyde dehydrogenases. This compound is known to block all RA synthesis in the zebrafish embryo (Perz-Edwards et al., 2001). In fact, DEAB-treated embryos show a more severe phenotype than the raldh2 zebrafish mutants neckless/no fin (nls/nof) because nls/nof embryos still possess residual RA activity (Perz-Edwards et al., 2001; Begemann et al., 2004). Additionally, DEAB also induces stronger phenotypes than the commonly used pan RAR-blocker BMS493 (Begemann et al., 2004).
Inhibition of RA Signaling During and After Gastrulation Causes Neural Crest Defects
We blocked RA signaling during gastrulation (between 4–10 hr post-fertilization (hpf)) by incubating embryos in 50 μM DEAB, which led to the loss of rhombomere 5 (r5), a reduction of postotic mesoderm, and loss of pectoral fins, small otic vesicles, and heart edema, effectively phenocopying RA-blocked embryos described by other authors (Fig. 2A,B; Begemann et al., 2004). Treatment with this concentration of DEAB also caused the loss of insulin expression, which has been shown to depend on RA (data not shown, Stafford and Prince, 2002). Lower concentrations of DEAB (10 μM) result in incomplete loss of r5, whereas higher concentrations cause non-specific defects and high mortality (data not shown).
To investigate a possible influence of RA signaling on pharyngeal arch development as the arches undergo morphogenesis, embryos were treated with 100 μM DEAB starting at 16, 20, and 24 hpf until fixation at 30 hpf. This concentration resulted in pharyngeal arch phenotypes without inducing mortality or strong overall alterations in morphology. At 30 hpf, late blocked embryos are characterized by smaller otic vesicles and heart edema (Fig. 2C).
RA Deficiency During Gastrulation Affects Postotic Neural Crest Cells
In Tg (foxd3:gfp) 30-hpf zebrafish embryos, foxd3 is expressed in neural crest cells that have migrated into five pharyngeal arches (Odenthal and Nusslein-Volhard, 1998; Gilmour et al., 2002). In 30-hpf Tg (foxd3:gfp) embryos treated with DEAB between 4–10 hpf, only neural crest cells in the first two arches are GFP-positive, while neural crest cells in the more posterior arches appear to be absent (Fig. 3A,B). To test if neural crest cells are indeed absent in the posterior arches, or whether they are present but have down-regulated foxd3, we performed in situ hybridization experiments with genes known to be expressed in cranial neural crest cells. One such gene is the homeobox containing transcription factor dlx2, which is expressed in all migratory and post-migratory cranial neural crest cells, as well as in the diencephalon and the hypothalamus (Akimenko et al., 1994). In 30-hpf untreated embryos, neural crest cells in five distinct pharyngeal arches are labeled, while in blocked embryos only cells in the first two pharyngeal arches are visible (Fig. 3C,D). Another marker for neural crest cells is the basic helix-loop-helix transcription factor hand2. hand2 is expressed in the neural crest cells surrounding the mesodermal core of the arches and in the midline heart tube at 30 hpf in control embryos (Yelon et al., 2000; Miller et al., 2003). Embryos treated with DEAB during gastrulation lack expression in pharyngeal arches 3–5 and formation of the heart tube is disrupted; however, neural crest cells are clearly present in the anterior two pharyngeal arches (Fig. 3E,F). Since the neural crest markers foxd3, dlx2, and hand2 are absent from pharyngeal arches 3–5, we conclude that neural crest cells are indeed missing in the posterior pharyngeal arches in embryos treated with DEAB between 4–10 hpf. To investigate whether neural crest cells in pharyngeal arches 1 and 2 differentiate normally, we performed Alcian Blue cartilage stainings at 96 hpf (Piotrowski et al., 1996). Cartilages derived from neural crest cells of the first and second pharyngeal arches are reduced but present in RA-depleted embryos, while cartilages of pharyngeal arch 3 are reduced and cartilages of arches 4–5 are absent (Fig. 3G,H). The data show that neural crest cells that have migrated into the arches of RA deficient embryos still differentiate into cartilage.
Disruption of RA Signaling Post-Gastrulation Does Not Affect Neural Crest Migration Into the Pharyngeal Arches
To examine the effects of loss of RA signaling post-gastrulation on how neural crest cells populate the pharyngeal arches, Tg(foxd3:gfp) embryos were treated with DEAB from 16–30 hpf. Compared to untreated embryos, the GFP signal in neural crest cells of the mandibular (1st) and hyoid (2nd) arch of treated embryos is normal. However, neural crest cells in the fourth and fifth pharyngeal arches are not separated from one another (Fig. 3A,I). The presence of neural crest cells in the posterior pharyngeal arches is also confirmed by dlx2 and hand2 expression (Fig. 3C,J,E,K). dlx2 expression is absent in pharyngeal arch 5, which could be explained by a lack of neural crest cells in arch 5. However, we do not observe TUNEL-positive cells in the third neural crest stream. Alternatively, dlx2 might be downregulated in these cells, a phenomenon we also observed in vgo/tbx1 mutants, which lack posterior endodermal pouches (Piotrowski and Nusslein-Volhard, 2000). Even though neural crest cells are present in the posterior arches of embryos treated with DEAB between 16–30 hpf, they do not differentiate into cartilage (Fig. 3G,L), suggesting that the endodermal pouches, which are important for cartilage differentiation, might not have formed properly (described below).
Loss of RA During Gastrulation Affects Hindbrain Segmentation Possibly Causing Postotic Neural Crest Defects
To test whether the lack of neural crest cells in the pharyngeal arches is a reflection of neural crest migration defects, we analyzed several markers of migratory neural crest cells. In untreated embryos, neural crest cells commence migration at around 6 somites (som, Schilling and Kimmel, 1994). At around 16 som, three distinct streams are visible when labeled with dlx2 (Fig. 4A, s1–s3). In embryos treated with DEAB during gastrulation, the two anterior streams are present but the third one consists of few cells. Cells in this small third stream migrate anteriorly and ventral to the otic vesicle where they fuse with the second stream (Fig. 4B, arrowhead). No neural crest cells are migrating into pharyngeal arches 4–5. To test if postotic neural crest cells (in the third neural crest stream) are specified but then undergo cell death, we performed TUNEL staining in 14 som control and embryos treated with DEAB between 4–10 hpf (Fig. 4C,D). In embryos treated with DEAB during gastrulation, we do not observe TUNEL-positive cells posterior to the otic vesicle, suggesting that the absence of postotic neural crest cells is due to a failure in their specification. Surprisingly, neural crest cells are dying in the second stream (Fig. 4D, s2, arrow). This finding is unexpected, since second arch neural crest cells still form second arch cartilages, such as the ceratohyal cartilages (Fig. 3H). Future studies are necessary to determine if other second arch neural crest derivatives are affected in these embryos.
Since neural crest cells arise from specific rhombomeres, it is possible that the postotic neural crest defects are secondary to defects in hindbrain segmentation. Therefore, we examined the expression of genes that mark specific rhombomeres in the hindbrain. In untreated embryos, krox20 is expressed specifically in r3 and r5 at 24 hpf (Fig. 4E; Oxtoby and Jowett, 1993). In embryos treated with DEAB between 4–10 hpf, the expression of krox20 in r5 is greatly reduced or lost (Fig. 4F). In 30-hpf untreated embryos, hoxa2 is expressed in r2–5, as well as in neural crest cells in the second and more posterior arches (Fig. 4G; Prince et al., 1998). In treated embryos, the hoxa2 expression domain in the posterior hindbrain is slightly expanded and the posterior boundary is diffuse. Additionally, the expression in pharyngeal arches 3–5 is strongly reduced (Fig. 4H). These results suggest that the rhombomere boundaries posterior to r5 do not form properly and that neural crest cells normally migrating from these hindbrain levels are probably not specified (Fig. 4H).
In contrast, in embryos treated with DEAB, post-gastrulation neural crest cells appear to migrate normally and no increase in cell death is observed by TUNEL staining in either neural crest cells or the endoderm (Fig. 4I,J). Also, hindbrain segmentation is normal, since krox20 is expressed correctly in r3 and r5 in DEAB-treated embryos (Fig. 4K). Likewise, hoxa2 does not display changes in the hindbrain expression pattern (Fig. 4L).
In summary, these results demonstrate that RA depletion during gastrulation probably causes a loss of the postotic neural crest stream, whereas RA loss at later stages of development only affects the separation of postotic neural crest cells into distinct pharyngeal arches. Since in untreated embryos the endodermal pouches separate the pharyngeal arches, we investigated whether inhibition of RA synthesis post-gastrulation disturbs endodermal pouch formation.
RA Signaling Is Required for Endodermal Pouch Morphogenesis
To determine whether RA plays a role in endodermal pouch morphogenesis post-gastrulation, we blocked RA signaling with DEAB between 16 and 30 hpf. Subsequently, the embryos were fixed and processed with the antibody ZN8 (previously called ZN5), which recognizes DM-GRASP, a cell adhesion molecule of the immunoglobulin superfamily (Trevarrow et al., 1990; Fashena and Westerfield, 1999). ZN8 labels the endodermal pouches as well as sensory ganglia and nuclei (Fig. 5A). Zebrafish develop six endodermal pouches separating seven pharyngeal arches. In control embryos, endodermal pouches develop in a temporal fashion from anterior to posterior and by 30 hpf five pouches have formed (Fig. 5A; Crump et al., 2004). When we apply DEAB from 16–30 hpf and subsequently fix the embryos, we observe that only three of the five endodermal pouches have formed when compared to the control siblings (Fig. 5A,B). We did not expect to see defects in pouches 1 and 2 because they are known to be insensitive to RA (reviewed in Mark et al., 2004). However, we did expect to observe defects in endodermal pouch 3. This discrepancy probably results from how quickly endogenous RA is depleted and novel RA synthesis is prevented by DEAB application.
Since the endodermal pouches develop from anterior to posterior, we aimed to determine the temporal requirement of RA during pouch morphogenesis. We blocked RA signaling at different time points post-gastrulation between 16–30 hpf. In untreated embryos, the first endodermal pouch appears to be fully formed at around 18 hpf (our data; Crump et al., 2004; Holzschuh et al., 2005), the second at approximately 22 hpf, the third at 24 hpf, the fourth at 28 hpf, and the fifth at 30 hpf. Application of DEAB between 20–30 hpf results in an underdeveloped pouch 4 and pouch 5 is missing (Fig. 5C), whereas blocking of RA signaling between 24–30 hpf results in the loss of only pouch 5 (Fig. 5D).
The number of cartilaginous elements formed in the treated embryos reflects the loss of posterior pouches. The more pouches are present, the more pharyngeal arch cartilages form, which is not surprising given that it is well understood that signals from the endoderm are important for cartilage differentiation (Hall, 1980; Piotrowski and Nusslein-Volhard, 2000; Couly et al., 2002; David et al., 2002; Crump et al., 2004). The blocking experiments do not allow us to determine the precise stage at which the pouches require RA for their morphogenesis because of the time that it takes for RA to become completely depleted in the embryo. However, our data demonstrate that there is a specific temporal requirement for RA within pouches 3–5 as they are forming.
RA Does Not Play a Role in Pharyngeal Endoderm Specification
Because RA has been shown to be important in cardiac progenitor and pancreas specification, we aimed to determine if RA also affects the specification of the more anterior pharyngeal endoderm and endodermal pouches (Stafford and Prince, 2002; Keegan et al., 2005). We inhibited RA signaling during gastrulation between 4–10 hpf and compared the resulting phenotypes with embryos treated post-gastrulation. To better resolve the endodermal phenotypes resulting from RA inhibition, we labeled the pharyngeal endoderm with nkx2.3, an ortholog of the Drosophila homeobox gene tinman (Lee et al., 1996). In 36-hpf control embryos, nkx2.3 labels endodermal pouches 2–6 (Fig. 6A) but is not detected in the first pouch. In addition, nkx2.3 is expressed in the ectoderm of the pharyngeal arches (Fig. 6A, black arrowhead).
We observed that inhibition of RA signaling during gastrulation and during later stages of development leads to similar endodermal phenotypes, the only difference being that embryos treated at earlier stages possess fewer endodermal pouches than the later treated ones. In embryos treated during gastrulation, the number of endodermal pouches is reduced to two, whereas embryos treated anytime between 16–36 hpf form between 2–4 pouches (Figs. 5, 6B,C). Irrespective of when the embryos are treated with DEAB, the length of the ectodermal domain is not shortened when compared to control embryos (Fig. 6A–C; white arrowheads).
In embryos treated with DEAB during or post-gastrulation, the remaining posterior endodermal pouches are thicker than in control embryos and more nkx2.3-positive cells are located close to the midline (Fig. 6B, white asterisk; C–E). This raises the possibility that these nkx2.3-positive cells represent endodermal pouch cells that are not able to migrate laterally to form the pouches. Therefore, the loss of endodermal pouches upon DEAB treatment possibly presents a combination of loss of endodermal pouch cells and a defect in lateral migration of existing endodermal pouch cells.
The fact that pouch morphogenesis appears to be affected if RA signaling is blocked during gastrulation raises the question of whether DEAB does get washed off efficiently. To test whether DEAB can be washed out, we treated embryos with DEAB between 10–36 hpf and compared their phenotype with embryos that only had been treated between 10–14 hpf and subsequently raised in embryo medium (Fig. 6D,E). If DEAB could be washed out efficiently and normal RA signaling would resume, we would expect to lose one endodermal pouch in embryos treated for only 4 hr (this is approximately the time that it takes for a single pouch to form). However, embryos treated between 10–14 hpf only show maximally one more endodermal pouch compared to embryos that have been treated between 10–36 hpf, arguing that the effects of DEAB are not efficiently reversed after washing off the drug (Fig. 6D,E). This result suggests that the loss of endodermal pouches 3–5 that we are observing in 36-hpf embryos treated during gastrulation might represent defects caused by the continuing absence of RA post-gastrulation (see below).
To uncover if RA has any additional effects on endodermal pouch specification, we also treated embryos between 4–9 hpf and fixed the embryos at 100% epiboly. Subsequently, we performed in situ hybridization with several genes known to be expressed in pouch endoderm, such as her5 (Fig. 6F,G; Tallafuss and Bally-Cuif, 2003), foxa3, which is expressed in endoderm and axial mesendoderm (Warga and Kane, 2003), as well as sox17, a pan-endodermal marker (data not shown; Alexander and Stainier, 1999). We did not detect any obvious changes in the number of cells expressing these genes (Fig. 6F,G), suggesting that specification of the pouch endoderm is probably not affected by RA during gastrulation.
Conclusive evidence that RA is not affecting pouch endoderm specification is provided by experiments in which we took advantage of the fact that the effects of DEAB can be counteracted by exogenous application of RA. When we treat embryos between 4–10 hpf simultaneously with DEAB and RA at 10−7 M, normal endodermal pouch development proceeds and the expression of nkx2.3 is indistinguishable from control embryos (Fig. 6H,I). To uncover a role for RA specifically during gastrulation, we treated embryos with DEAB between 4–10 hpf and subsequently added exogenous RA between 10–36 hpf. If DEAB is affecting endoderm specification, we should not be able to rescue endodermal pouch morphogenesis later on. Yet this is not what we observe. Compared to embryos that have only been treated with DEAB for 6 hr during gastrulation and only formed 2 pouches (Fig. 6B), embryos that were subsequently soaked in RA at 10−7 M developed 4–5 pouches and resemble control embryos (Fig. 6H,J). These results indicate that endodermal pouches develop normally in the absence of RA during gastrulation. Importantly, neither development of the insulin-expressing pancreas, which is dependent on RA during gastrulation (Stafford and Prince, 2002), nor neural crest migration into the posterior pharyngeal arches can be rescued by subsequent application of RA (Fig. 6J–M). This demonstrates that inhibition of RA synthesis, specifically during gastrulation, does not affect endodermal pouch development and, therefore, does not play a role in pharyngeal endoderm specification.
To test if RA is required as individual pouches develop, we disrupted RA signaling at different time points post-gastrulation for 4-hr periods and subsequently soaked the embryos in RA until fixation at 30 and 36 hpf. Irrespective of whether we block RA signaling between 12–16 hpf (Fig. 7A–C) or between 18–24 hpf (Fig. 7D–F), treated embryos possess one less endodermal pouch when compared to control embryos. Spacing between the forming endodermal pouches is not affected if RA signaling is blocked during a 4-hr period (Fig. 7C,F). Interestingly, loss of RA starting at 12 hpf results in the shortening of the postotic region between the pharynx and the insulin-expressing pancreas (Fig. 7A,B; Begemann et al., 2001), whereas loss of RA at 18 hpf does not affect the length of this area (Fig. 7D,E). Further experiments are required to determine which tissues are affected and by what mechanism RA influences the development of the postotic region.
The posterior endodermal pouches cannot be distinguished from each other and we, therefore, cannot determine if one pouch is skipped, or if endodermal pouch development is arrested in the absence of RA. However, we believe the loss of RA is not simply arresting pouch development, since embryos that are exposed to RA after DEAB treatment exhibit enlarged second pouches, suggesting that loss of RA might have also affected the segmentation process. Future experiments are necessary to determine if RA plays a role in establishing endodermal pouch segmentation or whether it is involved in segmentation-independent processes, such as cell migration.
The Pharyngeal Endoderm Consists of Two Different Subpopulations
RA is known to affect patterning of tissues along the embryonic a/p axis (Gavalas and Krumlauf, 2000; Grapin-Botton, 2005). To test whether disruption of RA signaling leads to anteriorization (lengthening of the pharyngeal endoderm with respect to more posterior endoderm, which causes endodermal pouch defects), we performed in situ hybridization with foxa2, a transcription factor, which labels the more medial endoderm rather than the pharyngeal pouches (Fig. 8A). Surprisingly, neither blocking of RA signaling during gastrulation nor during pharyngula stages causes changes in the length of the medial endoderm (Fig. 8B–D). We normalized the data by calculating the ratio between the length of the pharyngeal endoderm (distance “a”) divided by the length of the pharyngeal endoderm plus the length of the brain expression (distance “b”; Fig. 8A, red letters). A t-test supports our conclusion that the means of the ratios between treated and control embryos are not significantly different.
Since we also did not detect any changes in the extent of the pharyngeal ectoderm (Fig. 6A–C), we conclude that RA does not affect patterning of the entire pharynx but only the segmentation and/or morphogenesis of the endodermal pouches. The dataalso suggest that the pharyngeal pouch endoderm and the more medial pharyngeal endoderm present two distinct populations of pharyngeal endoderm and that only the pouch endoderm is responsive to RA. Thishypothesis is supported by the fact that the pouch endoderm and the more medial endoderm express different sets of genes.
The effect of RA loss on hindbrain, neural crest, heart, and pharyngeal arch development has been the subject of a growing number of publications (reviewed in Maden et al., 1991; Gavalas and Krumlauf, 2000; Trainor and Krumlauf, 2001; Mark et al., 2004). Nevertheless, such studies have been performed in a variety of model organisms at different developmental stages and with varying degrees of RA synthesis inhibition, complicating integration of the results reported.
Complete depletion of RA, as judged by the severity of resulting hindbrain phenotypes is observed in Vitamin A–deficient (VAD) quail and rat embryos. Zebrafish embryos blocked with DEAB, an inhibitor of RA synthesis, or AGN193109, a pan retinoic acid receptor (RAR) blocker (Maden et al., 1991; Quinlan et al., 2002; Begemann et al., 2004; Linville et al., 2004), also display the effects of complete RA depletion. Additionally, the effects of RA on organ development have been tested by supplying exogenous RA instead of depleting it (Kalter and Warkany, 1961; Shenefelt, 1972; Fantel et al., 1977; Davis and Sadler, 1981; Scambler, 2000). However, it is now hypothesized that RA excess and deficiency act on different developmental pathways (Mark et al., 2004). To completely block RA, we therefore treated embryos with DEAB, which phenocopies VAD quail and rat embryos but provides us with the advantage of temporal control of RA signaling blockade during development. To test if RA plays distinct roles at different stages of pharyngeal arch development, we performed a series of blocking experiments during gastrulation and pharyngeal arch morphogenesis.
RA Is Required for Morphogenesis and Segmentation of the Pharyngeal Arches
Based on studies performed in amphioxus, mice, quail, and zebrafish, we now understand that RA is crucial for posterior endodermal pouch development (reviewed in Mark et al., 2004). However, these studies did not determine whether RA is involved in specification, morphogenesis, or patterning of the endodermal pouches.
Since RA is also known to affect patterning processes along the a/p axis during development, the question arises if RA affects patterning of the pharyngeal region (Wendling et al., 2000; Quinlan et al., 2002). Based on the normal ectodermal expression domain of nkx2.3 and the normal position of the most posterior medial nkx2.3-positive cells in treated embryos (Fig. 6A,B, black asterisk), we believe that patterning of the pharyngeal region as a whole is not affected by RA. Our data suggest that endodermal pouch cells either have lost migratory capabilities or they have lost the positional and segmental information of where within the pharynx to migrate laterally.
Our results also have conclusively determined that RA is not required for pouch endoderm specification but is required for proper morphogenesis or segmentation of endodermal pouches 3–6, because disruption of RA signaling at different time points post-gastrulation only affects the newly forming pouches. Interestingly, if we block RA signaling at several time points post-gastrulation for 4-hr periods, spacing between pouches is not affected but one less pouch forms. This phenotype can be explained by a requirement of RA for pouch morphogenesis, e.g., cell migration and/or RA might be a component of the segmentation clock. This second hypothesis is supported by our finding that the second pouch in DEAB-treated embryos is enlarged, a possible reflection of segmentation defects (Fig. 6B–E). However, as yet, the presence of a segmentation clock characterized by cycling gene expression has not been described in the developing pharynx. Clearly, more experiments are required to determine if RA plays a direct role in setting up segmentation of the endodermal pouches. Since RA is involved in somite segmentation and influences the size of the somites (Moreno and Kintner, 2004), it is likely that RA is one of the key molecules regulating pharynx segmentation as well.
Loss of RA During Gastrulation Causes Neural Crest Defects, Whereas Loss of RA Post-Gastrulation Causes Endodermal Defects
Classically, neural crest cells have been thought of as a tissue that patterns the pharyngeal arches with the endoderm playing a passive role in this process (reviewed in Graham and Smith, 2001; Trainor and Krumlauf, 2001; Graham, 2003; Graham et al., 2004; Mark et al., 2004). We now understand that the endoderm plays a more important role in pharyngeal arch morphogenesis and patterning than previously thought (Veitch et al., 1999; Piotrowski and Nusslein-Volhard, 2000; Couly et al., 2002; Graham et al., 2004; Mark et al., 2004). Since neural crest cells and the endoderm interact extensively during pharyngeal arch development, we aimed to determine the effects of RA deficiency on the development of both of these tissues during gastrulation and pharyngeal arch morphogenesis.
Our experiments have shown that complete loss of RA signaling during gastrulation causes a loss of postotic neural crest cells, which is probably caused by misspecification of the posterior rhombomeres that normally give rise to the postotic neural crest stream. Such a loss of the postotic neural crest stream is also observed in VAD quail embryos (Maden et al., 1996). Labeling with rhombomere-specific markers shows that the rhombomeres posterior to r4 are not specified when RA is blocked during gastrulation in either the zebrafish or in VAD quail embryos (Fig. 4E,F; reviewed in Maden et al., 1996; Gavalas and Krumlauf, 2000; Begemann et al., 2004; Linville et al., 2004; Mark et al., 2004; Maves and Kimmel, 2005). In contrast, when embryos are allowed to develop normally until about 16 hpf (when the endodermal pouches begin to form) and then RA synthesis is inhibited, the hindbrain is properly segmented and the postotic neural crest cells migrate normally into pharyngeal arches 3–5. However, in these embryos, even though neural crest cells are present in the posterior pharyngeal region, the posterior endodermal pouches that normally separate neural crest cells in arches 4–6 fail to develop. Based on similar observations in other model systems, it was concluded that loss of RA primarily affects the endodermal pouches and only secondarily causes neural crest defects (Dupe et al., 1999; Wendling et al., 2000; Mark et al., 2004).
Our results demonstrate that the underlying causes for craniofacial defects caused by misregulation of RA differ depending on when during development RA signaling is disrupted. Loss of RA specifically during gastrulation causes pharyngeal arch defects because postotic neural crest cells are not specified, even though endodermal pouch development is not affected. If RA signaling is disrupted post-gastrulation, then the posterior endodermal pouches fail to develop, which secondarily leads to neural crest defects. Therefore, in model systems in which RA signaling is disrupted continuously, beginning at fertilization, the resulting pharyngeal arch phenotypes are a combination of the loss of postotic neural crest cells during gastrulation and the failure of posterior endodermal pouch morphogenesis during somitogenesis stages.
The Pharynx Contains Two Distinct Populations of Endoderm
Previous studies of the tbx1 mutant van gogh (vgo) have suggested that the medial pharyngeal endoderm and the more lateral pouch endoderm present distinct cell populations. This is based on the fact that the medial and lateral endoderm express different sets of genes, and that their development depends on different molecular mechanisms, since a mutation in tbx1 only affects the pharyngeal pouch endoderm (Piotrowski and Nusslein-Volhard, 2000; Warga and Stainier, 2002; Piotrowski et al., 2003). The same is true for fgf8-;fgf3-morpholino injected embryos, which lack endodermal pouches but possess a normal medial endoderm (Crump et al., 2004). Our current experiments on the function of RA during endodermal pouch formation support the notion that the pharyngeal endoderm consists of at least two subpopulations, one of which is sensitive to RA while the other is not. Irrespective of whether we block RA signaling during gastrulation or pouch morphogenesis, we only observe defects in the lateral pouch endoderm, whereas the medial endoderm develops normally. This raises the interesting question of when and where the endodermal pouch cells become specified and thus become different from the rest of the pharyngeal endoderm. Two studies have reported lineage analysis on ventral foregut cells in chick and mouse embryos and a beautiful fate map study of the entire zebrafish endoderm has been performed (Warga and Nusslein-Volhard, 1999; Kirby et al., 2003; Tremblay and Zaret, 2005). However, our present studies indicate that more detailed lineage analysis experiments of the pharyngeal endoderm will be necessary to elucidate the origin and regionalization of the two pharyngeal endodermal subpopulations defined by RA sensitivity.
Other Signaling Pathways Involved in Pharyngeal Endodermal Pouch Formation
Members of almost all signaling pathways are expressed in the pharyngeal arches (Gerhart, 1999). The challenge lies in elucidating in which tissues these pathways are acting, whether they play distinct roles during different phases of pharyngeal arch development and, most importantly, how these signaling pathways interact with each other. So far, a failure in the initiation of endodermal pouch development has been described in zebrafish and mouse embryos with mutations in fgf3, tbx1, and pbx (Popperl et al., 2000; Jerome and Papaioannou, 2001; Lindsay, 2001; Merscher et al., 2001; Piotrowski et al., 2003; Herzog et al., 2004; Manley et al., 2004). In addition, fgf8 is involved in endodermal pouch development through its interaction with fgf3 (Crump et al., 2004). Other genes, such as pax9, shh, and foxi1 appear to be involved at later stages for endodermal pouch maintenance and growth. Disruption of these genes results in gene expression changes in the endodermal pouches (Peters et al., 1998; Nissen et al., 2003; Moore-Scott and Manley, 2005).
The phenotypes observed in Tbx1 and Fgf signaling-impaired zebrafish suggest that both pathways play similar roles in the segmentation and lateral migration of endodermal pouch cells (our results; David et al., 2002; Crump et al., 2004). We believe that, similarly to RA, fgf3 and tbx1 are not involved in the specification of the pouch endoderm because (1) the anterior two pouches form normally in tbx1-deficient embryos, (2) her5 expression, which labels prospective pouch endoderm at the end of gastrulation, is not reduced in tbx1-depleted embryos (unpublished results), and (3) tbx1 and fgf3 single mutant embryos still express fgf3 in the unsegmented pouch endoderm, arguing that the endodermal pouch cells are present but do not migrate laterally to form pouches (our unpublished data, Herzog et al., 2004).
Nevertheless, there is one striking difference between the functions of Fgf on the one hand and RA and Tbx1 signaling on the other in endodermal pouch development. While Fgf signaling is required for the formation of all endodermal pouches, including the two anterior pouches, RA and Tbx1 only influence the development of endodermal pouches 3–7. Crump et al. studied endodermal pouch morphogenesis in arches 1 and 2 and, with the help of elegant transplantation experiments, showed that proper segmental expression of fgf3 in the hindbrain and expression in the lateral mesoderm between 10–14 hpf is crucial for the development of these pouches (Crump et al., 2004). As such, they propose that hindbrain segmentation likely does not influence segmentation of the posterior pharyngeal arches.
In support of this hypothesis, our studies of inhibition of RA synthesis during zebrafish embryogenesis show that posterior endodermal pouch morphogenesis is disrupted even though the hindbrain is segmented normally. Therefore, Fgf and RA signaling must either be acting directly in the posterior endodermal pouches or functioning in the surrounding mesenchyme or ectoderm of the pharyngeal arches. In the case of RA, analysis of chimeric mouse embryos with mutant cells in which one allele was replaced with a RAR-gamma ligand-deficient receptor functioning as a dominant-negative RAR demonstrated that RA acts directly in the endoderm and not in the surrounding mesenchyme (Iulianella and Lohnes, 2002). In such chimeric embryos, contribution of wild-type cells to the endoderm always leads to normal development of the pharyngeal arches, even if the neighboring pharyngeal arch mesoderm consists almost completely of mutant cells. This experiment demonstrated that RA signaling is necessary in the endoderm for proper formation of the pharyngeal arches. In the case of Fgf and Tbx1 signaling, it still needs to be determined in which tissue they are required for posterior pouch segmentation. Fgfs are possibly required in the arch mesenchyme, as is the case for Fgfs in trachea development in flies, cecal development in mice, and lung budding in mammals (Chuang and McMahon, 2003; Ghabrial et al., 2003; Desai et al., 2004).
Based on our present study and other recent findings, we hypothesize that RA, Tbx1, and Fgf signaling are interacting during pharyngeal pouch development (Abu-Issa et al., 2002; Frank et al., 2002; Vitelli et al., 2002). This hypothesis is supported by the observation that Tbx1- and Raldh2-deficient mice and zebrafish show strikingly similar phenotypes (our unpublished data; Vermot et al., 2003). In addition, interactions between RA and Fgf signaling are crucial during the development of several other organs and tissues (Kudoh et al., 2002; Diez del Corral et al., 2003; Novitch et al., 2003; Moreno and Kintner, 2004; Maves and Kimmel, 2005; Vermot and Pourquie, 2005). For example, Fgfs and RA play opposing roles during boundary formation in somitogenesis in the trunk, raising the interesting question of whether segmentation of the pharyngeal endodermal pouches in the head relies on similar interactions between these two signaling pathways.
Fish Maintenance and Fish Strains
Zebrafish were raised and maintained under standard conditions (Westerfield, 1993). To slow down the rate of embryonic development, embryos were incubated at 22°C in a cooling incubator (EchoTerm IN20, Torrey Pines Scientific). Embryos were staged as described in Kimmel et al. (1995). Wild-type stocks used were TÜ (Tübingen) and AB* (Eugene, Oregon). Neural crest cells were visualized in the Tg(foxd3:gfp) transgenic line (Gilmour et al., 2002).
RA and RA Synthesis Inhibitor Treatments
Pharmacological inhibition of retinoic acid synthesis was performed by incubating live embryos in their chorions in embryo medium supplemented with DEAB (4-(Diethylamino) benzaldehyde; Sigma-Aldrich; no. D86256) at final concentrations of 50–100 μM under standard embryo culture conditions in the dark. DEAB stock solutions were diluted in DMSO (Fisher) and stored at a concentration of 10 mM at −20°C. As controls, half of the embryos were treated at the same time in 0.1% DMSO only. No defects were observed in control embryos suggesting that DMSO is not affecting embryogenesis (data not shown). Exogenous application of RA was performed by incubating the embryos in a solution of RA at 10−7 M in embryo medium containing 0.1% DMSO in the dark.
In Situ Hybridization
Embryos were dechorionated with forceps or by using a 1-mg/ml pronase solution for 8–10 min and then fixed in 4% PFA overnight at 4°C. After fixation, embryos were dehydrated in an ascending methanol series and stored in 100% methanol at −20°C. Embryos were rehydrated for 5 min in 50% and 30% methanol/PBST at room temperature (RT) and washed 2 × 5 min in PBST. To bleach pigment, embryos were incubated for 7–9 min in bleaching solution (10% H202, 5% formamide, 60% PBST) at RT and washed 3 × 5 min in PBST. For permeabilization, embryos older than shield stage were incubated with Proteinase K (10 μg/ml in PBST) at RT (24–27 hpf ∼7 min, 28–32 hpf ∼9 min) and refixed in 4% PFA for 20 min at RT. Subsequently, embryos were washed 3 × 5 min in PBST at RT, transferred into HYB+ solution, and prehybridized for 12 hr at 63°C. The remaining steps were performed as described (Piotrowski and Nuesslein-Volhard, 2000). For flat mounting, the yolk was dissected away, the embryos placed ventral side up on a microscope slide and covered with a cover slip. Subsequently, the samples were dehydrated in 100% methanol and rinsed with a 2:1 mixture of benzylbenzoate:benzylalcohol. Specimens were photographed on a Zeiss Axioskop 2 using an Axiocam camera. Probes used were hoxa2 (Prince et al., 1998), krox20 (Oxtoby and Jowett, 1993), foxa2/axial (Strahle et al., 1993), fgf3 (Kiefer et al., 1996), dlx2 (Akimenko et al., 1994), nkx2.3 (Lee et al., 1996), insulin (Stafford and Prince, 2002), and hand2 (Yelon et al., 2000). Measurements of the lengths of the medial pharyngeal endoderm were performed in Adobe Photoshop and the units on the y-axis present arbitrary numbers.
Embryos were fixed at different stages of development in 4% PFA overnight at 4°C. After fixation, embryos were dehydrated through an ascending methanol series (25, 50, 75, and 100% for 5 min each) and stored in 100% methanol at −20°C. Subsequently, they were washed two times in PBST (PBS, 1% Tween 20) at RT and washed once with distilled water for 1 hr at RT. For permeabilization, embryos were incubated for 7 min in 100% cold acetone at −20°C. After washing with PBST, samples were blocked for at least 1 hr in blocking buffer (PBST, 10% lamb serum) at RT. The primary antibody was diluted in blocking buffer (ZN8, 1: 200; anti-GFP, 1:400, Molecular Probes) and incubated overnight at 4°C or 4 hr at room temperature.
Specimens were washed four times in blocking buffer at RT and incubated with the secondary antibody in blocking buffer at 4°C overnight or for 4 hr at RT (anti-mouse-IgG [HRP], 1:500, Novus Biologicals; Alexa fluorophores, 1:400, Molecular Probes). To reduce background, embryos were washed several times with PBST. Horseradish peroxidase amplification was performed according to standard protocols (Piotrowski and Nuesslein-Volhard, 2000). The ZN-8 antibody developed by B. Trevarrow was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa, Dept. of Biological Sciences.
Alcian Blue Cartilage Staining
Embryos were fixed in 4% PFA overnight at 4°C, dehydrated in 50% MeOH and stored in 100% MeOH. Embryos were placed in Alcian blue staining solution (0.15% Alcian Blue [Sigma Alcian Blue 8 GX], 50% EtOH, 0.1M HCL, pH 1) overnight at RT on the shaker. The next day, the specimens were rinsed repetitively with 100% EtOH until the liquid in the tube became clear. Subsequently, the embryos were rehydrated with 50% PBS/50% EtOH into 1× PBS. Embryos were then digested in 0.05% trypsin in 1× PBS for at least 2 hr at RT. The digest was stopped by washing in 1× PBST 2 to 3 times, 5 min per wash. To bleach the embryos, they were placed in 3% hydrogen peroxide, 1% KOH for 8–10 min until pigments were gone.
Embryos were washed several times in 1× PBST and cleared in 80% glycerol to stop the reaction.
Detection of Apoptotic Cells
To detect cells undergoing programmed cell death, we performed TUNEL (terminal deoxynucleotidyl transferase [TdT]-mediated deoxyuridinetriphosphate [dUTP] nick end-labelling). Control and DEAB treated embryos were fixed overnight at 4°C in 4% paraformaldehyde and TUNEL was performed using the ApopTag Red In Situ Apoptosis Detection kit (Chemicon, Temecula, CA).
We thank Gretchen King and the staff of the Centralized Zebrafish Facility for excellent fish care; Rachel Warga, Eric Ross, and Susan Chapman for fruitful discussions, and Alejandro Sanchez Alvarado for critically reading the manuscript. In addition, D. Kopinke would like to thank Thomas Bosch for his support.