Zebrafish limb development is triggered by a retinoic acid signal during gastrulation



Studies in mouse and zebrafish show that vertebrate forelimb development is initiated by retinoic acid (RA). An RA signal leads to transcription of tbx5 in forelimb precursors which is necessary and sufficient for limb development. However, the timing of the RA signaling event has remained controversial as have source tissue and tissue interactions. We have thus determined the contribution of RA to zebrafish pectoral fin development at different developmental stages. Specifically, an early gastrula stage RA signal triggers the process that leads to determination of tbx5-expressing limb precursors, while a later somitogenesis stage RA signal maintains these precursors. Preceding the lack of tbx5-expressing limb precursors in RA deficient zebrafish embryos, aldh1a2 and cyp26a1 expression domains are distorted along the gastrula margin suggesting that positional values in the ventrolateral mesodermal anlagen are affected. We propose that limb precursor determination requires RA dependent specification of lateral plate territories during gastrulation. Developmental Dynamics 240:1116–1126, 2011. © 2010 Wiley-Liss, Inc.


Vertebrate limbs develop from limb buds that arise from the lateral plate mesoderm. Limb buds are autonomous structures and limb growth and patterning is under the control of signals from within rather than from outside the buds (Tabin and Wolpert,2007; Towers and Tickle,2009). In contrast, determination of the territories where limb bud formation will occur, so called limb fields, depends on patterning processes that are less well understood (Capdevila and Izpisua Belmonte,2001; Mercader,2007; Allard and Tabin,2009). Embryologically, chick wing fields were defined in the pre-bud lateral plate mesoderm by their capacity to form a wing bud when grafted to ectopic positions in the lateral plate of a host embryo (Kieny,1960; Pinot,1970; Saunders and Reuss,1974; Michaud et al.,1997). Molecularly, the “direct initiator of limb bud formation” (Agarwal et al.,2003) is expression of the T-box transcription factor tbx5 in the precursors of mouse forelimb, chick wing and zebrafish pectoral fin buds (Begemann and Ingham,2000; Saito et al.,2002; Agarwal et al.,2003) which is both necessary and sufficient for forelimb formation (Ahn et al.,2002; Garrity et al.,2002; Ng et al.,2002; Rallis et al.,2003; Agarwal et al.,2003; Takeuchi et al.,2003; Minguillon et al.,2005). Thus, to understand limb precursor determination, it is necessary to understand the processes that lead to tbx5 expression.

A signaling molecule known to act upstream of tbx5 during forelimb/pectoral fin development is retinoic acid (RA; Begemann et al.,2001; Mic et al.,2004; Gibert et al.,2006). RA has long been known to be able to influence patterning of regenerating and developing limbs (Maden,1982; Tickle et al.,1982) while its involvement in limb initiation has been suggested only in recent years (Grandel et al.,2002; Mic et al.,2004; Gibert et al.,2006; Mercader et al.,2006; Zhao et al.,2009). RA is synthesized from its precursor retinal that is maternally supplied to the zebrafish egg (Costaridis et al.,1996). This oxidation is catalyzed by aldehyde dehydrogenase 1a2 (Aldh1a2), the transcription of which is restricted to the mesoderm in early embryos, while the action of RA is further limited by its degradation through a cytochrome P450 enzyme, Cyp26a1. Mouse and zebrafish aldehyde dehydrogenase 1a2 (aldh1a2) mutants that are unable to synthesize retinoic acid during early development do not develop limbs/fins and do not show tbx5 expression, while exogenous application of RA to aldh1a2 mutant embryos can rescue tbx5 expression and promote limb/fin development (Niederreither et al.,1999;2002; Begemann et al.,2001; Grandel et al.,2002; Mic et al.,2004; Gibert et al.,2006; Zhao et al.,2009).

The precise timing of the requirement for RA has remained subject to debate because the competence of the prospective limb/fin bud mesoderm to respond to an RA signal with fin/limb precursor formation extends from gastrulation to mid-somitogenesis stages. It is important to clarify this timing issue because it will provide a key to the specific RA source and the tissues that need to interact during the earliest steps of the limb induction process; e.g., epidermal–mesenchymal interactions in the embryonic flank are known to be at work in the limb field immediately preceding limb bud outgrowth. They involve Fgf10 in the mesenchyme and fgf8 in the overlying epidermis (Ohuchi et al.,1997; Sekine et al.,1999; Norton et al.,2005). It is tbx5 that is required to initiate mesenchymal fgf10 expression (Agarwal et al.,2003; Fischer et al.,2003).

But while expression of Tbx5 in the lateral plate of the chick is the earliest known maker expressed in wing precursors, it is detected only 3 hr after wing field determination is observable by the classical lateral plate graft (Saito et al.,2002). Using an explant strategy and transplants of the prospective wing mesoderm into the more permissive environment of the spinal cord of a host embryo, Saito and coworkers demonstrated that the prospective wing precursors are specified with respect to their later Tbx5 expression and prospective skeletal pattern already 16 hr before determination is testable by the classical lateral plate grafting assay (Saito et al.,2002). Thus, the limb specification process likely starts significantly earlier than Tbx5 expression in limb precursors. But when and where does RA come into play? While all studies agree that the competence of the lateral plate mesoderm to respond to RA with limb/fin development extends up to mid-somitogenesis in the mouse and fish, different conclusions have been reached on the timing of the endogenous RA signaling event that leads to limb/fin precursor determination and, consequently, to tbx5 expression: Determination of forelimb/pectoral fin precursors has been proposed to require RA signaling from the margin and/or the paraxial mesoderm of the zebrafish embryo either before somitogenesis (Grandel et al.,2002) or from trunk somites during early somitogenesis stages (Gibert et al.,2006; Mercader et al.,2006; Zhao et al.,2009). Importantly, Zhao and coworkers showed results of rescue experiments in mouse aldh1a2 mutants that suggest lack of RA signaling in the prospective limb precursors and early forelimb buds during early somitogenesis (Zhao et al.,2009), a result compatible with limb precursor determination before somitogenesis. It is noteworthy in this context that raldh1a2 and its antagonist cyp26a1 are expressed in mouse, chick, and zebrafish embryos already during gastrulation stages (Fujii et al.,1997; Niederreither et al.,1997; Swindell et al.,1999; Begemann et al.,2001; Grandel et al.,2002; Kudoh et al.,2002; Dobbs-McAuliffe et al.,2004).

We thus reinvestigated the timing of the initial RA signaling event that leads to fin precursor determination and compared zebrafish embryos mutant for aldh1a2 (nlsu11; the no-fin allele of neckless, Grandel et al.,2002) to embryos treated with an inhibitor of RA production. We find a repeated requirement of RA signaling during the fin determination process. During somitogenesis stages, fin precursors rely on continued RA signaling for their maintenance, but also acquire a distinct function of controlling growth of the fin bud after somitogenesis. Importantly, we find that endogenous RA signaling acts already during gastrulation stages upstream of fin specific tbx5 expression. In agreement with a function of RA in the gastrula stage mesoderm, we demonstrate that aldh1a2 and cyp26a1 are regulated by RA already during this time. We propose that RA signaling acts repeatedly, and already during gastrulation stages specifies the fin precursors in the prospective lateral plate mesoderm.


RA Signaling Is Necessary for Fin Precursor Specification

As reported previously, nlsu11 (aldh1a2) mutants lack pectoral fins at early larval stages (Begemann et al.,2001; Grandel et al.,2002; Fig. 1A,B). Mutants can already be recognized from the 10-somite (s) stage on by the lack of the tbx5 expression domain lateral to somites 1–3 which marks fin precursors (Fig. 1C–H), suggesting involvement of RA signaling in the process of pectoral fin precursor specification before the 10s stage.

Figure 1.

Phenotypic analysis of nlsu11 mutant and sibling embryos and larvae. A–H: Scanning electron micrographs (A,B) and stereomicroscopic images (C–H). Dorsal views, anterior is to the left. A: At 3 days postfertilization (dpf), wild-type siblings show pectoral fins apposed to the yolk. B: No pectoral fins are seen in most nlsu11 mutants (arrows in A and B). C:tbx5 expression marks the initially contiguous heart (arrowhead) and pectoral fin (asterisk) primordia in wild-types at the 10s stage. E: The fin precursor specific tbx5 expression domain lateral to somites 1–3 (asterisk) has separated from the heart field domain (arrowhead). G: Pectoral fin buds retain tbx5 expression at 32 hr (asterisk). D,F: A fin precursor specific tbx5 expression domain is lacking at all stages in nlsu11 embryos (asterisks). H: This prefigures the later observed lack of pectoral fin buds (asterisk).

Temporal Requirement of RA Signaling During Fin Development

To test the precise temporal requirement of RA signaling during pectoral fin precursor determination in zebrafish, we applied a pharmacological inhibitor of RA synthesis starting during gastrulation to mid-somitogenesis stages and continued the treatments until 24 hours postfertilization (hpf) or 32 hpf. We then monitored presence or absence of tbx5-expressing fin precursors or buds, respectively. As an RA synthesis inhibitor, we used 10−5M diethylaminobenzaldehyde (DEAB) that was previously reported to abrogate tbx5-expressing fin precursors at 24 hr, just before fin bud formation, when applied throughout somitogenesis (Gibert et al.,2006). We confirmed these results: blocking RA synthesis with DEAB throughout somitogenesis is sufficient to delete tbx5 expression and fin bud formation at 24 hpf or 32 hpf, respectively (Fig. 2A,B). As expected, expression of tbx5 in fin precursors/buds at 24 hpf or 32 hpf becomes detectable again when treatments are initiated at or after the 6s stage (Fig. 2A,B). However, we find that tbx5 expression and size of these buds are reduced, effects that become less pronounced as inhibitor treatments start at progressively later somitogenesis stages (Fig. 2B).

Figure 2.

Application of 10−5M diethylaminobenzaldehyde (DEAB) to zebrafish embryos. A:tbx5 positive fin precursors are absent at 24 hpf in embryos treated from shield, 75% epiboly and tail bud stage but are detectable if treatments are initiated at the 6-somite (s) stage; dorsal view; anterior is to the left. B: DEAB treatments throughout somitogenesis lead to loss of tbx5 expression and fin bud development at 32 hours postfertilization (hpf). Later treatments lead to reduction of tbx5 expression and fin bud size at 32 hpf; lateral view, anterior is to the left. C: DEAB treatments during early to midsomitogenesis reduce tbx5 expression and fin bud size at 32 hpf; lateral view, anterior is to the left. D: Schematics of inhibitor treatments and results.

Our findings thus suggest that RA signaling influences fin precursor/bud development during somitogenesis as shown previously (Grandel et al.,2002; Niederreither et al.,2002; Mic et al.,2004; Gibert et al.,2006; Zhao et al.,2009); however, the influence of RA on fin precursors persists well into somitogenesis stages. It is thus unclear whether RA functions in fin induction during a time window between tail bud and the 6s stages as suggested by Gibert et al. (2006).

To more directly test the contribution of RA signaling to fin development during early somitogenesis, we confined our inhibitor treatments to early somitogenesis stages only. We selectively treated zebrafish embryos with DEAB between tail bud stage, and 6s, 8s, 10s, 12s, and 14s stages. We then monitored fin bud development and tbx5 expression at 32 hpf (Fig. 2C). These inhibitor treatments overlap with and extend beyond the window of competence during which the lateral plate of aldh1a2 mutants is able to react to an exogenous RA signal with initiation of fin development (Grandel et al.,2002; Gibert et al.,2006).

These more confined DEAB treatments, cause small fin buds at 32 hpf that decrease in size with incubation time. However, tbx5-expressing precursors are present in all cases. Strength of tbx5 expression is progressively reduced though, the longer the treatments last (Fig. 2C). This experiment thus yields the complementary result to the experiment described above (compare Fig. 2B to 2C). This suggests that application of DEAB during early somitogenesis stages alone does not cause the late loss of fin precursors. Instead, we find a persistent requirement for RA signaling throughout somitogenesis to support the 32 hr limb bud regarding its growth and tbx5 expression.

Expression of aldh1a2 and cyp26a1 Indicate a Transcriptional Response to Altered RA Levels in the Gastrula Mesoderm

RA signaling, as monitored by aldh1a2 expression, occurs already during gastrulation, and thus much earlier than somitogenesis stages in fish, chick, and mouse. Along with other lines of evidence (see the Introduction section), this raises the possibility that RA signaling might also act significantly earlier on the development of fin precursors. We thus sought to directly examine the consequence of loss of RA signaling on the prospective lateral plate mesoderm during gastrulation, and tested ventrolateral hypoblast markers after RA inhibition. While several neuroectodermal genes have been reported to be RA controlled during gastrulation (Grandel et al.,2002; Kudoh et al.,2002; Maves and Kimmel,2005; White et al.,2007) only the weak paraxial expression domain of cyp26a1 in the epi- and hypoblast straddling the head trunk border by mid-gastrula stages has been reported to depend on local RA levels (White et al.,2007; Fig. 3A–F).

Figure 3.

A–K: Perturbation of cyp26a1 (A–F) and aldh1a2 (G–K) expression domains in the gastrula hypoblast upon 10−5M diethylaminobenzaldehyde (DEAB) treatments from 30% epiboly until the stage of fixation. Animal is to the top and dorsal to the right (A,B,D,E,G,H) or shown in dorsal view (J,K) and in parasagittal sections (C,F). A–F: DEAB treatments lead to an absence of the paraxial cyp26a1 domain at 75% and to a notable reduction of the marginal hypoblast expression domain (compare expression domains in A,D,C to B,E,F, arrowheads). G,H,J,K: DEAB treatments from 30% epiboly on reveal an enhanced signal intensity of the aldh1a2 expression domain ventrolaterally at 90% epiboly and tail bud stage (compare G,J to H,K). I: The graph shows the relative enhancement of marginal aldh1a2 expression at 90% epiboly of 10 experimental embryos (red) compared to 10 control embryos (blue) in one experiment; asterisks show significant differences in staining in controls vs. experimental embryos.

We thus investigated whether this is true for cyp26a1 expression in the gastrula hypoblast in general. We also asked whether aldh1a2 might be regulated in a corresponding way.

We tested the hypothesis that RA-mediated control of cyp26a1 and aldh1a2 transcription operates in the hypoblast already during gastrulation after treating embryos from late blastula stages on with DEAB, and find evidence for RA dependent gene expression. Upon inhibitor treatments, the marginal domain of cyp26a1 decreases in strength and width by 75% epiboly and the paraxial subdomain of the cyp26a1 expression located in the dorsal hypoblast is lost as a consequence of this treatment (Fig. 3A–F; cf. White et al.,2007). In a complementary manner, aldh1a2 staining increases in strength and width toward the ventral side of the gastrula (Fig. 3G–K), which is obvious at the end of gastrulation (Fig. 3J,K). We observe this expansion by 90% epiboly, which is quantifiable, when measuring gene expression levels in DEAB-treated and untreated control embryos (Fig. 3G–I). In two independent experiments, signal strength of the inhibitor-treated and control embryos differed significantly along the dorsoventral axis (Mann-Whitney Exp.1: U = 90, n1 = n2 = 10, P < 0.001; Exp.2: U = 93, n1 = n2 = 10, P < 0.001). The altered expression domains of aldh1a2 and cyp26a1 indicate those regions in the embryo that react immediately to reduced RA levels. The prospective lateral plate mesoderm where pectoral fin precursors are known to reside (Keegan et al.,2004; Warga and Nusslein-Volhard,1999) is among the affected territories. Thus, altered aldh1a2 and cyp26a1 expression directly reflect an early patterning influence of the loss of RA signaling within the prospective lateral plate mesoderm during gastrulation.

RA Signaling During Gastrulation Triggers the Fin Precursor Determination Process

Because RA signaling influences patterning within the prospective lateral plate mesoderm already during gastrulation we wished to observe the fate of fin precursors as early as possible after DEAB treatments to see if the early effects of the inhibitor on the gastrula hypoblast would be succeeded by an early lack of fin precursors. To this end, we initiated RA inhibitor treatments at different points in time between shield and the 6s stages, and monitored tbx5 expression in the lateral plate as early as the 10s stage. At this stage, the posterior part of the tbx5 expression domain lateral to somites 1–3 represents the pectoral fin precursor population (Ahn et al.,2002).

The short treatment intervals and early read out of fin precursor formation necessitate characterization of the kinetics of DEAB effects on RA signaling. We thus performed time course experiments during gastrulation (Fig. 4A) and early somitogenesis (Fig. 4B). We fixed embryos after different DEAB incubation periods and examined the expression of cyp26a and hoxb5a, two bona fide RA target genes, as direct read outs of RA signaling (Loudig et al.,2000; Oosterveen et al.,2003; cf. the analysis of cyp26a1 expression in experimental embryos above). We also performed a washout of DEAB after 2 hr of incubation during gastrulation and investigated hoxb5a expression in embryos at the 10s stage (Fig. 4C).

Figure 4.

Kinetics of diethylaminobenzaldehyde (DEAB) effects on retinoic acid (RA) target gene transcription after incubation of embryos with 10−5M DEAB for different periods of time during gastrulation (A), somitogenesis (B), and after inhibitor wash out (C). Stages of inhibitor incubation are given on the lower left, duration in hours (h) on the lower right side of each frame. A: During gastrulation, inhibitor treatments for 2.75 h, 2 h, and 1 h reduce the ventral marginal domain and the paraxial domain of cyp26a expression (arrowheads; dorsal is to the right). B: Application of DEAB for 8 h, 4 h, or 1 h before the 10s stage strongly down-regulate hoxb5a expression (dorsal view, anterior is to the left). Lack of RA synthesis manifests itself at least 1 h after DEAB application (A,B). C: Wash-out of DEAB at 75% epiboly after 2 h of treatment does not affect hoxb5a expression at 10s (dorsal view, anterior is to the left).

Upon application of DEAB during gastrulation or somitogenesis, we find marginal and paraxial cyp26a1 expression and hoxb5a expression noticeably down-regulated already after 1 hr of incubation, respectively, indicating that 1 hr is the maximum time period after which DEAB treatments effectively interfere with RA signaling (Fig. 4A,B). Washing away DEAB at mid-gastrulation, after 2 hr of treatment, allowed for wild type levels of hoxb5a at the 10s stage, indicating reestablishment of RA signaling after the wash out (Fig. 4C).

We performed inhibitor experiments from shield to the 10s stage or from 75% epiboly to the 10s stage and observed that tbx5-expressing pectoral fin precursors were missing at the 10 s stage. In contrast, inhibitor treatments that started at tail bud stage or later, did not lead to the loss of fin precursors at the 10s stage (Fig. 5A), while the latter treatments strongly down-regulated hoxb5a transcription (Fig. 4B). The inhibitor treatments in this experiment (Fig. 5A) lasted for 8–2 hr. Notably, a 2-hr time window of inhibiting RA synthesis during gastrulation seemed sufficient to cause a loss of tbx5 expression at the 10s stage. To corroborate this finding, we applied the inhibitor for 2 hr at different time intervals between shield and the 10s stage. Indeed, inhibition of RA synthesis for 2 hr during gastrulation is sufficient to interfere with the regular onset of fin precursor-specific tbx5 expression at the 10s stage (Fig. 5B), while DEAB treatments beginning at tail bud stage or later have no effect on the extent of the tbx5 domain (Fig. 5B). Notably, a 2-hr inhibitor treatment during gastrulation only, which is sufficient to block development of tbx5-expressing fin precursors at the 10s stage (Fig. 5B), does not cause general loss of RA-regulated hoxb5a expression (Fig. 4C). In contrast, inhibitor treatments after tail bud stage cause down-regulation of the RA target hoxb5a at the 10s stage (Fig. 4B) but do not interfere with tbx5 expression (Fig. 5A,B). This underscores the significance of a gastrula stage RA signal to trigger the process of fin precursor determination lateral to somites 1–3 at the 10s stage.

Figure 5.

Application of 10−5M diethylaminobenzaldehyde (DEAB) to zebrafish embryos before the 10-somite stage; myoD, tbx5 double-stainings at the10s stage; dorsal views, anterior is to the left. A: DEAB is applied from successively later stages onward; early fin precursors are absent if treatments include gastrula stages but are present if treatments start after gastrulation. B: Two-hour inhibitor treatments: Fin precursors are absent if treatments are conducted during gastrulation but are present if treatment are conducted during early somitogenesis. C: Schematics of inhibitor treatments and results. Red boxes indicate the time window of fin precursor specification.

RA Signaling During Somitogenesis Is Able to Trigger a Delayed Fin Precursor Determination Process

Inhibitor-treated embryos lacking the gastrula stage RA signal do not develop obvious fin buds by 32 hpf, which is 6 hr after they arise in controls (Fig. 6A). Nevertheless, on the gene expression level, they do show a few fin precursors with reduced tbx5 expression by this time (Fig. 6B). Thus, on the one hand, the gastrula stage inhibitor applications demonstrate that the embryo fails to execute the regular pectoral fin developmental program upon these treatments (Fig. 5A,B). On the other hand, they reveal a persistent competence of the lateral plate to initiate fin development with a delay in response to the endogenous postgastrulation RA signal (Fig. 6B). In agreement with this, the application of exogenous RA to nls or inhibitor-treated embryos from the tail bud stage on has been reported to rescue fin precursors and fin development (Grandel et al.,2002; Gibert et al.,2006). The results of our experiments with either an early or a late read out, suggest that, in the unperturbed embryo an RA signal acts before the end of gastrulation, to initiate the process that leads to fin precursor determination and tbx5 expression at the 10s stage but that a reestablished RA signal after gastrulation may still be sufficient to trigger fin precursor development.

Figure 6.

Response of the lateral plate upon a delayed endogenous retinoic acid (RA) signal; lateral views. A: Lack of fin buds in embryos at 32 hours postfertilization (hpf) upon application of 10−5M DEAB from 30% epiboly to tail bud stage. B: In situ hybridization reveals some tbx5 expression in the mesoderm of the presumptive fin bud territory at the same time.


Retinoic acid acts early on during limb/fin development (Grandel et al.,2002; Mic et al.,2004; Gibert et al.,2006; Zhao et al.,2009). However, the developmental stage during which RA influences pectoral fin development for the first time has remained controversial. We previously suggested that RA signaling acts before the end of gastrulation in zebrafish pectoral fin precursor determination (Grandel et al.,2002). In contrast, Gibert et al. (2006) proposed that the RA signal required for fin precursor determination acts during early somitogenesis and involves RA signaling from anterior somites to the lateral plate mesoderm. The experimental approach, used in both studies, relies on inhibition of RA signaling or synthesis in embryos during defined time windows. Both studies show absence of late fin precursors or early fin buds if inhibitor treatments include gastrulation stages. However, inhibitor treatments during early somitogenesis yield differing results, depending on inhibitor concentration. Application from tail bud stage on causes loss of fin precursors at 24 hpf if high inhibitor concentrations are used (Gibert et al.,2006), whereas low inhibitor concentrations do not interfere with fin bud formation (Grandel et al.,2002).

These experimental differences are at the basis of the divergent conclusions of the two studies. However, the experimental results can also be viewed as complement to each other. We suggest that the results obtained with different inhibitor concentrations reflect different functional contexts of RA signaling during gastrulation vs. early somitogenesis that involve higher vs. lower endogenous RA levels, respectively. With the experiments reported above, we aim to scrutinize this proposal and to reconcile the divergent findings.

Experimentally, we made use of DEAB to inhibit RA synthesis. We find DEAB effective in down-regulating transcription of bona fide RA targets during gastrulation and early somitogenesis stages within 1 hr of application. During our study, DEAB was applied to embryos for at least 2 hr, which ensures that the treatments lead to pronounced down-regulation of RA signaling during the incubation period.

Our initial experiments show that inhibition of RA synthesis restricted to the first 6 hr of somitogenesis does not lead to loss of fin buds, which is seen if embryos are DEAB treated for 14 hr (to 24 hpf) or 22 hr (to 32 hpf), but rather affects their growth. From this result, we conclude that RA signaling acts to maintain fin precursors during somitogenesis stages. In the light of the down-regulation of RA target gene expression after even shorter DEAB treatments during early somitogenesis, our result suggests that RA signaling during the first 6 hr of somitogenesis is not necessary for pectoral fin precursor determination in spite of the competence of the lateral plate to respond to an RA signal with limb precursor determination during that time.

To address this possibility more directly, we undertook experiments involving an earlier read out of fin precursor determination. Because of the maintenance function of RA on fin precursors during somitogenesis, we analyzed tbx5 expression as early as the 10s stage lateral to somites 1–3 where it marks the initial fin precursor population (Ahn et al.,2002). This enabled us to distinguish a failure of early determination from the subsequent loss of fin precursors due to failed maintenance. Inhibition of RA synthesis for 2 hr during gastrulation is sufficient to inhibit development of tbx5-expressing fin precursors at the 10s stage. In contrast, inhibition of RA synthesis for 2 hr after gastrulation does not affect development of tbx5-expressing fin precursors at the 10 s stage. hoxb5a, a direct RA target, shows the opposite behavior upon these DEAB treatments, suggesting that an RA signal during gastrulation initiates the process that leads to fin precursor determination lateral to somites 1–3 at the 10s stage.

Nevertheless, in the absence of RA signaling during gastrulation, supplementation of RA before the 10s stage can trigger fin development in an experimental situation (Grandel et al.,2002; Gibert et al.,2006). Similarly, we find the endogenous RA production capable to establish a very reduced fin precursor population in a retarded manner in such embryos, indicating that the lateral plate mesoderm remains competent to respond to an RA signal beyond the timing of the endogenous signal. Such extended competence is well known in the context of vertebrate limb development (e.g., Cohn et al.,1995). Given that the function of RA during somitogenesis lies in the maintenance of the already determined fin precursors, this may enable the recapitulation of the missing earlier, RA-dependent inductive step. A similar example of such plasticity is provided by the late rescue of the isthmic fold demarcating the junction between optic tectum and cerebellum in the brains of acerebellar (fgf8) mutant embryos. Acrylic beads soaked in recombinant Fgf protein are able to rescue the wild-type structure of this brain region, although the defect in the mutant originates much earlier (Reifers et al.,1998; Jaszai et al.,2003).

As evidence for an RA requirement in that part of the hypoblast fated to become the pectoral fin primordium we show altered gene expression domains in RA depleted embryos. Expression of cyp26a1 and aldh1a2 has been shown to be feedback-controlled by RA during somitogenesis (Begemann et al.,2001; Dobbs-McAuliffe et al.,2004), which has been explained as a direct effect of RA on the respective promoters (Dobbs-McAuliffe et al.,2004; Loudig et al.,2005; Elizondo et al.,2009). We show that RA-depleted embryos show similar effects on cyp26a1 and aldh1a2 expression in the hypoblast already during gastrulation, suggesting that RA already exerts control on gene expression during gastrulation.

These findings suggest that RA signaling influences pattern formation in the hypoblast. In embryos where RA levels are perturbed during gastrulation, the processes of proper precursor selection, patterning, and determination of other mesendodermal organs have been reported to fail in the case of heart-, pronephros-, and pancreas-precursors, respectively (Stafford and Prince,2002; Keegan et al.,2005; Cartry et al.,2006; Wingert et al.,2007). We wish to point out that the experimentally determined time window of fin precursor determination overlaps with the time window during which RA signaling affects hypoblast gene expression and with the determination phase of other mesendodermal organs during gastrulation. Our results indicating that an RA signal triggers the fin precursor determination program during gastrulation are consistent with a recent report showing RA-dependent pectoral fin precursor determination before the 8s stage (Waxman et al.,2008). We propose that RA is involved in the specification of fin precursors at a distinct location in the prospective lateral plate in the zebrafish gastrula (Fig. 7A). Loss of the gastrula stage RA signal hence leads to the failure to acquire the fin specific fate and thus, as a consequence, to the failure to initiate the fin precursor determination program in these cells later on, as visualized by the failure to express tbx5 lateral to somites 1–3 at the 10s stage (Fig. 7B). RA acts beyond the determination phase on fin precursor maintenance which is seen in embryos that later loose the early tbx5-positive fin precursors upon continued RA-inhibitor treatment (Fig. 7C). A delayed endogenous RA signal is able to trigger fin precursor fate in a delayed manner, but an externally protruding fin bud is not observed at 32 hpf. This underscores the importance of the gastrula stage RA signal to ascertain initiation of the normal fin bud developmental program (Fig. 7D). The available fate map data suggest that the origin of pectoral fin precursors lies within the gene expression domains that we have show to be affected by RA signaling at late gastrula stages (Warga and Nusslein-Volhard,1999; Keegan et al.,2004). The establishment of late stage gastrula fate maps is clearly desirable, not only with respect to the pectoral fins, but also with respect to lateral plate derivatives in general to substantiate the idea of distinct precursor populations at this developmental stage.

Figure 7.

Model. A: Under the influence of retinoic acid (RA) (graded shading), fin precursors (FP, encircled) are specified in the late gastrula hypoblast which become determined to follow a fin fate and express tbx5 (red) lateral to somites 1–3 during somitogenesis. They finally undergo fin bud morphogenesis during pharyngula stages. B: In the absence of RA at all developmental stages, fin precursor determination and bud morphogenesis are not observed. C: Embryos receiving an RA signal during gastrulation initiate the process that leads to fin precursor determination but fail to maintain them. D: Embryos receiving a delayed endogenous RA signal from somites and intermediate mesoderm (trunk) resurrect a weakly tbx5-positive territory in the lateral plate in a delayed manner that fails to undergo regular fin bud morphogenesis.

We finally wish to point out that better knowledge of the developmental context during which RA signaling first affects fin precursor determination also has bearings on our appreciation of the general structure of the vertebrate ground plan. From the perspective of our early-trigger-to-determination view, vertebrate limbs cannot be considered as mere “appendages” that reflect the organization of the body axis in any way, but as independent organs with an independent structure of their own.


Fish Maintenance, Mutants

Zebrafish were kept under standard laboratory conditions at approximately 27°C (Westerfield,2000; Brand et al.,2002). Embryos were obtained by natural matings and raised during the first 5 days at approximately 28.5°C. The necklessu11 (nlsu11) allele of the aldh1a2 gene has previously been described as the no-fin mutant (Grandel et al.,2002).

Whole-Mount In Situ Hybridization

In situ hybridizations were done as previously described (Reifers et al.,1998). Wild-type expression patterns have been described: tbx5 (Begemann and Ingham,2000; Ahn et al.,2002), myoD (Weinberg et al.,1996). Thick sections were prepared with sharpened tungsten needles. Embryos were mounted in glycerol.

Pharmacological Treatments

DEAB (diethylaminobenzaldehyde, Fluka) is a competitive reversible inhibitor of aldehyde dehydrogenases (Russo,1997; Begemann et al.,2004). A stock solution of 10−2M DEAB in dimethyl sulfoxide (DMSO) was prepared and stored at +4°C. Embryos were treated in the dark in a final dilution of 10−5M DEAB in E3. Controls were placed into 10−5M DMSO in E3. Experimental and control embryos were manually dechorionated before treatments and then placed into the appropriate solution in agarose-coated Petri dishes (V = 20 ml). The inhibitor concentration was calculated with reference to the total liquid volume of the dishes (E3 + agarose). The dishes were filled several hours before use to permit equilibration.

Measurements of Expression Levels

Expression levels of aldh1a2 in wt and 10−5M DEAB-treated embryos were measured. Images of lateral views of embryos stained for aldh1a2 expression were taken at ×32 magnification using an Olympus MVX10 microscope. These were analyzed with the ImageJ-133 software package. Signal intensity was converted into a grey value at each point along the ventrodorsal axis of the embryo resulting in lowest values at the ventral side and highest values on the dorsal side.


We thank Steve Wilson, Victoria Prince, and Eric Weinberg for reagents; Thomas Kurth (CRTD Electron Microscopy facility), Marika Fischer, and Katrin Sippel (Fish facility), Isabell Wenzel, Michaela Geffarth, and Anja Machate for excellent technical assistance; Alexander Picker, Matthias Nowak, Christian Bökel, and Gerrit Begemann for discussion and comments on an earlier version of the manuscript. M.B. was funded by the Deutsche Forschungsgemeinschaft and the European Union.