Conserved RARE localization in amphioxus Hox clusters and implications for Hox code evolution in the vertebrate neural crest

Authors

  • Hiroshi Wada,

    Corresponding author
    1. Seto Marine Biological Laboratory, FSERC, Kyoto University, Wakayama, Japan
    2. Graduate School of Life and Environmental Sciences, University of Tsukuba, Tsukuba, Japan
    3. PRESTO, JST, Kawaguchi, Japan
    • Graduate School of Life and Environmental Sciences, University of Tsukuba, Tsukuba 305-8572 Japan
    Search for more papers by this author
  • Hector Escriva,

    1. UMR 5161 du CNRS, INRA LA 1237, Laboratoire de Biologie Moléculaire de la Cellule, IFR128 BioSciences Lyon-Gerland, Ecole Normale Supérieure de Lyon, Lyon, France
    Search for more papers by this author
  • Shicui Zhang,

    1. Department of Marine Biology, Ocean University of Qingdao, Qingdao, P. R. China
    Search for more papers by this author
  • Vincent Laudet

    1. UMR 5161 du CNRS, INRA LA 1237, Laboratoire de Biologie Moléculaire de la Cellule, IFR128 BioSciences Lyon-Gerland, Ecole Normale Supérieure de Lyon, Lyon, France
    Search for more papers by this author

Abstract

The Hox code in the neural crest cells plays an important role in the development of the complex craniofacial structures that are characteristic of vertebrates. Previously, 3′ AmphiHox1 flanking region has been shown to drive gene expression in neural tubes and neural crest cells in a retinoic acid (RA)-dependent manner. In the present study, we found that the DR5-type RA response elements located at the 3′ AmphiHox1 flanking region of Branchiostoma floridae are necessary and sufficient to express reporter genes in both the neural tube and neural crest cells of chick embryos, specifically at the post-otic level. The DR5 at the 3′ flanking region of chick Hoxb1 is also capable of driving the same expression in chick embryos. We found that AmphiHox3 possesses a DR5-type RARE in its 5′ flanking region, and this drives an expression pattern similar to the RARE element found in the 3′ flanking region of AmphiHox1. Therefore, the location of these DR5-type RAREs is conserved in amphioxus and vertebrate Hox clusters. Our findings demonstrate that conserved RAREs mediate RA-dependent regulation of Hox genes in amphioxus and vertebrates, and in vertebrates this drives expression of Hox genes in both neural crest and neural tube. This suggests that Hox expression in vertebrate neural crest cells has evolved via the co-option of a pre-existing regulatory pathway that primitively regulated neural tube (and possibly epidermal) Hox expression. Developmental Dynamics 235:1522–1531, 2006. © 2006 Wiley-Liss, Inc.

INTRODUCTION

A complex craniofacial structure is a vertebrate characteristic that accompanied the evolution of neural crest cells (Gans and Northcutt,1983; Shimeld and Holland,2000). These cells play a central role in the differentiation of the pharyngeal arches by producing connective tissues of a specific morphology that depends on the anterior-posterior origin of neural crest cells. It has been believed that neural crest cells perform an instructive role, by determining the identity of each pharyngeal arch (LeDouarin and Kalcheim,1999; Noden,1986). This idea is consistent with the presence of a Hox code in the neural crest cells and is further supported by the fact that there is abnormal development of pharyngeal arches in knock-out mice with Hox genes (Chisaka and Capecchi,1991; Chisaka et al.,1992; Lufkin et al.,1991; Rijli et al.,1993). However, recent studies have suggested that endodermal tissue in the branchial region also plays an important role in pharyngeal arch patterning. Veitch et al. (1999) showed that regionalization occurs within individual pharyngeal arches even in the absence of the neural crest cells. Moreover, Couly et al. (2002) found that grafting small pieces of endoderm altered skeletal development in a manner that depended on the axial level of the graft. However, this only occurred when they grafted endoderm onto the pharyngeal arch, where neural crest cells do not express Hox. Collectively, these observations suggest that pharyngeal arch patterning is governed by signals from both the neural crest cells and endoderm (Santagati and Rijli,2003).

Traditionally, it was assumed that ascidians and amphioxus lack a neural crest because no cells migrate from the neural tube during embryogenesis. However, a growing body of evidence suggests that the epidermis abutting the neural tube has a pattern of gene expression resembling the vertebrate neural crest (e.g., Dll and Pax3/7) (Holland et al.,1996; Wada et al.,1997). Therefore, it would appear that the neural crest originated from the dorsal midline epidermis, with some contributions from the dorsal neural tube (Meulemans and Bronner-Fraser,2004; Wada,2001). Wada (2001) postulated that the true vertebrate neural crest evolved via the acquisition of two novel cell properties: elaboration of the capacity to migrate and acquisition of anterior-posterior (AP) positional information. However, recent observations suggest there is a migratory cell population resembling neural crest cells in the urochordate Ecteinascidia turbinata (Jeffery et al.,2004). Therefore, migration may not be a feature novel to the vertebrate neural crest cells, although the homology between migrating cells of E. turbinata and the vertebrate neural crest cells is arguable. With regard to AP positional information, the evolution of the vertebrate Hox code was a crucial event in the evolution of vertebrate craniofacial structure. Although urochordates seem to have lost the typical colinearity between genomic organization and expression pattern (Ikuta et al.,2004), amphioxus retains a single Hox gene cluster organization (Garcia-Fernandez and Holland,1994). Colinear Hox gene expression was observed in the neural tube and epidermis of amphioxus (Schubert et al.,2004; Wada et al.,1999). Schubert et al. (2005) also indicated that Hox1 is involved in pharyngeal region anterior-posterior patterning. In the ascidian Ciona intestinalis, a Hox gene expression pattern was observed that respects the anterior-posterior axis in the neural tube, epidermis, and endoderm (Ikuta et al.,2004). Therefore, the Hox code operates in the ectoderm and endoderm of both ascidians and amphioxus.

It is reasonable to assume that the Hox code in the vertebrate neural crest cells originated in the ectoderm, either the neural tube or epidermis. Therefore, we can hypothesize three possible and not exclusive scenarios for the Hox code origin in the neural crest cells. First, the epidermal expression of Hox genes in amphioxus suggests that the regulatory machinery for the vertebrate neural crest cells originated from the epidermis of a chordate ancestor (Schubert et al.,2004) (Fig. 1A). This may be most consistent with the idea that the neural crest originated from dorsal epidermal cells of an ancestral chordate (Wada,2001; Meulemans and Bronner-Fraser,2004). In addition, AP2 is involved in mouse Hoxa2 neural crest expression (Maconochie et al.,1999). AP2 homologues are expressed in epidermal cells of amphioxus (Meulemans and Bronner-Fraser,2002), although it is not clear whether AP2 is required for the expression of their Hox genes in the epidermis. Second, it is also possible that neural crest expression evolved when ancestral vertebrates acquired novel regulatory systems (Fig. 1B). The observation that the anterior expression limits of certain vertebrate Hox genes in the neural crest cells differ from the limits within the neural tube supports both of these scenarios. For example, Hoxa2 is expressed in the neural tube up to rhombomere (r) 2, but neural crest cells migrating from r2 do not express Hoxa2 (Prince and Lumsden,1994; Tümpel et al.,2002). Finally, the third possible scenario involves the fact that the dorsal neural tube and neural crest share a common precursor (Fig. 1C) (Bronner-Fraser and Fraser,1988,1989; Collazo et al.,1993; Serbedzija et al.,1994). Hox expression might have evolved solely because the neural crest inherited the Hox expression encoded in the neural tube or neural plate. This is consistent with the fact that the Hoxb1 autoregulatory element is sufficient for neural crest expression (Pöpperl et al.,1995). In this case, the difference between the anterior expression limit in the neural tube and crest cells might result from different maintenance systems for the two cell populations (Trainor et al.,2002).

Figure 1.

Possible evolutionary scenarios for the origin of the Hox code of the vertebrate neural crest. A: Scenario 1: Neural crest expression was inherited from epidermal expression in ancestral chordates. B: Scenario 2: A novel cis-regulatory system evolved for vertebrate neural crest expression. C: Scenario 3: The neural crest Hox code evolved simply through transfer of the neural tube Hox code to the neural crest when cells migrated from the neural tube.

To examine the molecular mechanisms that underlie Hox code evolution in the vertebrate neural crest cells, Manzanares et al. (2000) studied the genomic DNA regulatory potential of amphioxus (Branchiostoma floridae) in mouse and chick embryos. They found that element 1A, a 2.5-kb genomic DNA fragment from the 3′ AmphiHox1 flanking region, could drive reporter gene expression in both the vertebrate neural tube and crest cells, while element 3B, a 6-kb genomic DNA from the 5′ AmphiHox3 flanking region, drives reporter gene expression only in the neural tube. In this study, we further investigated the cis-regulatory activity of elements 1A and 3B of Branchiostoma floridae. In particular, we focused on whether the regulatory elements involved in neural crest expression differ from those in the neural tube. We found that the DR5-type RA response elements located at the 3′ AmphiHox1 flanking region are necessary and sufficient for reporter gene expression in both the neural tube and crest of vertebrates. Therefore, the cis-regulatory elements for neural tubes and crest cells are not separable. We confirmed that DR5 from chick Hoxb1 is also capable of driving the same expression. We also found that AmphiHox3 possesses a DR5-type RARE in its 5′ flanking region, indicating that the location of the DR5-type RARE is conserved in amphioxus and vertebrate Hox clusters. The Hox code origin in the vertebrate neural crest cells can be explained partly by RA control system transfer from a neural tube of an ancestral chordate to the neural tube and crest in the vertebrate lineage, and partly by novel regulatory systems, such as AP2, and an inhibitory signal from the isthmus.

RESULTS

Mapping cis-Elements in Element 1A

To test whether neural crest expression regulatory elements differ from those for the neural tube, we electroporated deletion constructs of the AmphiHox1 element 1A from Branchiostoma floridae genome. Using deletion assays from both the 5′ and 3′ ends of element 1A, we observed that a DR5 type (direct repeats spaced by five nucleotides) RARE (DR5-1A) is necessary for reporter expression in both chick neural tubes and crest cells (Fig. 2). When electroporating constructs 1A1, 1A2, 1A6, and 1A7, we observed a strong lac-Z signal in both the neural tube and neural crest cells, and the signal was restricted to the post-otic level (posterior to the boundary between r6 and r7). We analyzed more than ten embryos for each construct, and all the embryos tested showed the same staining pattern. Occasionally, we observed some patchy anterior staining; however, this staining was much weaker than that in the post-otic level, and easy to distinguish. In construct 1A3, the 5′ region upstream from DR5-1A was truncated, but the DR5-1A sequence driving reporter expression was intact in both the neural tube and neural crest cells. In contrast, construct 1A4, which lost half the DR5 sites, lacked reporter activity (Fig. 2B). Similarly, construct 1A8, which lost most of the 3′ region downstream from DR5-1A, but in which the intact DR5-1A was retained, was sufficient to drive the expression. But further deletion prevented the activity (1A9; Fig. 2B). Simple deletion of DR5-1A (1Aδdr5) completely abolished the reporter expression.

Figure 2.

Transgenic analyses of the 3′ AmphiHox1 flanking region. A: Genomic fragments used for transgenic analyses in chick embryos. B: Nucleotide sequence of DR5-type retinoic acid (RA) response elements (DR5-1A, boxed) and surrounding regions. 1A3 and 1A8 contain DR5-1A and drive reporter expression in both the neural tube and crest. In contrast, 1A4, which contains only a half site of RAR/RXR binding sequence, and 1A9, which lacks DR5-1A, do not drive reporter expression. Deleting DR5-1A from element1A1 (1Aδdr5) results in no expression in the neural tube or crest.

To test whether DR5-1A is sufficient for gene expression, we assessed activity of two tandemly linked DR5-1A elements in the reporter vector (see Experimental Procedures section). As shown in Figure 3A, DR5-1A was sufficient for reporter gene expression in both the neural tube and crest cells (21/21 were positive for both the neural tube and crest cells). We marked cells that received the DR5-1A reporter construct by co-electroporating the DR5-1A reporter construct with a GFP expression vector driven by a CMV-IE enhancer (pCAGGS-GFP) (Niwa et al.,1991; Momose et al.,1999). The co-electroporation experiment confirmed that DR5-1A-driven lac-Z was expressed only posterior to the otic vesicle (Fig. 3A) even though neural tube and crest cells had been electroporated both anterior and posterior to the otic vesicle (Fig. 3D). Therefore, GFP-positive cells anterior to the otic vesicle received a reporter construct, but reporter lac-Z was not activated within these cells.

Figure 3.

Reporter expression driven by the DR5-type RARE from the 3′ AmphiHox1 flanking region. A: Histological detection of lac-Z activity. B,C: In situ hybridization for lac-Z mRNA in transgenic embryos. D–F: GFP signal driven by a CMV-IE enhancer in the embryos shown in A–C, respectively. Lac-Z enzymatic activity and mRNA expression was observed at HH stage 13–14, stage 12–13, respectively. Most of the cells that received electroporated plasmids produced a GFP signal. Note that we observed GFP signals beyond the otic vesicle level in neural tube and neural crest cells, particularly those from rhombomere 4. Signal from electroporated DNAs were observed in B (arrows). In each panel, arrowheads indicate the otic vesicle; nc: neural crest cells. All embryos are oriented as anterior to the left.

The progeny of individual cells within the dorsal neural tube contribute to both the neural crest and tube (Bronner-Fraser and Fraser,1988,1989; Collazo et al.,1993; Serbedzija et al.,1994). Since it is possible to detect lac-Z even after transcription ceases, the lac-Z reporter enzymatic activity detected in the neural crest might have originated from the lac-Z expression that occurred before cells were committed to the neural tube or crest. We tested this possibility by examining lac-Z reporter gene transcription. As shown in Figure 3B,C,E, and F, we detected lac-Z mRNA in both the neural tube and migrating neural crest cells. Since in situ hybridization detects the presence of both electroporated DNA and lac-Z mRNA (arrows in Fig. 3B), we arrested staining before DNA signal emergence. Since the late signal occurred in a pattern identical to the control GFP signal driven by the CMV-IE enhancer, it had to be from the DNA. The in situ hybridization signal was clear at the post-otic level in both of these regions (neural tube: 15/15; neural crest cells: 12/15, Fig. 3B,C,E,F). DR5-1A appears to have cis-regulatory activity that can drive reporter gene expression in the neural tube and crest cells, and this activity is restricted to the post-otic level. The weakness of lac-Z transcription in neural crest cells compared with the signal in the neural tube suggests that the maintenance of transcription requires additional cis-elements.

DR5 in the 3′ Flanking Region of Chick Hoxb1

There have been several studies of RARE cis-regulatory activity of the mouse RAR-β gene in mice and zebrafish. These studies indicated that RARE of RAR-β drives gene expression primarily in the posterior hindbrain, anterior spinal cord, and retina (Mendelsohn et al.,1991; Rossant et al.,1991; Balkan et al.,1992; Perez-Edwards et al.,2001). Although weak reporter gene expression was observed in zebrafish branchial arches, no expression was reported in migrating neural crest cells (Perz-Edwards et al.,2001). Since there are no clear statements made about the absence of signals in migrating neural crest cells, it remains possible that the relatively weak expression in neural crest cells was overlooked in the above studies. Therefore, we tested whether DR5 from the 3′ flanking region of chick Hoxb1 (Langston et al.,1997) can drive similar expression in the neural tube and crest cells. Figure 4A clearly indicates that chick DR5 drives reporter expression in both the neural tube and neural crest cells, and this expression is restricted to the post-otic level (10/10 are positive in both the neural tube and crest cells), as seen in the DR5-1A from AmphiHox1. We detected in situ hybridization for lac-Z mRNA in both the neural tube and crest cells (neural tube: 12/12; neural crest cells: 8/12, Fig. 4B,C). As seen for the DR5-1A from AmphiHox1, however, transcription seems to be reduced in neural crest cells. This suggests that although RARE can activate the initiation of Hox1 transcription, its maintenance may require additional cis-elements.

Figure 4.

Reporter expression driven by DR5-type RARE from the 3′ flanking region of chick Hox1. A: Histological detection of lac-Z activity. B,C: In situ hybridization for lac-Z mRNA in transgenic embryos. Note that some anterior epidermal cells (A and C) are positive for lac-Z mRNA. We found that the basal promoter (the human β-globin promoter) can drive strong background expression in epidermal cells even without DR5. This anterior epidermal expression is thus due to non-specific activity of the basal promoter. D–F: GFP signal in the embryos shown in A–C, respectively. Lac-Z enzymatic activity and mRNA expression was observed at HH stage 13–14, stage 12–13, respectively. Note that we observed GFP signals beyond the otic vesicle level in neural tube and neural crest cells, particularly those from rhombomere 4. In each panel, arrowheads represent the otic vesicle; nc: neural crest cells. All embryos are oriented as anterior to the left.

Mapping cis-Elements in Element 3B

Manzanares et al. (2000) showed that element 3B in the 5′ flanking region of AmphiHox3 also drives reporter expression in the neural tube. However, while element 1A drives expression in both the neural tube and crest cells, element 3B only drives reporter expression in the neural tube. Thus, we mapped cis-elements from element 3B (Fig. 5A). Although the cis-regulatory activity of element 3B is weaker than that of element 1A in chick embryos, reporter expression is restricted to the post-otic level in the neural tube in a similar way to element 1A. In contrast to Manzanares et al. (2000), we detected lac-Z expression in neural crest cells as well as in the neural tube (11/13 were positive in both the neural tube and crest cells, Fig. 5C). Occasionally, embryos only showed a GFP signal in the ventral neural tube. We probably did not observe a lac-Z signal in neural crest cells of these embryos because reporter DNA was not electroporated into the neural crest. This may explain why the lac-Z signal in the neural crest cells was missing in Manzanares et al. (2000). Deletion assays from both the 5′ and 3′ ends of element 3B also indicated that a 0.3-kb genomic region containing a DR5-type RARE (DR5-3B) is necessary for reporter expression in both the neural tube and crest cells (Fig. 5A). Deletion of DR5-3B sequence from construct 3B1 or constructs 3B3: 3B1δdr5, and 3B3δdr5, respectively (Fig. 5A, B), did not completely abolish reporter expression, although it did drastically reduce the signal (Fig. 5D). The presence of weak cis-elements around DR5-3B sequence could explain this; indeed, we found RARE consensus sequences at the 3′ flanking side of DR5-3B (Fig. 5B). Nevertheless, DR5-3B is obviously responsible for most reporter expression by element 3B.

Figure 5.

Transgenic analyses of the 5′ AmphiHox3 flanking region. A,B: Genomic fragments used for transgenic analyses in chick embryos. DR-5 type RARE (DR5-3B) is boxed, and two other single RARE consensus sequences are underlined. C: Histological detection of lac-Z reporter expression driven by 3B4. Patchy signal in anterior hindbrain is non-specific staining that is not reproduced in other embryos. D: Lac-Z expression driven by 3B3δdr5. E,F: GFP signal in the embryos shown in C, D, respectively. Lac-Z enzymatic activity was observed at HH stage 13–14. Arrowheads indicate the otic vesicle; nc: neural crest cells. All embryos are oriented as anterior to the left.

Electrophoresis Mobility Shift Assay (EMSA)

Recently, Yu et al. (2004) pioneered the production of transient transgenic amphioxus embryos using microinjection. However, there are still some technical limitations to the production and use of transgenic amphioxus: amphioxus eggs can only be obtained in a few places worldwide, and even if it were possible to control spawning easily, it would still be limited to two months of the year (Fuentes et al.,2004). Consequently, we decided to use an electrophoresis mobility shift assay (EMSA) to test whether the cis-element is really functional in amphioxus. The EMSA indicated that DR5 sequences from both element 1A and element 3B can specifically compete in a concentration-dependent manner for interaction between a consensus DR5 sequence and amphioxus RAR and RXR recombinant proteins (Fig. 6) (Escriva et al.,2002). Since both DR5s exhibited specific binding to the amphioxus RAR/RXR heterodimeric complex, they are likely functional in the amphioxus genome. Since the cis-regulatory activity of DR5-3B is weaker than that of DR5-1A, we tested whether DR5-1A and DR5-3B exhibited different affinities for binding the amphioxus RAR/RXR heterodimer. However, we could not detect a significant difference in the affinity of the two DR5 elements for the amphioxus recombinant RAR/RXR complex (Fig. 6).

Figure 6.

Electrophoretic mobility shift assay. A: We synthesized AmphiRAR and AmphiRXR in vitro and allowed them to bind to a 32P-labeled consensus DR5 probe (CGA TTT GAG GTC ACC AGG AGG TCA CAC AGT) in the EMSA (lines 2–15). Lane 1, unprogrammed (pSG5) reticulocytes, was a control. We added increasing amounts of unlabelled oligonucleotides corresponding to the same DR5 consensus element, DR5 from element 1A of AmphiHox1, and DR5 from element 3B from AmphiHox3 (over the probe) as competitors. We also added a non-specific element (NS) in a molar excess of 100× (over the probe) as a competitor control. We repeated the experiment three times and quantified the retarded bands on a Storm 860 (Molecular Dynamics) apparatus (bottom). B,C: A regression plot of binding percentage versus competitor concentration shows that there were no significant differences in affinity between the different DR5 elements used.

Cis-Regulatory Activity of Elements 1A and 3B in Ascidian Embryos

In contrast to amphioxus embryos, it is possible to obtain ascidian (Ciona intestinalis) embryos easily at any time of the year, and it is possible to obtain a large number of transgenic embryos using electroporation (Corbo et al.,1997). Therefore, we surveyed cis-regulatory activities of amphioxus genomic fragments in the 3′ end of the amphioxus Hox cluster, including elements 1A and 3B. Except for element 2B, which is described elsewhere (Wada et al.,2005), none of the DNA fragments analyzed by Manzanares et al. (2000) appeared to drive reporter expression in Ciona embryos. Since the human globin promoter can drive specific expression in Ciona when combined with element 2B or a 100-bp genomic DNA fragment of the 3′ AmphiHox2 flanking region (part of element 2B: Wada et al.,2005), this inability cannot be explained by the functional inability of the basal promoter. In addition, DR5 from element 1A does not drive expression in Ciona when combined with either the human β-globin promoter or CiHox1 promoter (data not shown).

DISCUSSION

Conservation of RARE in Chordate Hox Clusters

Marshall et al. (1996) showed that RA regulates transcription of several vertebrate Hox genes, especially anterior Hox genes located in the 3′ region of the clusters. In most cases, RA regulation is mediated directly through a RARE. Interestingly, Mainguy et al. (2003) showed that RARE locations were conserved in the four vertebrate Hox clusters, suggesting that these RAREs were present in a unique ancestor of these Hox complexes. These conserved RAREs were found in the 3′ flanking region of Hox1, 5′ flanking region of Hox3, and both sides of Hox4. In addition, the RARE nucleotide sequences in each cluster location were highly conserved (Mainguy et al.,2003; Table 1). In this study, we identified RAREs in amphioxus in the 3′ Hox1 flanking region and 5′ Hox3 flanking region, and these are functional during vertebrate embryogenesis. These results are consistent with previous observations that AmphiHox1 and AmphiHox3 expression are sensitive to treatment with ectopic RA or an RA antagonist (Schubert et al.,2004). Since these RAREs can specifically interact with a recombinant RAR/RXR amphioxus complex, they are likely to be functional during amphioxus embryogenesis. Therefore, it appears that the RAREs in the 3′ flanking region of Hox1 and 5′ flanking region of Hox3 are conserved in both vertebrate and amphioxus Hox clusters, and they were likely present in the Hox cluster of an ancestral chordate. Chambeyron and Bickmore, (2004) suggested that RA is involved in chromatin de-condensation and the nuclear reorganization of vertebrate Hox. The conserved nature of RAREs in amphioxus Hox clusters suggests that RAs regulate amphioxus Hox genes in a similar manner.

Table 1. Comparison of the Nucleotide Sequence of DR5 Type RARE
3′ of vertebrate Hox1GGTTCA (n5) AGTTCA
3′ of amphioxus Hox1GGGTCA (n5) CGGTCA
5′ of vertebrate Hox3GGTTCA (n5) AGTTCA
5′ of amphioxus Hox3GGGTCA (n5) AGGACA
3′ of vertebrate Hox4(A/G)GTTCA (n5) AGGACA

In contrast to the conservation of vertebrate RARE sequences, the corresponding amphioxus RAREs have distinct nucleotide sequences of consensus binding sites (Table 1). It may be worth noting that the nucleotide sequence of DR5-3B is more similar to the conserved sequence of RARE in the 3′ flanking region of vertebrate Hox4, although the DR5-3B locates about 40 kb away from AmphiHox4 (Garcia-Fernandez and Holland,1994). Since it has been suggested that nucleotide differences of consensus binding sequences of transcriptional factors may be involved in the fine regulation of gene expression (Stathopoulos and Levine,2002), the distinct nucleotide sequence of the RAREs for each amphioxus Hox gene may contribute to fine Hox gene transcriptional regulation. Indeed, switching RARE sequences in mouse Hox genes resulted in distinct reporter expression in the neural tube (Gould et al.,1998; Nolte et al.,2003). Since the cis-regulatory activity of DR5-3B is weaker than that of DR5-1A, we tested whether DR5-1A and DR5-3B exhibited different affinities for binding the amphioxus RAR/RXR heterodimer. However, we did not detect significant differences between them. Escriva et al. (2002) found that TR2/4 can compete with RARE for cis regulation. Differential affinities to TR2/4 may explain differences in the anterior expression boundary. As Chambeyron and Bickmore (2004) suggested, colinear expression of Hox genes may be controlled at a higher level, such as chromosome organization, although no information is yet available on the chromatin regulations in amphioxus.

In contrast to aspects of RARE conservation between amphioxus and vertebrates (i.e., chromosome organization, interchangeability), RAREs cannot drive expression in embryogenesis of the ascidian Ciona intestinalis. Since the human β-globin promoter can drive specific expression when combined with element 2B of amphioxus Hox (Wada et al.,2005), the functional inability of the basal promoter (the human β-globin promoter) cannot explain this inability in Ciona. Ciona Hox1 expression responds to exogenous RA (Nagatomo and Fujiwara,2003), so it is likely that RA is also involved in transcriptional regulation of Ciona Hox genes. RA may regulate Ciona Hox genes through a distinct mechanism. Indeed, the 3′ CiHox1 flanking region is very short; another gene (cDNA Cluster03704, AK113647, which shows homology to mouse F-box and leucine-rich repeat protein 15 protein [Fbxl15]) is located just 1.5 kb from the 3′ end of CiHox1. We found no sequence similar to RARE in the region. Rather, the 5′ upstream region and second intron of Ciona Hox1 has cis-regulatory activity for epidermis and neural tube reporter expression, respectively, and the reporter expression is sensitive to exogenous RA (Wada et al., unpublished data). We think the unique regulatory mechanisms of Ciona Hox by RA is a secondarily derived state, which may have to do with the fact that cluster organization in Ciona Hox genes was disrupted. It seems unlikely that the RAREs in the 3′ Hox1 flanking region and in 5′ Hox3 flanking region have evolved independently in amphioxus and vertebrates. As far as we know, there is no evidence that RA is involved in Hox gene regulation of echinoderms or hemichordates. Our preliminary search for RARE in sea urchin Hox cluster identified a putative DR5-type RARE (CGTTCA-n5- TGTTCA) in 3.3 kb downstream from homeodomain coding reagion of Hox1 (http://www.hgsc.bcm.tmc.edu/blast/?organism=Spurpuratus), although we are not certain whether it is involved in Hox transcriptional regulation.

Evolution of the New Hox Code in Vertebrates

The present study demonstrated that AmphiHox1 and chick Hox1 RAREs are sufficient to activate reporter gene expression in both the neural tube and crest cells posterior to the otic vesicle. This suggests that Hox1 gene expression in the neural crest cells reflects a transfer of regulatory information from the neural tube. Since element 3B can also drive expression in both the neural tube and crest cells, it is likely that the transfer of regulatory information to neural crest cells is applicable to Hox genes controlled by RA. Although no RARE has been found for vertebrate Hox2 (Mainguy et al.,2003), cross-regulation between Hox genes is responsible for Hoxb2 expression (Ferretti et al.,2000). In addition, mutation in the 3′ Hoxa1 DR5 affects Hoxa2 expression (Dupe et al.,1997). Therefore, a transfer of regulatory information from the neural tube might explain the evolution of anterior Hox gene neural crest expression. However, traditionally it has been believed that Hox gene expression in the neural crest cells is regulated independently from the neural tube. Two observations support this idea. First, the anterior boundaries of some Hox genes in the neural crest cells differ from those in the neural tube: for example, Hoxa2 is expressed up to level r2, but not in the neural crest cells migrating out of r2 (Prince and Lumsden,1994). Trainor et al. (2002) claimed that this absence of Hox expression from r2 crest cells is due to a repressive isthmus signal. Although Hoxa2 expression is activated both in the neural tube and neural crest of r2, the absence of Hoxa2 expression from r2 crest cells is due to a neural crest–specific repressive signal. Second, there are some neural crest–specific cis-elements of Hox expression. Hoxa2 possesses r4 crest–specific cis-elements regulated by AP2 and other factors (Maconochie et al.,1999). However, Tümpel et al. (2002) demonstrated that the AP2 consensus binding sequence is found only in mammals, and not in chick or shark Hoxa2. Therefore, AP2 involvement in neural crest development may have been acquired later in vertebrate evolution, specifically after the mammalian lineage diverged from reptiles. Furthermore, auto- or cross-regulatory elements can activate Hoxb1 and Hoxb2 expression in the r4 neural tube and crest cells (Pöpperl et al.,1995; Ferretti et al.,2000). This is consistent with initial Hox activation in the neural crest and tube being regulated by the same system. Therefore, the initiation of the Hox code in the neural crest was set up by the RA-response system that originated from the RA-responsive system in the neural tube of an ancestral chordate. This Hox code is still plastic, and thus needed to be maintained by AP2 or auto- or cross-regulatory systems. In addition, because of the plasticity, expression of some genes can still be inhibited by the signal from ithsmus. Those maintanance and inhibitory signals may be novel systems of vertebrates.

In order to determine the evolution of the neural crest Hox code, we must also find out how DR5 elements are involved in transcriptional regulation of amphioxus Hox. Schubert et al. (2004) demonstrated that AmphiHox1, 3, and 4 expression are sensitive to exogenous RA in both the neural tube and epidermis. We are currently investigating how RAREs are involved in expression of the neural tube and epidermis of ascidian and amphioxus embryos. We have confirmed that in Ciona, epidermal expression is regulated by a cis-regulatory element distinct from those for neural tube expression (Wada et al., unpublished data).

In conclusion, we propose that the new neural crest Hox code evolved using a combination of Scenarios 2 and 3 (Fig. 7). It may be partly explained by transference of the RA response system of a chordate ancestor to vertebrate neural tissues that included the developmental progenitors of the neural tube and crest. Therefore, the initial Hox code was set in neural tissue before lineage segregation occurred between the neural tube and crest (Scenario 3: Fig. 1C). However, the neural crest expression needs to be maintained, and is further refined in certain vertebrate lineages by additional regulatory systems, such as auto- or cross regulatory system, AP-2, and inhibitory signals from the isthmus (Scenario 2: Fig. 1B) (Trainor et al.,2002). Assuming that ancestral chordates possess distinct cis-regulatory elements for epidermis from that for neural tube, as in present ascidians (Wada et al., unpublished data), cis elements for neural tube expression may have been primarily responsible for the innovation of neural crest Hox code, because the initial Hox code was set in neural tissue before lineage segregation occurred between the neural tube and crest. However, the cis-regulatory mechanisms of Hox gene expression in epidermis of ascidians and amphioxus must be analyzed to determine the contributions from the factors explained in Scenario 1.

Figure 7.

Schematic illustration of the evolutionary scenarios for the Hox code origin in the vertebrate neural crest. The RA response system of a chordate ancestor was transferred to vertebrate neural tissues, including developmental progenitors of the neural tube and crest. However, neural crest expression was needed to be maintained by an auto- or cross-regulatory system and AP2, and was further refined by additional regulatory systems, such as inhibitory signals from the isthmus.

EXPERIMENTAL PROCEDURES

Construction of Plasmids

We tested the activity of cis-acting regulatory elements of genomic DNA from amphioxus (Branchiostoma floridae) in a reporter expression vector that contained an SV40 polyadenylation signal and the minimal human β-globin promoter linked to a bacterial beta-galactosidase gene (lac-Z) (Manzanares et al.,2000). Deletion constructs were made either using Exonuclease III and mung bean nuclease, or by amplifying truncated fragments using specific primers. We produced two tandemly linked DR5 from element 1A and from chick Hoxa1 (Langston et al.,1997) using the following oligonucleotides, and inserted them into the above-mentioned lac-Z expression vector and complementary oligonucleotide: the DR5 from element 1A was 5′-ggaagcttGGGTCATCTAT-CGGTCAggGGGTCATCTATCGGTCAagcttcc-3′ and the DR5 from chick Hoxb1 was 5′-ggctcgagCAGGTTCACACAAAGTTCAGccGGTTCACACA- AAGTTCAGCaagcttcc-3′ (sequences from the amphioxus genome are underlined, and RAR/RXR binding consensus sequences are shown in bold. Lowercase characters are sequences inserted for restriction digestion or spacing).

Electroporation

We electroporated chick embryos, as described by Nakamura et al. (2000). Briefly, this involved injecting 1 mg/ml plasmid DNA into the closed neural tube at the eight- to ten-somite stage (Hamburger and Hamilton [HH] stage 9–10; 33–36 hr), and applying five electric pulses (20-mV square pulses at 50-ms intervals) through electrodes spaced 4 mm apart; we used an Electrosquareporater T820 (BTX, Genetronics) during this process. Embryos were processed for lac-Z staining or in situ hybridization after being cultured for 12–18 hr following electroporation (HH stage 12–14; 45–50 hr). In situ hybridization of chick embryos followed the process as detailed by Nieto et al. (1996). We tested more than ten embryos for each construct in the reporter assays in chick embryos. Ciona electroporation essentially followed the method described by Corbo et al. (1997): a 50-mV square electric pulse was applied for 20 ms in a 4-mm-wide cuvette using the Electrosquareporater T820 (BTX, Genetronics).

Electrophoretic Mobility Shift Assay (EMSA)

We translated proteins (AmphiRAR and AmphiRXR) in vitro using a TNT reticulocyte lysate kit (Promega, Madison, WI) and allowed them to bind to a 32P-end-labeled 30-bp oligonucleotide in a reaction buffer containing 10% glycerol, 10 mM HEPES, 30 mM KCl, 4 mM spermidine, 0.1 mM EDTA, 0.25 mM dithiothreitol (DTT), and 1 mM Na2HPO4 (pH 7.9). We added single-stranded salmon sperm DNA (1 μg) and poly(dI-dC) (0.4 μg). The reaction products were run on a 5% native acrylamide gel. Where indicated, a 10- to 100-fold (10, 25, 50, and 100) molar excess of 30-bp unlabeled oligonucleotides (a consensus DR5, the AmphiHox1 DR5, the AmphiHox3 DR5, and a non-specific probe) were added as competitors. To confirm that AmphiRAR and AmphiRXR proteins were properly produced in the in vitro system, we added [35S]methionine to the translation mixture in parallel side-reactions of protein analyses, which we performed under standard conditions.

Acknowledgements

We thank Yoshiko Takahashi for her kind instruction of chick electroporation and for providing us with pCAGGS-GFP. We also thank Robb Krumlauf for providing us reporter constructs with β-globin minimal promoter; Seb Shimeld, Linda Holland, and Michael Schubert for critical reading of the manuscript; and Miguel Manzanares, Nobue Itasaki, and Peter Holland for their comments on a previous version of the manuscript. This work is supported by the CNRS, MENRT, and Association pour la Recherche contre le Cancer (ARC) to V.L., and by grants from Nissan Science Foundation, Kato Memorial Bioscience Foundation, and Grants-in-Aid from the Ministry of Education, Science, Sports and Culture, Japan (12026219, 13045020, 14034228) to H.W.

Ancillary