Retinoic acid signaling acts via Hox1 to establish the posterior limit of the pharynx in the chordate amphioxus

In developing vertebrates and their sister group, the cephalochordates(amphioxus), the vitamin-A derivative retinoic acid (RA) specifies regional identities along the anterior/posterior axis(Dupé et al., 1999; Dupé and Lumsden, 2001; Escriva et al., 2002; Holland and Holland, 1996; Matt et al., 2003; Schubert et al., 2004). Although RA patterns all three tissue layers, most studies have focused on its role in patterning the neuroectoderm and mesoderm(Allan et al., 2001; Blumberg et al., 1997; Dupé and Lumsden, 2001; Gavalas and Krumlauf, 2000; Grandel et al., 2002). The role of RA in endodermal morphogenesis was long ignored because it was believed that the effect of exogenous RA on the vertebrate pharynx was due solely to mispatterning of neural crest migrating into the branchial arches(Alexandre et al., 1996). However, it is now known that the influence of neural crest-derived mesenchyme is subordinate to patterning within the pharyngeal endoderm itself(Couly et al., 2002; Graham, 2003; Le Douarin, 1982; Mark et al., 2004). Ablation of neural crest in the chick does not inhibit formation of the pharyngeal arches and pouches (Veitch et al.,1999). Conversely, in the zebrafish, function of tbx1 in the pharyngeal endoderm is required for normal development of neural crest-derived pharyngeal structures(Piotrowski et al., 2003; Piotrowski and Nüsslein-Volhard,2000). Moreover, studies in amphioxus, which lacks definitive neural crest, leave no doubt that pharyngeal patterning is mediated primarily by the endoderm (Escriva et al.,2002; Holland and Holland,1996).

Throughout the chordates, RA signaling specifies anterior/posterior position of pharyngeal structures such as the gill slits (aquatic chordates)or pharyngeal arches and pouches (non-aquatic chordates). Excess RA prevents formation of the gill slits (branchial basket) in tunicates(Hinman and Degnan, 2000). In amphioxus the pharynx is absent; the mouth (thought to be a modified gill slit) and gill slits do not form and the pharyngeal endoderm remains thin(Escriva et al., 2002; Holland and Holland, 1996). Conversely, in embryos treated with a RA antagonist, the pharynx with its thickened endoderm is expanded posteriorly(Escriva et al., 2002). Similarly, in vertebrates, excess RA prevents pharyngeal development in lampreys and causes fusion of the first two branchial arches in gnathostomes(Kuratani et al., 1998; Lee et al., 1995; Mulder et al., 1998), while decreased RA signaling has the opposite effect, expanding pharyngeal structures posteriorly. Consequently, in vertebrates with reduced RA signaling, the first pharyngeal pouch and the first two pharyngeal arches are normal, the second pouch is expanded posteriorly and the remaining pouches do not form (Niederreither et al.,2003; Quinlan et al.,2002; Wendling et al.,2000). Similarly, in amphioxus, the mouth is enlarged and the gill slit primordia are either elongated or absent altogether; presumably a low level of RA signaling is essential for gill slit penetration(Escriva et al., 2002).

RA signaling in chordates is directly mediated by the RA receptors (RARs),that heterodimerize with the retinoid X receptors (RXRs)(Laudet and Gronemeyer, 2001)In general, vertebrates have three RARs and three RXRs, whereas amphioxus and tunicates have one each (Bertrand et al.,2004; Nagatomo et al.,2003). In chordates other than tunicates(Ishibashi et al., 2003), RARs are autoregulated, and their expression generally reflects the level of RA signaling. In vertebrates, RAR gene expression is generally high in the foregut endoderm (Matt et al.,2003; Smith,1994), while in amphioxus, AmphiRAR is most intensely expressed in the middle third of the endoderm, just posterior to the mouth and the first three gill slits (Escriva et al., 2002).

The effects of altered RA signaling on endodermal expression of RARs are similar in amphioxus and vertebrates. For example, in the mouse, treatment with an RA agonist induces ectopic expression of RARβ in the first two pharyngeal pouches (Matt et al.,2003). Correspondingly, in amphioxus, excess RA expands expression of AmphiRAR into the anteriormost pharyngeal endoderm, while treatment with an RA antagonist downregulates RAR(Escriva et al., 2002; Wendling et al., 2000). However, little is known of the molecular mechanisms downstream of RAR/RXR that underlie pharyngeal patterning. In addition to RARs, only a few genes are known that exhibit altered expression in the pharyngeal endoderm in response to altered levels of RA signaling. In vertebrates, these include Hoxa1 and Hoxb1, expressed in the caudal pharynx, Pax1 and Pax9 expressed in pharyngeal pouches 1-3 and 1-4,respectively, and Fgf3 and Fgf8 expressed in the endoderm of the pharyngeal arches and caudal-lateral pharynx respectively(Neubüser et al., 1995; Wallin et al., 1996; Wendling et al., 2000). Expression of the single Pax1/9 gene in amphioxus is also affected by increased RA (Holland and Holland,1996), while in tunicates, expression of Otx in the pharynx is decreased by RA treatment(Hinman and Degnan, 2000).

Hox1 genes in both vertebrates and amphioxus are direct targets of RA signaling (Arcioni et al.,1992; Balmer and Blomhoff,2002; Manzanares et al.,2000; Ogura and Evans,1995). Expression of Hoxa1 and Hoxb1 in the pharyngeal endoderm of vertebrates is expanded by treatment with RA or an RA agonist, while treatment with an RA antagonist or mutation of the RA response element (RARE) markedly decreases Hoxa1 expression and eliminates that of Hoxb1 (Alexandre et al.,1996; Li and Lufkin,2000). Ectopic expression of Hoxa1 results in a similar phenotype to that found with application of RA. However, since loss of Hoxa1 and Hoxb1 affects hindbrain patterning and migration of neural crest into the pharyngeal arches(Gavalas et al., 1998; McClintock et al., 2002; Pasqualetti et al., 2001; Rossel and Capecchi, 1999),some authors reasoned that the effects of altered expression of these genes on pharyngeal patterning are probably due primarily to abnormal neural crest(Gavalas et al., 1998; Rossel and Capecchi, 1999). Others, however, have emphasized a more direct role of Hoxa1 and Hoxb1 in mediating RA signaling in the pharyngeal endoderm(Matt et al., 2003; Wendling et al., 2000).

Amphioxus is particularly useful for deciphering the molecular mechanism whereby RA signaling in the endoderm regulates anterior/posterior patterning of the pharynx, because it lacks neural crest and has single genes for RAR, Hox1 and most other endodermal markers. Moreover, at the neurula stage, the pharynx of the small, transparent embryos consists of only two cell layers - an inner endoderm and an outer ectoderm. The pharynx is asymmetrical with the first three gill slits forming ventrally on the right in an anterior/posterior series and the mouth, thought to be a modified gill slit,on the left (Fig. 1). Metamorphosis, resulting in a bilaterally symmetrical adult, occurs at 9-11 gill slits.

Previously, we showed that a high level of RA signaling establishes the posterior limit of the amphioxus pharynx, a relatively low level being required for gill slit formation (Escriva et al., 2002). In the present work, we investigate the mechanism whereby a high level of RA signaling in the middle third of the endoderm sets the posterior limit of the amphioxus pharynx. We first characterize the effects of both increased and decreased RA signaling on the spatio-temporal expression of 11 genes in the pharyngeal endoderm. Since these results reveal a new domain of AmphiHox1 in the endoderm that is congruent with that of AmphiRAR, we then tested whether AmphiHox1 mediates the effects of RA in establishing the posterior limit of the pharynx by knocking down its function with an antisense morpholino-oligonucleotide. Our results suggest a hierarchy of pharyngeal markers with anterior/posterior limits regulated by RA signaling and show that AmphiHox1 mediates the role of RA in establishing the posterior extent of the pharynx, but not the role of RA in gill slit penetration. We thus propose that a RAR-Hox1 gene hierarchy regulates endoderm fate by promoting posterior foregut/midgut formation.

Adult amphioxus (Branchiostoma floridae) were collected in Old Tampa Bay, Florida, during summer and induced to spawn by electric stimulation(Holland and Yu, 2004). After fertilization, the embryos were raised in filtered seawater at 25°C. At the late blastula stage, RA (dissolved in DMSO), the RA antagonist BMS009(dissolved in DMSO) or DMSO alone was added to the embryonic cultures to a final concentration of 1 × 10^-6 M(Escriva et al., 2002; Holland and Holland, 1996). After hatching at the early neurula stage, embryos were transferred to untreated seawater. All the DMSO-treated control embryos developed normally. Embryos were fixed for in situ hybridization at various developmental stages(Holland et al., 1996).

In situ hybridizations were performed as previously described(Holland et al., 1996). Clones used as templates for riboprobes were as follows: AmphiPax1/9(U20167) (Holland et al.,1995); AmphiNotch (Y12539)(Holland et al., 2001); AmphiWnt3 (AF361013) (Schubert et al., 2001); AmphiNodal (AY083838)(Yu et al., 2002); AmphiHox1 (AB028206) and AmphiOtx (AF043740) (both provided by J. Garcia-Fernàndez and P. W. H. Holland); AmphiIslet(AF226616) (provided by W. R. Jackman); AmphiFoxA2(HNF3β) (Y09236), AmphiPitx (AJ438768) and amphioxus hedgehog (AmphiHh) (Y13858) (all three provided by Sebastian M. Shimeld). After in situ hybridization, the embryos were first photographed as whole mounts and subsequently counterstained in Ponceau S, embedded in Spurr's resin and prepared as sections for light microscopy(Holland et al., 1996).

Microinjection of amphioxus eggs was as described by(Holland and Yu, 2004). Unfertilized eggs were injected with either the control morpholino(5′-CCTCTTACCTCAGTTACAATTTATA-3′) or one specific for AmphiHox1 from B. floridae(5′-ATTCTTGCCGTGTCCATTTGCTCCA-3′) (Gene Tools, Philomath, OR,USA). Approximately 2 pl of a solution containing 15% glycerol, 2 mg/ml Texas Red dextran (Molecular Probes, Eugene, OR, USA) and 500 μM morpholino was injected. The morpholinos were heated to 65°C for 5 min prior to use. After injection, the eggs were fertilized and fixed at either the late neurula stage (24-30 hours) or the early larval stage (36-40 hours). Fixed embryos showing clear fluorescence of the Texas Red dextran were analyzed by in situ hybridization (Holland et al.,1996).

For in vitro translation, the AmphiHox1 coding region cDNA was amplified by PCR and cloned into the pCS2+ vector(Rupp et al., 1994; Turner and Weintraub, 1994). In vitro translation was with the TnT Quick Coupled Transcription/Translation System. 200 ng of plasmid DNA containing the AmphiHox1 coding region was assayed together with different amounts of control or AmphiHox1-specific morpholino (100 ng, 500 ng, 1000 ng or 5000 ng). After the reactions, the samples were subjected to electrophoresis on a 12%polyacrylamide gel and transferred to a nitrocellulose membrane. AmphiHox1 protein was detected by the Transcend Non-Radioactive Translation Detection Systems (Promega, Madison, WI, USA).

To investigate how RA establishes the posterior limit of the amphioxus pharynx, we first determined the expression of 11 endodermal markers in amphioxus embryos treated with RA or the RA antagonist BMS009 at final concentrations of 1 × 10^-6 M. The effects of altered RA signaling on the spatio-temporal expression of these genes revealed which genes are downstream of RAR/RXR and their approximate hierarchy. These markers included AmphiRAR (Escriva et al., 2002), AmphiHox1(Holland and Holland, 1996), AmphiWnt3 (Schubert et al.,2001), AmphiPax1/9(Holland et al., 1995), AmphiPitx (Boorman and Shimeld,2002b), AmphiNotch(Holland et al., 2001), AmphiNodal (Yu et al.,2002), AmphiOtx(Williams and Holland, 1996), AmphiIslet (Jackman et al.,2000), AmphiFoxA2 (HNF3β)(Shimeld, 1997) and AmphiHh (Shimeld,1999). In normal embryos, three of these (AmphiRAR,AmphiHox1 and AmphiWnt3) are expressed in the middle third of the endoderm at the early neurula stage, three (AmphiPax1/9,AmphiPitx and AmphiNotch) have expression limited to the pharyngeal endoderm from an early stage, and five (AmphiNodal, AmphiOtx,AmphiIslet, AmphiFoxA2 and AmphiHh) are initially expressed throughout the length of the endoderm (Figs 2, 3, 4, 5, 6, 7). The limits of expression are measured from the anterior end of the embryo and given as a percentage of the total body length.

Although expression of AmphiHox1 in ventrolateral regions of amphioxus embryos was thought to be entirely ectodermal(Wada et al., 1999), frontal sections show that there is also a corresponding endodermal domain(Fig. 2C,G,K). Endodermal expression of AmphiHox1 generally parallels that of AmphiRAR(see Fig. S1 in the supplementary material)(Escriva et al., 2002),although, especially at early stages, the anterior limit of expression of AmphiHox1 (Fig. 2A) is somewhat posterior to that of AmphiRAR(Escriva et al., 2002). By 20 hours, expression domains of AmphiHox1 and AmphiRAR in the endoderm are approximately congruent and remain similar during later development (Fig. 2B,D)(Escriva et al., 2002). For example, at 20-24 hours of development, the anterior limits of AmphiRAR and AmphiHox1 in the endoderm are 38% and 40%respectively. Treatment with RA shifts expression of AmphiHox1anteriorly in the endoderm (Fig. 2E-H). Correspondingly, treatment with BMS009 downregulates expression of AmphiHox1 and shifts expression posteriorly by an additional 10% of the length of the embryo(Fig. 2I-L). Overlapping expression of AmphiHox1 and AmphiRAR plus the presence of a RARE in the regulatory region of AmphiHox1(Manzanares et al., 2000), to which AmphiRAR/AmphiRXR binds in vitro (H. Escriva, H. Wada and V. Laudet,unpublished), suggest that AmphiHox1 is a direct target of RA signaling in the endoderm as well as in nerve cord and mesoderm.

At the mid-neurula stage, AmphiWnt3 transcription begins in the ventral endoderm just posterior to the pharynx(Fig. 2M). The onset of expression is later than that of either AmphiRAR or AmphiHox1 and appears to be earlier in RA-treated embryos and a little later in BMS009 embryos than in controls(Fig. 2M,O,Q). The anterior limit of AmphiWnt3 expression (38% at 20 hours) is similar to that of AmphiHox1 (40% at 20 hours) and is also shifted anteriorly by RA (to 20% at 20 hours) and slightly posteriorly by BMS009 (to 43% at 20 hours). In addition, BMS009 broadens the domain posteriorly(Fig. 2O-R). However, by 30 hours of development, AmphiWnt3 is almost completely downregulated in the endoderm of RA-treated larvae (Fig. 2P), although whether it is a direct target of RA signaling remains to be determined.

AmphiPax1/9 and AmphiPitx are expressed in the pharyngeal endoderm with posterior limits at the mid-neurula stage (51% and 54%respectively) overlapping the anterior limits of AmphiHox1 and AmphiRAR (compare Fig. 2A-D with Fig. 3A,B,G,H; Fig. 7)(Holland et al., 1995; Yasui et al., 2000). Expression of both genes is reduced where the first gill slit will form(Fig. 3A,B,G,H). Expression of AmphiPax1/9 does not change throughout the neurula and early larval stages (Fig. 3B). After 36 hours, expression of AmphiPitx becomes limited to the mouth and to Hatschek's anterior left diverticulum, the precursor of Hatschek's pit, the homolog of the adenohypophysis (Fig. 3I, double arrowhead). At the midneurula stage, AmphiNotch is broadly expressed in the pharyngeal endoderm with a posterior limit (35%) somewhat rostral to those of AmphiPax1/9 and AmphiPitx (Fig. 3P). Expression of AmphiNotch is reduced where the first gill slit will form (Fig. 3P,Q)(Holland et al., 2001). By the early larval stage (30 hours), the posterior limits of AmphiNotch,AmphiPax1/9 and AmphiPitx are approximately the same (44-49%;compare Fig. 3Q with Fig. 3B,H).

Treatment with RA at the gastrula stage shifts the posterior limit of all three genes anteriorly in the endoderm. This shift is not marked at 18-20 hours; the posterior limits shift only from 54% in controls to 50% with RA for AmphiPax1/9, 51% to 40% for AmphiPitx and 35% to 30% for AmphiNotch. However, by 30 hours, the difference is obvious with the posterior limits of AmphiPax1/9 and AmphiNotch shifting to 17% in RA-treated larvae and that of AmphiPitx to 0%. RA-treatment also eliminates the zones of reduced expression where the gill slits would normally have formed (Fig. 3C,D,J,R,S). BMS009 has the opposite effect - the pharyngeal expression domains are expanded posteriorly (to about 71% for AmphiPax1/9, 60% for AmphiPitx and 53% for AmphiNotch at 20 hours; Fig. 3E,F,M-O,T,U). Expression of AmphiPitx in Hatschek's anterior left diverticulum is not affected by changes in RA signaling(Fig. 3I,L,O). Although the level of expression of AmphiPax1/9 is not obviously affected by levels of RA signaling (Fig. 3C-F), AmphiPitx and AmphiNotch appear to be downregulated by RA and upregulated by BMS009(Fig. 3J-O,R-U).

The relatively early response of these three genes to altered levels of RA signaling that together with AmphiWnt3, they are comparatively high up in the hierarchy of the RA signaling pathway. Their expression patterns and response to altered levels of RA signaling indicate that that high levels of RA signaling suppress AmphiPax1/9, AmphiPitx and AmphiNotchexpression in the middle third of the endoderm. Moreover, the posterior limits of AmphiPax1/9 and AmphiPitx are just posterior to the anterior limits of AmphiRAR and AmphiHox1, suggesting that AmphiRAR acting via AmphiHox1 (see below) may set the posterior limit of expression of these genes as well as the anterior/posterior extent of the endodermal domain of AmphiWnt3.

During amphioxus development, AmphiNodal and AmphiOtx are normally expressed throughout all or most of the length of the endoderm at the early to mid-neurula stage (Fig. 4A-C,M). For both genes, expression is reduced ventrally in the primordia of the first two gill slits (Fig. 4A,C,M). By the late neurula (24 hours), expression of AmphiNodal becomes largely restricted to the anterior endoderm(posterior limit at 42%), although there is still weak expression in the midgut and hindgut (Fig. 4C). Expression of AmphiOtx similarly becomes restricted to the pharyngeal endoderm, but later than that of AmphiNodal. By the early larval stage (30 hours), AmphiOtx remains expressed in the pharyngeal endoderm (Fig. 4N), but AmphiNodal is largely downregulated(Fig. 4D)(Williams and Holland, 1996; Williams and Holland, 1998; Yu et al., 2002). RA treatment at the gastrula stage has little effect on endodermal expression of either gene at the mid-neurula stage (Fig. 4E,O). However, by 24 hours, RA treatment restricts the endodermal expression of AmphiNodal to the anterior pharynx (posterior limit at 37%; Fig. 4G). Expression is downregulated somewhat sooner in RA-treated embryos than in controls(Fig. 4D,H). The effect of RA on the posterior limit of AmphiOtx is not apparent until the early larva. At 30 hours, the pharyngeal expression domain of AmphiOtx is reduced and the posterior limit is shifted anteriorly compared to controls(posterior limit at 26%; Fig. 4N,P). BMS009 has the opposite effect. The posterior limits of the strong pharyngeal expression domains of both AmphiNodal and AmphiOtx are shifted posteriorly at the late neurula and early larval stages respectively (Fig. 4I,K,L,Q,R), and downregulation of AmphiNodal in the endoderm is delayed (Fig. 4L). Since, unlike AmphiPax1/9, the anterior/posterior extent of expression of both genes in the endoderm is not regionalized at the mid-neurula stage and is affected rather late by RA and BMS009 treatments, it is likely that they are farther downstream than AmphiPax1/9 in the hierarchy of RA signaling.

AmphiNodal and AmphiPitx together with AmphiHhare the only known amphioxus genes with pharyngeal expression limited to the left side of the endoderm (Fig. 4B) (Shimeld,1999; Yasui et al.,2000; Yu et al.,2002). However, neither RA nor BMS009 induces expression of AmphiNodal (Fig. 4F,J), AmphiPitx or AmphiHh (data not shown) on the right side of the endoderm in amphioxus. We conclude that RA signaling does not control left/right asymmetry of the amphioxus pharynx.

The endodermal expression domain of amphioxus AmphiIslet at the mid-neurula stage is similar to that of AmphiOtx but does not extend as far posteriorly (74% versus 91%; Fig. 5A) (Jackman et al.,2000). However, unlike AmphiOtx, AmphiIslet remains expressed along much of the length of the endoderm through the early larval stage (Fig. 5B,C). As with AmphiOtx, RA treatment inhibits downregulation of AmphiIsletwhere the first gill slit would normally have formed(Fig. 5D-F). However, neither RA- (Fig. 5D,F) nor BMS009 treatment (Fig. 5G-I) has a marked effect on the posterior limit of expression.

In control embryos, AmphiFoxA2 is expressed throughout the pharynx at the mid- to late-neurula stages (Fig. 5J) (Shimeld,1997). Unlike AmphiOtx and AmphiIslet,AmphiFoxA2 remains expressed where the first gill slit will form,although expression is reduced anteriorly in the pharyngeal endoderm(Fig. 5J-L). The only evident effect of altered RA signaling on endodermal expression of AmphiFoxA2is that in RA-treated embryos and larvae, it is not downregulated where the gill slits and mouth would have formed(Fig. 5M-O), while in larvae treated with BMS009, AmphiFoxA2 is largely downregulated in an expanded region of the anterior endoderm(Fig. 5P-R).

AmphiHh is weakly expressed throughout the length of the endoderm on the left side (Fig. 5S,T)(Shimeld, 1999). Expression is particularly high anterior to the mouth(Fig. 5S). As development proceeds, AmphiHh becomes upregulated around the first gill slit(Fig. 5T). Altered RA signaling does not substantially affect either the left/right asymmetry or the anterior/posterior extent of endodermal expression. However, treatment with RA reduces the size of the strong expression domain in the anteriormost pharyngeal endoderm at the mid-neurula and almost completely downregulates endodermal expression by the early larval stage(Fig. 5U,V). In contrast, while BMS009 does not alter the expression domain of AmphiHh substantially,it does appear to upregulate the gene somewhat(Fig. 5W,X). Failure of altered levels of RA signaling to change the posterior limit of expression of AmphiIslet, AmphiFoxA2 and AmphiHh suggests that they are involved in specification of posterior foregut/midgut structures as well as pharyngeal structures. This is not surprising in light of the roles of their homologs in specification of the foregut and its derivatives in vertebrates(Chen et al., 2004; Gauthier et al., 2002; Yuan and Schoenwolf,2000).

To test whether AmphiHox1 mediates RA signaling in setting the posterior limit of the pharynx, we knocked-down AmphiHox1 function by injection of an antisense morpholino oligonucleotide. In vitro translation showed that the AmphiHox1-specific morpholino effectively blocks translation of AmphiHox1 mRNA (see Fig. S2 in the supplementary material). Injected embryos were fixed at late neurula and early larval stages and hybridized with riboprobes for three genes: AmphiHox1,AmphiPax1/9 and AmphiOtx. Although the AmphiHox1-specific morpholino does not affect expression of AmphiHox1 in the nerve cord (where expression of AmphiOtx is expanded posteriorly), expression of AmphiHox1 in the endoderm is shifted somewhat posteriorly as in animals treated with the RA antagonist BMS009 (anterior limit shifted from 30% to 36%; Fig. 6A,B). In addition,pharyngeal expression of both AmphiPax1/9(Fig. 6C,D) and AmphiOtx (Fig. 6E-H)is expanded posteriorly in embryos and larvae injected with the AmphiHox1-specific morpholino (posterior limits changed from 37% to 67% and 43% to 55%, respectively), showing that both genes are downstream of AmphiHox1 in the RA signaling hierarchy. However, gill slits form normally in embryos injected with the AmphiHox1-specific morpholino(Fig. 6D,H). These results indicate that like RA antagonist treatment, the injection of an AmphiHox1-specific morpholino expands the pharyngeal region posteriorly. We conclude that AmphiHox1 probably mediates RA signaling in the amphioxus endoderm to establish the posterior limit of the pharynx, but probably does not mediate the role of RA in gill slit penetration.

The present research has taken advantage of the uncomplicated body plan and relatively unduplicated genome of amphioxus to begin elucidating the developmental mechanism whereby RA signaling in the posterior foregut/midgut endoderm establishes the posterior limit of the pharynx. AmphiHox1(Manzanares et al., 2000) and AmphiRAR, like their vertebrate homologs, are probably both direct targets of RA signaling in the endoderm. Although the regulatory elements of AmphiRAR have not been studied, we have previously shown that AmphiRAR is strongly upregulated by RA and downregulated by the RA antagonist BMS009 (Escriva et al.,2002) (and Fig. S1 in the supplementary material). Moreover, the effects of altered RA signaling on expression of AmphiRAR are much like those on expression of vertebrate RARs, which are known to be autoregulated (Sucov et al.,1990). Both AmphiRAR and vertebrate RARβare expressed in the posterior foregut endoderm(Escriva et al., 2002; Mollard et al., 2000) and are ectopically expressed in the pharyngeal endoderm in embryos treated with RA and/or RA agonists (Wendling et al.,2000). Conversely, RA antagonists downregulate expression of RARs in the endoderm of both vertebrates and amphioxus(Escriva et al., 2002). These parallels suggest that AmphiRAR, like vertebrate RARs, is probably autoregulated in the pharyngeal endoderm.

Similarly, the AmphiHox1 gene contains a RA response element(RARE) 3′ of the coding region(Manzanares et al., 2000) to which the AmphiRAR/AmphiRXR heterodimer can bind in vitro (H. Escriva, H. Wada and V. Laudet, unpublished). In addition, the effects of altered RA signaling on expression of AmphiHox1 are similar to those on vertebrate Hoxa1 and Hoxb1, which are also expressed in the posterior part of the pharyngeal endoderm (Matt et al., 2003; Wendling et al.,2000). These genes also contain RAREs, required to direct expression of Hoxa1 and Hoxb1 reporter constructs to the foregut (Huang et al., 1998). Not surprisingly, as with amphioxus, treatment of mouse embryos with a pan-RAR antagonist eliminates or greatly reduces pharyngeal expression of Hox1 genes(Matt et al., 2003). Conversely, treatment with RA or RAR agonists results in an anterior shift of Hox1 gene expression in the pharynx(Escriva et al., 2002; Wendling et al., 2000). Taken together these data clearly suggest that the general shape of the RAR-Hox1 hierarchy is conserved in vertebrates, but has become more complex due to gene duplications early in vertebrate evolution that resulted in three RAR and three Hox1 paralogs.

Our results show that blocking function of AmphiHox1 expands the amphioxus pharynx to the same extent as inhibiting RA signaling and demonstrate an approximate hierarchy of downstream genes(Fig. 7). In our model(Fig. 8), RA signaling activates AmphiHox1, which is co-expressed with the RA receptor AmphiRAR in the middle third of the endoderm. AmphiHox1 in turn represses AmphiPax1/9 and AmphiOtx expression posterior to the pharynx.

The order of genes in the hierarchy downstream of AmphiRAR and AmphiHox1 is suggested by the time in development at which expression of endodermal markers becomes restricted to the pharynx(Fig. 7). Expression of AmphiPax1/9, together with AmphiPitx and AmphiNotchis restricted to the pharynx early in development suggesting that they are high in the hierarchy of downstream genes. In fact, AmphiPax1/9,which first turns on at the early neurula stage, is never expressed posterior to the pharyngeal region (Holland et al.,1995), suggesting that AmphiPax1/9 is likely to be very high up in the gene network downstream of AmphiRAR and AmphiHox1. In contrast, expression of AmphiOtx and AmphiNodal becomes restricted to the pharynx relatively late(Fig. 7). For example,expression of AmphiOtx does not become restricted to the pharyngeal endoderm until the late neurula (Williams and Holland, 1996; Williams and Holland, 1998), and it is only at this time that altered levels of RA signaling affect the anterior/posterior limit of expression. Thus, AmphiOtx is likely to be much farther downstream than AmphiPax1/9 in the RA-/Hox1-signaling pathway. Together, our results suggest the scenario in Fig. 8in which AmphiHox1 is a direct target of RA signaling(Manzanares et al., 2000) and in turn sets the posterior limit of the pharynx by repressing expression of pharyngeal markers in the endoderm of the posterior foregut/midgut. Although not yet unequivocally demonstrated, it is likely that AmphiRAR and AmphiHox1 turn on specific markers of the posterior foregut/midgut endoderm, such as AmphiWnt3 (Fig. 2).

Comparison with vertebrates suggests that the model in Fig. 8 may also apply to anterior/posterior patterning of the pharyngeal endoderm in vertebrates. In vertebrates, as in amphioxus, Hox1 genes (Hoxa1 and Hoxb1)are expressed in endoderm of the foregut and extreme caudal end of the pharynx(Frasch et al., 1995; Huang et al., 1998; Wendling et al., 2000). Pharyngeal expression of both genes is severely decreased by treatment with an RA antagonist (Wendling et al.,2000). Moreover, as in amphioxus embryos treated with RA,treatment of mouse embryos with an RA agonist induces strong ectopic expression of both Hoxa1 and Hoxb1 in the anterior pharyngeal endoderm (Matt et al.,2003). This suggests that in vertebrates as well as amphioxus,Hox1 genes mediate the effect of RA in establishing the posterior limit of the pharyngeal endoderm.

The targets of Hox1 genes in the pharyngeal endoderm of vertebrates have not been described. However, they are probably very much the same as in amphioxus since the pharyngeal endoderm of both amphioxus and vertebrates expresses similar suites of genes. Vertebrates have two homologs of AmphiPax1/9, Pax1 and Pax9. Both are expressed throughout the length of the pharyngeal endoderm that will give rise to the pharyngeal pouches and later in the endoderm of the definitive pharyngeal pouches themselves (Müller et al.,1996; Ogasawara et al.,2000). As in amphioxus embryos treated with BMS009, reduced RA signaling extends expression of Pax1 posteriorly(Dupé et al., 1999; Quinlan et al., 2002; Wendling et al., 2000). Conversely, RA treatment reduces Pax1 expression in the endoderm of the third pharyngeal pouch in the mouse in connection with a greatly reduced third pharyngeal arch or fusion of the third and fourth arches(Mulder et al., 1998). Altered levels of RA signaling have similar effects on Pax9. In the mouse,the domain of Pax9 expression in the second pouch is expanded in embryos treated with a pan-RA antagonist, and expression where the third pouch would normally form is nearly eliminated(Wendling et al., 2000). Mouse knockouts of both RARα and RARβ have a somewhat less severe phenotype, but even so, Pax1 expression in the third pouch is reduced (Dupé et al.,1999). Together, these results suggest that in vertebrates, as in amphioxus, high levels of RA signaling may activate RAR and Hox1 expression in the endoderm and that Hox1 expression in turn represses, directly or indirectly, transcription of Pax1/9 genes in the endoderm posterior to the pharynx.

Similarly, expression of Otx genes in the pharyngeal endoderm is common to tunicates and vertebrates as well as amphioxus. The effects of loss of Hox1 gene function on Otx expression in chordates other than amphioxus has not been studied. However, RA treatment has a similar effect on Otxexpression in these organisms as in amphioxus. In ascidian tunicates,reduction of the pharynx in RA-treated embryos correlates with reduced expression of Otx in the pharynx(Hinman and Degnan, 2000). In addition, in vertebrates, expression of Otx genes in the anterior mesendoderm and later in the pharyngeal endoderm of the first pharyngeal pouch(Blitz and Cho, 1995; Tomsa and Langeland, 1999) is lost in embryos treated with RA (Bally-Cuif et al., 1995; Simeone et al.,1995).

Homologs of the remaining pharyngeal markers we have identified with their posterior limits set by a high level of RA signaling are also expressed in the pharyngeal endoderm of vertebrates and other chordates, although the effects of RA signaling on expression of these genes in vertebrates is not known. For example, expression of AmphiPitx in the endoderm around the mouth and Hatschek's anterior left diverticulum, the homolog of the adenohypophysis(Yasui et al., 2000), is comparable to that of Pitx2 in the pituitary and Pitx1 in the stomodaeum and rostral foregut endoderm in the mouse and chick(Lanctot et al., 1997). Similarly, in the lamprey, Pitx genes are expressed in the stomodaeum,neurohypophyseal duct and pharyngeal endoderm among other locations(Boorman and Shimeld, 2002a). In larval tunicates, Pitx is also expressed in the nascent pharynx(Boorman and Shimeld,2002b).

Vertebrate Notch genes are expressed in the pharynx as in amphioxus,although their roles in pharyngeal development are not well understood. For example, in the mouse, Notch2 is expressed in the anterior part of the first branchial arch (Williams et al.,1995), but it is not known if this is in the endodermal portion or not. Notch1 is also expressed in the epibranchial placodes associated with branchial arches 1-3, near the fourth arch, in neural crest migrating into first and second arches (Williams et al., 1995) and in the thymus(Weinmaster et al., 1991; Weinmaster et al., 1992). Notch genes are also expressed in the developing pancreas and the lung, which are both endodermal derivatives (Kim and Hebrok, 2001; Lammert et al.,2000; Post et al.,2000). In the zebrafish blastula, Notch signaling appears to regulate the number of endodermal cells; overexpression reduces the number of cells expressing the endodermal marker foxa2(Kikuchi et al., 2004). However, whether Notch genes have a later role in development of the pharyngeal endoderm is unknown.

Expression of Nodal in amphioxus and vertebrates is also similar. In amphioxus, AmphiNodal is expressed at the dorsal lip of the blastopore in the early gastrula and throughout the length of the endoderm at the mid-neurula stage, subsequently becoming restricted to the pharyngeal endoderm (Yu et al., 2002). In vertebrates, Nodal expression in mesendodermal precursors is required for endoderm formation, in particular for the foregut endoderm where it is upstream of Pitx2 (Faucourt et al., 2001; Tam et al.,2003). Taken together, these data suggest that the gene networks specifying anterior endoderm are similar in amphioxus and vertebrates, and that in both, a RAR/Hox1 signaling cascade determines fore/midgut identity and restricts expression of anterior endodermal genes to the pharynx. However, in vertebrates, extensive gene duplications have evidently conferred added complexity on these gene networks.

AmphiWnt3 is expressed ventrally in the endoderm just posterior to the pharynx with anterior/posterior limits coinciding approximately with those of AmphiRAR and AmphiHox1. Like AmphiRAR and AmphiHox1, AmphiWnt3 expression is shifted anteriorly by RA, and expanded posteriorly by BMS009. AmphiWnt3 is expressed relatively late in the gut and is therefore probably acting downstream of RA signaling. In vertebrates, Wnt3a is expressed in the vertebrate foregut endoderm of the chick (Theodosiou and Tabin,2003), and is downregulated by RA in an embryonic carcinoma cell line (Katoh, 2002) as well as in the tail bud (Shum et al.,1999). This suggests that a role of Wnt3a in regionalization of the gut in the chick may have its antecedents in an amphioxus-like ancestor. However, whether there is cross-talk between RA signaling and WNT signaling in patterning the vertebrate foregut remains to be demonstrated.

Specification of left/right position involves the evolutionarily conserved series of Shh, Nodal and Pitx2(Cooke, 2004). In the vertebrates, high concentrations of RA randomize heart looping and can induce bilateral expression of Nodal and Pitx2 on the right side(Chazaud et al., 1999; Smith et al., 1997; Wasiak and Lohnes, 1999). However, expression of Shh is not affected(Smith et al., 1997), and it appears to act either in parallel to or downstream of RA signaling(Tsuki et al., 1999). In amphioxus, AmphiHh, AmphiNodal and AmphiPitx are all expressed on the left side of the pharyngeal endoderm. Expression of these genes on the left side of the body is not affected by altering the levels of RA signaling. Thus RA signaling is not required for establishment of left/right asymmetry in amphioxus, or apparently, in tunicates(Hinman and Degnan, 1998), and its role in left/right asymmetry may be a vertebrate innovation.

Our results show that amphioxus is particularly advantageous for understanding developmental mechanisms and that it can serve as a simplified model for comparable patterning in vertebrate embryos. Because many amphioxus genes are present in single copies (including RAR and the Hox genes),functional knockdowns are relatively easy to interpret. Moreover, since amphioxus lacks definitive neural crest, the model we present for patterning of the pharynx applies unequivocally to the endoderm, thereby giving insights into the separate roles of the endoderm and neural crest in pharyngeal patterning of vertebrates. It is likely that similar regulatory cascades involving Hox1-mediated RA signaling help to direct pharyngeal patterning in both amphioxus and vertebrates. In addition, in vertebrates, the evolution of neural crest evidently led to the elaboration of novel pharyngeal structures,which were superimposed on the already existing pharyngeal patterning intrinsic to the endoderm.

A role for Hox genes in patterning the endoderm is widespread in the animal kingdom (Brunschwig et al.,1999; Irvine and Martindale,2000; Marty et al.,2001). Recent evidence suggests that the RAR genes may be more ancient than previously thought, having been secondarily lost in Drosophila and nematodes (Bertrand et al., 2004). Thus, endodermal patterning by RAR/Hox1 may not be limited to chordates, and the model we present here for regionalization of the amphioxus endoderm may provide a framework for understanding endodermal patterning in a wide spectrum of bilaterian animals.

We thank J. M. Lawrence (University of South Florida, Tampa, FL, USA) for providing laboratory space during the amphioxus breeding season and H. Gronemeyer (University of Strasbourg, France) for generously providing BMS009. This research was supported by NSF IBN 00-78599 to N.D.H. and L.Z.H.,NAG2-1585 from the National Aeronautics and Space Administration to L.Z.H and by grants from the MENRT, the CNRS, the ARC and the Région Rhône-Alpes to V.L. M. Schubert is supported by a European Community Marie Curie Fellowship.