Retinoic acid stage-dependently alters the migration pattern and identity of hindbrain neural crest cells

Vitamin A and its derivatives, collectively referred to as retinoids, play important roles in vertebrate development and in the differentiation of a wide variety of cell types (reviewed in Roberts and Sporn, 1984 and Morriss-Kay, 1992). In par-ticular, retinoic acid (RA), a biologically active retinoid, exerts a broad spectrum of effects.

It is well known that an excess or deficiency of retinoids induces abnormal morphology in mammalian embryos (reviewed in Morriss-Kay, 1992). When rodent embryos are exposed to excess RA during or shortly before neurulation, the preotic hindbrain region is shortened in relation to other head structures (Morriss, 1972; Morriss and Thorogood, 1987). Marshall et al. (1992) showed that RA administration to pregnant mice at 7.5 days post coitum (d.p.c.) induces poster-iorization in the body axis, i.e., it alters the anterior structure in the hindbrain region into the posterior one. In contrast, treatment with RA at later stages causes different types of mal-formations, e.g., reduction and/or fusion of the first and second branchial arches that eventually form the maxilla, mandible, and external and middle ears (Goulding and Pratt, 1986; Webster et al., 1986).

Most of the craniofacial structures in vertebrates, including the branchial arches and cranial ganglia, contain cells derived from the cranial neural crest (reviewed in Le Douarin, 1982; Morriss-Kay and Tan, 1987; Noden, 1988). The hindbrain neural crest cells, which are thought to be prepatterned or imprinted with positional information (Noden, 1988), migrate to the adjacent branchial arches (I-IV) in a segmental manner (Lumsden et al., 1991; Serbedzija et al., 1992; Matsuo et al., 1993; Osumi-Yamashita et al., 1994). As a parallel event to such migration, a segmented rhombomere structure appears along the anterior-posterior axis, with these rhombomeric boundaries denoting cellular compartments that restrict cell lineage (Fraser et al., 1990; Birgbauer and Fraser, 1994).

In the mouse, Hox genes have spatially restricted patterns of expression that appear before the formation of rhom-bomeric morphology and later map to specific rhombomeric boundaries (Graham et al., 1989). These genes are also expressed in the branchial arches in segment-specific patterns (Hunt et al., 1991a,b; Wilkinson et al., 1989); hence sug-gesting that the prepatterned information contained in hindbrain crest cells is associated with the expression pattern of Hox genes (reviewed in Hunt and Krumlauf 1991; Krumlauf, 1993).

Of particular interest, based on observations of Hox marker gene expression and nerve morphology, Marshall et al. (1992) suggested that RA treatment of mouse embryos at 7.5 d.p.c. alters the identity of rhombomeres 2/3 (r2/3) to that of r4/5, as well as altering the identity of the trigeminal nerve to that of the facial nerve. However, since the cranial ganglia contain neurons derived from the hindbrain crest and their nerves innervate into the corresponding branchial arches that are populated by the hindbrain crest cells, a question remains as to whether or not the migration pathway of neural crest cells emerging from the altered hindbrain origins can be changed from a standpoint of cell lineage.

Another interesting phenomenon reported by Brown et al. (1992) is that in rat embryos treated with RA from the head-fold stage (a period occurring after the treatment mentioned above), the Hox B2 gene, whose expression is normally detected in the second branchial arch and posterior ones, is expressed segmentally in the posterior part of the FBA. It is not known, however, if neural crest cells within the FBA mix or remain distinct.

These two questions led to the present study which elucidates the stage-dependent RA effects on craniofacial pattern formation by investigating the alteration in the migration pathway of cranial neural crest cells from a stand-point of cell lineage and identity. Towards this end, we treated rat embryos in vitro with RA at specific developmental stages and used the vital dye 1,1-di-octadecyl-3,3,3’3’-tetramethylin-docarbocyanine perchlorate (DiI) to monitor the migration of neural crest cells from the anterior hindbrain (the presump-tive r1 and r2) and the preotic hindbrain (the presumptive r3 and r4). In addition, to confirm the identity of the hindbrain brachial arches and cranial ganglia, we performed immunos-taining of CRABP I and neurofilament on sections and whole-mounts, respectively.

Our results clearly show that RA influences craniofacial development in a stage-dependent manner and that it causes alteration in the identity of branchial arches, cranial ganglia and rhombomeres at morphological and molecular levels. Regarding the migration pathways of hindbrain crest cells, RA treatment at the early stage induced ectopic migration of anterior hindbrain crest cells; thereby suggesting from a cell lineage perspective that RA transforms the regional identities of these crest cells.

Wistar-Imamichi rat embryos were surgically explanted from anes-thetized mothers at 9.0 and 9.5 d.p.c. (early and late neural plate stage, respectively; plug day; 0). The time course of the whole embryo culture is shown in Fig. 1. Whole embryos were cultured according to a previously described method (Matsuo et al., 1993; Osumi-Yamashita et al., 1994). Briefly, embryos were placed in 15 ml culture bottles containing 3 ml of culture media consisting of 100% immedi-ately centrifuged rat serum with 2 mg/ml glucose. The culture bottles were attached to a rotator drum and rotated at 20 revs/minute and 37°C while being continuously supplied with a gas mixture, i.e., 5% O₂/5% CO₂/90% N₂ for the first 36 or 24 hours for early and late stage embryos, respectively, and subsequently with 20% O₂/5% CO₂/75% N₂ for the remaining culture period. The flow rate of gas was increased as necessary.

All-trans RA (Sigma) was dissolved in dimethyl sulfoxide (DMSO) to make a stock solution (2×10^-4 M). A 3 μl aliquot was added to 3 ml of culture medium so that the final RA concentration was 2×10^-7 M. Embryos were exposed to RA for 6 hours and further cultured in fresh medium until they developed to the desired stage (Fig. 1). Vehicle-control embryos were exposed to the same amount of DMSO for 6 hours. After being treated with RA or vehicle, embryos were washed several times with Tyrode’s solution and transferred to fresh medium for further culture. Because DMSO did not affect normal development of the embryos, inject-control embryos were not treated with the vehicle. The RA concentration, treatment duration and stage at treatment were determined from preliminary experiments, being optimally adjusted so that RA induced to a maximum extent the ter-atogenic effect in branchial arches.

Microinjection of DiI was performed according to the methods pre-viously described (Matsuo et al., 1993; Osumi-Yamashita et al., 1994). Briefly, micropipettes (internal diameter 10 μm) were filled with DiI (Molecular Probes, Inc.) at 0.25% (w/v) in 100% dimethyl-formamide. Embryos that developed to the 3-to 7-somite stage were selected for DiI labeling (Fig. 1). Focal injections of DiI were performed at the anterior hindbrain (the presumptive r1 and r2) and the preotic hindbrain (the presumptive r3 and r4) in order to label pre-migratory neural crest cells. All successfully injected embryos were immediately transferred into culture bottles and incubated as described above.

After culturing, embryos were assessed for heart beat, yolk-sac and whole body blood circulation, yolk-sac diameter, crown-rump length, somite number and general morphology. To ascertain if the culture period itself affected embryonic development, preliminary experi-ments were carried out in which cultured embryos were compared with those allowed to develop in utero for the same length of time. Based on the resultant somite number and general morphology, embryos that were cultured for 60 hours from the early stage and 48 hours from the late stage appeared similar to ones developed in utero (data not shown).

Distribution of DiI was photographed (Kodak PKL200) using whole-mount embryos placed under a fluorescent microscope with a rhodamine filter set (Zeiss). Some embryos were fixed in 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS) for 2-4 hours, frozen in OCT compound (Miles Inc.), and serially sectioned (14 μm) at -20°C. These sections were also subjected to fluorescent microscopy. Data from the microscope were readable on Nikon Coolscan set and images were processed using Adobe Photoshop software (Adobe system, Inc.) on a Macintosh Quadra 840 AV computer.

For scanning electron microscopic (SEM) observations, embryos were fixed in 0.05 M cacodylate buffer containing 1% PFA and 1.5% glutaraldehyde (pH 7.35-7.45) for 2 hours at room temperature and rinsed with 0.1 M sucrose in 0.05 M cacodylate buffer. The samples were postfixed with 1% osmium tetroxide in 0.05 M cacodylate buffer, dehydrated in graded ethanol, critical point dried (Critical Point Dryer, Hitachi), sputter coated with Au-Pd (HCP-2, Hitachi) and then observed (JSM-T200, JEOL).

For histological observations, embryos were fixed in Bouin’s solution, dehydrated with ethanol series and embedded in paraffin. Serial frontal sections (5 μm) were mounted on glass slides, stained with hematoxylin-eosin, observed under a light microscope (Olympus) and photographed (Fujichrom, RDP).

The employed antibody was raised against synthetic peptides corre-sponding to residue 68-81 of bovine CRABP I. The specificity of the antibody to CRABP I is well characterized in rat tissue (Eriksson et al., 1987), and that it has no detectable cross-reactivity to CRABP II (Dencker et al., 1990, 1991; Maden et al., 1991, 1992).

For detection of CRABP I protein, frozen sections of cultured embryos were prepared as described above, quenched with hydrogen peroxide, blocked with normal horse serum and processed overnight for application of the antibody against CRABP I at a dilution of 2 μg/ml and 4°C. Immunoreactivity was detected using an ABC kit (Vector Laboratorie). Diaminobenzidine (DAB) was used to visualize the reaction product and the sections were counterstained with methyl green.

As the primary antibody we used 2H3, an anti-neurofilament mono-clonal antibody, which labels 166×10³M_r neurofilament raised from adult rat brain membranes. As the secondary antibody, horseradish peroxidase (HRP)-conjugated anti-mouse IgG (NGF 825, Amersham) was used.

For the whole-mount immunostaining, the embryos were further cultured for 9 hours until they developed to the 28-somite stage, and prepared as previously described (Sundin and Eichele, 1990; Kuratani and Eichele, 1993) using the following modification. Cultured embryos were fixed overnight in 4% PFA in PBS at 4°C and stored in methanol at -20°C. The samples were subsequently treated with DMSO/methanol and 2% Triton X-100, and then washed with TST (Tris-HCl buffered saline: 20 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.5% Triton X-100). Next they were treated with 1% periodic acid solution to block endogenous peroxidase activity, sequentially blocked with 5% dry non-fat milk in TST (TSTM), and afterwards incubated overnight at room temperature with 2H3 antibody in spin-clarified TSTM (1:50-100) containing 0.1% sodium azide. Following washing with TST, the samples were treated with HRP-conjugated secondary antibody in TSTM (1:200) overnight at room temperature. Finally, they were preincubated with Tris-HCl-buffered saline (TS) containing DAB (250 μg/ml) for 1 hour, and then reacted with the same concentration of DAB/TS containing hydrogen peroxide (0.05%) at 0°C for 20 minutes. The reaction was stopped by rinsing the samples with 30% glycerol in distilled water. To obtain trans-parency, the stained embryos were transferred through a graded series of up to 80% glycerol in distilled water containing a trace amount of thymol.

The hindbrain region of embryos clearly showed two types of segmental structures after culturing, i.e., a branchial arch structure and a rhombomeric component. In the control embryos cultured from 9.0 or 9.5 d.p.c., the first branchial arch, including the maxillary and mandibular prominences, was larger than the second and posterior arches (Fig. 2A). In contrast, RA induced two typical craniofacial abnormalities depending on stages. RA treatment at the early stage reduced the size of the first branchial arch, including a decrease in the maxillary and mandibular prominences (Fig. 2B; Table 1). This arch did not show the typical bifurcated shape, but instead looked like the second branchial arch. In contrast, the corre-sponding late-stage-treated embryos (Fig. 2C) had their first and second branchial arches fused together. Although a fused branchial arch (FBA) was usually observed to be bilaterally situated, it was unilaterally situated in a few embryos (Table 2).

The hindbrain morphology also showed interesting differences. The frontal sections of the control embryos had well-developed segmental rhombomeres (Fig. 2D). Although the early- and late-stage RA-treated embryos had a normal total number of rhombomeres, the region with rhombomeres (r) 1-3 was often compact and asymmetric in the early-stage RA-treated embryos (Fig. 2E), while only the r3-4 region was compact in the late-stage ones (Fig. 2F).

Moreover, the late-stage-treated embryos had their otocyst position shifted rostrally, being due to a shortened preotic hindbrain region (Fig. 2C,F). This was not observed in the early-stage-treated embryos, which instead showed a small midbrain and shortened anterior hindbrain (Fig. 2B). In both the RA-treated groups, some embryos had an open neural tube and/or microcephaly (Tables 1, 2).

The effects of RA on general embryonic development were not as severe, yet statistically significant differences were found in comparison to the control (injected or not injected) and vehicle control groups regarding yolk-sac diameter, crown-rump length and number of somites (Tables 3, 4). The toxicity of DiI labeling was evaluated by comparing the injected and uninjected embryos within the control groups and RA-treated groups. Since the parameters for embryonic devel-opment did not significantly differ within these groups, this indicates the labeling by microinjection had no effect on general development of embryos (Tables 3, 4).

It should be noted that the early-stage treatment (6 hours) embryos showed very similar craniofacial morphology results to those previously described in which pregnant mice (7.5 d.p.c.) were treated (Marshall et al., 1992). In fact, in additional experiments that exposed the early-stage embryos with RA for only 3 hours, similar abnormalities were obtained, although compaction and asymmetry of the rhombomeres did not occur (data not shown).

In the control groups, immunoreactivity of CRABP I was not detected in r1 or r3 and was weak in both r2 (Fig. 3A) and the mesenchyme of the first branchial arch (Fig. 3D). In contrast, strong immunoreactivity occurred in r4-7 (Fig. 3A) and in the second and posterior arches (Fig. 3D), as well as in the fron-tonasal mass; similar to previous observations in mouse embryos (Maden et al., 1992; Gustafson et al., 1993).

The expression of CRABP I in the rhombomeres was not different in the late-stage-treated embryos, i.e., the immunore-action was only strong in r4-6 (Fig. 3C). In the FBA, the posterior region showed intense CRABP I reactivity (Fig. 3F), while being weak in its anterior part; thereby suggesting that the neural crest cells do not freely mix despite the fusion of the arches. The early-stage-treated embryos, however, induced up-regulation of CRABP I immunoreactivity in r1-3, as well as in the mesenchyme of the first branchial arch (Fig. 3B,E), i.e., neural crest cells in their first branchial arch and their associ-ated rhombomeres (r1-3) were strongly stained like those in the second and posterior arches and r4-7.

Whole-mount immunostaining with anti-neurofilament antibody 2H3 clearly visualized the morphology of the trigem-inal, acousticofacial, and more posterior ganglia and their nerve branches in the control embryo at the 28-somite stage (Fig. 4). The ophthalmic branch arose from the trigeminal ganglion extending rostrally toward the eye primordium, while the maxillomandibular branches extended to the first branchial arch. Also the acousticofacial nerve innervated to the second branchial arch and the otocyst (Fig. 4A).

In late-stage RA-treated embryos at the same stage, com-pression of the preotic hindbrain region resulted in the fusion of the trigeminal and acousticofacial ganglia near their proximal portions, though the typical features of the ganglia and branching patterns were retained (Fig. 4C). In contrast, these ganglia remained separated in early-stage RA-treated embryos with the same number of somites. Interestingly, however, the morphology of the trigeminal ganglion was clearly altered in this case; the ophthalmic branch innervated the eye region as in the control, whereas the typical axonal pro-jections into maxillary and mandibular regions had vanished (Fig. 4B). The morphology of the facial nerve and its ganglion did not change in early-stage RA-treated embryos.

In situ labeling was carried out at the 3-to 7-somite stage based on results by Tan and Morriss-Kay (1985) and Osumi-Yamashita et al. (1994) which clarify when the emigration of hindbrain crest cells begins. The premigratory neural crest cells were labeled at the anterior hindbrain (similar to rhombomere A in Bartelmez and Evans, 1925, corresponding to presump-tive r1 and r2) and preotic hindbrain (rhombomere B, the pre-sumptive r3 and r4).

Crest cells labeled in the anterior hindbrain of the control groups were observed in the first branchial arch and the trigem-inal ganglion, while those labeled at the preotic hindbrain were distributed in the second branchial arch and acousticofacial ganglion (Table 5; Fig. 5A-D). Such segmental migration patterns are identical to those observed in previous studies (Serbedzija et al., 1992; Matsuo et al., 1993; Osumi-Yamashita et al., 1994). Neuroepithelial cells labeled at the anterior hindbrain distributed in r1 and r2, while those labeled at the preotic hindbrain in r3 and r4 (Fig. 5E,F). Labeled neuroep-ithelial cells were sometimes observed crossing over the rhom-bomeric boundaries in the dorsal unsegmented region (arrows in Fig. 5E,F).

In 18/18 cases of the early-stage RA-treated embryos, crest cells labeled in the anterior hindbrain ectopically migrated into the second branchial arch and/or the acoustico-facial ganglion in addition to their original destinations, i.e., the first branchial arch and the trigeminal ganglion (Fig. 5G-I; Table 5). Of importance, labeled neuroepithelial cells derived from the anterior hindbrain were observed in r1 and r2, but not in r3 or r4 (Fig. 5G-J), which suggests that the exit point of anterior hindbrain crest cells remains unchanged and that these crest cells directly migrate into the ectopic regions. This is typically indicated in the dorsal view of the embryo since it simultaneously shows labeled neu-roepithelial cells in r1 and r2 and labeled crest cells in the acousticofa-cial ganglion (arrowheads in Fig. 5J). In this embryo, some labeled cells were also observed in the dorsal midline corresponding to r3 but, because this region was completely unsegmented, it should not be con-sidered to be a rhombomere. The migration pathways of the crest cells derived from the preotic hindbrain (presumptive r3 and r4) were not much different from the control group’s, although 4 of 12 embryos showed ectopic migration from the preotic hindbrain into the trigeminal ganglion (Table 5). A particular issue of interest in these observations is that the preotic sulcus may possible serve as a lineage restriction in the early-stage RA-treated embryos.

Surprisingly, the migration patterns were segmental in the late-stage RA-treated embryos even though FBA occurred. Labeled cells from the anterior hindbrain populated the anterior half of the FBA and trigeminal ganglion (Table 5, Fig. 5K,M), whereas those from the preotic hindbrain populated the posterior half of the FBA and acousticofacial ganglion (Fig. 5L,N). This behavior is noteworthy because cells from the different prorhombomeres (anterior and preotic hindbrain, i.e., rhombomere A and B) did not mix in the FBA of these embryos, despite the fact they have adjacent positions in the same arch. Like the control groups, no cells originating from different neuroepithelial areas were observed to be mixed in hindbrain axial level. Moreover, the labeling patterns in rhom-bomeres were the same as those in control embryos even though they showed the compacted rhombomeres; i.e., labeled cells at the anterior hindbrain distributed in r1 and r2, while those at the preotic hindbrain in r3 and r4 (Fig. 5O,P). Conse-quently, the segmental migration pattern of the late-stage embryos was not affected by the RA treatment, nor were the regional properties affected of neural crest cells originating from different hindbrain regions; suggesting that the treatment did not change their identities.

Although RA treatment induced compaction of rhombomeres in the late-stage embryos and sometimes in the early-stage ones, their number was the same in both groups. These results are in agreement with the effects of RA treatment on frog, fish and mouse embryos (Papalopulu et al., 1991b; Holder and Hill 1991; Marshall et al., 1992), which suggest that such changes are in common with vertebrates. Morriss-Kay et al. (1991) and Wood et al. (1994) observed unsegmented-type rhombomere malformation in maternally treated mouse embryos, which we did not observe in cultured rat embryos: a discrepancy that may be due to using different experimental systems.

In all the RA-treated embryos, we found an altered structure of the branchial arches in addition to the rhombomeric defects. Early-stage treatment specifically affected the first branchial arch, since it appeared to have a similar size and structure to the second branchial arch. Marshall et al. (1992) reported a corresponding finding in mouse embryos that were maternally treated with RA at a similar stage. In the late-stage-treated embryos, however, another type of arch defect typically occurred, i.e., fusion of the first and second branchial arches (FBA). The treatment also concomitantly induced an anterior shift of the otocysts caused by shortening of the anterior hindbrain region. A similar FBA has also been observed in RA-treated mouse embryos (Goulding and Pratt, 1986; Webster et al., 1986).

We also found differential effects of RA on cranial ganglia formation. Whole-mount immunostaining of neurofilament in early-stage-treated embryos clearly demonstrated that the trigeminal ganglion changed its morphological structure, though it developed independently from the acousticofacial ganglion. A similar situation has been reported by Marshall et al. (1992) who found that the trigeminal nerve adopted a facial nerve identity with respect to the location of nerve cell bodies and their axonal projection after RA administration. In contrast, in our system, the late-stage treatment induced a fusion of trigeminal and acousticofacial ganglia. It should be noted that, genetic disruption of Krox-20 produces a similar fusion of these ganglia in mouse embryos (Schneider-Manoury et al., 1993; Swiatek and Gridley, 1993). However, such fusion seemed to be due to the loss of r3 (Schneider-Manoury et al., 1993; Swiatek and Gridley, 1993), which is likely to be caused by a different developmental mechanism than ours since r3 exists in our late-stage-treated embryos.

Taken together, the effects of RA on embryonic patterning are obviously stage-dependent. The treatment at the early neural plate stage induces malformation specifically in the first branchial arch and trigeminal ganglion, which may correspond to the transformation of the anterior hindbrain in RA-treated mouse embryos as shown by Marshall et al. (1992). In contrast, later-stage treatment caused entirely different phenotypes, i.e., fusion of the first and second branchial arches and that of the cranial ganglia.

It is well known that the migration of hindbrain crest cells is segmental in normal development, i.e., those derived from the anterior hindbrain migrate to the first branchial arch and never mix with the pathway of the preotic hindbrain crest cells migrating to the second arch (Lumsden et al., 1991; Serbedz-ija et al., 1992; Matsuo et al., 1993; Osumi-Yamashita et al., 1994).

However, early-stage RA treatment altered such segmental migration patterns of the hindbrain crest cells by inducing ectopic pathways in all embryos investigated. Anterior hindbrain crest cells posteriorly migrated to the second branchial arch and acousticofacial ganglion, as well as to their original destination, the first arch and trigeminal ganglion. Conversely, the preotic hindbrain crest cells had their pathways anteriorly altered to the trigeminal ganglion, which is normally occupied by anterior hindbrain crest cells (Fig. 6B).

In the ectopic migration of anterior hindbrain crest cells, most of neuroepithelial cells, including premigratory crest cells, did not move posteriorly over the preotic sulcus, i.e., labeled anterior hindbrain neuroepithelial cells were restricted in r1 and r2. Thus, RA treatment at the early neural plate stage induces distortion of the neural tube, though it does not alter the location of the exit points of the neural crest cells, instead only altering the identity of the anterior hindbrain into that of the preotic hindbrain before migration begins. In other words, RA treatment causes premigratory and/or migrating anterior hindbrain crest cells to have a common identity with the preotic hindbrain crest cells, and then these altered anterior hindbrain crest cells ectopically migrate directly to the second branchial arch and/or the acousticofacial ganglion. Regarding the crest cells emerging from the preotic hindbrain in early RA-treated embryos, they migrated to the second branchial arch (ventral structure), and to the trigeminal and acousticofacial ganglia (dorsal structure), though they were never observed in the first branchial arch. One possible explanation is that a discrepancy exists in the emigrating time of crest cells derived from the anterior hindbrain and preotic hindbrain, i.e., the former emigrate earlier than the latter (Tan and Morriss-Kay, 1985) and occupy the first arch before the latter enter it. Further support of this possibility is that a ventral-to-dorsal migration pattern is a natural feature of crest cells (Serbedzija et al., 1992; Osumi-Yamashita et al., 1994).

In contrast, the late-stage RA treatment did not change the normal segmental migration from different hindbrain crests in the FBA (Fig. 6C). The anterior half of the FBA was populated by anterior hindbrain crest cells, whereas the posterior half by preotic ones. Cranial ganglia were also segmentally populated, i.e., from the anterior hindbrain to the trigeminal ganglion and from the preotic hindbrain to the acousticofacial ganglion, despite the fact that both ganglia were fused.

These results clearly indicate that the migration pattern of the hindbrain neural crest cells is stage specifically affected by exogenous RA because segmental hindbrain crest cell migration is perturbed by the early treatment but not by the late one. We should point out that this is the first evidence showing that RA treatment alters the hindbrain identity by affecting the lineage of hindbrain crest cells.

Another important conclusion drawn here is that the function of the branchial arches is not solely to separate different pop-ulations of hindbrain crest cells. We base this on the fact that these crest cells did not freely mix in the FBA of the late-stage RA-treated embryos, whereas in the early-stage ones, changing the identity of the cells allowed them to mix or take alternate pathways in the branchial arches. Thus, migration of the neural crest cells into specific arches is not a completely passive process, but instead somewhat dependent on the ability of the cells to interpret migratory cues.

The observed stage-dependent effects of RA on the migration pattern of neural crest cells are consistent with the results of Conlon and Rossant (1992) in that the expression of Hox-B1 (Hox 2.9) and Hox-B2 (Hox 2.8) genes are responsive to RA in the early neural plate stage, though by the early somite stage, expression within the neural tube no longer occurs.

RA treatment of mouse embryos at the preheadfold stage (roughly corresponding to the early neural plate stage in the present experiment) induces the anteriorly shifted expression of Hox-B1 and Hox-B2 and Krox 20 genes in rhombomeres and branchial arches (Morriss-Kay et al., 1991; Conlon and Rossant, 1992; Marshall et al., 1992; Wood et al., 1994). Segment-specific expression patterns of these genes in rhom-bomeres and branchial arches are suggested to impose a regional identity in prospective craniofacial structures (Hunt et al., 1991a,b). Therefore early stage-RA treatment possibly induces a homeotic transformation in the identity of rhom-bomeres and branchial arches with respect to gene expression.

In contrast, RA treatment at the late neural plate stage may not change the regional identities of the branchial arches and rhombomeres that have already been established by the time of the treatment. Supportive evidence of this is provided by Brown et al. (1992) who found that Hox-B2 expression only takes place in the posterior part of the FBA in cultured rat embryos treated with RA from the late-head-fold stage.

In accordance with the stage-dependent alteration of gene expression described above, we found that CRABP I distribu-tion was changed by the early treatment, but not altered by the late one. In normal development, CRABP I is intensely expressed in r4-6 and in the second and posterior arches, while it is not specifically detected in r1 and r3, and is weak in r2 as well as in the first arch (see our results; Maden et al., 1992; Gustafson et al., 1993; McKay et al., 1994). Thus CRABP I should be regarded as another suitable marker for identifying the hindbrain regions. Early RA treatment was found to upreg-ulate CRABP I in the first branchial arch and r1-3 to a similar degree as that in the second branchial arch and r4. This obser-vation corresponds well with the altered expression of Hox and Krox-20 genes in RA-treated embryos (Morriss-Kay et al., 1991; Marshall et al., 1992; Wood et al., 1994), and provides molecular evidence pointing towards the stage-dependent effects of RA affecting the rigional character of the hindbrain and their crest cells.

Taken together, our results and those of other systems (Marshall et al., 1992; Wood et al., 1994) suggest that RA mediates head development in vertebrates by controlling cell lineage as well as through specific gene expression. We do know that RA can activate Hox genes in vitro in a colinear manner (Simeone et al., 1990, 1991; Papalopulu et al., 1991a,b). In fact, Hox-B1 gene has RARE at its 3’ end, with mutation analysis of RARE showing that it is essential for establishing early Hox-B1 neural expression (Marshall et al., 1994). In addition, RA is known to induce both Hox-B1 and Hox-A1 genes through the RARE elements (Studer et al., 1994). Most recently, it was shown that RA activation of Hox-A1 gene may have created a part of the RA-induced pheno-types, i.e., when the Hox-A1 construct is injected into trans-genic mice it induces Hox-B1 expression in r2, leading into the transformation of the identity of r2 to r4 (Zhang et al., 1994). Thus, RA can induce rapid changes in Hox gene expression, which alters the regional identity of both the hindbrain neu-roepithelium and crest cells. This in turn would alter the migration patterns of the hindbrain crest cells and result in the abnormal patterning of their derivatives. It may be too premature at present to conclude that the altered morphology of the first branchial arch and trigeminal ganglion is the trans-formation as seen in the experiments of Marshall et al. (1992). However, the altered expression of CRABP I in the branchial arches and hindbrain and changes in migration patterns of the hindbrain crest cells may possibly imply that RA can also cause transformation in the rostral branchial region. Further experimental studies on the mechanisms involved in the trans-formation of hindbrain and branchial arches should contribute to a better understanding of craniofacial development in ver-tebrates.

Sincere gratitude is extended to Dr Robb Krumlauf for his valuable comments. We also thank Dr Shigeru Kuratani for helpful suggestions and technical advise on whole-mount immunostaining. The 2H3 antibody was obtained from the Developmental Studies Hybridization Bank maintained by the Department of Pharmacology and Molecular Sciences at Johns Hopkins University School of Medicine Baltimore, MD, and the Department of Biology at the University of Iowa, Iowa City, IA, under contract number NO-HP-2-3144 from the NICHD. This work was supported by a Grant-in Aid for Scientific Research from the Japanese Ministry of Education, Science and Culture (No. 05557078, 06671805).