Expression of a homeobox gene product in normal and mutant zebrafish embryos: evolution of the tetrapod body plan

Transcription

1 Development 109, (1990) Printed in Great Britain The Company of Biologists Limited Expression of a homeobox gene product in normal and mutant zebrafish embryos: evolution of the tetrapod body plan ANDERS MOLVEN 1 *, CHRISTOPHER V. E. WRIGHT 2!, RUTH BREMILLER 1, EDDY M. DE ROBERTIS 2 and CHARLES B. KIMMEL 1 1 Institute of Neuroscience, University of Oregon, Eugene, OR 97403, USA 2 Department of Biological Chemistry, University of California, Los Angeles, CA , USA * Present address: Laboratory of Biotechnology, University of Bergen, PO Box 3152, Arstad, N-5029 Bergen, Norway t Present address: Department of Cell Biology, Vanderbilt University, Nashville, TN 37232, USA Summary An antibody was used to detect antigens in zebrafish that appear to be homologous to the frog homeodomaincontaining protein XlHbox 1. These antigens show a restricted expression in the anteroposterior axis and an anteroposterior gradient in the pectoral fin bud, consistent with the distribution of XlHbox 1 protein in frog and mouse embryos. In the somitic mesoderm, a sharp anterior limit of expression coincides exactly with the boundary between somites 4 and 5, and the protein level fades out posteriorly. A similar, graded expression of the antigen is seen within the series of Rohon-Beard sensory neurons of the CNS. We also immunostained the mutant spt-1 ('spadetail'), in which the trunk mesoderm is greatly depleted and disorganized in the region of XlHbox 1 expression. The defects stem from misdirected cell movements during gastrulation, but nevertheless, newly recruited cells that partially refill the trunk mesoderm express the antigen within the normal span of the anteroposterior axis. This finding suggests that the mutation does not delete positional information required for activation of the XlHbox 1 gene. Key words: Brachydanio rerio, homeodomain protein, embryogenesis, spadetail mutant, neural development, fin bud. Introduction The homeobox was initially discovered in Drosophila melanogaster as a conserved DNA sequence in genes of the Antennapedia and bithorax complexes (McGinnis etal. 1984; Scott and Weiner, 1984). The growing family of identified Drosophila homeobox-containing sequences now includes genes from the maternal effect, segmentation and homeotic classes, in addition to genes involved in specification of cell fates during later stages of embryogenesis. The homeodomain (the 60 amino acids encoded by the homeobox) mediates sequencespecific binding to DNA (Miiller et al. 1988), indicating that homeoproteins generally function as transcription factors. This hypothesis has gained support from the recent cloning of several transcription factors, which turned out to contain homeodomains (Herr et al. 1988). Homeoboxes are widely distributed throughout the animal kingdom, from nematodes to humans. In vertebrates, most is known about murine homeobox genes related to the Drosophila Antennapedia sequence. More than 25 such 'Hox' genes have thus far been isolated, and their restricted temporal and spatial expression patterns are highly suggestive of a regulatory role during embryogenesis (reviewed by Holland and Hogan, 1988; Wright et al. 1989a). The genomic organization of vertebrate homeobox genes is intriguingly similar to that of Drosophila homeotic genes (Boncinelli et al. 1988; Graham et al. 1989). The murine Hox genes are organized into at least four large gene clusters, and, as in Drosophila, the position of a gene within a cluster reflects its relative domain of expression along the anteroposterior axis of the embryo (Gaunt et al. 1988; Graham etal. 1989; Duboule and Dolle", 1989). Growing evidence suggests that homeobox gene products serve similar developmental functions in flies and vertebrates (Wright et al. 1989b; M. Kessel and P. Gruss, personal communication). A distinct feature of the vertebrate Antennapedialike genes is their high degree of conservation between distantly related species. For example, definite homologues of murine Hox genes have been found in chicken (Wedden et al. 1989), frog (Harvey et al. 1986; Fritz et al. 1989) and zebrafish (Nj0lstad et al. 1988a, 1990). Also the order of these genes within the clusters is probably conserved (Fritz et al. 1989; Kappen et al. 1989; Nj0lstad et al. 1990). Comparative studies on the expression of the same homeobox genes in different

2 280 A. Molven and others vertebrates might be fruitful in unravelling the function of these genes during vertebrate evolution. In the present study, an antibody originally made against the Xenopus laevis homeodomain protein XlHbox 1 was used to characterize expression in normal and mutant zebrafish embryos. XlHbox 1 was the first vertebrate homeobox gene to be isolated (Carrasco et al. 1984). It is under the control of two promoters, resulting in a long and a short version of the translation product (Cho et al. 1988). The two proteins display different anterior borders of expression in both frog and mouse embryos (Oliver et al. 1988fl). Furthermore, in these animals XlHbox 1 is expressed in forelimb bud mesoderm in an anteroposterior gradient (Oliver et al ). The preparation of antibodies against various XlHbox 1 domains has been described previously (Oliver et al. 1988a). Due to strong amino acid conservation between XlHbox 1 and its- mammalian homologues (Simeone et al. 1987; Sharpe et al. 1988), it is not surprising that these antibodies crossreact between species. The present paper shows that antibodies against XJHbox 1 react with antigens in zebrafish embryos. The staining pattern is generally the same as for frog and mouse, including a gradient in the pectoral fin/anterior limb bud. This implies a fundamental role for an XlHbox 1-like gene in CNS and mesoderm differentiation and during appendage formation in all vertebrate embryos. By a double-labeling technique, we have been able to identify some of the positive cells in the spinal cord as early sensory neurons. The XlHbox 1 antigen is expressed in approximately correct axial locations in spt-1, a zebrafish mutation producing severe deficiencies in the trunk mesoderm (Kimmel et al. 1989). This result suggests that the mutation does not affect position-dependent mechanisms that regulate homeobox gene expression along the anteroposterior axis in both the CNS and mesoderm. Materials and methods Embryos and mutants Zebrafish were maintained and bred essentially as described in Stuart et al. (1988). Developmental age is given as h, hours after fertilization at 28.5 C, the temperature of incubation. Staging was by morphological criteria, including counting somites in embryos between 12 and 24 h (Hanneman and Westerfield, 1989). Prior to fixation the chorions were removed with forceps. Wild-type embryos are heavily pigmented at 48 h, and therefore we selected embryos homozygous for the 'golden' mutation, gol-l(bl) (Streisinger et al. 1986) for the staining experiments performed at that stage. These mutants are sparsely pigmented at 48 h, but otherwise appear to develop identically to wild-type embryos. The 'spadetail' embryos were homozygous for the mutation spt-1 (blo4) (Kimmel et al. 1989). Antibodies As mentioned above, the XlHbox 1 protein is present in frog and mouse embryos in short and long forms, which have different spatial expression patterns. Whether this will turn out to be the situation in zebrafish is not clear. Antibody B of Oliver et al. (198&i), which detects both XlHbox 1 proteins, but not other homeodomain proteins, reacts strongly in zebrafish embryos and was used in all the immunolocalization experiments reported here. Antibody C, which recognizes only the long protein (Oliver et al. 1988fl), also reacted in zebrafish. In a preliminary analysis of embryos stained as whole mounts, antibody C produced a pattern indistinguishable from that of antibody B. This could be due to low resolution of subtle differences in anteroposterior borders by the whole-mount procedure (as is the case with Xenopus embryos; C.V.E.W. and E.M.D.R., unpublished observations). Alternative explanations are that in zebrafish both antibodies cross-react with epitopes present only in the long protein or thatfishexpress only the long protein version of the XlHbox 1 gene. Clarification of this issue will require further experimentation, in particular the isolation of zebrafish cdna clones. Antibody B is an affinity-purified rabbit antisera derived from a fusion protein of /S-galactosidase coupled to 142 amino acids of XlHbox 1 /V-terminal sequence, some of which are shared by the long and the short proteins. The preparation of the fusion protein and antibody, and descriptions of the specificity of the antibody are included in Oliver et al. (1988a). Antibody B was generally diluted 1:50 for staining of whole embryos and 1:25 for sections. Non-specific staining was monitored by substituting antibody B with an 'out-of-frame' antibody. The latter was made against a fusion protein in which a segment of XlHbox 4 DNA (encoding 56 amino acids) was cloned out-of-frame to )3-galactosidase (Cho et al. 1988). When this control antibody was used, no staining could be detected, showing that the patterns described are specific for XlHbox 1-related antigens. The preparation and characteristics of the monoclonal antibody zn-12 used in the double-labeling experiment are described in Metcalfe et al. (1990). Immunostaining of whole embryos Embryos were fixed for 4-6h at 4 C in 0.1M sodium phosphate buffer (ph7.3) containing 4% sucrose, 0.15 mm CaCl 2 and 4% paraformaldehyde. They were rinsed in 0.1M phosphate buffer, rinsed in distilled water and then submerged in acetone at 20 C for 7min. The embryos were rinsed in distilled water once more, then in 0.1M phosphate buffer, and treated with 2% normal goat serum in BDP (1% BSA, 1% DMSO in PBS) for 30min to block non-specific binding sites. Following blocking, the primary antibody (diluted in BDP) was allowed to react with the tissue for 5 h at room temperature or overnight at 4 C. The embryos were washed in BDP for 2h with at least 4 changes of solution, and secondary antibody (goat anti-rabbit IgG) diluted 1:500 in BDP was added. Reaction conditions and washing were as for the primary antibody. The embryos were then treated for 5 h at room temperature or overnight at 4 C with rabbit peroxidase-anti-peroxidase complex diluted 1:500 in BDP. After thorough washing in 0.1M phosphate buffer, the embryos were stained for peroxidase activity. They were presoaked for 15min in 50 mm phosphate buffer (ph7.3) containing 0.05 % diaminobenzidine (DAB) and 1 % DMSO, followed by addition of H 2 O 2 to a final concentration of 0.004%. The staining reaction was generally complete in lomin. Finally, the embryos were rinsed in 0.1M phosphate buffer, rinsed in distilled water, dehydrated through an alcohol series, cleared in methyl salicylate and mounted between two coverslips with Permount (Fisher Scientific Company), which allowed the embryos to be observed from both the top and bottom.

3 Expression of a homeodomain protein in zebrafish 281 Sectioning of stained embryos After treatment with methyl salicylate, the stained wholemount embryos were infiltrated with a solution containing 1 part methyl salicylate and 1 part resin (Epon-Araldite) overnight at room temperature. They were then placed in pure resin (2 changes of solution, 2h each) and embedded in fresh resin. The blocks were polymerized overnight at 60 C. Ralph knives were used to cut 10 //m sections on an MT-1 ultramicrotome. The sections were dried onto a glass slide and mounted in Permount. Double-labeling experiment Embryos were fixed as usual, rinsed 3x5min in 0.1M phosphate buffer containing 4% sucrose and 0.15 HIM CaCl 2, and embedded in agar blocks (1.5% agar in 5% sucrose, heated to 40 C). The blocks were allowed to cool to room temperature, sunk overnight in 30% sucrose at 4 C, frozen in liquid nitrogen, and sectioned at 16^m in a cryostat at 20 C. The sections were picked up on gelatin-subbed slides and dried at room temperature. An avidin-biotinylated-hrp system (Vector Laboratories) was used for immunostaining. The sections were washed 4x 5 min with PBS both initially and between each step. For the first reaction, antibody B (diluted 1:50) was applied overnight at 4 C. The sections were washed, treated with biotinylated secondary antibody (goat anti-rabbit IgG, 1:200 dilution) for 30 min, washed, treated with avidin-biotinylated-hrp (60min, 1:100 dilution) and stained with a DAB solution (0.1 M Tris-HCI, ph7.6, 0.04% DAB, 0.003% H 2 O 2 ) for 10 min in the presence of 0.3% Ni(NH 4 ) 2 (SO 4 ) 2-6H 2 O. The heavy metal ions give the reaction product a dark grey color. After washing thoroughly with PBS, the second primary antibody, zn-12, was applied in a 1:500 dilution and the sections were incubated at room temperature for 5h. Biotinylated secondary antibody (horse anti-mouse IgG, 1:200 dilution) and avidin-biotinylated-hrp were employed as described above. The DAB reaction mixture now contained 0.05% DAB, 50mM phosphate buffer and 0.004% H 2 O 2, resulting in a brown product. Finally, the sections were dehydrated through an alcohol series, cleared in xylene and mounted in Permount. Results The expression patterns of antigens reacting with anti- XIHbox 1 antibodies in zebrafish embryos are presented in developmental sequence. For the sake of brevity, we designate both the zebrafish and the frog antigens as 'XIHbox 1'. We emphasize, however, that we have not definitely shown that the zebrafish antigen is in fact the homologue of the frog protein (see Discussion). Onset of expression Previous studies (Eiken etal. 1987; Nj0lstad etal. 1988a, 1988/?) have shown that transcription of zebrafish hox genes is first detected about h, the time when the first somites are being formed. To determine the onset of XIHbox 1 expression, we started with 10 h embryos and monitored the staining every second hour until clearly positive cells were visible. Very weak staining was first detected at 14h, when of the eventual 30 somites have formed, in cells that seem to represent an epithelium lining somite 5 and possibly 6 (not shown). Expression above background could not be detected elsewhere, including the CNS. At this and at all later stages, we found the labeling confined exclusively to cell nuclei, as expected for a homeodomain protein. This finding thereby strongly supports our contention that the antibody is detecting a homeobox gene product in zebrafish, as it does in mouse and frog. In 16h embryos (14-15 somites), the staining of somitic mesoderm is more prominent, although still somewhat weak (Fig. 1A). Cells of the lateral surfaces of somites 5, 6 and probably somite 7 are labeled, in an anteroposterior gradient with respect to the strength of the signal. The anterior limit of expression is very sharp, and coincides exactly with the boundary between somites 4 and 5. A few XIHbox 1-expressing cells are now also present in the CNS (Fig. 1A). They lie in a subregion of the spinal cord, in segments 2 through 6, as delimited from the positions of the adjacent somites (see Hanneman et al. 1988). Accordingly, the anterior border of initial expression in the CNS is more anterior than in the somitic mesoderm ('out-of-register' expression, see Discussion). At this stage (16h), XIHbox 1 staining is also detected along a specific length of the notochord (Fig. IB). Notochord cells have a distinctive location and morphology, and it is clear that their nuclei, and not those of some notochord-associated cells, contain the marker. The notochord is labeled between the level of somite 8 and somites 13 or 14. The positive region does not overlap with the region of expression in the somites (more anterior) and the CNS (more anterior still). This is an example of out-of-register expression by separate derivatives, axial and paraxial, of the same germ layer. Expression in 22 h embryos For a more detailed analysis of XIHbox 1 expression, we chose embryos at 22 h, by which time somite formation is nearly finished (about 26 out of the final 30 somites are present). At this stage the embryos are still unpigmented, they have obtained a fish-like appearance, the primary organs have formed and several cell types have begun to differentiate. Fig. 2A and B shows that the somitic staining is stronger by 22 h, although the positive region is essentially the same as six hours earlier, including the sharp anterior cut-off at the boundary between somites 4 and 5. Labeling is strongest in somite 5 and fades out over somites 6 and 7 with no clear posterior limit (Fig. 2B). In the somites, most of the expressing cells are located ventrally (see also Fig. 7A,B). The rounded nuclei and the close packing of the positive cells strongly suggest that they are not muscle fibers, many of which have differentiated by this stage. Staining is also observed in the mesoderm lateral to the somites and in the region where the pectoral fin will develop (treated separately below). In 22 h embryos, antibody B labels many more spinal cord cells than at the earlier stage and the staining intensity is higher (Fig. 2A,B). Judged by comparison

4 282 A. Molven and others to somite level, the anterior limit of CNS expression remains the same, at segment 2. Posteriorly the staining extends to segments 8 or 9, that is, two to three segments posterior to that detected earlier. Probably, the general increase in the intensity of XlHbox 1 expression now permits the detection of cells that were too weakly labeled to be visualized at the 16 h stage. The XlHbox 1 antigen appears to be a useful marker for the analysis of the nervous system. The main population of positive spinal neurons consists of two lateral rows of cells on each side of the midline (Fig. 2C). Their position and large, rounded nuclei suggest they are developing neurons, a number of which are known to be differentiating by 22 h (see Hanneman et al. 1988). The most dorsal members of this population are early sensory neurons (see below). We have not yet identified the others, but they appear to be located dorsal to the positions of primary motoneurons at this stage (Eisen et al. 1986; Myers et al. 1986), and could be interneurons. The second type of labeled cells in the spinal cord are small and cuboidal, and located midventrally, i.e. in the floor plate. The positive floorplate cells are easily seen in whole mounts of h embryos (not shown), or in transverse sections of older stages (Fig. 3C). In 22 h embryos, staining of notochord cells is no longer visible. Expression in 48 h embryos We have also characterized XlHbox 1 expression in 48 h embryos. At this relatively late stage, the zebrafish embryo is capable of swimming movements and has Fig. 1. Immunolocalization of XlHbox 1 antigens in 16 h embryos (whole mounts). (A) Trunk region. The lateral surfaces of somites 5 (s5), 6 and possibly 7 are labeled. The anterior limit of mesodermal expression coincides exactly with the boundary between somites 4 and 5. XlHbox 1 is also expressed in a few cells in the spinal cord (arrowheads). Anterior is to the left in this and in all other whole-mount views and sagittal sections. (B) Expression in the notochord. This view is more posterior than A and at a different focal plane. Somites 8 and 13 are denoted by s8 and sl3, respectively. Scale bar: (A and B) 50/.wn. Fig. 2. Immunolocalization of XlHbox 1 antigens in 22 h embryos (whole mounts). (A) Whole zebrafish embryo stained with anti-xlhbox 1 antibody. (B) Expression in the somites and the spinal cord. This view is a close up of the trunk region of A. XlHbox 1 antigens are detected in essentially the same region as in 16h embryos. Although somewhat out of focus in this picture, one can see labeling of the region where the fin bud will form. Abbreviations: fr, fin-forming region; sc, spinal cord; s5, somite 5; y, yolk sac. (C) Expression in the spinal cord (dorsal view). Positive cells lie in two rows, one on each side of the midline. Somite 5 is indicated by brackets. Scale bars: (A) 200/an; (B andc) 50;/m. well-differentiated muscular, circulatory and nervous systems. Fig. 3 shows that at 48 h a new population of XlHbox 1-positive neurons located more anteriorly has become apparent. These neurons are positioned at the level of, and also somewhat anterior to, the first muscle segment (Fig. 3A), that is, at the level of the hindbrain/spinal cord junction (see Hanneman et al. 1988). The anterior border of staining is sharp, and the labeled cells have a very restricted spatial distribution: they are confined to 3B 3C rb ^- i^ Fig. 3. [mmunolocalization of XlHbox 1 antigens in 48h embryos (Nomarski optics). (A) Sagittal section of the posterior hindbrain and anterior spinal cord. A 'new' positive population of neurons has turned on expression at this stage (arrow), msl is muscle segment 1. (B) Transverse section at the level of the first muscle segment. XlHbox 1-expressing neurons are indicated by arrows. (C) Transverse section at the level of the fourth muscle segment showing the different types of positive cells in the spinal cord: floor plate cells (fp) and lateral neurons (In). The dorsalmost of the latter are early sensory neurons (rb, Rohon-Beard cells), n is notochord. Scale bar: (A,B and C) 50 jan.

5 1A 1B

6

7 Expression of a homeodomain protein in zebrafish 283 Fig. 4. Dorsal cells expressing XIHbox 1 are Rohon-Beard neurons. (A) A sagittal section (25 h) showing the positive, dorsal neurons as a row of single, separated cells. Abbreviations: n, notochord; sc, spinal cord; v, ventral parts of somites. (B) A sagittal section labeled with both anti-xlhbox 1 antibody (black) and the monoclonal antibody zn-12 (brown). Dark nuclei are seen inside brown membranes (arrowheads), confirming that the dorsal XIHbox 1-expressing neurons belong to the Rohon-Beard population. Scale bars: (A) 100^m; (B) 25 pirn. Fig. 5. XIHbox 1 is expressed during formation of the fish pectoral fin. (A) The fin-forming region as it appears when stained with anti-xihbox 1 antibody at 25 h (whole mount). Expression is strongest at the anterior basis of the region (arrowheads). (B) Immunolocalization of XIHbox 1 antigens in the pectoral fin bud at 48 h (horizontal section). The concentration of protein is highest in the anteroproximal region. XIHbox 1 is also expressed in nuclei of the epidermal layer of the bud. Scale bar: (A and B) 25/im. Fig. 6. Immunolocalization of XIHbox 1 antigens in 22 h embryos bearing the spt-1 mutation (whole mounts). (A) Whole spt-1 embryo stained with anti-xihbox 1 antibody. The characteristic spadetail is indicated by an arrow and the cell-depleted trunk region where XIHbox 1 is expressed, is framed. (B) Close up of the trunk region. The normal expression pattern of the somitic mesoderm is disrupted. Compare with the wild-type embryo in Fig. 2B. (C) Dorsal view of the trunk region. In the spinal cord, XIHbox 1 expression is not severely disturbed. Compare with the wild-type embryo in Fig. 2C. Scale bars: (A) 400 jun; (B and C) 50 ^m. a small patch, crescent-shaped in the transverse view, on each side of the midline (Fig. 3B). Their nuclei are round and generally smaller than the other labeled neurons in the spinal cord (Fig. 3C). This fact, together with the late onset of expression and the specific localization argue that the anterior XIHbox 1-positive cells make up a neuronal class distinct from the lateral neurons described above. In the mesoderm at 48 h, we can no longer see labeled cells in the ventral part of the somite-derived tissue. However, staining is present in mesenchymal cells surrounding both the main blood vessels and the pronephric ducts (not shown). Judged by segment level and with the exception of the new, anterior population of positive neurons, the regions of ectodermal and mesodermal XIHbox 1 expression have not significantly changed in 48 h embryos as compared to 22h embryos. Graded expression of XIHbox 1 antigen in Rohon- Beard neurons In sagittal sections, the most dorsal XIHbox 1-positive neurons are prominent as a row of single cells (Fig. 4A). Their localization and their large, rounded nuclei suggested that they might correspond to the early sensory neurons known as Rohon-Beard cells (Myers et al. 1990). To test this hypothesis we performed a double-labeling experiment where sagittal sections were first stained with the anti-xihbox 1 antibody and then with the monoclonal antibody zn-12. The latter recognizes the cell-surface HNK-1/L2 epitope (Metcalfe et al. 1990), which is expressed very strongly and specifically by the Rohon-Beard cells in the early spinal cord (Myers et al. 1990). In Fig. 4B, the XIHbox 1-expressing nuclei (black) are seen surrounded by zn- 12-labeled cell membranes (brown), consistent with these cells being Rohon-Beard neurons. An apparent anteroposterior gradient of XIHbox 1 labeling is present within this cell population (Fig. 4A). The anteriormost stained nuclei are labeled most prominently, and expression fades posteriorly in a graded manner, however continuing posteriorly past the more ventral neurons that express the marker. The anterior limit of expression lies at the level of somite 2 at 25 h, but Rohon-Beard cells are known to continue anteriorly into the hindbrain region (Myers et al. 1990). Accordingly, XIHbox 1 is not visibly active in the most anterior Rohon-Beard cells. Similarly, Rohon-Beard cells are also present, but unlabeled with anti-xihbox 1 antibody, in the most posterior spinal segments. Expression in the fin bud One notable feature of XIHbox 1 expression in mouse and frog embryos is that the protein is present as a gradient in the forelimb bud mesoderm (Oliver et al ). Because the tetrapod forelimb is presumed to have evolved from the pectoral fin of early bony fishes, we examined whether the antigen could be found in the pectoral fin bud of fish as well. In zebrafish at 29 h, the fin bud itself becomes visible, as a condensation of lateral mesodermal cells at the level of somites 2 and 3. Ten hours before a distinct fin bud is formed, cells in the region where it will arise stain with anti-xihbox 1 antibodies (at 19h, not shown). In 25h embryos, labeling of this population is strong (Fig. 5A) and the more anterior cells are labeled more intensely. Fig. 5B shows that in 48 h embryos XIHbox 1 is expressed in a striking pattern in the mesoderm of the fin bud. Expression is maximal in the anterior and proximal region of the fin mesoderm. In addition, the protein is present in nuclei of the epidermal layer of the fin bud. This pattern of expression is in perfect agreement with what has been reported for frog and mouse embryos (Oliver et al ). Expression in the spadetail mutant In zebrafish embryos homozygous for the lethal mutation spt-1 ('spadetail'), cells that should normally form trunk mesoderm fail to migrate to the correct position during gastrulation (Kimmel et al. 1989), and move to the tail. This produces a major disruption of mesodermal segmentation in the trunk region. Our observations show that XIHbox 1 is expressed in a trunk region that is affected by spt-1. Because morphogenesis and patterning is disrupted in the region where XIHbox 1 is normally present, we wondered whether the mutation might disturb expression of this homeodomain protein. Homozygous spt-1 embryos, stained at 22h, express XIHbox 1 in approximately the normal region of the body (Fig. 6A). Close observation reveals that XIHbox

8 284 A. Molven and others 7C rm Fig. 7. Immunolocalization of XlHbox 1 antigens in 22 h wild-type and spt-1 embryos (transverse sections, Nomarski optics). (A) A section of a wild-type embryo, at the level of somite 5. The section goes through the central part of the somite. Staining is prominent in cells in the ventral part of and lining the somites. Positive nuclei are also seen in the spinal cord and in the lateral mesoderm. Some positive cells (arrowheads) surrounding the notochord might represent migrating neural crest cells. These cells begin to migrate in the trunk at about 17 h, and occur in this position, but only along the central part of each somite (G. Bobrowicz and J. S. Eisen, personal communication). Expression of XlHbox 1 in zebrafish neural crest cells would be consistent with observations in frog and mouse embryos, where anti-xlhbox 1 antibodies are known to stain neural crest derivatives (Cho et al. 1988; Oliver et al. 1988a). (B) A section adjacent and posterior to A. The positive cells around the notochord are absent as expected for neural crest cells, otherwise the staining pattern is essentially the same as in A. (C) A section of the trunk region of a spt-1 embryo. As in the wild type, positive cells are located laterally in the spinal cord (thin arrows). Although the trunk region has been depleted of the mesodermal cells that normally should form somites, XlHbox 1-expressing cells (thick arrows) are lining up in a position reminiscent of the situation in the wild type. Abbreviations: lm, lateral mesoderm; n, notochord; pd, pronephric ducts; rm, remaining mesodermal cells; sc, spinal cord. Scale bars: (A,B and C) 50f<m. 1 protein is present both in the somitic mesoderm (Fig. 6B) and in the neurectoderm (Fig. 6C). As compared to the normal pattern, the labeling of somitic mesoderm is seriously disorganized in the spt-1 embryo (Fig. 6B). This disturbance is expected because the trunk is depleted of mesodermal cells. When one considers how severely spt-1 affects the trunk region, it is perhaps surprising that some of the remaining mesodermal cells express XlHbox 1 at all. The positive cells are even positioned correctly around the ventral and lateral surface of the cell-depleted paraxial mesoderm that normally forms the somites (Fig. 7C, compare with the wild-type embryo in Fig. 7A,B). Moreover, the anterior and posterior borders of XlHbox 1 expression in the mesoderm appear approximately normal in the spt-1 mutant (Fig. 6). In the spinal cord of spt-1 mutants, the positive nuclei lie in two lateral rows (Fig. 6C) as in wild-type embryos. Apparently, the severe mesodermal defects in the spt-1 embryo do not disturb the basic pattern of XlHbox 1 expression in the CNS. Discussion A zebrafish homologue of the frog XlHbox 1 gene There are several lines of evidence suggesting that a gene closely related to frog XlHbox 1 is present in the zebrafish genome. The immunolocalization experiments described in this paper show that when zebrafish embryos are stained with antibodies against XlHbox 1, restricted regions of the spinal cord, the segmented and unsegmented mesoderm, and the fin bud are labeled. The spatial expression pattern is thus in principle similar to that in frog and mouse embryos. The existence of a zebrafish XlHbox 1 homologue is not surprising when one considers how extensively this gene has been conserved in other vertebrates. For example, the human cdna clone c8 (Simeone et al. 1987) encodes 128 out of 152 amino acids identical to the short version of the frog protein (including a completely identical homeodomain) and of the 24 changes, many are conservative. XlHbox 1 homologues have also been described for the mouse (Hox-6.1: Sharpe et al. 1988, now renamed Hox-3.3: Schughart et al. 1989) and for the newt (NvHboxl: Savard et al. 1988). Strong sequence conservation between homeobox genes of distantly related vertebrate species might be a general phenomenon. For example, convincing relationships are found between the mouse Hox-2.1 /Hox- 2.2 genes and their zebrafish homologues (Nj0lstad et

9 Expression of a homeodomain protein in zebrafish 285 al. 1988a, 1990), and between the region of an engrailed-like fish homeobox and the mouse gene En-2 (Fjose et al. 1988). A sequence homology of more than 90%, which is the case for mouse and fish Hox-2.1 proteins (when conservative amino acid changes are included), is remarkable considering that these species have been separated for possibly more than 200 million years. Moreover, evidence is accumulating that not only the sequences, but also the way the homeobox genes are organized into clusters is primarily the same (Kappen et al. 1989; Nj0lstad et al. 1990). This is promising for the use of a simple vertebrate such as zebrafish as a genetically accessible model system (Kimmel, 1989) for homeobox gene function in higher animals. A candidate for zebrafish XlHbox 1 has recently been isolated by Nj0lstad et al. (1990). Their gene hox[zf-61] is apparently closely related to Hox-3.3; the C-terminal regions are 81 % identical at the amino acid level, there is only one amino acid substitution in the homeodomain, and a splice site has been conserved. Furthermore, hox[zf-6l\ is located on a genomic lambda clone, which also contains a homeobox sequence very similar to Hox-3.4. This clustering and the high amino acid similarity strongly argue that hox[zf-61] is the zebrafish homologue of the mouse gene Hox-3.3 and therefore of Xenopus XlHbox 1. Final proof of this relationship must, however, await the determination of the N- terminal sequence of hox[zf-61]. The pattern of XlHbox 1 antigen expression in zebrafish has many similarities but also some differences with previous studies on this gene (Sharpe et al. 1988; Oliver et al. 1988a). The antibody used here reacts with both long and short XlHbox 1 proteins in frog and mouse embryos (Oliver et al. 1988a). In higher vertebrates, the anterior somites express the short protein, while in zebrafish a sharp border is present in somite 5. This could be explained if the antibody recognized preferentially the long protein in zebrafish due to higher conservation at the N-terminal domain, or if the short protein were absent in fishes. Clarification of this issue will require the isolation of zebrafish cdna clones. The in situ hybridization studies of Sharpe et al. (1988) in the mouse did not utilize probes that would distinguish long and short protein transcripts. They found, however, mrna throughout the entire length of the spinal cord. This is not the case for antibody staining of XlHbox 1 proteins, which are always undetectable in the caudal spinal cord. It will be difficult to sort out these discrepancies until localization studies with zebrafish probes are carried out. Another approach that may yield information on the fine regulation of this gene, is the expression in zebrafish embryos of reporter gene constructs containing the short and long protein promoters. For initial studies, we plan to use heterologous XlHbox 1 promoters where detailed molecular characterization has already been performed (Simeone et al. 1987; Cho et al. 1988). XlHbox 1 as a molecular marker in zebrafish embryogenesis Most vertebrate homeobox genes analyzed thus far are active in spatially restricted regions of the neurectoderm and mesoderm (Holland and Hogan, 1988; Wright et al. 1989a). Accordingly, their patterns of expression suggest a role in establishing or responding to positional values along the anteroposterior axis of the embryo. The distribution of XlHbox 1 antigens in zebrafish embryos is consistent with this view. A speciality of the zebrafish is the early transient expression in the notochord. This has not been reported for the frog or mouse, although in some preparations weak reaction with mouse notochord has been noted (G. Oliver, personal communication). Expression in the somitic mesoderm appears highly dependent on the anteroposterior position of the cells. The anterior boundary is sharp and coincides with a segment boundary. Somite 4 is entirely negative while expression decreases in a graded fashion in somites 6 and 7. Notably, XlHbox 1 expression seems to be graded across each somite; within one somite, nuclei at the anterior edge are most strongly labeled. The high level of expression in somite 5 may serve as a useful marker, for example in cell transplantation studies. Moreover, because transgenic zebrafish can be generated (Stuart et al. 1988, 1990), one can envisage that fusions of reporter genes to promoters such as that of XlHbox 1 will in the future provide zebrafish lines with marked cell populations. XlHbox 1 may be useful also as a molecular marker for the Rohon-Beard cells. This well-characterized neuron population pioneers a particular pathway through the spinal cord that later neurons follow (Kuwada, 1986). They are among the first neurons to differentiate in the spinal cord (Myers et al. 1990) and can therefore be studied when the CNS is developmentally early and very simple. The Rohon-Beard neurons exhibit a graded decrease in intensity of staining from anterior to posterior over the domain of XlHbox 1 expression (Fig. 4A), even though their cell nuclei are well separated from each other. From our results, it is clear that a single neuronal cell type can differentially express potential regulatory genes according to location along the neural axis. Perhaps in both the Rohon-Beard cells and in the somites, the XlHbox 1 gene somehow senses positional information along the body axis that affects its level of expression. Lack of alignment of CNS and mesodermal expression An interesting problem that homeobox gene markers might help to resolve is whether there are two independent sets of anteroposterior positional values in the mesoderm and in the CNS, or if the initial domains of the body plan are determined by a single axial system. In Xenopus embryos, the mesoderm and CNS have similar anterior and posterior borders of XlHbox 1 expression ('in-register' expression; De Robertis et al. 1989; Wright et al. 1989a). Because transplantation studies in amphibia have shown that the mesoderm determines the anteroposterior polarity of the neural

10 286 A. Molven and others plate (Spemann, 1938), it has been proposed that the mesoderm might transmit to the CNS similar positional values that turn on the XlHbox 1 expression in both germ layers (De Robertis et al. 1989). In the day 13 mouse embryo, Hox genes, including the XlHbox 1 homologue, are expressed out of register; i.e. the anterior border of mesodermal activity is posteriorly displaced with respect to that of the CNS (e.g. Gaunt et al. 1988). Nevertheless, in situ hybridization studies at earlier stages have shown that murine Hox genes initially can be expressed at the same segmental levels in the mesoderm and ectoderm (Gaunt, 1987; Murphy etal. 1989). In the zebrafish, XlHbox 1 antigen is first detected in the somites followed two hours later by the first staining of the CNS. XlHbox 1 expression in the CNS is, from the earliest moment of appearance, out of register with that in the mesoderm (Fig. 1A). It is possible that this is because we are detecting protein and not RNA, or that cell movements and growth during embryogenesis could have displaced the regions of XlHbox 1 expression in the fish and mouse embryos. However, tracing the movements of the progenitors of somite and CNS cells suggests that, in the zebrafish, the XlHbox 1-positive cells are never in register, even at earlier times (Kimmel et al. 1988). Thus, two independent positional information systems may be present that control XlHbox 1 expression in the mesoderm and in the CNS, respectively. In situ hybridization studies at early stages similar to those that have been carried out in the mouse may help to resolve this issue. The spt-1 mutation The spt-1 mutation acts autonomously in prospective somitic mesodermal cells to misdirect their gastrulation movements (Ho et al. 1989) and leaves the trunk region deficient of mesoderm at the stages when trunk somites normally form (from 11 h to 18h). Trunk mesoderm is eventually partially repopulated during later development. Despite the severe mesodermal disturbance, XlHbox 1 antigen is expressed in approximately the appropriate anteroposterior domains both in the CNS and in the mesoderm. This finding indicates that the mutation does not affect the positional information mechanisms to which hox genes respond. Since the mutation acts directly in prospective somitic mesoderm and not in neurectoderm (Ho et al. 1989), correct expression in the spinal cord might be expected if early interactions with somitic mesoderm are unessential for the specification of the CNS domain. Accordingly, the observation suggests that expression in each germ layer is patterned independently (see also the discussion above). Furthermore, because the spt-1 disturbance in cell movements occurs during gastrulation, finding later expression in the mutant mesoderm implies that the positional cues activating XlHbox 1 in mesoderm must be present (or persist) after somitogenesis is normally well underway. The mesodermal components of some trunk segments eventually form in spt-1, even though they are imperfect. We speculate that XlHbox 1 plays an important role in position-specific control of their production. The mesodermal cells that express the marker at 22 h may have originated by expansion of the few mesodermal cells present in the spt-1 trunk at earlier stages, or they may have come from some ectopic location. We note that normally a separate homeodomain protein, of the engrailed class, is expressed in another specific subset of cells in each zebrafish somite (Patel et al. 1989). In contrast to the observed expression of XlHbox 1, regulative development is not seen in spt-1 mutants when the cells expressing the engrailed antigen are studied; in this case labeling cannot be detected in the trunk region (K. Hatta and C.B.K., in preparation). Apparently, as judged by appearance of antigens from two homeodomain proteins, some classes of cells can be restored in the mutant segmented mesoderm, while others cannot. On the origin of the vertebrate forelimbs The vertebrate limb has been a favorable system for study of pattern formation. Observations by Ross Harrison on the morphogenesis of the fin buds of fish led to transplantation studies in newt forelimb development (Harrison, 1918), giving rise to the principle that cells are determined very early to become particular organs. The limb field, a region of cells that lies lateral to the axis, becomes committed to forelimb formation long before there is any overt sign of morphological differentiation (Harrison, 1918; Huxley and De Beer, 1934). Recent studies showed that homeobox proteins form gradients of expression in the limb bud. XlHbox 1 forms an anteroposterior and proximodistal gradient in the forelimb bud (but not in the hindlimb bud) mesoderm of frog and mouse embryos (Oliver et al. 1988b). A second gene, Hox-5.2, forms a complementary gradient of opposite polarity (Oliver et al. 1989). In the murine limb bud, genes of the Hox-5 complex are activated in a specific temporal and spatial order according to their position in the cluster (Doll6 et al. 1989). Consistent with this, a molecular link between morphogenetic fields, gradients and homeobox genes has been proposed (Oliver et al. 1988ft). XlHbox 1 may also play a role in amphibian forelimb regeneration (Savard et al. 1988). Because the tetrapod forelimb is presumed to have evolved from the pectoral fin of early bony fishes, it was of interest to investigate XlHbox 1 expression in the pectoral fin. Staining in the pectoral fin-forming region can actually be detected as early as 19 h, about ten hours before the fin bud is itself recognizable. By 48 h, the fin bud has a pattern of expression very much like that in the forelimb buds of frogs or mice. This suggests that the gradient of XlHbox 1 expression evolved before the appearance of the tetrapod limb. Gradients of homeobox proteins may have basic functions in the organogenesis of all vertebrates. The new molecular markers that are now becoming available will clearly be useful for the study of comparative vertebrate development.

11 Expression of a homeodomain protein in zebrafish 287 We thank Judith Eisen, Monte Westerfield and Pal Nj0lstad for comments on the manuscript. We are grateful to Rebecca Andrews for supplying us with the gol-1 mutants and to Don A. Kane for the spt-1 mutants. The work was supported by grant NS17963 and HD22486 to C.B.K. and NIH HD to E.M.D.R. A.M. was supported by the Norwegian Research Council NTNF and was. a recipient of a Wigeland Scholarship from the American-Scandinavian Foundation. C.V.E.W. was a Senior Fellow of the ACS (California Division). References BONCINELLI, E., SOMMA, R., ACAMPORA, D., PANNESE, M., D'ESPOSITO, M., FAIELLA., A. AND SIMEONE, A. (1988). Organization of human homeobox genes. Human Reprod. 3, CARRASCO, A. E., MCGINNIS, W., GEHRING, W. J. AND DE ROBERTIS, E. M. (1984). Cloning of an X. laevis gene expressed during early embryogenesis coding for a peptide region homologous to Drosophilu homeotic genes. Cell 37, CHO, K. W. Y., GOETZ, J., WRIGHT, C. V. E., FRITZ, A., HARDWICKE, J. AND DE ROBERTIS, E. M. (1988). Differential utilization of the same reading frame in a Xenopus homeobox gene encodes two related proteins sharing the same DNAbinding specificity. EMBO J. 7, DE ROBERTIS, E. M., OLIVER, G. AND WRIGHT, C. V. E. (1989). Determination of axial polarity in the vertebrate embryo: Homeodomain proteins and homeogenetic induction. Cell 57, DOLLfi, P., IZPISUA-BELMONTE, J.- C, FALKENSTEIN, H., RENUCCI, A. AND DUBOULE, D. (1989). Coordinate expression of the murine Hox-5 complex homoeobox-containing genes during limb pattern formation. Nature. 342, DUBOULE, D. AND DOLLE\ P. (1989). The structural and functional organization of the murine HOX gene family resembles that of Drosophila homeotic genes. EMBO J. 8, ElKEN, H. G., NJ0LSTAD, P. R., MOLVEN, A. AND FjOSE, A. (1987). A zebrafish homeobox-containing gene with embryonic transcription. Biochem. Biophys. Res. Comm. 149, EISEN, J. S., MYERS, P. Z. AND WESTERFIELD, M. (1986). Pathway selection by growth cones of identified motoneurones in live zebra fish embryos. Nature 320, FJOSE, A., ElKEN, H. G., NJ0LSTAD, P. R., MOLVEN, A. AND HORDVIK, I. (1988). A zebrafish engrailed-like homeobox sequence expressed during embryogenesis. FEBS Lett. 231, FRITZ, A. F., CHO, K. W. Y., WRIGHT, C. V. E., JEGALIAN, B. G. AND DE ROBERTIS, E. M. (1989). Duplicated homeobox genes in Xenopus. Devi Biol. 131, GAUNT, S. J. (1987). Homoeobox gene Hox-1. 5 expression in mouse embryos: earliest detection by in situ hybridization is during gastrulation. Development 101, GAUNT, S. J., SHARPE, P. T. AND DUBOULE, D. (1988). Spatially restricted domains of homeo-gene transcripts in mouse embryos: relation to a segmented body plan. Development 104 Suppl., GRAHAM, A., PAPALOPULU, N. AND KRUMLAUF, R. (1989). The murine and Drosophila homeobox gene complexes have common features of organization and expression. Cell 57, HANNEMAN, E., TREVARROW, B., METCALFE, W. K., KIMMEL, C. B. AND WESTERFIELD, M. (1988). Segmental pattern of development of the hindbrain and spinal cord of the zebrafish embryo. Development 103, HANNEMAN, E. AND WESTERFIELD, M. (1989). Early expression of acetylcholinesterase activity in functionally distinct neurons of the zebrafish. J. comp. Neurol. 284, HARRISON, R. G. (1918). Experiments on the development of the forelimb of Amblystoma, a self-differentiating, equipotential system. J. exp. Zool. 25, HARVEY, R. P., TABIN, C. J. AND MELTON, D. A. (1986). Embryonic expression and nuclear localization of Xenopus homeobox (Xhox) gene products. EMBO J. 5, HERR, W., STURM, R. A., CLERC, R. G., CORCORAN, L. M., BALTIMORE, D., SHARP, P. A., INGRAHAM, H. A., ROSENFELD, M. G., FINNEY, M., RUVKUN, G. AND HORVTTZ, H. R. (1988). The POU domain: a large conserved region in the mammalian pit-i, oct-1, oct-2, and Caenorhabditis elegans unc-86 gene products. Genes Dev. 2, Ho, R. K., KANE, D. A. AND KIMMEL, C. B. (1989). Cell transplantation in the zebrafish embryo: is the spt-1 mutation cell autonomous? Soc. Neurosci. Abstr. 15, 809. HOLLAND, P. W. H. AND HOGAN, B. L. M. (1988). Expression of homeo box genes during mouse development: a review. Genes Dev. 2, HUXLEY, J. S. AND DE BEER, G. R. (1934). The Elements of Experimental Embryology. Cambridge: Cambridge University Press. KAPPEN, C, SCHUGHART, K. AND RUDDLE, F. H. (1989). Two steps in the evolution of Antennapedia-class vertebrate homeobox genes. Proc. natn. Acad. Sci. U.S.A. 86, KIMMEL, C. B. (1989). Genetics and early development of zebrafish. Trends Genet. 5, KIMMEL, C. B., SEPICH, D. S. AND TREVARROW, B. (1988). Development of segmentation in zebrafish. Development 104 Suppl., KIMMEL, C. B., KANE, D. A., WALKER, C, WARGA, R. M. AND ROTHMAN, M. B. (1989). A mutation that changes cell movement and cell fate in the zebrafish embryo. Nature 337, KUWADA, J. Y. (1986). Cell recognition by neuronal growth cones in a simple vertebrate embryo. Science 233, MCGINNIS, W., LEVINE, M. S., HAFEN, E., KUROIWA, A. AND GEHRING, W. J. (1984). A conserved DNA sequence in homoeotic genes of the Drosophila Antennapedia and bithorax complexes. Nature 308, METCALFE, W. K., BASS, M. B., TREVARROW, B., MYERS, P. Z., CURRY, M. AND KIMMEL, C. B. (1990). The pattern of expression of the L2/HNK-1 carbohydrate in embryonic zebrafish. Submitted to Development. MCLLER, M., AFFOLTER, M., LEUPIN, W., OTTING, G., WOTHRICH, K. AND GEHRING, W. J. (1988). Isolation and sequence-specific DNA binding of the Antennapedia homeodomain. EMBO J. 7, MURPHY, P., DAVIDSON, D. R. AND HILL, R. E. (1989). Segmentspecific expression of a homoeobox-containing gene in the mouse hindbrain. Nature Ml, MYERS, P. Z., EISEN, J. S. AND WESTERFIELD, M. (1986). Development and axonal outgrowth of identified motoneurons in the zebrafish. J Neurosci. 6, MYERS, P. Z., METCALFE, W. K. AND KIMMEL, C. B. (1990). Rohon-Beard and trigeminal sensory neurons share a common pattern of early development in the zebrafish. Submitted to Development. NJ0LSTAD, P. R., MOLVEN, A., HORDVIK, I., APOLD, J. AND FjOSE, A. (1988fl). Primary structure, developmentally regulated expression and potential duplication of the zebrafish homeobox gene ZF-21. Nucl. Acids Res. 16, NJ0LSTAD, P. R., MOLVEN, A., ElKEN, H. G. AND FjOSE, A. (19886). Structure and neural expression of a zebrafish homeobox sequence. Gene 73, NJ0LSTAD, P. R., MOLVEN, A., APOLD, J. AND FjOSE, A. (1990). The zebrafish homeobox gene hox-2.2: transcription unit, potential regulatory regions and in situ localization of transcripts. EMBO J. 9, OUVER, G., WRIGHT, C. V. E., HARDWICKE, J. AND DE ROBERTIS, E. M. (1988a). Differential antero-posterior expression of two proteins encoded by a homeobox gene in Xenopus and mouse embryos. EMBO J. 7, OLIVER, G., WRJGHT, C. V. E., HARDWICKE, J. AND DE ROBERTIS, E. M. (1988b). A gradient of homeodomain protein in developing forelimbs of Xenopus and mouse embryos. Cell 55, OLIVER, G., SIDELL, N., FISKE, W., HEINZMANN, C, MOHANDAS, T, SPARKES, R. S. AND DE ROBERTIS, E. M. (1989).

12 288 A. Molven and others Complementary homeo protein gradients in developing limb buds. Genes Dev. 3, PATEL, N. H., MARTIN-BLANCO, E., COLEMAN, K. G., POOLE, S. J., ELLIS, M. C, KORNBERG, T. B. AND GOODMAN, C. S. (1989). Expression of engrailed proteins in arthropods, annelids and chordates. Cell 58, SAVARD, P., GATES, P. B. AND BROCKES, J. P. (1988). Position dependent expression of a homeobox gene transcript in relation to amphibian limb regeneration. EM BO J. 7, SCHUGHART, K., PRAVTCHEVA, D., NEWMAN, M. S., HUNIHAN, L. W., JIANG, Z. AND RUDDLE, F. H. (1989). Isolation and regional localization of the murine homeobox-containing gene Hox-3.3 to mouse chromosome region 15E. Genomics 5, SCOTT, M. P. AND WEINER, A. J. (1984). Structural relationships among genes that control development: Sequence homology between the Antennapedia, Ultrabithorax, and fushi tarazu loci of Drosophila. Proc. nasn. Acad. Sci. U.S.A. 81, SHARPE, P. T., MILLER, J. R., EVANS, E. P., BURTENSHAW, M. D. AND GAUNT, S. J. (1988). Isolation and expression of a new mouse homeobox gene. Development 102, SlMEONE, A., MAV1L1O, F., ACAMPORA, D., GlAMPAOLO, A., FAIELLA, A., ZAPPAVIGNA, V., D'ESPOSITO, M., PANNESE, M., Russo, G., BONCINELU, E. AND PESCHLE, C. (1987). Two human homeobox genes, cl and c8: Structure analysis and expression in embryonic development. Proc. natn. Acad. Sci. U.S.A. 84, SPEMANN, H. (1938). Embryonic Development and Induction. New Haven: Yale University Press. STREISINGER, G., SINGER, F., WALKER, C, KNAUBER, D. AND DOWER, N. (1986). Segregation analyses and gene-centromere distances in zebrafish. Genetics 112, STUART, G. W., MCMURRAY, J. V. AND WESTERFIELD, M. (1988). Replication, integration and stable germ-line transmission of foreign sequences injected into early zebrafish embryos. Development 103, STUART, G. W., VIELKIND, J. R., MCMURRAY, J. V. AND WESTERFIELD, M. (1990). Stable lines of transgenic zebrafish exhibiting reproducible tissue-specific transgene expression patterns. Development (in press). WEDDEN, S. E., PANG, K. AND EICHELE, G. (1989). Expression pattern of homeobox-containing genes during chick embryogenesis. Development 105, WRIGHT, C. V. E., CHO, K. W. Y., OLIVER, G. AND DE ROBERTIS, E. M. (1989a). Vertebrate homeodomain proteins: families of region-specific transcription factors. Trends Biochem. Sci. 14, WRIGHT, C. V. E., CHO, K. W. Y., HARDWICKE, J., COLLINS, R. H. AND DE ROBERTIS, E. M. (19896). Interference with function of a homeobox gene in Xenopus embryos produces malformations of the anterior spinal cord. Cell 59,