In early embryonic development, the brain is divided into three main regions along the anteroposterior axis: the forebrain, midbrain and hindbrain. Through retroviral insertional mutagenesis and chemical mutagenesis experiments in zebrafish, we have isolated mutations that cause abnormal hindbrain organization and a failure of the midbrain-hindbrain boundary (MHB) to form, a region that acts as an organizer for the adjacent brain regions. The mutations fail to complement the spiel-ohne-grenzen (spg) mutation, which causes a similar phenotype, but for which the affected gene is unknown. We show through genetic mapping, cloning of the proviral insertion site and allele sequencing that spg mutations disrupt pou2, a gene encoding the Pou2 transcription factor. Based on chromosomal synteny, phylogenetic sequence comparison, and expression and functional data, we suggest that pou2 is the zebrafish ortholog of mouse Oct3/Oct4 and human POU5F1. For the mammalian genes, a function in brain development has so far not been described. In the absence of functional pou2, expression of markers for the midbrain, MHB and the hindbrain primordium (pax2.1, wnt1, krox20) are severely reduced, correlating with the neuroectoderm-specific expression phase of pou2. Injection of pou2 mRNA restores these defects in spg mutant embryos, but does not activate these markers ectopically, demonstrating a permissive role for pou2. Injections of pou2-morpholinos phenocopy the spg phenotype at low concentration, further proving that spg encodes pou2. Two observations suggest that pou2 has an additional earlier function: higher pou2-morpholino concentrations specifically cause a pre-gastrula arrest of cell division and morphogenesis, and expression of pou2 mRNA itself is reduced in spg-homozygous embryos at this stage. These experiments suggest two roles for pou2. Initially, Pou2 functions during early proliferation and morphogenesis of the blastomeres, similar to Oct3/4 in mammals during formation of the inner cell mass. During zebrafish brain formation, Pou2 then functions a second time to activate gene expression in the midbrain and hindbrain primordium, which is reflected at later stages in the specific lack in spg embryos of the MHB and associated defects in the mid- and hindbrain.
The mechanisms involved in the formation of the various subdivisions of the brain share many common elements among vertebrates, making it possible to generalize information learned in one organism to all vertebrates. Because zebrafish can be used effectively in large-scale classical genetics, there is great potential to use zebrafish to discover new factors involved in the early brain development of all vertebrates.
In vertebrates, regional identity within the neurectoderm is established during gastrulation. Originally it was believed that the dorsal gastrula organizer (Spemann, 1938; Nieuwkoop, 1973; Kodjabachian and Lemaire, 1998; Solnica-Krezel, 1999), which establishes dorsoventral (DV) patterning through BMP, Noggin and Chordin signaling (Hammerschmidt et al., 1996; Bauer et al., 1998; Fekany-Lee et al., 2000), is also responsible for the anteroposterior (AP) determination of the neurectoderm. However it has since been shown by studying several zebrafish mutants that the AP axis is not disrupted in the absence of many of these signals (Barth et al., 1999; Fekany-Lee et al., 2000). Thus, how the AP axis of the neurectoderm is established in the early embryo remains unknown.
Cell fates of the neurectoderm are in large part demarcated by a sequence of gene expression patterns that are established early in the developmental program. In zebrafish, otx1 and otx2 are first observed at about 65% epiboly in what will ultimately become the fore- and midbrain (Li et al., 1994; Mori et al., 1994). fgf8, wnt1 and pax2.1 (pax2a – Zebrafish Information Network) appear next, defining the mid-hindbrain region (Krauss et al., 1991a; Kelly and Moon, 1995; Lun and Brand, 1998; Reifers et al., 1998; Rhinn and Brand, 2001). Slightly later, pax6 expression is evident in both the forebrain and hindbrain regions (Krauss et al., 1991b; Amirthalingam et al., 1995). Thus, the neuroepithelium is divided into a minimum of three domains with unique molecular identity before any brain structures are visible under the microscope. However, many aspects of how this regionalization is established remain elusive.
The boundary region that separates the midbrain and the hindbrain (MHB, also known as isthmus), serves as an organizer for midbrain and cerebellum development. Grafts of isthmic tissue induce ectopic midbrain and/or cerebellar tissue in chick embryos, depending on the location of the graft (Martinez et al., 1991; Puelles et al., 1996; Rhinn and Brand, 2001; Wurst and Bally-Cuif, 2001). In zebrafish, large-scale genetic screens have identified many mutations that affect early developmental stages (Driever et al., 1996; Haffter et al., 1996; Amsterdam et al., 1999; Talbot and Hopkins, 2000). Among the isolated mutations, three genes affect the MHB region in a specific manner when mutated. The acerebellar (ace) mutation inactivates the zebrafish fgf8 gene (Brand et al., 1996; Reifers et al., 1998). no isthmus (noi) mutations are in the zebrafish pax2.1 gene (Brand et al., 1996; Lun and Brand, 1998). Mutations in the third gene, spiel-ohne-grenzen (spg), cause lack of the cerebellum and the isthmus, with a variably enlarged tectum and ear defects (Schier et al., 1996; Driever et al., 1997), a phenotype resembling that of acerebellar mutants. We have identified the gene altered by spg mutations as pou2, which encodes a POU homeobox transcription factor.
POU homeobox transcription factors have been described in different species, and function during nervous system development. Four original transcription factors Pit-1, Oct-1, Oct-2, and Unc-86 share a region of homology that defines a unique class with a bipartite DNA-binding domain that consists of a POU-specific domain (POUS) and a homeodomain (POUH) (for a review, see Ryan and Rosenfeld, 1997).
In two ongoing genetic screens, one using pseudotyped retroviruses as an insertional mutagen (Lin et al., 1994; Gaiano et al., 1996; Amsterdam et al., 1999; Talbot and Hopkins, 2000) and the second a chemical mutagenesis screen in haploid embryos (C. Klisa, N. Morita, G. R., W. Huttner and M. B., unpublished), mutations were isolated that mimicked the spg phenotype. We demonstrate here, by complementation testing and genetic mapping that the newly isolated mutations fail to complement each other and are allelic to spg. A chemically induced allele and the proviral integration in the insertional mutation map to and disrupt the pou2 gene (Takeda et al., 1994; Hauptmann and Gerster, 1995), which encodes a class V homeodomain transcription factor with a POU domain (Ryan and Rosenfeld, 1997). Injection of antisense morpholinos phenocopy the spg phenotype. Based on the syntenic chromosomal position and phylogenetic sequence comparisons, we suggest that pou2 is orthologous to the mammalian Oct3/Oct4/Pou5f1 genes. We show that pou2 expression is crucial to proper formation and development of the zebrafish midbrain, isthmus and hindbrain primordia, in keeping with its specific expression in these territories. We also demonstrate that the expression of pou2 is essential for the early onset of the MHB, establishing pou2 as a regulator upstream of pax2.1 in formation of the mid-hindbrain boundary. In the accompanying paper, we also report a detailed phenotypic analysis of spg function in early brain development, leading us to propose that spg/pou2 acts to confer regionally specific competence to respond to Fgf8 signaling during brain development (Reim and Brand, 2002).
MATERIALS AND METHODS
Normal practices were used for fish rearing and mating (Westerfield, 1995; Brand et al., 2002). Mutant carriers were identified by PCR or random intercrosses. Mutant embryos were obtained by inbreeding heterozygous carriers. Morphological features and time of development at 28.5°C were used to stage embryos.
Linker-mediated PCR was used to identify the integration responsible for the mutant phenotype and to isolate the adjacent genomic DNA. Briefly, DNA was isolated from either tail biopsy from the adult carrier fish or whole phenotypic embryos digested with MseI and ligated to linkers made from annealed primers with the sequences 5′-GTAATACGACTCACTATAGGGCACGCGTGGTCGACGGCCCGGGCTGGT-3′ and 5′-TAACCAGCCCAACTTCGAATTAAA-3′. PCR was performed using the oligonucleotides 5′-GTAATACGACTCACTATAGGGC-3′ and 5′-CGTGTATCCAATAAACCCTCTTGC-3′. A nested PCR reaction was performed using the primers 5′-ACTATAGGGCACGCGTGGT-3′ and 5′-AACAAGGAAGTAGCAAAGCTGCTC-3′. The samples were run on a 7% non-denaturing polyacrylamide gel, stained with Ethidium Bromide and destained with distilled water. The single band common to all the carrier adults and all the mutant embryos was excised from the acrylamide gel, eluted re-amplified and sequenced. Further genomic sequence was obtained using linker-mediated PCR from wild-type genomic DNA using primers based on the sequence obtained from the first amplification.
Linkage of the proviral integration to the mutant phenotype was established using PCR. DNA from phenotypically mutant embryos or phenotypically wild-type embryos was isolated and PCR was performed using the oligos 5′-ACGTCTTCTTTTAAGGAGACTCTGACTA-3′, 5′-ACTCACATCCTGAGGGTTCTCG-3′ and 5′-CGTGTATCCAATAAACCCTCTTGC-3′. These primers differentiated between homozygous wild type, heterozygous and homozygous mutants for the proviral locus. Inverse PCR was performed as described previously Becker et al. (Becker et al., 1998). Genetic linkage of the spg locus to the pou2 locus was carried out by a PCR-based mapping method (Knapik et al., 1996) showing co-segregation of the spg phenotype with the SSR (simple sequence repeat) marker z13467. The mapping marker serves as a PCR primer. Wild-type and spge713 DNA of haploid embryos was used. Haploid embryos were generated as described (Westerfield, 1995).
Sequencing of mutant alleles
cDNA from homozygous mutant spg and control wild-type sibling embryos was used to amplify the coding region of the mutant spg alleles, then subcloned and sequenced. Surprisingly, spgm216 and spge713 showed the same base change, causing a Lys to Pro exchange in the POU-specific domain. The result was reproduced many times. Both mutations were induced in different genetic backgrounds in different laboratories, but ultimately derive from a common Oregon AB background. We recovered spge713 from a mutagenized strain that was genetically marked with the pigment mutation golden; moreover, spge713 was initially linked to a second unrelated mutation that was induced on the same chromosome and that we subsequently removed by recombination. The same base change in pou2 was found on the chromosomes before and after the recombination event. We therefore think it unlikely that spge713 arose owing to contamination of our golden stock that was used in the mutagenesis. Because the two mutations carry the same molecular change, and also share a number of closely linked SNIPs, the mutation may have been present as a polymorphism in the AB wild-type population that was mutagenized; spge713 and spgm216 would then be re-isolates of a background mutation that was present in the original AB wild-type population at a low frequency. Alternatively, a mutation in this position could generate a particularly strong, and hence easily detected, phenotype (for an example of two identical mutations of Pax2 arising in mouse and human) (Favor et al., 1996). Until this issue is resolved, we will treat the two mutations as separate alleles.
In situ hybridization and antibody staining
In situ hybridization and antibody detection were performed using protocols described previously. Probes used were for pou2 (Takeda et al., 1994), pax2.1 (previously pax[zf-b]) (Krauss et al., 1991a), krox20 (egr2 – Zebrafish Information Network) (Oxtoby and Jowett, 1993) and pax6 (Li et al., 1994). Anti-acetylated tubulin antibody (Sigma) and 3A10 monoclonal antibody (Furley et al., 1990) were also used. Distribution of co-injected lacZ mRNA is visualized by staining with anti-β-gal antibody (Promega, 1:500) after in situ hybridization (Dornseifer et al., 1997).
RNA and morpholino injections
RNA was transcribed from plasmid pCS2+ (Rupp et al., 1994), using conventional methods, and injected as described previously (Reifers et al., 1998). Briefly, embryos were harvested from matings of wild-type fish 10 minutes after the appearance of the first fertilized eggs. Embryos were put in 20 ml of Holtfreter’s solution (Westerfield, 1995) and several drops of 30 mg/ml pronase was added. At the first sign of chorion separation, the embryos were washed five to six times with Holtfreter’s solution. Embryos were placed on a agarose ramp under Holtfreter’s solution and RNA is injected into the cell cytoplasm from the one- or two-cell stage. Embryos were then moved to 24-well dishes filled with Holtfreter’s solution and placed at 28.5°C for 8-24 hours. Morpholinos were designed and synthesized by Gene-Tools, LLC (Corvallis, OR). The sequence of the anti-pou2 morpholino-1 is 5′-CGCTCTCTCCGTCATCTTTCCGCTA-3′ (underlined bases were exchanged to generate the 4 bp mismatch control morpholino, Mo1-ctrl), the morpholino-2 is 5′-TTCAAACAAGAAAGCGTAAAGACTG-3′. The sequence of the control morpholino is 5′-CCTCTTACCTCAGTTACAATTTATA-3′ (Gene Tools).
Mutagenesis screens yield new alleles of spg
In the course of a large-scale, insertional mutagenesis screen for mutations that affect embryonic development (Amsterdam et al., 1999), we isolated a mutant carrier family Hi349. The mutation has several different phenotypes that display significant variation in their expression. The earliest visible phenotype is a loss of the fold at the midbrain hindbrain boundary (MHB), which is clearly visible after the first 24 hours of development (Fig. 1). Phenotypes range from a reduction of the cerebellum to a complete deletion of the MHB area without visible change in the anterior tectum or the posterior hindbrain regions. Other phenotypes include mis-shapen and smaller otic vesicles [commonly containing only a single otolith (Fig. 1B)], a variable defect in the length of the tail (Fig. 1D) and a significant reduction in the escape reflex response to touch (not shown). Two additional alleles were isolated in a haploid ENU mutagenesis screen for mutants affecting development of the MHB organizer, as initially judged by loss of fgf8 expression at the MHB (N. Morita, C. Klisa, G. R., W. Huttner and M. B., unpublished). Homozygotes for the mutation e713, which causes lack of fgf8 expression, lack the MHB during pharyngula stages. A similar, albeit weaker disruption was found in the mutant e68, which retains a partially formed MHB (Fig. 1C). The similarity of the mutant phenotypes to the mutant spg isolated in an earlier chemical mutagenesis screen (Schier et al., 1996) suggested that we had isolated new alleles of this mutation. In complementation test crosses with the previously isolated MHB mutants noi, ace and spg (Brand et al., 1996; Schier et al., 1996), Hi349, e713 and e68 all failed to complement spgm216, and are therefore new alleles of spg. From this point onwards, the mutations will be known as spghi349, spge713 and spge68. For further phenotypic analysis we concentrated on the strong alleles spghi349or spge713, which give very similar phenotypes when homozygous.
We used antibodies against acetylated tubulin at 30 hours postfertilization (hpf) on whole mutant embryos homozygous for spghi349or spge713 to examine the architecture of the axonal scaffold in the brain. The segmented appearance of the wild-type scaffold is severely disrupted in the hindbrain of mutant embryos (Fig. 1E,F) and the cerebellar commissure at the MHB is missing, probably because of the missing MHB, whereas other parts of the brain appear unaffected in the mutant embryos (not shown). Therefore, in addition to the MHB, overall hindbrain development is also disrupted in spg mutants. In wild-type embryos, monoclonal antibody 3A10 clearly stains the Mauthner neurons in whole-mount staining at 30 hpf. In the mutant embryos, the Mauthner neurons, which normally form in rhombomere 4 of the hindbrain, are completely absent (Fig. 1G,H), suggesting that very specific disruption of cell types also occurs in the hindbrain of spg mutants.
spg mutations are in the zebrafish pou2 gene
Four out of 14 mating pairs from family had offspring that showed the mutant phenotype. Tail biopsies from each spghi349 heterozygous carrier fish were taken and the genomic DNA was isolated. Using linker-mediated PCR and a primer specific to proviral sequences, we identified a single proviral insert that was present in both parents of each pair whose offspring showed the mutant phenotype, but was never present in both parents in pairs that did not display the phenotype (data not shown). The fragment containing the genomic DNA that was 3′ of the identified insertion site was excised from a polyacrylamide gel and sequenced. The genomic DNA showed that the provirus had integrated into an exon of the zebrafish pou2 gene, a transcription factor identified previously (Takeda et al., 1994; Hauptmann and Gerster, 1995). Using Southern blot analysis and inverse PCR, the genomic DNA adjacent to the 5′ side of the virus was also isolated and it showed that the DNA on the other side of the provirus continued in the same pou2 exon. The proviral integration was 875 nucleotides from the starting methionine in the cDNA sequence, 37 amino acids into the POU-specific domain (Fig. 2A). As this integration disrupts the POU-specific domain, and truncates the protein ahead of the entire POU homeodomain, it is likely that this integration generates a null allele.
Independently, we mapped one of the chemically induced alleles, spge713, and found it to be linked to the same chromosomal location as pou2 (Fig. 2B,C) (Geisler et al., 1999). z13467 is a marker located on chromosome 21 at 32.4 cM, very close to pou2. This marker gives different size bands, depending on the genetic background, and we found no recombinants between this marker and the spg mutant phenotype among 39 haploid embryos obtained from a heterozygous carrier female (21 spg and 18 wild-type; Fig. 2B), showing that spg is linked genetically to pou2. To determine the molecular defect in the ENU alleles, we used pou2-specific primers to amplify the coding region of the ENU alleles. Sequencing revealed a single amino acid exchange in spgm216 and spge713 from Lys to Pro in helix one of the POU homeodomain, which most probably disrupts the function of the protein. Both mutations were induced on a similar genetic background (AB), but were found in independent mutagenesis experiments with strains that carried different genetic markers, suggesting that this mutation may have been present in the AB background at a low frequency (see also Materials and Methods). No amino acid exchange was found in the coding region of the weak allele spge68, which we hypothesize might therefore affect the regulatory sequences of pou2.
Once the molecular nature of spg alleles had been identified, we designed primers that distinguished between homozygous wild-type, homozygous mutant and heterozygous embryos for the proviral insertion. Nineteen phenotypically mutant embryos and, 19 phenotypically wild-type embryos were tested with these primer sets, 19/19 mutant embryos were homozygous for the Pou2 insertion while 0/19 phenotypically wild-type embryos where homozygous at that locus (data not shown). In situ hybridization of embryos from spghi349 carrier parents using pou2 sequences as the probe showed 22 of 78 embryos with a severe reduction in staining starting at 40-50% epiboly (not shown); expression is totally abolished at the end of gastrulation (see below, Fig. 4E). A reduction of pou2 mRNA is also observed in homozygous spgm216 and spg e713 embryos, but slightly later from the beginning of somitogenesis onwards (not shown). These observations argue that Pou2 is involved in feedback regulation of its own expression, that it already functions before the onset of gastrulation in all embryonic cells and that spghi349 is the strongest allele.
Previous studies of pou2 had not resolved the relationship between zebrafish pou2 and mammalian Pou genes. Phylogenetic sequence comparison with the full-length sequence furthermore shows that these genes fall into the same class V subfamily of POU transcription factors (Fig. 2E). In the conserved POU domain, Pou2 is 74% identical and 89% similar to murine Oct3, and in the in the homeodomain 71% and 84%, respectively. Outside these functional domains there is little conservation, giving lower values over the entire amino acid sequence (30% identity, 40% similarity). Importantly, the chromosomal position of pou2 shows that it is located in an area that is syntenic with mammalian chromosomes, and which contains the Oct3/Oct4/Pou5f1 genes in mice and humans (Fig. 2D) (Woods et al., 2000). Furthermore, mouse Oct3/Oct4 is expressed in the neural plate at the right time, and injection of mouse Oct3/Oct4 rescues pax2.1 expression in spg homozygotes (Reim and Brand, 2002). We therefore suggest that pou2 is the ortholog of the mammalian Oct3/Oct4/Pou5f1 genes.
Pou2 is required for pax2.1 and krox20 expression
To analyze the spg phenotype further, we performed in situ hybridization on embryos collected from matings of two spghi349 or spge713 heterozygous adult carrier fish. Using krox20 and pax6, we found disruptions in the normal organization of the hindbrain at very early stages of development. Krox20 expression normally starts as a stripe of expression in the presumptive rhombomere 3 (Fig. 3A). Later, at the three-somite stage, two parallel bands of expression define the location for rhombomeres 3 and 5 (Fig. 3C). In spghi349 homozygous embryos, the expression of krox20 shows a marked reduction even at the very first sign of krox20 staining, and in particular the shape of the expression domain becomes pointed and smaller close to the midline (Fig. 3B,D). pax6 also shows alterations in expression. The normal regions of expression for pax6 are in the forebrain and hindbrain separated by the MHB that is devoid of expression (Fig. 3E). In spghi349 homozygous embryos, the forebrain domain of pax6 expression expands posteriorly and nearly fuses with its hindbrain expression domain, creating one large domain of expression by the 10 somite stage (Fig. 3F).
Using pax2.1, a marker for the presumptive midbrain and MHB, showed that one quarter of the embryos had a strong reduction of pax2.1 expression (Fig. 3H), as reported previously (Schier et al., 1996). Because of the slight variability associated with the spg phenotype, we wanted to ascertain that the early defective gene expression is specific to homozygous spg embryos. We collected individual embryos either lacking pax2.1 expression or with wild-type expression and isolated their genomic DNA. PCR analysis confirmed that the embryos without pax2.1 expression were homozygous for the spghi349 integration (Fig. 2F). Induction of pax2.1 was unaffected in other areas such as the otic placode (not shown).
To understand the mechanism by which Pou2 activates pax2.1 transcription, we examined the relative distribution of these genes during formation of the midbrain and hindbrain primordia by whole-mount double ISH. Both Takeda et al. (Takeda et al., 1994) and Hauptmann and Gerster (Hauptmann and Gerster, 1995) reported that at tailbud- to two-somite stages these genes occupy non-overlapping expression domains in the midbrain and anterior hindbrain primordium, respectively, but earlier stages of neural development were not examined. We find that at 80%-90% epiboly, the Pou2 domain is initially broader and encompasses the pax2.1 domain at its anterior end (Fig. 4F). Pou2 is therefore available to regulate pax2.1 in the same cells at this stage, which may explain why the midbrain is affected in spg mutants. As development of the embryo progresses, the overlap of expression reduces in size, until at the tailbud stage, the majority of the pou2 expression lies in the anterior hindbrain primordium (Fig. 4G), posterior to the pax2.1 domain. As described above, pou2 is required at this stage for expression of krox20. otx2 and gbx1 are involved in forebrain, midbrain and hindbrain development, respectively, and determine the site of MHB formation, including pax2.1 expression (K. Lun and M. B., unpublished). Expression of otx2 and gbx1 is initially normal in spghi349 homozygotes (Reim and Brand, 2002), arguing that spg specifically affects activation of pax2.1 and krox20 expression, without disrupting earlier neuroectodermal gene expression.
Rescue of gene expression at the MHB
Previous reports have suggested that misexpression of pou2 had no effect on the developing embryo (Takeda et al., 1994). We studied pou2 function further by injecting 120 pg pou2 mRNA into one side of two-cell stage embryos from a clutch of eggs derived from spghi349 or spgm216 heterozygous parents. Before fixation, the injected embryos looked morphologically normal. The embryos were fixed at either the tailbud or the three-somite stage and stained for pax2.1. Injection of pou2 mRNA rescued the reduced pax2.1 expression at the MHB of mutant embryos in all embryos (Fig. 5B,D,F,H; n=447). In addition, pax2.1 expression was slightly upregulated and broadened in its dorsoventral extent in both the wild-type and spg embryos on the injected side of the embryos (Fig. 5A-C), perhaps owing to a slight shortening of the injected embryos. In a similar fashion, we were able to rescue the expression of the MHB markers wnt1 and her5, as well as the rhombomere marker krox20, by pou2 overexpression (Fig. 5E-H; data not shown). This demonstrates that pou2 injection rescues the loss of pax2.1 expression and of the other markers, which are normally reduced in spg mutants, adding further proof that the loss of pou2 is what is causing the spg mutant phenotypes. Importantly, pax2.1, wnt1 and krox20 are not expressed outside their normal domain of expression on the injected side in either mutant or wild-type embryos, nor is pax2.1 expression in the intermediate mesoderm affected in these embryos (Fig. 5A-H). This suggests that the induction of these genes by pou2 is specific to the midbrain and hindbrain region, and that pou2 is necessary but not sufficient for their expression. In addition to rescuing expression of MHB- and hindbrain markers, higher doses of injected pou2 mRNA cause a slight overall morphologically aberrant development; the basis for this phenotype currently not clear. The morphogenetic defects may be due to nonspecific effects from higher than normal early expression, specific effects on cell migration and axis formation during gastrulation, a slight dorsalizing potency or a combination of all.
Morpholino-induced knock-down of pou2 function
pou2 mRNA is initially maternally supplied to the embryo and then supplemented with zygotic expression (Fig. 4) (Takeda et al., 1994; Hauptmann and Gerster, 1995). Recently, antisense inhibition using oligos with morpholino chemistry has proven to be extremely effective in eliminating protein translation in zebrafish embryos (Nasevicius and Ekker, 2000). To determine if maternally contributed pou2 message has a role, we injected two different pou2 antisense morpholinos into wild-type embryos (Table 1) (morpholino 2 required a higher concentration to work than morpholino 1). At low doses (2-5 ng, depending on the morpholino), the injections phenocopy the spg phenotype. Injected embryos displayed a loss of cerebellar structures and a disorganization of the hindbrain, as well as an ear and tail phenotype that is similar to spg mutant embryos (compare Fig. 6B with Fig. 1B,C). This effect is highly specific, because control morpholinos with randomized or a four bp mismatch sequence had no effect, and similar phenotypes were not observed with other morpholinos in our laboratory. Moreover, embryos injected with morpholinos show a strong reduction of pax2.1, wnt1 and krox20 at the end of gastrulation (Fig. 6G-L), whereas expression of otx2 and gbx1 at 60-70% of epiboly and the tailbud stage is never affected by morpholinos (not shown), thus reproducing also other phenotypic traits of spg homozygous mutants.
When injected with intermediate doses (4-5 ng and 8 ng, respectively, depending on the morpholino) of pou2 antisense morpholinos, embryos arrested between 50% and 70% of epiboly. At high doses of pou2 antisense morpholino (6 ng), all embryos stopped cell division and morphogenesis during blastula stages (sphere-dome stage; Fig. 6D; Table 1), which we assume also reflects inactivation of the maternally supplied mRNA. The cells remained large and failed to migrate around the yolk normally, even 10 hours post-injection. By 12 hours, the blastoderm detaches from the yolk and none of the embryos survived until the next day. Embryos injected with a control morpholino displayed none of this behavior (Fig. 6C). This effect is similar to the effect caused by the injection of an RNA encoding a truncated version of the protein (Takeda et al., 1994).
To determine the specificity of the morpholinos, we attempted to rescue the developmental arrest by co-injecting 16 ng morpholino 2 at the one-cell stage together with 100 pg pou2 mRNA(-5′UTR), which is devoid of the 5′UTR and hence is not recognizable by morpholino 2 (Fig. 6F). In contrast to morpholino 2, which exclusively interferes with sequences within the 5′UTR, morpholino 1 can also bind within the coding region of pou2 mRNA. Embryos co-injected with 8 ng morpholino 1 and 100 pg pou2 mRNA(-5′UTR) are released from the pre-gastrula arrest, but the rescue is less efficient compared with the experiment carried out with morpholino 2 (Fig. 6E). The reversion of the morpholino induced phenotype by pou2 mRNA confirms that the pre-gastrula arrest is a specific consequence of the ‘knockdown’ of maternal and early zygotic pou2 message. Further analysis of this phenotype is currently under way.
We have demonstrated that spiel-ohne-grenzen mutations affect the gene for the POU domain transcription factor Pou2, and have presented evidence that pou2 is orthologous to the mammalian Oct3/Oct4/Pou5f1 genes. We have examined pou2 function through mutant analysis, morpholino-inhibition and pou2 mRNA injections. pou2 expression is essential for proper development of the midbrain, hindbrain, ear and body axis. The normal organization of the midbrain and hindbrain, including correct onset of pax2.1 and krox20 expression, is significantly disrupted in spg mutants, and may cause secondarily the defects in ear development. Many of the structures in the MHB are reduced or absent. Our data suggest two main periods of requirement for spg function. The first requirement is during the early cleavage period, and may be similar to the known function of the orthologous mammalian genes during differentiation and subsequent morphogenesis of the inner cell mass. Analysis of the second, tissue-specific period reveals a novel function of this gene during early brain regionalization in development of the midbrain-hindbrain organizer and the hindbrain, where it functions as a regulator of pax2.1 and krox20 expression. In the accompanying study, we examine the brain phenotype of spg mutants further and present evidence that pou2 exerts these effects through regulating competence to respond to Fgf8 signaling (Reim and Brand, 2002).
Nature of the spg alleles
We have focussed our phenotypic analysis on two strong alleles, spghi349 and spge713, which give essentially indistinguishable phenotypes. Two pieces of evidence point to the spghi349 allele as being a null allele. First, the integration disrupts an exon with a 6000 base-pair proviral integration. The disruption is located in the middle of the POUs domain, essentially disrupting both DNA binding domains. Second, in situ hybridization with pou2 antisense RNA probes on mutant embryos displayed a strong reduction in signal of gene expression, both for the insertion allele and for spge713. Pou2 may therefore be involved in positive-feedback regulation of its own expression. An alternative possibility, that the insertion or mutation destabilizes pou2 mRNA, is less likely to be correct, because pou2 expression in tail somites at 26 hours is normal in homozygous spghi349 mutants (data not shown). The loss of ubiquitous pou2 expression in spghi349 (see Fig. 4D,E) is already apparent before onset of gastrulation, around the time of normal onset of zygotic transcription. This has two important consequences for interpretation of pou2 function: (1) prior to onset of gastrulation, pou2 apparently functions in all cells of the embryo, raising the question whether, and if so how, this function contributes to the brain phenotype of the mutants; (2) homozygotes for the spghi349 allele most probably lack all zygotic pou2 function, but should still have normal maternal contribution. Thus, the phenotype of spg mutants is likely to be the result of the embryo using the substantial maternal wild-type pou2 RNA/protein during the early stages of development, and only after the embryo switches to zygotic expression can the effects of the mutation be seen.
The role of pou2 in early embryonic development
The evidence presented suggests that there are two different functions for pou2 in the developing embryo, an early general role, and later in development, a distinct and neuroectoderm-specific role. Maternal pou2 message is localized to the oocyte cortex, and is restricted to the animal pole in the freshly laid oocyte (Howley and Ho, 2000). In the zygote, the message then becomes restricted to the deep layer cells that will become the actual zebrafish embryo (Hauptmann and Gerster, 1995). In situ hybridization data suggest a transition from maternal to zygotic pou2 expression between 30% and 40% of epiboly. During gastrulation, pou2 mRNA is still present throughout the epiblast, but becomes gradually restricted towards the end of gastrulation. We found that between 80% and 100% epiboly, the expression domain comprises both the midbrain- and hindbrain primordium, and only later does it become restricted to the hindbrain primordium (Fig. 4B,C) (Takeda et al., 1994).
Based on the syntenic chromosomal position and the phylogenetic sequence comparison (Fig. 2), we have argued that zebrafish pou2 and murine Oct3/Oct4 are orthologous. Several additional arguments support orthology, demonstrating a similar role for zebrafish pou2 and murine Oct3/Oct4:
(1) Both our morpholino results and the overexpression of the truncated t-pou2 by Takeda et al. (Takeda et al., 1994) suggest that the early role of spg/pou2 may be to maintain the cells in an undifferentiated, rapidly dividing state. In the absence of functional pou2 protein, either through antisense inhibition or through squelching with the truncated protein, the cells stop dividing and gastrulation is arrested. This role is very reminiscent of the orthologous murine POU transcription factor Oct4 (also known as Oct3 and Pou5f1) which is thought to maintain cells of the inner cell mass in a pluripotent state (Schöler et al., 1989; Okazawa et al., 1991; Nichols et al., 1998).
(2) Precursor cells of the enveloping layer of the zebrafish embryo are the first lineage to differentiate during early cleavage stages (Kimmel et al., 1995), and as they do so, they loose pou2 expression (Hauptmann and Gerster, 1995). Murine Oct4 similarly acts during an early differentiation step: Oct4 expression is restricted to the pluripotent inner cell mass (ICM) of the developing mouse blastocyst, but is excluded from the extra-embryonic cells, and serves to maintain ICM in an undifferentiated state. Cells that lose Oct4 expression become trophectoderm, and ICM cells require Oct4 to maintain Fgf4 expression (Nichols et al., 1998).
(3) Mouse ES cells overexpressing Oct4 produce primitive endoderm and mesoderm (Niwa et al., 2001). We find that loss of pou2 function causes an early failure of endoderm differentiation, as seen with the marker sox17, which suggests that pou2/Oct4 could act more generally as a lineage switch in endoderm formation (Reim and Brand, 2002).
(4) Oct4–/– mouse embryos arrest at the expanded blastocyst stage (Nichols et al., 1998), a stage that may be analogous to the arrest shown in the morpholino injected zebrafish embryos, and controls activity of the adhesion modulator osteopontin at pre-implantation stages, which is thought to be crucial for controlling migration of mouse hypoblast cells at blastocyst stages (Botquin et al., 1998). Speculatively, at the pre-gastrula stage Pou2 may act in a similar way to Oct4, by preventing the dividing cells from restrictions in fate and hence, migration, allowing the cell number to expand enough for the embryo to develop normally. Using the spg mutants and pou2 morpholino inhibition, we currently examine these issues in more detail.
(5) During early brain development, Oct4 is expressed in the murine neural plate, and injection of murine Oct4 mRNA into spg mutants rescues the spg mutant phenotype (Reim and Brand, 2002).
The role of pou2 in AP patterning of the neuroectoderm
The nature of the spg phenotype suggests that pou2 also has a more specific function in later embryonic development. One would not anticipate the specific deletion of a brain structure if the role of pou2 were simply to maintain pluripotency throughout the embryo; in mice a brain-specific role has not been reported. However, Oct4 is also expressed at the appropriate early neural plate stages in mice, and injection of mouse Oct4 rescues the loss of pax2.1 expression normally seen in spg mutant embryos (Reim and Brand, 2002), suggesting that Oct4 may have a similar function in mice. In zebrafish, the refinement in expression during the transition from 80% epiboly to 100% epiboly is key to the observed phenotype. By the completion of epiboly, pou2 expression is restricted to a T-shaped region. Our analysis shows that this region of expression actually corresponds initially to the presumptive midbrain and hindbrain (Fig. 4F,G), and only later to the presumptive hindbrain and the developing neural tube. As development progresses, the expression of pou2 becomes more and more restricted, first to the second and fourth rhombomeres and the end of the tail, then by the seven-somite stage, expression is seen only in the tip of the tail (Fig. 4C,D) (Hauptmann and Gerster, 1995). Diffuse and slight expression in the di-, mes- and metencephalon can still be seen at 28 hpf, and expression in the tail is detected until 34 hpf (Hauptmann and Gerster, 1995). This pattern of expression directly correlates with many of the observed phenotypes. At 30 hpf, the hindbrain architecture is severely affected, and our evidence suggests that the defects originate at much earlier stages. Indeed, already during activation of the first hindbrain marker genes, such as krox20 and pou2 itself, it is evident that the hindbrain primordium is not properly specified. The segmentation and organization of the seven rhombomeres becomes less distinct and often it is not possible to recognize all of them. A particularly interesting observation is that the Mauthner neurons are absent in spghi349 homozygous embryos. The cell body for the Mauthner neuron resides in r4 and in situ hybridization shows a concentration of pou2 signal in r4 at the three to five somite stage (Fig. 4C,D) which suggests pou2 may have important cell autonomous functions in establishing specific cell fates in r4. A perhaps similar situation has been observed in the determination of the NB4-2 neuroblast lineage in Drosophila (Yang et al., 1993; Bhat and Schedl, 1994). Pdm-1 and Pdm-2 are POU domain proteins expressed at high levels in the ganglion mother cell (GMC-1). As the level of Pdm-1 and Pdm-2 drops, the GMC-1 cell divides to form an RP2 motoneuron and a sibling cell. When Pdm-1 or Pdm-2 is overexpressed, the GMC-1 cell divides without a drop in Pdm level which generates two new GMC-1 cells. The level of Pdm then drops, and a duplication of the RP2 motoneurons results. The absence of both Pdm-1 and Pdm-2 in mutants completely prevents the formation of RP2 motoneurons.
The most interesting observed phenotype of spg mutants is the absence of the MHB. As previously shown for the slightly weaker allele spgm216 (Schier et al., 1996), we have shown that for spghi349 and spge713, in the absence of the pou2 gene, the expression of pax2.1 at the MHB is severely affected. The first observed stage of regionally specific pou2 expression directly correlates with the first observed expression of pax2.1 (Krauss et al., 1991a; Lun and Brand, 1998). In zebrafish embryos, pax2.1 transcripts are first seen during late gastrulation. Pou2 expression initially overlaps pax2.1 expression and transplantation results suggest that this induction is cell-autonomous (Reim and Brand, 2002). Importantly, otx2 and gbx1 staining are not affected in spg mutant embryos, showing that the MHB patterning process upstream of pax2.1 is not affected in a general way (Reim and Brand, 2002). In combination with the injection results presented here, this suggests that the absence of pax2.1 in spg mutants is a direct result of the loss of pou2 and not a secondary defect from a general disorganization of the neuroectoderm, and that pou2 is an important positive upstream regulator of pax2.1. This is also consistent with our observation that injection of pou2 mRNA into the zygote can rescue the expression of pax2.1 staining in the MHB. We also observe that the absence of pou2 expression negatively affects the expression of wnt1 and krox20, suggesting that the role of pou2 in neurectoderm patterning is not exclusively to induce pax2.1 expression, but extends to activating gene expression in the hindbrain. Further experiments addressing the function of spg/pou2, as well as the similarity of the brain phenotype of spg and acerebellar/fgf8 mutant embryos (Reifers et al., 1998), demonstrate an important role for Pou2 in mediating competence to respond to Fgf8 signaling (Reim and Brand, 2002).
Injection at higher levels of pou2 also has other effects on embryonic development. High levels of pou2 mis-expression slightly upregulate and expand pax2.1 expression slightly in the dorsoventral direction at the MHB. However, pax2.1 staining is not seen in other areas of pou2 mis-expression and the endogenous expression of pax2.1 in other parts of the embryo is unresponsive to mis-expression of pou2, showing that pou2 is necessary but not sufficient to induce pax2.1 expression.
pax6 expression in early zebrafish embryos normally first appears at approximately 10 hpf in the presumptive forebrain region. A second region of expression in the hindbrain appears shortly thereafter. pax6 expression is normally completely excluded from the midbrain and MHB. In spg mutants, forebrain expression of pax6 expands into where the MHB should be and actually fuses with the posterior domain of expression forming one large pax6 domain (Schier et al., 1996; Reim and Brand, 2002). This suggests that in the wild-type environment, the midbrain or MHB is inhibiting the expansion of pax6 expression, an observation also made in mouse double knockouts of Pax2/Pax5. Additionally, in both mice and chicken, pax6 can inhibit pax2 expression when overexpressed (Mastick et al., 1997; Schwarz et al., 1999; Matsunaga et al., 2000), making it likely that inhibitory signals from each domain are reinforcing the sharpness of the boundaries.
The role of pou2 in ear development
spg mutant embryos typically show a slightly smaller otic vesicle. However, unlike the situation in the MHB, pax2.1 expression in the otic placode appears unaffected in the mutant embryos. It is therefore unlikely that the ear defect is a result of a failure to induce proper expression of pax2.1 in the ear. There are several examples of zebrafish mutations that affect the hindbrain initially and as a consequence of that hindbrain defect, display defects in ear development (Moens et al., 1996; Mendonsa and Riley, 1999). Several studies have also implicated signals emanating from the hindbrain in the induction of the otic placode (Gallagher et al., 1996; Mahmood et al., 1996; Groves and Bronner-Fraser, 2000; S. Léger and M. B., unpublished). Therefore it is likely that the defects in the ear seen in the spg mutation are a result of the primary defects in the hindbrain.
We thank the members of the Hopkins and Brand laboratories for support and thoughtful discussion, especially Christiane Klisa and Noriyuki Morita for conducting the ENU mutagenesis screen and Muriel Rhinn for genetic complementation tests of mutant carrier fishes. Furthermore we thank Carl-Philipp Heisenberg and Florian Raible for critical reading of the manuscript. This work was supported by grants from the Deutsche Forschungsgemeinschaft and the European Union (Bio4-CT98-0309, QLRT-2000-02310) to M. B. and from the National Institutes of Health and Amgen to N. H.