Sequence and expression pattern of pax-6 are highly conserved between zebrafish and mice

Many features of developmental control genes are highly conserved among organisms as diverse as nematodes and humans (Boncinelli et al., 1988; Biirglin et al., 1991; Dressier and Gruss, 1988; Duboule and Dolle, 1989; Graham et al., 1989; Kessel and Gruss, 1990; Scott et al., 1989). Most of the these genes are members of multigene families that share conserved domains such as homeo-boxes, paired-boxes, zinc-fingers, or POU-boxes (Bopp et al., 1986; Dressier et al., 1988; Dressier and Gruss, 1988; Gehring and Hiromi, 1986; Herr et al., 1988; Kessel and Gruss, 1990; Struhl, 1989; and references therein). Recently evidence has been obtained for an essential function during vertebrate development for some of these genes (Balling et al., 1988, 1989; Chisaka and Capecchi, 1991; Cho et al., 1991; Harvey and Melton, 1988; Kessel et al., 1990; McMahon and Bradley, 1990; Ruiz i Altaba and Melton, 1989; Wright et al., 1989). The structural and functional conservation of these genes suggests that despite different kinds of cell movements during gastrulation and neurulation, vertebrates share a common genetic program of development (Dressier and Gruss, 1988; Kimmel, 1989).

One particularly interesting class of genes is the pax gene family which presently contains eight members in vertebrates (Burri et al., 1989; Dressier et al., 1988; Walther et al., 1991). All pax genes analysed thus far are expressed during early development of the mouse in a temporally and spatially restricted manner (Kessel and Gruss, 1990). With the exception of pax-1, pax genes are expressed in the developing nervous system along the entire length of the hindbrain and spinal cord and in specific parts of the brain. The various pax genes are expressed in specific dorsoventrally restricted domains (Deutsch et al., 1988; Dressier et al., 1990; Goulding et al., 1991; Jostes et al., 1991; Nomes et al., 1990; Walther and Gruss, 1991). Certain features of their expression patterns, like expression of pax-2 and pax-8 in the embryonic kidney and the developing eye (Dressier et al., 1990), suggest that pax genes may function in inductive interactions. To examine the degree of evolutionary conservation of the pax genes and as a first step towards analyzing their functions, we isolated pax genes from the zebrafish, Brachydanio rerio. Here we report the sequence and embryonic expression pattern of the zebrafish pax-6 gene. Our comparative analysis with its murine homolog demonstrates that this gene has been more conserved both in sequence and expression pattern than the wnt-1 (Molven et al., 1991), engrailed-related (eng) (Fjose et al., 1988; Holland and Williams, 1990) or hox genes (Njplstad et al., 1990).

All embryos used in this study were obtained from the Oregon AB line. Embryos and zebrafish were maintained as described previously (Westerfield, 1989). Embryos were staged by hours postfertilization at 28.5°C (h).

The following fragments were used to screen the cDNA library: a 2 kb Saw HI pax-3 cDNA fragment (Goulding et al., 1991), a 487 bp EcoRI fragment from a pax-2 cDNA, pC31A, containing paired-box sequences (Dressier et al., 1990; provided by H. Fickenscher), a genomic 320 bp HindIII - Xbal fragment containing paired-box sequences of pax-4 (Walther et al., 1991), and a 1.7 kb EcoRI pax-6 cDNA fragment (Walther and Gruss, 1991). Subclones of cDNA n108 (Fig. 1A) cloned into the EcoRV site of the Bluescript vector (Stratagene) were used for Northern-blot analysis and in situ hybridization (Fig. 1A). Probe A is a 436 BamHI - EcoRI fragment from the 5′ end. Probe B is a 42 bp oligonucleotide with the sequence of the paired-box insertion of zipax-6b (synthesized by the Biotechnology laboratory, University of Oregon). Probe C is a 140 bp HindIII - Ndel fragment from the 3′ end of the coding region. A 250 bp Sau3A cDNA fragment for eng-1 and an 890 bp EcoRI cDNA fragment for eng-2 were used for in situ hybridization (Ekker and Westerfield, unpublished data). Digoxigenin-labelled RNA-probes were synthesized with the Genius-Kit (Boehringer) according to the manufacturer’s specifications.

Approximately 1.6 × 10⁶ clones of a λZAP library prepared from 20 to 28 h zebrafish embyros (made by R. Riggleman and K. Heide, kindly provided by D. Grunwald) were screened under conditions of low stringency (43% formamide, 37°C) with probes from the mouse paired-box containing genes pax- 2, pax-3, pax-5 and pax-6 as described previously (Breier et al., 1988; Colberg-Poley et al., 1985). 140 clones were isolated and after three rounds of purification 80 clones remained positive. 21 of these clones were identified as cDNAs corresponding to the pax-6 gene by sequencing, restriction mapping and hybridization with each other at conditions of high stringency. Positive clones were excised from the phage by the protocol recommended by the manufacturer (Stratagene).

Both strands of overlapping subclones of the cDNAs were sequenced as double-stranded DNA using the sequenase kit (USB). Sequences were analysed using the GCG program package (Devereux et al., 1984).

For preparation of total RNA, embryos were lysed with 4 M guanidinium thiocyanate and the RNA was isolated with a CsCl gradient as described previously (Püschel et al., 1990). Poly(A)⁺ RNA was prepared from total RNA with the MicroFastTrack kit (Invitrogen). Northern blots were prepared according to standard procedures and hybridized at 42°C in 50% formamide, 5 × SSC, 0.5% SDS and 5 × Den-hardts. Washes were done as described previously (Breier et al., 1988; Colberg-Poley et al., 1985).

In situ hybridization to 8 μ m paraffin sections of embryos was done using standard procedures (Wilkinson and Green, 1990; M. Kessel, personal communication). Embryos were fixed overnight in 4% paraformaldehyde in PBS at 4°C and dehydrated in a graded alcohol series for 15 minutes at each step and embedded in paraplast plus. Hybridization of whole embryos with RNA probes was done based on established procedures (Tautz and Pfeifle, 1989; S. Schulte-Merker, personal communicaton). Embryos were fixed overnight in 4% paraformaldehyde in PBS and subjected to the following treatments at room temperature, if not stated otherwise: twice 5 minutes in PBS, twice 5 minutes in methanol, 1 hour to several days in methanol at – 20°C, 5 minutes in 50% methanol in PBST (1 × PBS, 0.1% Tween-20 (Biorad)), 5 minutes in 30% methanol in PBST, twice 5 minutes in PBST, 20 minutes in 4% paraformaldehyde in PBS, twice 5 minutes in PBST, 1 to 20 minutes (depending on the age of the embryos) in 50 μ g/ml proteinase K (Boehringer) in PBST, 5 minutes in PBST, 20 minutes in 4% paraformaldehyde in PBS, twice 5 minutes in PBST, 10 minutes in acetylation mix (100 mM triethanoamine, 1/400 volume acetic anhydride), twice 10 minutes in PBST. After this treatment, the embryos were transferred to Eppendorf-tubes with hybridization-mix (50% formamide, 5 × SSC, 0.1% Tween-20, 50 μ g/ml heparin (Sigma), 100 μ g/ml sonicated salmon sperm DNA (Pharmacia), 10 μ g/ml yeast tRNA (Sigma)) and prehybridized at 65°C for 1 to 4 hours. Afterwards the hybridization mix was replaced by new hybridization mix containing digoxigenin-labeled RNA probes at 1 to 2.5 ng/ μl final concentration and was then incubated overnight. An equal number of embryos, processed in parallel without hybridizing to an RNA probe, were used to preadsorb the anti-digoxigenin antibody. The probe was denatured for 5 minutes at 68°C and chilled on ice. After hybridization, the embryos were transferred to 50% formamide, 2 × SSC and washed for 1 hour at 65°C. The following washes were then used: three times 10 minutes in 2 × SSC at 37°C, 1 hour in PBST containing 20 μ g/ml RNAase A (Sigma) and 100 units/ml RNAase T1 (Boehringer) at 37°C, 10 minutes in 2 × SSC at 37°C, 1 hour in 50% formam-ide/2 × SSC at 65°C, 15 minutes in 2 × SSC at 55°C., twice 15 minutes in 0.2 × SSC at 55°C, 5 minutes in PBST. Embryos were incubated for 1 to 4 hours in PBS containing 2% normal sheep serum, 0.25% Tween-20 and 0.25% Triton X-100. In parallel, embryos were used to preabsorb the anti-digoxigenin antibody - alkaline phosphatase conjugate (Boehringer) in the same buffer at a dilution of 1 to 8000. The embryos were then incubated with the preadsorbed antibody overnight at 4°C. After four 30 minutes washes with the blocking solution at room temperature and three washes for 10 minutes in reaction buffer (100 mM Tris-HCl pH 7.9 or 9.5, 50 mM MgCl₂, 100 mM NaCl, 0.1% Tween-20, 1 mM levamisol (Sigma), the embryos were stained using 175 μ g/ml 5-bromo-4-chloro-3-indolyl phosphate and 337.5 μ g/mi nitroblue tétrazolium salt in reaction buffer for 5 to 24 hours.

No signal was observed with either method when using sense probes as a control (data not shown). Signals observed in whole mount in situ hybridization experiments were also seen when using sections.

We screened a cDNA library prepared from 20 to 28 h zebrafish embyros and isolated 21 clones of the zebrafish gene homologous to the murine pax-6 gene (Walther et al., 1991; Walther and Gruss, 1991) as shown by sequence analysis (see below). These 21 cDNAs fall into two main classes, which differ at their 3′ ends (I and II, Fig. 1A) based on restriction mapping and sequencing. The difference probably results from internal priming at an A-rich sequence at position 2130 of the sequence (Fig. 2) which is completely contained within clones extending further 3′. No poly(A) addition signal is found at this position. Two forms of cDNAs were found by sequencing the paired boxes of several cDNAs. One corresponds to the described murine cDNA (Walther and Gruss, 1991) and is named zfpax-6a in this paper (represented by clones n8, nl5, nl30, n9, nl30), the second, represented by two clones (n108, nl6), contained a 14 amino acid (42 bp) insertion after amino acid 45 of the paired box (termed zfpax-6b, Figs. 1A, 2). The longest open reading frame codes for a protein of 437 aminoacids which contains both a paired and a homeobox (see below). The complete sequence of 2.8 kb contains 500 bp 5′ nontranslated sequence and a 3′ nontranslated region of 950 bp. Northern analysis of poly(A)⁺ RNA from zebrafish embryos of 36 h shows a single transcript of approximately 3 kb size (Fig. 1B). Thus our cDNA clones are probably close to full length. Southern blot analysis showed that pax-6 is a single copy gene (A. Fritz and A.W.P., unpublished results).

Comparison of the conceptual translation of the longest open reading frame (ORF) to that of other isolated pax genes showed that the 21 isolated cDNA clones were most similar to the murine pax-6 gene. The coding sequences are 80% identical at the nucleotide level and 97% identical at the amino acid level. Two gaps of 3 and 1 amino acids, respectively, were introduced for the alignment (Fig. 3). The longest ORF of the zebrafish pax-6 gene contains 19 more amino acids at the amino terminus than its murine counterpart. The first ATG shows a higher degree of similarity to the Kozak consensus sequence than the second (Kozak, 1987). The homolog of the differentially spliced mRNA, zfpax-6b, was also found in the mouse. It contains an insert with an identical sequence at the identical position (Walther and Gruss, 1991). At position 82 of the paired domain, the fish gene contains a glycine instead of a serine; the homeodomain is identical. The high degree of conservation also extends outside these domains. Surprisingly, we found that the last 113 bp, including the poly(A) addition signal, are 90% identical to the corresponding sequence of the murine gene (nucleotides 2686 - 2799 in Fig. 2, Walther and Gruss, 1991). This region is separated from the coding sequence by 850 bp which show no appreciable amount of similarity to the murine homolog. Because of the high degree of sequence identity and the almost identical expression pattern (see below), we conclude that we have isolated the zebrafish homolog of the murine pax-6 gene. The partial sequence of pax[zf-a] recently described by Krauss et al. (1991) is identical to our sequence and thus, pax[zf-a] probably represents an independent isolate of the pax-6 homolog.

We analysed expression of the zebrafish pax-6 gene between 9 h and 3 days of development with in situ hybridization of RNA probes to paraffin sections and to whole embryos. Two different probes derived from sequences common to zipax-6a and zfpax-6b produced the same pattern. A probe specific for zfpax-6b, containing just the 42 bp insert in the paired box failed to detect expression at any stage analysed (data not shown), probably because of its small size. Thus, the isolation from a cDNA library is the only evidence for expression of this variant. We detected expression first around 10 h (100% epiboly, Fig. 4A-D). Shortly after epiboly is completed and before the tailbud is visible, an oval-shaped domain in the prospective prosencephalon expresses pax-6 (Fig. 4A-D). It contains a stripe of cells at its posterior margin which express higher levels of pax-6 than cells farther anterior. Shortly later, additional expression appears in the prospective rhombencephalon on both sides of the midline (Fig. 4A-D). By 14 h (10 - 11 somites; Figs. 4E,F, 5A-C.E) the anterior expression domain includes the optic vesicles, but excludes the optic stalk. Expression in the optic vesicle is highest in the lateral and posterior regions (Figs. 5A,E; 9A, B) which will invaginate to form the optic cup. The anteroposterior and dorso ventral boundaries of expression are identical in the prosencephalon and the optic vesicle (Figs. 4F, 9A,B and not shown). In addition, the overlying ectoderm, which is known in other species to interact with the optic vesicles to form the lens, contains pax-6 mRNA (Fig. 5E). Most of the hindbrain and spinal cord expresses pax-6 at this stage excluding only the most dorsal and ventral cells (Fig. 5D). Expression extends along the entire anteroposterior axis (Figs. 4E, 5A,B) and remains essentially the same through 20 h (Fig. 5F and data not shown).

At 24 h, cells in the ventral part of the hindbrain and spinal cord excluding the floorplate and cells adjacent to it express pax-6 (Figs. 5G, 6B-C). Expression extends anteriorly to the border between the met- and myelencephalon (Figs. 5F, 6A,B). The anterior boundary of expression in the hindbrain is crescent-shaped when viewed in horizontal sections (Fig. 5F) and extends to the middle of the Rol rhombomere near the midline (Fig. 6A,B). A second expression domain includes the diencephalon, parts of the telencephalon and the olfactory bulb (Figs. 5F, 6A,B). In additon, the eyes, including both the optic cup and the lens, express pax-6 (Fig. 6A, and not shown). Expression in the brain is more restricted at 32 h (Fig. 7A,B) and 36 h (Fig. 7C- H). The dorsal part of the diencephalon contains high levels of pax-6 transcripts up to the border between the di- and mesencephalon (Fig. 7A-H). Two distinct groups of cells at this border (Fig. 7C,D, arrow) and in the pituitary gland (Fig. 7A, B, arrow) also express pax- 6. These groups are contained within the broader regions that expressed pax-6 at earlier stages.

During the following 2 days of development, expression levels drop significantly in the spinal cord (data not shown) and become more restricted in the brain to smaller groups of cells in the di- and telencephalon (Fig. 8A-D). By 3 days, expression in the developing eye is restricted to the ganglion cell layer (Fig. 8D). The overall pattern of expression at this stage suggests that particular nuclei of the brain could specifically express pax-6. However, since specific brain regions of older zebrafish have yet to be identified, we did not unequivocally assign the detected signals to particular nuclei.

pax-6 is expressed in the prospective brain at a time when the eng genes are activated (Hatta et al., 1991). To determine the expression borders of pax-6 relative to the eng genes, we performed double-hybridizations to 14 h embryos with probes for pax-6 and eng-1 or eng-2. Both eng genes are expressed in the gap of pax-6 activity at the midbrain - hindbrain border (Fig. 9). The pax-6 and eng-1 expression domains are adjacent but separated by approximately three to five cells on each side of the eng-1 stripe (Fig. 8B, C). eng-2 is expressed in a narrower stripe in the middle of the eng-1 expression domain (Fig. 8A, D).

Sequence comparisons of the murine and fish pax-6 ORFs revealed a remarkably high identity of 97% at the amino acid level. This conservation is considerably greater than that of the fish homologs of hox-2.2, wnt-1, eng-1, eng-2 and the hox-7 genes which are all in the range of 70% identical to their murine counterparts (Fjose et al., 1988; Holland, 1991; Holland and Williams, 1990; Molven et al., 1991; Nj01stad et al., 1990; Ekker and Westerfield, unpublished data). More importantly, the conservation is seen throughout the entire coding sequence; whereas in the case of the hox or eng genes, only the homeoboxes and other discrete regions show more than 90% identity to their mouse homologs. In accordance with the conservation of the sequence, the expression pattern of pax-6 is also nearly identical in terms of tissue and regional specificity in the two species (Walther and Gruss, 1991). Both genes are expressed in the brain, the ventral part of the spinal cord, the olfactory bulb, the pituitary gland and the eye.

Both mice and zebrafish express a variant form of the pax-6 message, containing a 14 amino acid insertion in the paired domain. The complete conservation of this variant argues for its importance, although the functional consequences of this insertion are presently unclear. Recent biochemical analyses of the paired-box-containing genes showed that they are DNA-binding transcription factors (Chalepakis et al., 1991; Goulding et al., 1991; Treisman et al., 1991) as had been inferred previously from the presence of DNA-binding domains. Since the 14 amino acids are inserted into the paired domain they may affect DNA-binding abilities of the protein, thus providing an additional mode for developmental regulation.

We found an even more striking degree of conservation at the end of the 3′ non-translated region than in the coding sequence (89% compared to 80%). Although the function of this region is unknown, it is possible that it is involved in the regulation of expression, either at the transcriptional or posttran-scriptional level.

The high conservation of the pax-6 gene between mouse and fish, whose ancestors separated 300 million years ago, is remarkable and suggests that this gene serves an important function. In addition to the homeodomains and paired domains, another yet to be identified functional domain in the pax-6 protein could explain the high degree of homology in the C-terminal half.

pax-6 is expressed very early during neurulation in the zebrafish embryo. Initially in the brain (at 10 h) a broad, region expresses pax-6 including the prospective optic vesicle, the telencephalon and the diencephalon. It has been shown that the mesoderm specifies the regional identity of the overlying neuroectoderm in various vertebrates (Frohman et al., 1990; Gurdon, 1987; Hemmati et al., 1990; Sive et al., 1989). Expression of pax-6 could be an early response to this induction (Gurdon, 1987). pax-6 expression is first seen around the time epiboly is complete, at a time when other genes implicated in regional specification of the ectoderm, like the eng genes, are activated (Davis et al., 1991; Hatta et al., 1991). Our double labelling demonstrates that both eng-1 and eng-2 are expressed in the gap of pax-6 activity in the mes- and metencephalon. Interestingly, the eng-1 domain is not directly adjacent to, but separated from, pax-6 by several cells expressing neither gene. This could indicate a more complex subdivision of this region of the brain than previously suggested by analysis of Hox-2 and Krox-20 expression in the hindbrain. The Hox-2 and Krox-20 expression domains coincide with neuromere borders (Stem and Keynes, 1988; Lumsden, 1990; Wilkinson, 1990; and references therein) and are, therefore, directly adjacent to one another, whereas the eng and pax-6 domains are separated and seem to cross over neuromere borders.

pax, eng and other genes could form overlapping and complementary domains in the forebrain as they do in the hindbrain and spinal cord. Their early expression makes them good candidates for functioning in the regionalization of the forebrain in response to inductive signals from the mesoderm by regulating directly or indirectly other genes during the differentiation of neuronal structures, pax-6 expression in the brain becomes progessively more restricted during development. Thus pax-6 could, in addition, be involved in the specification of specific subsets of neurons, as has been shown in Drosophila for segmentation genes (Doe et al., 1988).

In the spinal cord, the initially broad region of expression becomes restricted to a ventral domain by 24 h. The notochord plays a key role in the dorsoventral patterning of the spinal cord (Yamada, 1990). In the mouse, the pax genes have overlapping and complementary expression domains in the spinal cord and are good candidates for genes responding to signals from the notochord and/or the floor plate. Recent evidence, that pax-gene expression is regulated by signals originating from the notochord (Goulding et al., unpublished data), supports this notion.

Two other sites of pax-6 expression, the pituitary gland and the optic vesicle and eye, are consistent with its role in induction. The floor of the diencephalon and Rathke’s pouch interact to form the pituitary gland and the optic vesicle interacts with the overlying ectoderm and induces the formation of the lens (Gurdon, 1987; and references therein). It remains to be determined if the same kind of interactions take place in the zebrafish. The isolation of the pax-6 gene provides a useful marker to study these processes. As the eye develops, pax-6 expression becomes progressively re-stricted; at 24 h both optic cup and lens express pax-6 whereas at 3 days only the ganglion cell layer shows pax-6 activity.

The expression pattern of pax-6 points to several developmental processes in which it may serve a regulatory function, pax-6 could be involved in the early regionalization of the brain and in dorsoventral patterning of the spinal cord. The expression in specific parts of the brain later in development suggests a function in specifying subsets of cells during neurogenesis. Thus, the pax genes may serve several different functions, as previously described for the segmentation genes in Drosophila (Doe et al., 1988). pax-6 is expressed predominantly in structures where inductive, cell-to-cell interactions regulate development. Thus, it may be activated by inductive signals and may, itself, be part, of a signaling pathway.

We thank C. Walther, H. Fickenscher, M. Goulding and M. Ekker for providing probes; R. Riggleman, K. Helde and D. Grunwald for providing the cDNA library; S. Schulte-Merker for providing his whole-mount in situ hybridization protocol; G.I. Kristjanson, E. Reinhard, T. Schilling and R. BreMiller for helpful discussions and suggestions; and S. Krauss et al. for a preprint of their unpublished manuscript. Special thanks are given to C. Walther for sharing unpublished results and for constant discussion. We also thank J. Eisen, C. Kimmel and E. Reinhard for helpful comments on the manuscript. This work was supported by the Max-Planck-Society, the NIH (grant HD22486) and the McKnight Foundation. A.W.P was supported by the BMFT and the EMBO.

The nucleotide sequences reported in this paper will appear in the EMBL, Gen Bank, and DDBJ nucleotide sequence databases under the accession number X63183.