ABSTRACT
We have studied developmental expression of zebrafish engrailed-like (Eng) antigens. Many cell types are reproducibly labeled by two antibodies that recognize the Eng homeodomain, but other cells are labeled by only one or the other, suggesting a hitherto unrecognized complexity of Eng proteins. Expression patterns vary remarkably according to cell type and location. In the undifferentiated primordia of the brain and of each myotome, expression by a stripe of cells spatially subdivides the primordium at a location where a morphological boundary forms later, suggesting expression may be required for development of the boundaries. Supporting this hypothesis, trunk myotomal cells that express Eng are absent in spt-1 mutant embryos, just where the myotomal boundaries fail to form. Another pattern is present in rhombomeres, pharyngeal arches, and the pectoral girdle. In each of these cases, cells (neuron, muscle, cartilage) generating a subset of a series of repeated elements selectively express Eng. These subsets then form specialized derivatives, suggesting Eng homeoproteins are involved in determining the specializations. Epidermal expression is present in the ventral half of the pectoral fin rudiment, precisely ‘compartmentalizing’ the fin. Neuronal cells at a certain dorsoventral level in each hindbrain and spinal cord segment selectively express Eng, suggesting segmental control of neuronal identity. Specific expression patterns are observed in taste buds, otic vesicles and teeth. Thus we propose that eng genes function in diverse cell types in zebrafish, but play selector roles that can be classified into a few basic types.
Introduction
Homeoproteins are DNA-binding proteins originally discovered in Drosophila that regulate cell fate and pattern formation (for review, see Ingham, 1988), and among them engrailed is one of the best studied (Kornberg et al. 1985; Fjose et al. 1985). Learning the detailed expression patterns of proteins provides an initial step towards understanding how they function, and what patterning features they may control.
Two engrailed genes are present in vertebrates, and have been examined in detail in the mouse. One of them, En-2 (Joyner et al. 1985), is expressed in a restricted region of the central nervous system (CNS). Expression appears in a prominent transverse stripe of cells at the future boundary between the midbrain and hindbrain, but before this boundary can be detected morphologically (Davis et al. 1988; Davidson et al. 1988). Expression continues at this location in the adult in restricted cells in the cerebellum and other nearby CNS cells (Davis et al. 1988). Deletion of En-2 function by targeted mutagenesis produces cerebellar deficiencies, revealing the requirement of the gene for normal development (Joyner et al. 1991). This pattern appears generally conserved among vertebrates (Gardner et al. 1988; Patel et al. 1989; Hemmati-Brivanlou and Harland, 1989; Davis et al. 1991; Kimmel, 1989). The second gene, En-1 (Joyner and Martin, 1987), is transcribed at the same location as En-2 in the embryo, and additionally at many other sites including ventrolateral cells in the hindbrain and spinal cord, limb bud epithelium, facial mesenchyme, dermatome and the anterior halves of the neural arches (Davidson et al. 1988; Davis and Joyner, 1988; Davis et al. 1991).
Two engrazZed-like genes have also been identified in zebrafish, eng-1 and eng-2 (collectively eng genes, see Njølstad et al. 1990 for conventions concerning nomenclature). Both are very similar to their counterparts in the mouse (Fjose et al. 1988; Holland and Williams, 1990; M. W. unpublished), and aspects of expression have been reported, eng-2, which corresponds to mouse En-2, is expressed specifically in the hindbrain/ midbrain stripe (Njølstad and Fjose, 1988). As revealed by labeling with the monoclonal antibody 4D9, which appears to recognize the homeodomains of both zebrafish Eng proteins, additional Eng expression is present in a subset of cells in each somite (Patel et al. 1989), and later in head mesenchymal precursors of two specific jaw muscles (Hatta et al. 1990).
In this report, we used 4D9 and a second antibody that recognizes the Eng homeodomain (αEnhb-1 ; Davis et al. 1991), to study Eng expression patterns during normal development from the gastrula to the adult. We correlated when and where expression occurs with such developmental processes as morphogenesis and cell differentiation in different organ systems. We show that a variety of cell types express Eng in diverse spatiotemporal patterns which may represent only a few bàsic types. Our studies reveal that the somitic cells previously described (Patel et al. 1989) develop as a special class of body wall muscle cells, termed muscle pioneers (Felsenfeld and Curry, in preparation). Using Eng labeling we discovered that muscle pioneers are specifically absent in the trunks of embryos bearing the spt-1 mutation, which disrupts muscle organization in the trunk (Kimmel et al. 1989). This finding supports a role of the muscle pioneers, and in particular their expression of Eng, in establishing the patterning of body wall muscle.
Materials and methods
Wild-type and spt-1 zebrafish
Zebrafish (Brachydanio rerio) embryos and larvae were obtained from natural crosses of a standard outbred laboratory strain. The abbreviations ‘h’ and ‘d’ stand for hours and days after fertilization with incubation at 28.5°C. We examined Eng expression patterns with the 4D9 monoclonal antibody throughout the entire bodies of fish from gastrula (beginning at 6h) through midlarval (21 d) stages of development, and in the adult (ca. 150 d) head between the eyes and ears. We examined labeling by αEnhb-1 at fewer stages (6h-28h, 3d, and 21 d).
Embryos homozygous for the recessive lethal ‘spadetail’ mutation, spt-l(b104) (Kimmel et al. 1989) were produced by crossing heterozygotes and were studied with the 4D9 antibody. For comparisons to wild type, we examined both genetically unrelated homozygous spt-1+ embryos as well as phenotypically wild-type siblings (including heterozygotes) of the mutants. No changes in Eng expression were found that could be attributed to a dominant effect of the mutation.
Immunocytochemistry
The 4D9 monoclonal antibody was generated against the Drosophila invected homeodomain (Patel et al. 1989; Coleman et al. 1987). The αEnhb-1 polyclonal antibody was generated against the mouse En-2 homeodomain and appears to recognize both engrailed-\ike gene products in several vertebrate species (Davis et al. 1991). We also used two monoclonal antibodies zn-5 (Trevarrow et al. 1990) and 3A10 (Furley et al. 1990; K. H. in preparation) for identification of cell types.
Zebrafish were processed for immunocytochemistry (Molven et al. 1990) either as whole-mounts (6 h to 3 d) or as 20 pm frozen sections (36 h to 150 d). The embryos were fixed for three to six hours at 4 °C and immunostained by the indirect peroxidase-anti-peroxidase method of Stemberger (1979). After staining, some of the whole mounts were embedded in Epon-Araldite and sectioned at 10pm. For the double label experiments, two antibodies were mixed together.
Western blotting
The chorions were removed from 36 h zebrafish embryos enzymatically (Westerfield, 1989). The embryos were transferred to a physiological saline (Westerfield et al. 1986) with 1mM ethylenediaminetetraacetic acid and 0.3 mM phenyl-methylsulfonylfluoride added. The yolks were removed by gentle trituration. After rinsing, the embryos were solubilized in 10% glycerol, 3.5 % sodium dodecyl sulfate (SDS), 5% β-mercaptoethanol in 63 mM tris hydroxymethyl aminomethane (Tris), pH6.8. The solution was boiled 5min and then centrifuged for 5 min. Proteins were separated by SDS-polyacrylamide gel electrophoresis (PAGE) and blotted to polyvinylidene difluoride sheets (Towbin et al. 1979). After blotting, the sheets were cut into strips that contained the equivalent of approximately 100 embryos each. The strips were soaked in TBS (3% dried milk in 500 mM NaCl, 20 mM Tris, pH 7.5) and then were washed in TBS with 0.05% Tween-20. The primary antibodies were added to this solution and the strips were incubated for four hours. After washing, secondary antibodies conjugated to alkaline phosphatase were applied for one hour, followed by additional washes and color development with p-nitrophenyl phosphate.
Results
Western blotting
The specificities of the two antibodies were tested by Western blot analysis of proteins solubilized from 36 h embryonic zebrafish. As shown in Fig. 1, the mono-clonal antibody, 4D9, recognized three bands at 47, 41, and 39 ×103Mr. The polyclonal antibody, αEnhb-1, recognized two of these bands, at 47X103 and 39x103Mr, and an additional smear which extended from about 38 to 44 ×103Mr.
The polyclonal and monoclonal antibodies recognize common and unique bands on Western blots. Western blots of proteins extracted from 36 h zebrafish embryos. Lane A, probed with 4D9. Lane B, probed with αEnhb-1. Lane C, control with no primary antibodies. The faint, lower molecular weight bands in lane A were inconsistently seen and probably represent degradation products. The migration of molecuar weight standards are indicated on the left.
The polyclonal and monoclonal antibodies recognize common and unique bands on Western blots. Western blots of proteins extracted from 36 h zebrafish embryos. Lane A, probed with 4D9. Lane B, probed with αEnhb-1. Lane C, control with no primary antibodies. The faint, lower molecular weight bands in lane A were inconsistently seen and probably represent degradation products. The migration of molecuar weight standards are indicated on the left.
The molecular weights of the highest and lowest bands are somewhat higher than predicted from the sequence of the two zebrafish genes; the Eng-1 protein is predicted to be 26 ×103Mr and the Eng-2 protein is predicted at 30 ×103. Mr(data not shown). These discrepencies between predicted and observed molecular weights are consistent with results obtained from Western blot analysis of Engrailed proteins in mouse, chick, and Xenopus where the predicted molecular weights are also lower (Davis et al. 1991). The discrepencies could result from post-translational modifications and/or anomalous mobilities during SDS-PAGE.
Complexity of Eng expression
The developmental stages and tissues that were examined in detail summarized in Fig. 2. We observed that the 4D9 and αEnhb-1 antibodies exclusively label cell nuclei, as expected for the localization of homeoprotein antigens. As described in detail below, the labeled cell types only partially overlapped: both antibodies labeled the same stripe of cells at the midbrain-hindbrain junction, the somitic cells previously described by Patel et al. (1989), and the jaw muscles described by Hatta et al. (1990). However, only 4D9 labeled a cartilage of the pectoral girdle, pectoral fin epithelium, teeth and the otic vesicle. Further, αEnhb-1 labeled cells in the hindbrain and spinal cord that were unlabeled by 4D9. These patterns argue that more than two eng gene products are present in zebrafish, and that they are expressed in a complex pattern in different cell types (see Discussion).
Temporal expression of Eng (recognized by 4D9 and αEnhb-1 labeling) differs in diverse tissues. Filled and open circles show the stages when expression was or was not observed, respectively. At the adult stage only the head was examined.
A stripe at the midbrain-hindbrain boundary
The first Eng expression to appear during development is a transverse stripe of cells across the brain primordium (Fig. 3A), as previously reported (Njølstad and Fjose, 1988; Patel et al. 1989). Expression begins at 10 –11 h, when the embryo has 1 –2 somites. The position of the stripe corresponds to one of the slight bumps, previously undescribed, that appear at the dorsal surface of the early neural primordium (Fig. 3B). The posterior border of expression is oriented in the transverse plane, and at early stages the anterior border is obliquely oriented, such that the stripe is wider dorsally than ventrally (Fig. 3B). Later the borders gradually become parallel. A deep furrow, forming the midbrain-hindbrain boundary, becomes visible in the middle of the stripe about nine hours later (at 20 h; Fig. 3C and D). The posterior limit of the stripe is near the center of the first hindbrain neuromere (rhombo-mere Rol; see Hanneman et al. 1988), similar to that observed in the mouse by Davis et al. (1991). The stripe’s anterior limit lies in the caudal midbrain. Both boundaries appear diffuse at the single-cell level, and fade gradually within a narrow transitional zone. In embryonic stages, the labeling is present in every CNS cell within the confines of the stripe, including the floor plate (Fig. 3D-F). Labeling continues through the adult stage, and becomes restricted to cells in particular groups including parts of the cerebellum, the optic tectum and certain midbrain tegmental nuclei (Fig. 3G-I; Fig. 4).
Expression of Eng (revealed with 4D9) in the developing central nervous system. Dorsal (A,C,D) and lateral (B) views of whole mounts. (A) The first detected expression of Eng at 10 –11 h. At this stage, the intensity of labeling is heterogeneous among cell nuclei in the CNS stripe. (B) 14h. The anterior boundary of the expressing stripe is inclined at this stage. At the dorsal surface, the expression corresponds to an oblique swelling. (C) 17 h. By this stage the neural tube is formed, but morphological subdivisions are not yet present. (D) 22 h. The stripe of Eng is now divided by the furrow separating the midbrain and hindbrain. (E,F) 28h. A single whole-mounted embryo photographed from the lateral aspect at different focal planes, from lateral to medial. E shows labeling of segmental subsets of cells within the anterior three segments (rhombomeres) of the hindbrain (arrow) and the precursor mesenchyme of two jaw muscles (levator arcus palatini and dilator operculi, arrowhead). Boundaries between midbrain (m) and hindbrain (h) and between rhombomeres are shown by bars. In F, note that the floor plate (arrow) is labeled within the CNS stripe at the midbrain-hindbrain junction. (G,H) 72 h. Horizontal sections of the brain at dorsal (G) and ventral levels (H). Labeling is bilaterally symmetrical in the.midbrainhindbrain stripe. In the Ro segments, a small cluster (arrow in G) and a large column of cells (arrow in H) can be seen. (I) 72 h. Sagittal section. In addition to the midbrain-hindbrain stripe, there are zones of weak expression in the forebrain (f). Bar=50 μm.
Expression of Eng (revealed with 4D9) in the developing central nervous system. Dorsal (A,C,D) and lateral (B) views of whole mounts. (A) The first detected expression of Eng at 10 –11 h. At this stage, the intensity of labeling is heterogeneous among cell nuclei in the CNS stripe. (B) 14h. The anterior boundary of the expressing stripe is inclined at this stage. At the dorsal surface, the expression corresponds to an oblique swelling. (C) 17 h. By this stage the neural tube is formed, but morphological subdivisions are not yet present. (D) 22 h. The stripe of Eng is now divided by the furrow separating the midbrain and hindbrain. (E,F) 28h. A single whole-mounted embryo photographed from the lateral aspect at different focal planes, from lateral to medial. E shows labeling of segmental subsets of cells within the anterior three segments (rhombomeres) of the hindbrain (arrow) and the precursor mesenchyme of two jaw muscles (levator arcus palatini and dilator operculi, arrowhead). Boundaries between midbrain (m) and hindbrain (h) and between rhombomeres are shown by bars. In F, note that the floor plate (arrow) is labeled within the CNS stripe at the midbrain-hindbrain junction. (G,H) 72 h. Horizontal sections of the brain at dorsal (G) and ventral levels (H). Labeling is bilaterally symmetrical in the.midbrainhindbrain stripe. In the Ro segments, a small cluster (arrow in G) and a large column of cells (arrow in H) can be seen. (I) 72 h. Sagittal section. In addition to the midbrain-hindbrain stripe, there are zones of weak expression in the forebrain (f). Bar=50 μm.
Expression of Eng (4D9) in the central nervous system in older zebrafish. (A) 72 h. Horizontal section showing the midbrain-hindbrain stripe at higher magnification. Heavily stained bilateral clusters of cells are present. (B-D) 21 d. Transverse sections through the midbrain and hindbrain. (B) The caudal midbrain. Strong expression is observed in the torus longitudinalis (tl), the valvula of the cerebellum (c), the nucleus lateralis valvulae (nlv), and in a subpopulation of cells in the optic tectum (ot). The large torus semicircularis (ts) is unlabeled. (C) Caudal to B in the main body of the cerebellum (c), which is mostly unlabeled this far caudally. Strong expression is present in the optic tectum and some nuclei (tentatively the nucleus isthmii (ni)). (D) Slightly caudal to the level of C. Two cell groups near the trigeminal nerve (V) express Eng at the location of the nucleus subeminentialis (nse) (Finger, 1978), a nucleus of interneurons that receives a projection from the cerebellum. Another possibility is that it may be a vestibular nucleus. For definitive identification of these nuclei, further anatomical experiments will be required. Bar=50 μm.
Expression of Eng (4D9) in the central nervous system in older zebrafish. (A) 72 h. Horizontal section showing the midbrain-hindbrain stripe at higher magnification. Heavily stained bilateral clusters of cells are present. (B-D) 21 d. Transverse sections through the midbrain and hindbrain. (B) The caudal midbrain. Strong expression is observed in the torus longitudinalis (tl), the valvula of the cerebellum (c), the nucleus lateralis valvulae (nlv), and in a subpopulation of cells in the optic tectum (ot). The large torus semicircularis (ts) is unlabeled. (C) Caudal to B in the main body of the cerebellum (c), which is mostly unlabeled this far caudally. Strong expression is present in the optic tectum and some nuclei (tentatively the nucleus isthmii (ni)). (D) Slightly caudal to the level of C. Two cell groups near the trigeminal nerve (V) express Eng at the location of the nucleus subeminentialis (nse) (Finger, 1978), a nucleus of interneurons that receives a projection from the cerebellum. Another possibility is that it may be a vestibular nucleus. For definitive identification of these nuclei, further anatomical experiments will be required. Bar=50 μm.
Segmentally arranged cells in the hindbrain and spinal cord
At 22 –24 h, a small number of cells positioned laterally in the hindbrain and spinal cord begin to express Eng. In contrast to the labeled stripe described above, the two antibodies reveal different, and only partially overlapping patterns. At every stage, 4D9-positive cells are strictly confined to the anterior three rhombomeres (Fig. 3E). They are at first somewhat scattered but concentrated in the rhombomere centers (see Trevarrow et al. 1990) in a pattern that is thus segmental and approximately bilaterally symmetrical. At 36h-48h, loose clustering of the cells in the second and third rhombomeres is more apparent, although the exact distributions of the labeled cells may differ among individual embryos. By 72 h, these cell groups have fused, the segmental distribution is gone and cells expressing Eng form new compact clusters; one large cell cluster in which the cell arrangement is columnar, and a smaller more dorsoanterior cluster (Fig. 3G&H; Fig. 4D). The groups are located dorsolateral to the trigeminal (fifth cranial) motor nuclei. Double staining experiments in 84 h embryos, in which cranial motoneurons were retrogradely labeled with horseradish peroxidase and Eng-positive cells were labeled with 4D9, suggested that the cells are not motoneurons, but dorsolateral to them (data not shown). The Engpositive cells are closely associated with the fibers of the trigeminal nerve. Further studies will be required to identify these cells (see the legend to Fig. 4 concerning their possible identification).
In contrast, αEnhb-1 labels CNS cells at an intermediate dorsoventral location through out the length of the hindbrain and spinal cord (Fig. 5). αEnhb-1 labels cells in the first three rhombomeres (where 4D9 labeling is observed) less intensely than cells more posterior. At 28 h, as revealed by double-labeling with both antibodies only the more lateral cells (near the brain’s pial surface) labeled by αEnhb-1 are also labeled by 4D9 (data not shown). Thus, some of the cells in these rhombomeres express a form of Eng that is recognized specifically by αEnhb-1 (See Discussion, Fig. 11).
Labeling pattern of αEnhb-1. (A) 22 h wholemount, dorsal view. The stripe at the midbrain and hindbrain is labeled similarly to 4D9. Boundaries between midbrain and hindbrain rhombomeres are indicated by bars. Cells in the first 3 rhombomeres (Rol-3) are unlabeled at this stage while cells caudal to these rhombomeres are strongly stained. Early expression in the spinal cord (arrows) shows a segmental pattern. Muscle pioneers (mp) and surrounding cells (See Fig. 6) are also labeled in the myotome and arranged segmentally. (B) Transverse section through the spinal cord at the same stage. The labeled cells (arrow) are near the pial surface and at an intermediate dorsoventral position. (C) 28 h whole-mount. The number of labeled cells has increased and the weaker staining extends to the anterior hindbrain segments. (D) Double staining of a 28 h embryo with αEnhb-1 and 3A10 which labels some reticular spinal neurons including Mauthner cells (Mcell). No cells are doubly stained. (E,F) Transverse sections through 28 h embryos at spinal cord (E) and hindbrain (F) levels. 3A10 labeled some axons running through the dorsolateral longitudinal axonal fascicles (DLF), which correspond to the lateral longitudinal fascicles (LLF) in the hindbrain, and the ventral-most medial longitudinal fascicles (MLF), (detailed staining pattern is in preparation). The αEnhb-1 positive cells (arrows) are closely associated with the dorsoventral level of DLF and LLF. Bar=50 μm.
Labeling pattern of αEnhb-1. (A) 22 h wholemount, dorsal view. The stripe at the midbrain and hindbrain is labeled similarly to 4D9. Boundaries between midbrain and hindbrain rhombomeres are indicated by bars. Cells in the first 3 rhombomeres (Rol-3) are unlabeled at this stage while cells caudal to these rhombomeres are strongly stained. Early expression in the spinal cord (arrows) shows a segmental pattern. Muscle pioneers (mp) and surrounding cells (See Fig. 6) are also labeled in the myotome and arranged segmentally. (B) Transverse section through the spinal cord at the same stage. The labeled cells (arrow) are near the pial surface and at an intermediate dorsoventral position. (C) 28 h whole-mount. The number of labeled cells has increased and the weaker staining extends to the anterior hindbrain segments. (D) Double staining of a 28 h embryo with αEnhb-1 and 3A10 which labels some reticular spinal neurons including Mauthner cells (Mcell). No cells are doubly stained. (E,F) Transverse sections through 28 h embryos at spinal cord (E) and hindbrain (F) levels. 3A10 labeled some axons running through the dorsolateral longitudinal axonal fascicles (DLF), which correspond to the lateral longitudinal fascicles (LLF) in the hindbrain, and the ventral-most medial longitudinal fascicles (MLF), (detailed staining pattern is in preparation). The αEnhb-1 positive cells (arrows) are closely associated with the dorsoventral level of DLF and LLF. Bar=50 μm.
As in the anterior rhombomeres, the cells labeled by αEnhb-1 in the more posterior rhombomeres and spinal cord appear segmentally organized (Fig. 5A,B). The segmental pattern is clearest shortly after expression begins; by 18 h in the caudal hindbrain and in several anterior segments of spinal cord, one to a few cells per hemisegment are lightly stained (data not shown). At 22 h this pattern expands anteriorly to rhombomere 3 (Ro3). In the spinal cord, more cells are labeled at this stage than at 18 h, and it is clear from the pattern that expression appears in an anterior-posterior sequence, such that younger segments contain one or two cells and older segments have additional cells, filling the space between the first ones (Fig. 5A,B). By 30 h (Fig. 5C) there are still segmentally arranged clusters of labeled cells in the rhombomeres, but in the spinal cord the labeled cells are in a nearly continuous longitudinal row and segmentation can no longer be discerned. This sequence, a segmental pattern giving way to a more columnar one, is similar to that described previously for the early development of spinal neurons located ventrally to the Eng-positive cells (Hanneman et al. 1988).
From the intermediate dorsoventral location of the labeled nuclei near the dorsolateral fascicle in the lateral mantle zone of the neural tube (Fig. 5E,F), it seems likely that the spinal Eng-positive cells are interneurons. Motoneurons are more ventral and Rohon-Beard sensory neurons are more dorsal. The nuclei of radial glial cells are located near the central canal in the spinal cord and the fourth ventricle in the hindbrain, not near the pial surface. Previously characterized hindbrain reticulospinal interneurons (Mendelson, 1986) and spinal CoPA interneurons (Kuwada et al. 1990) occupy positions near the cells expressing Eng, but are not identical to them, as revealed by double-antibody labeling with αEnhb-1 and 3A10, a monoclonal antibody that labels the identified interneurons selectively at the stages of interest (Fig. 5E,F). Thus, the specific identities of the Engexpressing cells remain unknown.
Muscle pioneers and neighboring cells
After somites form, a small subset of cells in each begins to expresses Eng, as in Fig. 6 and Patel et al. (1989). Expression starts at 13 –14h when the embryo has 8 –10 somites, and is highest at this stage in somites 5 and 6. Somites are added successively, about two per hour (Hanneman and Westerfield, 1989) in a posteriorly directed wave. At stages after Eng expression begins, we observed that the youngest (most posterior) somite to express Eng is always located two to four somites anterior to the somite most recently formed. Thus, except for the first several somites, in which expression seems to begin simultaneously, Eng expression in the somites develops in the same anterior-posterior wave in which somites form, and arises in a particular somite one to two hours after that somite forms.
Expression of Eng (4D9) in muscle pioneers and surrounding cells in the somites. (A) Nuclei expressing Eng at 17h are positioned at the anterior edge of the medial surface of the myotome at the level of the notochord (nt). The shape of one myotome is shown by the dotted line and the arrowheads indicate the dorsal (up) and ventral extends of the notochord, located deep to the myotome. (B) At 30 h the cells strongly expressing Eng are elongated and divide the myotome into dorsal and ventral halves. The cells surrounding the muscle pioneers also begin to express Eng weakly. (C) zn-5 labeling at 30 h, distinctively outlining the muscle pioneers. (D) Double staining of Eng and zn-5 antigens at 30 h, showing the cells strongly outlined by zn-5 contain nuclei (arrow) which strongly express Eng. Bar=50 μm.
Expression of Eng (4D9) in muscle pioneers and surrounding cells in the somites. (A) Nuclei expressing Eng at 17h are positioned at the anterior edge of the medial surface of the myotome at the level of the notochord (nt). The shape of one myotome is shown by the dotted line and the arrowheads indicate the dorsal (up) and ventral extends of the notochord, located deep to the myotome. (B) At 30 h the cells strongly expressing Eng are elongated and divide the myotome into dorsal and ventral halves. The cells surrounding the muscle pioneers also begin to express Eng weakly. (C) zn-5 labeling at 30 h, distinctively outlining the muscle pioneers. (D) Double staining of Eng and zn-5 antigens at 30 h, showing the cells strongly outlined by zn-5 contain nuclei (arrow) which strongly express Eng. Bar=50 μm.
We identified the somitic cells that express Eng as muscle pioneers (Figs 5A,B,E and 6). Muscle pioneers are the first muscle cells to differentiate within the myotome that develops from each somite, and they can be recognized by their characterisitic cell shape and their position at the center of the chevron-shaped myotome (Felsenfeld and Curry, in preparation). The muscle pioneers also express distinctively, but not uniquely, acetylcholine esterase activity (Hanneman and Westerfield, 1989) and an antigen recognized by the monoclonal antibody zn-5 (Fig. 6C), that is present transiently and apparently on the cell surfaces of a number of cell types in the embryo (Trevarrow et al. 1990; Hatta et al. 1991; and unpublished). The cells in the myotome that strongly express Eng also express the zn-5 antigen, as revealed by double-labeling experiments (Fig. 6D). Eventually each myotome has 2 –6 muscle pioneers, although the exact number varies.
Expression of Eng in the developing muscle pioneers permits characterization of their morphogenesis, including changes in both cell shape (see also Felsenfeld and Curry, in preparation) and arrangement. They first are epithelial in shape, and are positioned at the medial surface of the somite, facing the notochord, and at the anterior margin of the somite. The muscle pioneers change shape, first elongating in the posterior direction, so as to span the segment. Accompanying this change, the somite changes from a cuboidal shape to the characteristic chevron-shape of a myotome, with the muscle pioneers occupying the apex of the chevron. The elongated muscle pioneers then flatten in the horizontal plane, and move laterally to invade the myotome, finally dividing it into dorsal and ventral halves. At their location the horizontal myoseptum eventually forms as a connective tissue partition separating the dorsal and ventral muscle masses of each myotome.
Eventually (by 20 h in trunk segments) cells surrounding the muscle pioneers also express Eng (Fig. 6B). Their expression remains at a lower level than that of the muscle pioneers, such that the separate populations can be distinguished with either 4D9 or αEnhb-1. At least some of these other cells develop as muscle, as revealed by polarizing light (not shown). The immediate neighbors of the muscle pioneers are the first cells to follow them in expressing Eng.
Altered expression in trunk myotomes of spt-1 mutants
The correlation just described, between morphogenesis of the muscle pioneers and the formation of the horizontal subdivision of the myotome can be interpreted to mean that the muscle pioneers are themselves responsible for establishing this morphological dorsoventral boundary (see Discussion). A mutation, spt-l(b104), or ‘spadetail’, allows further examination of this proposal, since spt-1 severely perturbs somitogenesis in the trunk, and the myotomes that eventually form are disorganized and lack horizontal myosepta (Kimmel et al. 1989). Changes in the muscle pioneers, including their expression of Eng, in spt-1 might underlie the disturbance in organogenesis. At 22 h, Eng is strongly expressed in the trunk myotomes of wild-type embryos (Fig. 7A). Although some poorly formed myotomes are present in the trunk of spt-1 homozygotes at this stage, we could not detect Eng expression within them (Fig. 7B). Nor could we recognize the muscle pioneers by their characteristic shapes, or by labeling with zn-5 (data not shown). These observations suggest that the muscle pioneers fail to develop in the trunk segments of mutants rather than that they are present but fail to express Eng. The defect is specific to this population of muscle cells, since other muscles develop in the mutant trunk myotomes (Kimmel et al. 1989; Eisen and Pike, 1991).
Expression pattern of Eng (4D9) in spt-1 mutant embryos reveals that muscle pioneers are missing specifically in the trunk. (A) Wild-type sibling at 22 h. (B) spt-1 mutant at 22h. In the mutant at this stage, no labeling is observed within the trunk or tail. The midbrain-hindbrain stripe appears unperturbed. (C) spt-1 mutant at 30h. Muscle pioneers appear normal in the tail (arrowheads), but remain absent in the trunk, even though many other muscle cells have differentiated in the trunk by this stage (see text). Bar=100 μm.
Expression pattern of Eng (4D9) in spt-1 mutant embryos reveals that muscle pioneers are missing specifically in the trunk. (A) Wild-type sibling at 22 h. (B) spt-1 mutant at 22h. In the mutant at this stage, no labeling is observed within the trunk or tail. The midbrain-hindbrain stripe appears unperturbed. (C) spt-1 mutant at 30h. Muscle pioneers appear normal in the tail (arrowheads), but remain absent in the trunk, even though many other muscle cells have differentiated in the trunk by this stage (see text). Bar=100 μm.
Other aspects of Eng expression are normal in spt-1, in keeping with previous observations that the defects are limited to the trunk. In particular, somites in the tails of mutants form on schedule, and they give rise to normal-looking myotomes (Kimmel et al. 1989). The spt-1 tail myotomes have muscle pioneers that express Eng by 30 h (Fig. 7C). Recovery of muscle pioneers in the trunk fails to occur as late as 48 h, but we did occasionally see one or more weakly labeled myotomal cells that resembled those normally surrounding the muscle pioneers. The expression patterns in the head, including the stripe at the hindbrain-midbrain junction (Fig. 7B), the clusters in the Ro hindbrain segments, and the jaw muscle precursors (Hatta et al. 1990) are normal, spt-1 variably has pectoral fins, and, when the fins form, their expression of Eng is normal as determined by labeling mutants at 48 h (data not shown), compared to the wild type as described in the next section.
Ventral-anterior half of the pectoral fin
At 26 h a restricted set of epidermal cells in the anlagen of the pectoral fin begin to express Eng (Fig. 8A). At this stage, the fin bud is dome-shaped and Eng expression is limited to the ventral-anterior half of the fin bud. The boundary of the expression on the dome is straight and passes obliquely (slanting dorsoanteriorly) across the fin bud and divides it into ventral-anterior and dorsal-posterior halves. This boundary later follows the ridge of the flattening fin (Fig. 8B), which is the equivalent of the apical ectodermal ridge (AER) of other vertebrates (Wood, . 1982). This expression, revealed only by the 4D9 antibody and not by αEnhb-1 in zebrafish, becomes stronger and is most obvious at 2 to 3d, persisting at least until 4d. During this period, it remains restricted to half of the epidermis, an epithelial monolayer positioned between an outer epithelial periderm and underlying mesenchyme. The epidermal cell nuclei are uniformly labeled within the domain of expression, rather than in a graded fashion as for some other homeoproteins (see Discussion) and the boundary of expression at the AER is sharp. When fish swim, they appear to keep the Eng-positive surface of the pectoral fin down.
Expression of Eng (4D9) in the epidermis of the pectoral fin. (A) Lateral view of a whole-mount at 28 h. Early Eng expression is within the epidermis of the dome-shaped fin bud, and is limited to its ventral-anterior half. The apical epithelial ridge (AER) has not yet appeared; the folding that produces it occurs at the Eng expression boundary (arrowheads). (B) Horizontal section at 36 h. Strong expression is restricted to the ventral (anterior in this section) half of the epidermis. In the dorsal (posterior) half, little to no expression was observed. The arrowhead indicates the boundary within the AER. (C) Schematic representation of Eng expression during early development of the fin. Black areas represent the places where epidermal cell nuclei express Eng. The pectoral fin of the middle fish is flipped up to show the ventral surface, and that of the lower is flipped down to show the dorsal surface. Abbreviations; A, anterior; P, posterior; D, dorsal; V, ventral. Bar=50 μm.
Expression of Eng (4D9) in the epidermis of the pectoral fin. (A) Lateral view of a whole-mount at 28 h. Early Eng expression is within the epidermis of the dome-shaped fin bud, and is limited to its ventral-anterior half. The apical epithelial ridge (AER) has not yet appeared; the folding that produces it occurs at the Eng expression boundary (arrowheads). (B) Horizontal section at 36 h. Strong expression is restricted to the ventral (anterior in this section) half of the epidermis. In the dorsal (posterior) half, little to no expression was observed. The arrowhead indicates the boundary within the AER. (C) Schematic representation of Eng expression during early development of the fin. Black areas represent the places where epidermal cell nuclei express Eng. The pectoral fin of the middle fish is flipped up to show the ventral surface, and that of the lower is flipped down to show the dorsal surface. Abbreviations; A, anterior; P, posterior; D, dorsal; V, ventral. Bar=50 μm.
Cartilage of the pectoral girdle, sensory organs and teeth
Expression of Eng is evident in other tissues, including cartilage, sensory epithelia and teeth (Fig. 2). For example all of the cells of a single cartilage in the pectoral girdle (identified as the coracoid based on a generalized description of Teleostei by Harder, 1975) are labeled with 4D9 at 3 and 4d; other nearby cartilages are unlabeled (Fig. 9A). Sensory epithelial cells of some taste buds are variably labeled at 21 d and in adults (Fig. 9B). The sensory maculae of the otic vesicle are labeled at 21 d (weak expression begins at 3 d). In contrast to the labeling pattern in the taste buds, the labeling is in cells surrounding the area that contains the hair cells, not the sensory cells (hair cells) themselves, and forms a clear boundary within this epithelial structure (Fig. 9C). Epithelial cells in immature teeth at 21 d, express Eng. Some of these cells become mesenchymal in well-developed teeth (Fig. 9D). Other weak expression was also observed at the tips of the jaws, the tips of the gills, the roof of the mouth cavity, and some connective tissues (data not shown).
Expression of Eng (4D9) in diverse cell types. (A) 3d. Horizontal section. Cartilage of the pectoral girdle (coracoid, arrow). Cartilages of the gill arches are also shown in this section, but they are unlabled. (B) 5 months. Sensory epithelial cells of taste buds are labeled (arrow). (C) 3 weeks. Otic vesicle in a transverse section. Engexpression is restricted to cells that surround the sensory hair cells. Expression stops at a sharp boundary (arrowhead). (D) Teeth at 21 d in sagittal section. The epithelial cells in both the outer (ameloblast) and inner (odontoblast) layers of the developing teeth (left and center) and mesenchymal (odontoblast) cells in more mature teeth (right) express Eng. Bar=50 μm.
Expression of Eng (4D9) in diverse cell types. (A) 3d. Horizontal section. Cartilage of the pectoral girdle (coracoid, arrow). Cartilages of the gill arches are also shown in this section, but they are unlabled. (B) 5 months. Sensory epithelial cells of taste buds are labeled (arrow). (C) 3 weeks. Otic vesicle in a transverse section. Engexpression is restricted to cells that surround the sensory hair cells. Expression stops at a sharp boundary (arrowhead). (D) Teeth at 21 d in sagittal section. The epithelial cells in both the outer (ameloblast) and inner (odontoblast) layers of the developing teeth (left and center) and mesenchymal (odontoblast) cells in more mature teeth (right) express Eng. Bar=50 μm.
Discussion
We have examined labeling patterns in developing zebrafish obtained with two antibodies that recognize engrailed-like homeodomains. As so determined, Eng expression begins at separate stages (Fig. 2) in a variety of different cell types, and in different patterns in each tissue, suggesting that the zebrafish eng genes function in a number of developmental programs. However, as we discuss below, there seem to be a few ‘basic plans’ of expression that are shared by several cell types. These are interesting to consider with respect to the regulation of Eng during development.
Complexity of Eng homeoproteins in zebrafish
We were surprised to find that the two antibodies recognized only partially overlapping sets of cells. Both antibodies recognize epitopes within Engrailed-type homeodomains (Patel et al. 1989; Davis et al. 1991). Zebrafish, like other vertebrates, have at least two engrailed genes, eng-1 and eng-2 (Fjose et al. 1988; Holland and Williams, 1990). The deduced protein sequences of their homeodomains are nearly identical and the remainders of the proteins are also very similar (M.W., unpublished). Thus we expected both antibodies to label tissues identically, yet as summarized in Fig. 10 for one embryonic stage, they did not. Several explanations are possible to account for our findings. Presently we have no direct evidence that all the labeling we observe is due to Eng homeoproteins; some may come from cross-reacting, unrelated antigens. However, Eng expression is highly conserved among vertebrates (Patel et al. 1989), and many of the tissues that are labeled by one or the other antibodies were also labeled in in situ studies in the mouse (Davis et al. 1988; Davidson et al. 1988). This suggests that the antibodies are labeling Eng.
Schematic comparison of Eng expression revealed by 4D9 and αEnhb-1 at 30 h, and presented as a fillet such that the CNS is viewed from the dorsal aspect, and the myotomes and pectoral fin bud from the lateral aspect. Strongly labeled nuclei are shown by filled circles and more weakly labeled ones by shaded circles. Both antibodies similarly label the midbrainhindbrain stripe, the myo tomes, and two jaw muscle precursors (LAP & DO; see Hatta et al. 1990). In addition 4D9 specifically reveals expression in the anteriorventral pectoral fin and in a subset of neurons restricted to the 3 Ro rhombomeres. αEnhb-1 weakly labels the same cells and additionally more medial cells in the same segments, and strongly labels subsets of cells beginning in the next rhombomere (Mil) and continuing throughout the length of the hindbrain and spinal cord.
Schematic comparison of Eng expression revealed by 4D9 and αEnhb-1 at 30 h, and presented as a fillet such that the CNS is viewed from the dorsal aspect, and the myotomes and pectoral fin bud from the lateral aspect. Strongly labeled nuclei are shown by filled circles and more weakly labeled ones by shaded circles. Both antibodies similarly label the midbrainhindbrain stripe, the myo tomes, and two jaw muscle precursors (LAP & DO; see Hatta et al. 1990). In addition 4D9 specifically reveals expression in the anteriorventral pectoral fin and in a subset of neurons restricted to the 3 Ro rhombomeres. αEnhb-1 weakly labels the same cells and additionally more medial cells in the same segments, and strongly labels subsets of cells beginning in the next rhombomere (Mil) and continuing throughout the length of the hindbrain and spinal cord.
Alternatively more than two forms of Eng proteins may be present in zebrafish. Spinal cord cells, labeled selectively in zebrafish by αEnhb-1, specifically express En-1 in the mouse (Davis et al. 1988; Davidson et al. 1988; Davis et al. 1991). Fin bud cells, labeled here selectively by 4D9, are the homologues of mouse forelimb cells that also specifically express En-1. Thus the simplest explanation of our finding is that the zebrafish homologue of this gene, eng-1, codes a product present in several forms after some type of post-transcriptional modification. Alternatively, more than two eng genes might be present. This interesting issue, and whether we are detecting unrelated crossreacting angitens, might be resolved by in situ studies using probes selected to distinguish the eng-1 and eng-2 transcripts.
On the basis of Western blot analysis both antibodies, 4D9 and αEnhb-1, appear to recognize Eng proteins in zebrafish, although there are differences in the bands that each detects. The monoclonal antibody, 4D9, recognizes three distinct bands on Western blots. At least two of these bands probably correspond to the proteins encoded by the eng-1 and eng-2 genes since both gene products contain the epitope (Patel et al. 1989) recognized by the 4D9 antibody. The third band might correspond to an alternate form of one of the two gene products or a related gene product. Recognition of both eng gene products by the monoclonal antibody differs from results obtained in mouse and chick where only the En-2 gene encodes the 4D9 epitope (Davis et al. 1991). The polyclonal antibody, αEnhb-1, on the other hand, recognizes only two of these bands, one appearing as a smear. These results suggest that both antibodies recognize at least two common proteins, which possibly represent the primary eng gene products, and in addition each recognizes at least one other protein which is specific to the antibody.
Eng marks positions of future subdivisions of the brain and myotomes
In two independent systems, expression of Eng is associated with the subdivision of an organ primordium. In the CNS, the earliest expression occurs at the future midbrain-hindbrain boundary, as previously reported (Patel et al. 1989). The position of this stripe and its early expression are highly conserved among vertebrates: it appears when the zebrafish has 1 –2 somites, when the mouse has 1 somite and the chicken has 3 –4 somites (Davis et al. 1988), and before somites appear in Xenopus (Hemmati-Brivanlou and Harland, 1989).
Eng is expressed in a small population of cells in each somite (Patel et al. 1989). We found that these cells differentiate as muscle pioneers (Felsenfeld and Curry, in preparation) and occupy the position where the myotome splits horizontally to form dorsal and ventral muscle masses separated by a myoseptum. Thus, both in brain and myotomes, Eng expression occurs where a furrow appears later; its expression could be important in specifying the positions of the furrows.
We examined labeling in spt-1 mutant embryos to obtain evidence pertinent to this idea for one of the tissues, spt-1 autonomously disrupts gastrulation movements of trunk muscle precursors (Ho and Kane, 1990) producing myotomes that are poorly patterned and lack horizontal myosepta. We observed that they also lack muscle pioneers, the cells that normally express Eng at the position of this boundary. In the tail, where myotomes are properly patterned, Eng-expressing muscle pioneers are present, and another homeoprotein, the putative product of hox-3.3, is expressed in mutant trunk somites (Molven et al. 1990), revealing that the block of gene expression in this body region is specific to Eng. Thus, the absence of muscle pioneers in the poorly patterned trunk myotomes supports our argument that cells expressing Eng in the somite play an important role for the dorsoventral subdivision of the myotome. This role could be morphogenetic; we have seen these cells actively undergo changes in shape and arrangement where the boundary forms.
Expression in the fin may define ventral identity and a compartment in the epidermis
In the pectoral fin bud, before the folding and the AER appear, only the ventral-anterior half of the epidermis expresses Eng. The boundary is sharply coincident with the obliquely oriented AER. In the mouse, En-1 is expressed in the ventral half of the limb buds (Davis et al. 1991). Although the AER is oriented horizontally in the mouse, the expression domains relative to the AER are basically the same in the two species, suggesting that they may be generally conserved among vertebrates.
Several hox homeoproteins (Oliver et al. 1989; Dolle et al. 1989) have gradients of expression in the limb buds of drier vertebrates, and in the pectoral fin-bud in zebrafish (Molven et al. 1990). Eng expression is distinct in that it is restricted to the epidermis, and within its domain, appears uniform, not gradient-like. Thus, we suggest that Eng expression plays a unique role, important for defining differences between the dorsal and ventral regions of the fin and the boundary where the AER forms. It will be of interest to examine whether cell lineages are restricted at these compartment-like boundaries, as well as the possible interactions between Eng proteins in the epidermis and hox proteins in the epidermis or mesoderm of the bud.
Segmental patterning of Eng expression in the CNS
In overtly segmented regions of the CNS (Kimmel et al. 1988, 1991), we observed three patterns of Eng expression that relate to segmentation. First, the stripe of Eng labeling at the midbrain-hindbrain boundary may represent specification of a segment boundary. The stripe is approximately one-half segment wide on either side of the boundary, which may indicate the existence of a half-segment unit of patterning (see Trevarrow et al. 1990). Possibly, as in Drosophila, units defined by gene expression (parasegments, Martinez-Arias and Lawrence, 1985; Lawrence, 1989) are phase-shifted relative to segments defined morphologically (Hanneman et al. 1988). The finding that Eng expression corresponds to an early morphological swelling supports this speculation. Alternatively the stripe of expression might define a previously unrecognized rhombomere (see Lumsden, 1990).
In two locations Eng expression is at first segmental, but then the segmental patterning disappears. In particular, 4D9 labels isolated sets of cells in the first three rhombomeres, some of which subsequently appear to migrate together and form compact clusters. This morphogenetic behavior, aggregation and packing before cell differentiation, is very similar to that previously described for the mesodermal precursors of two jaw muscles that initially lie just outside of these same brain segments, and which also express Eng (Hatta et al. 1990).
The αEnhb-1 antibody reveals the other example where a segmental pattern is first present and then lost. Cells that express Eng initially appear in the centers of hindbrain segments (as defined relative to the rhombomere boundaries) and spinal segments (as defined relative to the positions of muscle pioneers beside the spinal cord; see Fig. 5A). Other cells then are added to the pattern, and obscure segmentation. It could be that Eng is specifying different kinds of interneurons that develop in a time-dependent fashion, beginning in the segment centers and extending towards the borders. We note that the very first neurons to develop, the so-called primary neurons (Kimmel and Westerfield, 1990), do not express Eng. This population includes hindbrain reticulospinal neurons (Mendelson, 1986) and spinal CoPA neurons (Kuwada et al. 1990) that closely neighbor the Eng-positive cells after the latter appear.
The observation that the labeling patterns of 4D9 and αEnhb-1 partially overlap in the developing brain suggests a way that specification of neuronal identity and the formation of functional neuronal groups, such as brain nuclei, may occur (Fig. 11). The cells stained by αEnhb-1 are likely to be restricted mainly by two types of positional information, dorsoventral information and segmentally repeating information in both the hindbrain and spinal cord, as discussed above. 4D9 labels only a subset of the cells labeled by αEnhb-1, found only in the anterior-most segments, and only the 4D9-labeled cells later rearrange to form the distinctive brain nuclei. Two additional informational signals could define early positions of these cells: information in the anterior-posterior axis would determine in which segments they appear, and information in the lumenopial axis would determine where in the segment they appear. Timing might be involved since, as described above, primary neurons do not express Eng, and since lumeno-pial gradients in neuronal generation are well known (reviewed in Jacobson, 1978). The identity of some of the brain nuclei thus may be partially specified by expression of nuclear proteins like Eng and dependent at early stages on a relatively simple, threeaxis system of coordinates such as shown in Fig. 11.
A model of the specification of brain nuclei dependent on the early coordinates along three axes and segmentation. In highly schematic drawings of Eng expression in the first 6 rhombomeres of hindbrain, both in transverse and lateral views, the single circles represent the cells labeled only by αEnhb-1, and double circles represent the cells labeled by both 4D9 and αEnhb-1. Cells are specified to express Eng according to their position along the dorsoventral (D-V) axis in a segmentally repeating fashion. Position along the anterior-posterior (A-P) axis would determine in which segments cells that can be double-labeled would appear, and information along the lumeno-pial (L-Pi) axis would determine which of the Eng-expressing cells they are. See Discussion.
A model of the specification of brain nuclei dependent on the early coordinates along three axes and segmentation. In highly schematic drawings of Eng expression in the first 6 rhombomeres of hindbrain, both in transverse and lateral views, the single circles represent the cells labeled only by αEnhb-1, and double circles represent the cells labeled by both 4D9 and αEnhb-1. Cells are specified to express Eng according to their position along the dorsoventral (D-V) axis in a segmentally repeating fashion. Position along the anterior-posterior (A-P) axis would determine in which segments cells that can be double-labeled would appear, and information along the lumeno-pial (L-Pi) axis would determine which of the Eng-expressing cells they are. See Discussion.
Eng may specify identities of many distinct cell types
We previously reported Eng expression in two specific jaw muscles (Hatta et al. 1990). Expression is present in their early mesenchymal precursors, as they associate with the growth cones of the motor axons which later innervate the muscles they form. Eng may thus play a role in specifying the identity of these muscles, a subset of those developing from pharyngeal arch mesoderm.
Similarly, in this paper we have shown a number of other examples where Eng is expressed in one or a few, but not all members of an iterated series, that then go on to develop as specialized parts of the series. Thus all of the cells that form one cartilage in the pectoral girdle express Eng, but cells forming other cartilages in the girdle, or in the nearby pharyngeal arches, do not. The expression in subsets of cerebellar cells, in midbrain nuclei, in muscle pioneers and as mentioned in the previous section, in rostral hindbrain segments may be other examples of such position-dependent specifications underlain by Eng. In each case, the pattern changes in a characteristic fashion along a single axis. The jaw muscle precursors, the cells in the rostral hindbrain, and the muscle pioneers of the myotomes all undergo morphogenetic rearrangements after they begin to express Eng. An early feature of such heterotypic specification could thus be activation of the genes that mediate morphogenesis, including expression of cytoskeletal proteins and adhesion proteins. These genes could be under the direct control of Eng.
ACKNOWLEDGEMENTS
We thank Thomas Kornberg, Kevin Coleman, Nipam Patel and Corey Goodman for the 4D9 antibody, Claytus Davis and Alexandra Joyner for the αEnhb-1 antibody, Jane Dodd for the 3A10 antibody, Bill Trevarrow for the zn-5 antibody, Thomas Schilling, Mamie Halpern, Judith Eisen, Alexandra Joyner, Nigel Holder, Anders Molven and Theresa Ellis for comments on the manuscript, Bettina Debu, Susan Pike, Kate Baraid, Andy Wood, Thomas Finger and especially Thomas Schilling for discussion and help in identifying cells that express Eng. Michele McDowell, Harry Howard and Reida Kimmel provided excellent technical assistance. Supported by NIH grant HD22486, NS21132 and a Naitoh Foundation Fellowship to K. H.