|Home | About | Journals | Submit | Contact Us | Français|
emx3 is first expressed in prospective telencephalic cells at the anterior border of the zebrafish neural plate. Knockdown of Emx3 function by morpholino reduces the expression of markers specific to dorsal telencephalon, and impairs axon tract formation. Rescue of both early and late markers requires low-level expression of emx3 at the one- or two-somite stage. Higher emx3 expression levels cause dorsal telencephalic markers to expand ventrally, which points to a possible role of emx3 in specifying dorsal telencephalon and a potential new function for Wnt/beta-catenin pathway activation. In contrast to mice, where Emx2 plays a major role in dorsal telencephalic development, knockdown of zebrafish Emx2 apparently does not affect telencephalic development. Similarly, Emx1 knockdown has little effect. Previously, emx3 was thought to be fish-specific. However, we found all three emx orthologs in Xenopus tropicalis and opossum (Monodelphis domestica) genomes, indicating that emx3 was present in an ancestral tetrapod genome.
Despite recent advances in our understanding of the role of signals from the anterior neural plate in specifying the telencephalon, relatively little is known about how these signals are interpreted by prospective telencephalic cells. Some factors expressed at the anterior edge of the neural plate, such as tlc, fgf3, and fgf8, mimic properties of the organizer activity produced by the first row of anterior neural plate cells in zebrafish, or the anterior neural ridge in other model vertebrates (Houart et al., 1998, 2002; Crossley et al., 2001; Echevarria et al., 2003; Walshe and Mason, 2003; Storm et al., 2006). In zebrafish, emx3, previously named emx1 (Morita et al., 1995; Houart et al., 1998; Derobert et al., 2002; Kawahara and Dawid, 2002), is one of the earliest transcription factors expressed in response to anterior neural plate organizer activity. However, although emx3 has been used extensively as a telencephalic marker, virtually nothing is known about its function.
Two paralogs of emx3, emx1 and emx2, have been identified in all currently characterized vertebrate genomes; emx3, however, has only been found in a shark and several teleost genomes, not in the tetrapod lineage (Derobert et al., 2002; Kawahara and Dawid, 2002). Because sharks are ancestral to bony fish and tetrapods, it is possible that emx3 was present in a common ancestor of fish and tetrapods and was subsequently lost in the tetrapod lineage (Derobert et al., 2002; Kawahara and Dawid, 2002).
In mice, Emx2 patterns the rostrocaudal axis of the cerebral cortex and is necessary for cortical radial layering (O’Leary et al., 2007). When Emx2 function is lost, positional information is shifted caudally and medially, and gene expression patterning and thalamocortical connectivity are affected. In addition, Emx2 knockout mice have small olfactory bulbs and poorly differentiated cortical cell layers. Expression of Emx1 and Emx2 is highest in dorsomedial structures such as the hippocampus and dentate gyrus. In these regions, cortical defects are most pronounced when gene function is lost (Yoshida et al., 1997; Mallamaci et al., 2000a; Bishop et al., 2002, 2003; Shinozaki et al., 2004). The reduced hippocampal structures in Emx2 knockout mice are thought to result from altered proliferation patterns (Tole et al., 2000; Heins et al., 2001; von Frowein et al., 2006). The role of Emx1 is less apparent, because Emx1 mutant mice have minor structural defects in the forebrain (Qiu et al., 1996; Yoshida et al., 1997). Homozygous mutant Emx1 mice are viable and behavioral tests showed impaired motor learning and reduced learning-induced hippocampal neurogenesis in adult mutants (Hong et al., 2007).
Zebrafish emx1 and emx2 are expressed in largely overlapping domains in the telencephalon and in nonoverlapping domains in the diencephalon. emx1 expression starts around the 17-somite stage (17.5 hr; hours postfertilization) in dorsal telencephalon and continues until after 24 hr (Kawahara and Dawid, 2002). emx2 is expressed in ventral diencephalon at early somite stages and starts to be expressed throughout the dorsal telencephalon by the 17-somite stage, with the exception of a small domain in the dorsomedial telencephalon (Morita et al., 1995). At bud stage (10 hr), the horseshoe-shaped expression domain of zebrafish emx3 (previously called emx1) and other genes at the anterior edge of the neural plate demarcate the prospective telencephalon (Wilson and Houart, 2004). Within this domain, emx3 is expressed in a two- to six-cell-wide stripe (Morita et al., 1995; Houart et al., 1998). During somitogenesis, emx3 mRNA is expressed in a gradient from high levels in dorsal telencephalon to low levels in ventral telencephalon and is absent from the most ventral third of the telencephalon (Morita et al., 1995). This expression pattern suggests a role of emx3 in dorsoventral patterning of the telencephalon.
In this study, we analyzed the roles of emx1, emx2, and emx3 in zebrafish forebrain development. We also identified an Emx3 ortholog in the gray short-tailed opossum (Monodelphis domestica) and in Xenopus tropicalis genomes. This is the first report of tetrapod genomes that contain all three emx genes. Our results suggest that emx3, but not emx1 or emx2, is necessary in a time- and dose-dependent manner for differentiation and axon tract formation of dorsal telencephalic neurons, and may involve activation of the Wnt/beta-catenin pathway. We find that the roles of zebrafish emx genes differ from those of their mouse orthologs.
To identify paralogy and orthology relationships with Emx proteins in other species, we searched several genome sequence databases for previously unpublished sequences encoding Emx3. The stickleback (Gasterosteus aculeatus) and tetraodon (Tetraodon nigroviridis) genomes contain predicted full-length genes encoding Emx1, Emx2, and Emx3 proteins. We also found three potential emx3 exon sequences in the elephant shark (Callorhinchus milii) genome, version 1.4x (Venkatesh et al., 2007). Although these putative emx3 exons do not lie on a contiguous stretch of DNA, they encode a putative full-length Emx3 protein with conserved intron positions and 10 of 11 amino acids unique to Emx3, compared with Emx1 and Emx2. Using translated BLAST (Altschul et al., 1997), we found a full-length Emx3 sequence encoded in the opossum (Monodelphis domestica) genome and a Xenopus tropicalis cDNA. The encoded proteins are 57% and 69% identical to dogfish Emx3, and contain 4 and 6 of 11 Emx3-specific residues (Derobert et al., 2002). The region between the amino terminal and homeodomain is most divergent from other Emx3 proteins. Both Emx1 and Emx2 are present in the Xenopus tropicalis genome. The opossum genome also contains a full-length predicted Emx2 gene and a partial predicted Emx1 exon 1 that groups with mammalian Emx1 sequences in a neighbor-joining tree of translated exon 1 sequences (data not shown). The opossum partial Emx1 exon1 sequence contains a frameshift and a stop codon, and may, therefore, be a pseudogene. These results suggest that both the opossum and Xenopus tropicalis genomes contain all three Emx paralogs, although the opossum emx1 gene may no longer be functional.
We aligned these new predicted Emx3 sequences with representative Emx sequences encoded by full-length genes or expressed sequence tags to calculate a neighbor-joining phylogenetic tree (Fig. 1A). The tree confirms three distinct groups of predicted Emx protein sequences with high bootstrap values (Fig. 1A). In addition to overall sequence conservation within the group, the sequences share most amino acids that were previously identified as unique to the Emx paralog groups (Derobert et al., 2002).
To investigate orthology relationships of emx1, emx2, and emx3 genes further and to analyze whether other tetrapod Emx3 genes have been lost (Derobert et al., 2002; Kawahara and Dawid, 2002), we compared conserved synteny of Emx genes with neighboring genes in assembled fish and tetrapod genomes (Fig. 1B–D). We identified four genes within 0.3 Mb of zebrafish emx1 and five genes within 1 Mb of zebrafish emx2, with shared synteny in tetrapod genomes. The orthologs of these genes were located within 3 Mb or less of the respective Emx ortholog on stickleback, tetraodon, chick, mouse, and human chromosomes (Fig. 1B,C). Sfxn5 is an immediate neighbor of Emx1 in all analyzed species including the opossum Emx1 gene fragment. Rab11fip2 is an immediate neighbor of Emx2 in all species analyzed except zebrafish, where this gene is located on a different chromosome. Within 2 Mb of emx3 in zebrafish, we found eight genes with conserved synteny in opossum and at least two other species including human, mouse, or chicken (Fig. 1D). Most of these genes are also close to emx3 in stickleback and tetraodon. The genes Emx3, Tmed9, and B4galt7 lie within a 50-kb region in all the genomes we analyzed. Thus, although Emx3 appears to be missing from chicken, human, and mouse genomes, the synteny of the region around emx1, emx2, and emx3 has been conserved throughout vertebrate genome evolution. These results are consistent with the hypothesis that Emx3 was present in a common vertebrate ancestor and has either been lost from several tetrapod sublineages, or has not yet been discovered in human, mouse, or chicken.
Based on their low bootstrap values and short branch lengths, Emx1 and Emx3 may be more closely related to each other than to Emx2. This interpretation predicts that we should find more conserved syntenic relationships between Emx3 and Emx1 than between these genes and Emx2. To test this hypothesis, we examined conserved synteny among Emx genes from different groups. We found, for example, that Nanos1 is syntenic with Emx2 in tetrapods but syntenic with emx1 in fish; stickleback has an additional nanos1 gene adjacent to emx3. smyd5 is closely linked to zebrafish and stickleback emx3, whereas the chicken, mouse, and human Smyd5 orthologs are linked to Emx1. Thus, syntenic relationships with these markers do not support the conclusion that Emx1 and Emx3 are more closely related to each other than to Emx2.
To complement previously published gene expression analyses (Morita et al., 1995; Houart et al., 1998; Kawahara and Dawid, 2002), we studied the mediolateral expression patterns of emx1, emx2, and emx3 in the telencephalon (Fig. 2). emx1 is expressed directly adjacent and dorsal to the olfactory organ, and is excluded from the ventricular zone. emx2 and emx3 are expressed throughout the dorsal telencephalon including the ventricular zone, with the strongest expression adjacent and dorsal to the olfactory organ. emx2 and emx3, but not emx1, are also expressed in the ventral diencephalon (Fig. 2A–C).
We tested whether knock down of emx1, emx2, or emx3 function with morpholino antisense oligonucleotides (MO) affects the regionalization and differentiation of the forebrain. We used a combination of two translation blocking morpholinos against each transcript because this combination resulted in the strongest phenotype for emx3MOs (Fig. 3D) and emx1MOs (Fig. 3B). To block morpholino toxicity (Robu et al., 2007), all experimental and control injections contained tp53MO (see the Experimental Procedures section). emxMO thus refers to injection of two translation blocking morpholinos and tp53MO unless otherwise indicated.
To assess the specificity of emx3MO, we injected single and combinations of emx3MO morpholinos and consistently obtained the same phenotypes (Fig. 3A,D,E). To assess the effectiveness of emx3MO, we identified two aberrantly spliced products by reverse transcriptase-polymerase chain reaction (RT-PCR) of embryos injected with emx3 splice blocking morpholinos: an exon1–exon3 fusion (Fig. 3H) that introduces a frameshift leading to an early stop codon in exon3, and exon 3 spliced to the cryptic splice site within exon 2 (Fig. 3I) that produces an in-frame deletion leading to loss of the second helix of the homeodomain. Because at least the exon1–exon3 out of frame fusion is likely not functional, and we because obtain a phenotype with these emx3 splice morpholinos, it is likely that this phenotype is due to knockdown of emx3. The ability to rescue emx3MO phenotypes with emx3 mRNA that had been modified so it was not complementary to emx3 translation blocking morpholinos (Fig. 6) provides further evidence that the observed emx3MO phenotypes are indeed due to knockdown of emx3. YFP expression in Tg(emx3:YFP)b1200 embryos is lost using translation blocking morpholinos but not with splice blocking morpholinos (data not shown), indicating that translation blocking morpholinos indeed knock down mRNA that is complementary to the MO sequence in its 5′ untranslated region. Together, these results suggest that the emx3 morpholino phenotypes we observe are specific and due to knockdown of emx3 function.
emx1MO-injected embryos showed a slight reduction of eomesa (tbr2; Fig. 3B), tcf4l (not shown), and atoh2b expression (not shown), whereas emx2MO injected embryos did not show changes in the expression patterns of eomesa (Fig. 3C) or any of the other genes tested for phenotypes with emx3MO. We confirmed the absence of phenotypes with an emx2 splice blocking morpholino that abolished a large portion of correctly spliced mRNA as determined by RT-PCR (data not shown) and led to aberrant out of frame splicing of exon 1 to exon 3 (Fig. 3G). Apart from having little or no effect on marker gene expression, coinjection of emx1MO and emx2MO together with emx3MO did not enhance the eomesa expression defect (Fig. 3F), nor any of the emx3MO phenotypes that we describe below (Fig. 4). We suggest that emx1 and emx2 have subtle, or perhaps, later functions in forebrain development, or that the markers we used did not detect their roles.
In embryos injected with emx3MO, several genes had smaller and less intense dorsal telencephalic expression domains than control embryos as tested by in situ hybridization (Fig. 4). These phenotypes were mostly subtle, but all of them were consistently present in >90% of emx3MO injected embryos (n > 18 embryos for each marker). Differences were apparent from the onset of expression of the respective genes in the telencephalon, which for most of the affected genes corresponds approximately to the 18-somite stage (18 hr). Affected markers include the proneural basic helix–loop–helix factor genes tcf4l, atoh2b, and neurod (Fig. 4A–C′), the elavl4 (HuD) RNA binding protein (Fig. 4D–D′), and the vesicular glutamate transporters slc17a6 and slc17a6l (Fig. 4G–H′). Their most strongly affected mRNA expression domains are located adjacent and dorsal to the olfactory placodes and may therefore include the prospective olfactory bulb, the region into which olfactory sensory axons project around 35 hr (Whitlock and Westerfield, 1998). Neuronal differentiation markers for telencephalic neurons that are more broadly expressed in the telencephalon, such as elavl3 (HuC) and reln (Fig. 4E–F′), appear unaffected by emx3MO, although in some embryos, the region directly ventral to the presumptive future olfactory placode appears to be less labeled in some embryos (e.g., as in Fig. 4E,E′ close to the letter t). Similarly, the transcription factors eomesa, dlx2a, lhx5, pax6a, emx1, fezf2 (Fig. 4I–N′), and foxp2, are reduced only in their dorsal telencephalic expression domain, but not in ventral telencephalon or diencephalon. emx2, emx3, and the zinc finger transcription factor, fezf2 (Fig. 4M–M′), are reduced only in the region of strong expression adjacent and dorsal to the olfactory placode, whereas the weaker expression domain that spans the dorsal telencephalon is normal. We observed no changes in expression for genes expressed in ventral telencephalon, such as fzd8a (Fig. 4N–N′), isl1, sfrp5, and the GABAergic neuronal markers gad1 and gad2 (not shown). Markers expressed in the most dorsal part of the telencephalon, such as wnt8b, lef1, and axin2, are also unaffected by emx3 morpholinos, as are the Wnt7b duplicates wnt7ba and wnt7bb (Fig. 4O–P′, and data not shown) that are expressed along the dorsal part of the telencephalic–diencephalic border. The fibroblast growth factor (FGF) pathway components with broad telencephalic expression patterns such as fgf8a (Fig. 4Q–Q′), spry4, and pea3 (not shown) are equally unaltered, as is foxg1 (bf1, Fig. 4R,R′) that is expressed throughout the telencephalon except the most dorsal tip. These unaffected gene expression domains show that the size of the telencephalon and the shapes of cells are normal at 24 hr in emx3MO-injected embryos. From these results, we conclude that emx3 knockdown impairs specification or differentiation of dorsal telencephalic neurons. However, the dorsal telencephalon, as such, is still specified, and loss of emx3 does not appear to affect specification of dorsal vs. ventral telencephalon.
After the 18-somite stage (18 hr), when most changes in marker gene expression become apparent in emx3MO injected embryos, telencephalic neurons extend axons through the supraoptic tract (SOT) into the diencephalon and then, later, across the ventral telencephalon, thereby forming the anterior commissure (AC; Chitnis and Kuwada, 1990; Wilson et al., 1990; Ross et al., 1992). To test for defects in axon formation or pathfinding, we labeled axons with antibodies against acetylated tubulin. By 35 hr, when 92% (n = 39) of control embryos had developed a well-formed AC, 53% (n = 51) of emx3MO-injected embryos had extended only five or fewer axons across the AC (Fig. 5A–D). The SOT of controls developed into a wide band of axons across the telencephalic surface with one slightly separate fascicle at its anterior end. Although the SOT was present after emx3 knockdown, it was thinner and split into two to four irregular fascicles (53%, n = 32; Fig. 5A,B). In addition, the olfactory nerve was defasciculated at its entry point into the telencephalon (Fig. 5C,D), where emx3 expression was also previously reported (Whitlock and Westerfield, 2000). At 2 and 3 days of development, emx3MO injected embryos had fewer presumptive olfactory glomeruli that we labeled with by Bodipy ceramide (Shanmugalingam et al., 2000), indicating that the olfactory bulb was not differentiating normally (Fig. 5E–F). Knockdown of emx1 and emx2 had no observable effects on axon tracts (not shown).
We tested whether cell proliferation defects may underlie the reduced number of differentiated dorsal telencephalic cells after emx3 knockdown. For example, progenitor cells may remain longer in a proliferative state instead of differentiating, potentially producing more proliferative cells or proliferation in ectopic locations. Alternatively, progenitor cells may fail to divide in the absence of emx3, producing fewer differentiated cells. To test whether cells keep proliferating instead of differentiating, or are dividing at ectopic locations at the time we find reduced marker gene expression, we examined emx3MO-injected embryos for changes in the quantity or position of mitotic cells using 5-bromo-2-deoxyuridine (BrdU) to label S-phase nuclei, and antibodies against Ser10-phosphorylated histone H3 (pH3) to label the condensed chromatin of mitotic cells. At 24 hr, the cells that incorporate BrdU during a 5-min incubation period are located in a three- to four-cell-wide area close to the ventricular zone in control and emx3 knockdown embryos; the ventricular zone is slightly expanded in some emx3MO injected embryos (Fig. 5G–H). Mitotic cells labeled by pH3 are located in the ventricular zone of the telencephalon, closer to the ventricular surface than BrdU-labeled S-phase cells, in both control and emx3MO injected embryos (Fig. 5G,H). We counted mitotic cells in the telencephalon in embryos double-labeled with pH3 and fluorescent in situ hybridization for foxg1 and emx3 mRNA to outline the telencephalon. We found no difference in the number of pH3-positive cells in a 54-μm-wide confocal stack in the dorsal or ventral halves of the telencephalon in embryos injected with emx3MO (27 ± 6 cells, 57 ± 6% of total pH3 cells, n = 33 embryos) compared with control embryos (26 ± 5 cells, 61 ± 8% of total, n = 31 embryos) (Fig. 5I–K). The stacks used for counting included all cells in the central ventricle between the left and right halves of the telencephalon. emx3 morpholino injection did not affect the uniform distribution of cells labeled by pH3 in the ventricular zone or at the telencephalon–diencephalon boundary (Fig. 5I,J). These data suggest that knockdown of emx3 has a very small, if any influence on cell division at the stage when gene expression patterns are most affected. Therefore, even though loss of emx3 function keeps cells from differentiating, it does not keep cells in a proliferative state, nor does it cause cells to proliferate at ectopic locations.
To rescue the emx3 morpholino phenotypes without disrupting early development, we conditionally activated green fluorescent protein (GFP)-tagged emx3 mRNA in heterozygous Tg(hsp70l:emx3-myc-GFP)b1202 transgenic embryos that had been injected with emx3MO. A 10-min heat shock at the one-somite or two-somite stage (10.5 hr), shortly after the onset of endogenous emx3 expression at bud stage (10 hr), that results in weak GFP fluorescence throughout the embryo in transgenic siblings, at least partially restores expression of tcf4l, HuD, and eomesa in 86–95% of embryos (n > 7 for each marker), compared with nontransgenic siblings that did not express GFP, without any apparent defects (Fig. 6 and data not shown). Heat shock of 30 min gave rise to embryos with small eyes and ventrally expanded markers of dorsal telencephalon (Fig. 6A,H), as described for overactivation of the canonical Wnt/beta-catenin pathway (van de Water et al., 2001). This phenotype is different from the loss of forebrain and eyes that is most frequently observed upon overactivation of the Wnt/beta-catenin pathway (Stachel et al., 1993; Kelly et al., 1995; Heisenberg et al., 2001; van de Water et al., 2001; Kim et al., 2002). We did not observe rescue of anterior commissure defects; however, anterior commissure was also lost in control embryos upon strong activation of the transgene, as indicated by high levels of GFP expression. This presumably indicates that axon outgrowth is too dose sensitive for this method of rescue. For embryos heat shocked at the five-somite stage (11.5 hr) or later, we observed neither rescue nor overexpression phenotypes (Fig. 6A,D). Heat shock at bud stage failed to induce uniform GFP expression (not shown). Our results indicate that emx3 is required before the five-somite stage for the expression of dorsal telencephalic markers, and that timing and dosage of Emx3 protein are critical requirements for normal function.
We tested whether gain of emx1, emx2, or emx3 function affects marker gene expression or axon phenotypes (Fig. 7). Injection of 12–25 pg of emx1, emx2, or emx3 mRNA per embryo at the one-cell stage strongly dorsalizes and posteriorizes embryos in a concentration-dependent manner. At bud stage, injected embryos are severely egg-shaped (Fig. 7E,M), the notochord is broadened as seen by flh expression (Fig. 7E), the forebrain and eye field markers pax6b (Fig. 7G) and rx3 (not shown) are reduced to a small spot at the animal pole or lost completely, and the posterior central nervous system (CNS) marker egr2b (krox20) is shifted anteriorly (Fig. 7G). At 35 hr, embryos lack eyes and have curled tails (Fig. 7B,C), and the gene expression domains of otx2 (Fig. 7I) and foxb1.2 (mar, not shown) are shifted anteriorly. All of these phenotypes are identical to those seen after activation of the Wnt/beta-catenin signaling pathway by LiCl treatment, after wnt mRNA injection, or in masterblind (axin1) mutant embryos (Fig. 7A–C,H; Stachel et al., 1993; Kelly et al., 1995; Heisenberg et al., 2001; van de Water et al., 2001; Kim et al., 2002). The Wnt/beta-catenin pathway inhibitor dkk1 (Shinya et al., 2000) and the Fgf pathway inhibitor spry4 (Fürthauer et al., 2001) both rescue loss of eyes and curled tails in emx1, emx2, emx3, and wnt8b mRNA injected embryos (Fig. 7H). These results suggest that ectopic emx mRNAs and Wnt/beta-catenin signaling affect the same mechanisms that regulate early dorsal–ventral patterning of the embryo and anterior–posterior development of the central nervous system. The rescue of this phenotype with an FGF inhibitor is expected because many aspects of beta-catenin–dependent dorsalization are mediated by activation of the FGF signaling pathway (Maegawa et al., 2006). Embryos injected with emx mRNA, however, do not resemble control embryos injected with fgf8 mRNA that are not egg-shaped at bud stage, do not lose eyes, and always have a deformed tail. Consistent with our finding that emx morpholino injection does not affect expression of fgf8, nor expression of the FGF downstream genes spry4 and pea3 during somitogenesis or at 24 hr, we consider it unlikely that emx3 is involved in activating FGF signaling. The effect of emx mRNA overexpression is probably not related to the endogenous function of emx in the early embryo because none of the emx genes are expressed before the end of gastrulation (data not shown). Nevertheless, the similarity of overexpression phenotypes both by heat shock and early mRNA expression raise the possibility that emx gene function during development of the telencephalon also involves Wnt/beta-catenin pathway activation.
We used the mRNA overexpression phenotypes to investigate whether Emx3 acts as a transcriptional activator or repressor. We injected mRNA encoding a fusion of the Emx3 homeodomain (emx3hd) with either the transcriptional activator domain of VP16 or the transcriptional repressor domain of Drosophila engrailed (en). The emx3hd-en fusion mRNA gave rise to the same phenotypes as emx3 mRNA, including egg-shaped embryos at bud stage (Fig. 7L,M), whereas embryos that injected with the emx3hd-vp16 fusion did not resemble embryos with overactivated Wnt/beta-catenin pathway at any stage, nor did they resemble emx3MO embryos (Fig. 7K and data not shown). These data suggest that emx3 acts as a transcriptional repressor, but do not exclude a different mode of action in the telencephalon.
Our analysis of emx3 mRNA overexpression indicated that emx3 could possibly play a role in cell signaling. emx3 might enhance signaling in the telencephalon in cells that secrete or receive signals, either cell-autonomously or non–cell-autonomously. We tested for a possible non–cell-autonomous role of emx3 by transplanting wild-type cells labeled with Alexa 555 dextran into emx3MO injected host embryos at blastula stages and analyzed the resulting chimeric embryos for atoh2b expression at 24 hr. We found that only uninjected donor cells express atoh2b mRNA when transplanted into the normal atoh2b expression domain, whereas cells derived from emx3MO injected host embryos never express atoh2b, even when immediately adjacent to the transplanted wild-type cells in the telencephalon or the olfactory placode (n = 7 embryos; Fig. 8A,B). In addition, mosaic, heat shock-induced over-expression of Emx3-myc-GFP in embryos injected with hsp70l:emx3-myc-GFP plasmid DNA induced the expression of emx1 cell-autonomously (Fig. 8C). Thus, we conclude that emx3 acts cell-autonomously in telencephalic cells. We suggest that, if emx3 activates a signaling pathway, it does so in signal receiving cells.
Our phylogenetic analysis of Emx protein sequences and shared syntenies confirms that the vertebrate Emx1, Emx2, and Emx3 genes constitute different paralog groups rather than local tandem duplicates. In addition, we identified all three Emx genes in Xenopus tropicalis and in the opossum (Monodelphis domestica), a marsupial mammal. In other vertebrate genomes that lack the Emx3 ortholog, we identified syntenic regions similar to opossum, suggesting that Emx3 was lost in these species. Opossum and Xenopus Emx3 provide the first evidence for Emx3 genes in tetrapods and rule out the possibility that emx3 was lost from the tetrapod lineage (Derobert et al., 2002; Kawahara and Dawid, 2002). We suggest that Emx3 is an ancient gene that was lost in some, and preserved in other vertebrate sublineages.
We analyzed emx3 gene function by morpholino knockdown. Our results indicate that emx3 functions specifically in the differentiation of dorsal telencephalic neurons. Expression of many different types of genes in this region, such as transcription factors, neurogenic genes, and neurotransmitter transporters, are reduced by emx3 knockdown (Figs. 2, ,3).3). These cells probably include but may not be limited to neurons of the prospective olfactory bulb, the region adjacent and dorsal to the olfactory placode into which the olfactory nerve axons later extend (Whitlock and Westerfield, 1998). Some of the affected genes have additional expression domains in ventral telencephalon or other areas of the brain, and these appear normal in emx3MO-injected embryos. Thus, the effects of emx3 seem to be specific to dorsal telencephalic development.
However, expression of most markers is not completely abolished, and emx3MO injection affects broadly expressed neuronal differentiation markers such as elavl3 and reln only to a small degree. These observations indicate that either the affected neurons constitute a small fraction of dorsal telencephalic neurons or that the affected neurons differentiate to some extent in emx3MO-injected embryos. We find no ectopic proliferation outside the ventricular zone, which indicates that the undifferentiated cells have exited the cell cycle. We therefore conclude that the major role of emx3 is to activate the expression of genes that are specific to the dorsal telencephalon. In accordance with this interpretation, we were able to activate expression of emx1 ectopically in cells expressing GFP-tagged emx3. The axonal defects and failure to form glomeruli are presumably consequences of aberrant and incomplete neural differentiation.
emx3 is one of the first genes expressed in the neural plate that demarcates prospective dorsal–ventral patterning of the telencephalon. However, phenotypes resulting from emx3 knockdown do not appear until the 18-somite stage (18 hr), whereas rescue of emx3MO injection phenotypes requires expression of emx3 mRNA at the one- or two-somite stage (10.5 hr). Because we cannot exclude the possibility that residual emx3 function from incomplete knockdown masks possible early phenotypes, or that earlier phenotypes are very subtle, it is possible that the late phenotypes are secondary consequences of earlier, undetected functions of Emx3. There are, however, other examples of genes expressed early that apparently function only later in development. fezf2, for example, starts to be expressed even earlier than emx3 in prospective telencephalon, and both mutant and morpholino injected animals lack specific subsets of monoaminergic neurons without displaying early patterning defects (Levkowitz et al., 2003; Jeong et al., 2007).
In this study, knockdown of emx3 alone produced a strong phenotype. We found a slight reduction in expression of few dorsal telencephalic differentiation markers in emx1MO-injected embryos, and we did not observe any phenotypes resulting from emx2 knockdown. We also found no evidence of synergy when knocking down multiple emx genes. It is possible that emx1 and emx2 morpholinos failed to knock down emx1 and emx2 translation sufficiently enough to produce a phenotype. Alternatively, emx1 and emx2 gene functions may be more subtle or required later in development.
Because zebrafish emx1 and emx2 play no obvious roles in patterning, it is possible that zebrafish emx3 supplies the functions provided by Emx1 and Emx2 in mouse. The reported mouse Emx knockout mutations affect the olfactory bulb and cortex, which correspond to the zebrafish pallium, the dorsal telencephalon (Wullimann and Mueller, 2004). In both mouse and fish, Emx loss-of-function affects only pallial structures, as indicated by altered gene expression patterns (Qiu et al., 1996; Yoshida et al., 1997; Guo et al., 2000; Mallamaci et al., 2000b; Bishop et al., 2002, 2003; Shinozaki et al., 2004). In Emx2 knockout mice, the BrdU-incorporating ventricular zone of the cortex is broader, indicating a more immature cortex (Mallamaci et al., 2000b; Tole et al., 2000; Bishop et al., 2003). Similarly, we find that knockdown of zebrafish emx3 produces a slight expansion of the ventricular zone and adjacent BrdU-positive layer, similar to the mouse Emx2 knockout phenotype. On the other hand, however, we find major differences between mouse Emx2 and zebrafish emx3 functions in regulating expression of other genes. fgf8 expression in zebrafish telencephalon is not affected by emx3 knockdown, whereas loss and gain of function studies of mouse Emx2 suggest that Emx2 represses Fgf8 (Fukuchi-Shimogori and Grove, 2003; Hamasaki et al., 2004). Similarly, dorsal Emx2 and ventral Pax6 repress each other in the mouse cortex (Muzio et al., 2002), whereas we find that emx3 positively regulates pax6a expression, although it is likely that the zebrafish pax6a expression domain in the telencephalon does not correspond to the Pax6 domain in the mouse cortex (Wullimann and Mueller, 2004). Mouse Emx2 positively regulates reelin expression (Mallamaci et al., 2000a), whereas we do not find zebrafish reln regulated by emx3. These comparisons suggest that mouse Emx genes may function in a genetic network that arose as an innovation during evolution of the mammalian cortex.
To rescue emx3MO induced phenotypes, emx3 mRNA needed to be expressed at low levels at the 1-somite or 2-somite stage (10.5 hr). Heat shock induced expression at higher levels, or only slightly earlier at bud stage, led to a marked ventral expansion of dorsal telencephalic markers and concomitant loss of eyes. This phenotype has been reported as a less frequent effect of Wnt/beta-catenin overactivation (van de Water et al., 2001) that is distinct from the usually observed loss of both forebrain and eyes (Stachel et al., 1993; Kelly et al., 1995; Heisenberg et al., 2001; van de Water et al., 2001; Kim et al., 2002). Our consistent observation of this otherwise rare phenotype may point to an interesting, time-sensitive effect of Wnt/beta-catenin signaling in which emx3 may play a role. Alternatively, this phenotype may be independent of Wnt/beta-catenin signaling and may reflect an ability of emx3 to expand prospective dorsal telencephalon at the expense of prospective eye field independently of Wnt/beta-catenin signaling. Currently, we cannot distinguish between these possibilities. In either case, sensitivity to overexpression ends around the five-somite stage (11.5 hr). emx3 function thus requires specific expression levels and timing.
In accordance with an involvement of the Wnt/beta-catenin pathway in Emx function, emx1, emx2, or emx3 overexpression phenotypes induced by injection of mRNA at the one-cell stage are similar to overactivation of the canonical Wnt pathway (Stachel et al., 1993; Kelly et al., 1995; Heisenberg et al., 2001; van de Water et al., 2001; Kim et al., 2002). The overexpression phenotypes are indistinguishable from the effects of wnt8b mRNA overexpression. Because the homeodomain-engrailed fusion, but not the homeodomain-VP16 fusion, phenocopies emx3 mRNA, emx3 may act as a transcriptional repressor, at least in this context.
Additional information is needed to link the physiological role of emx3 in telencephalic development conclusively to Wnt/beta-catenin signaling, for example, by identifying a Wnt/beta-catenin downstream target in the telencephalon that also requires emx3 function. In mice, Emx2 regulates many components of the canonical Wnt/beta-catenin pathway, including Wnt7b, Wnt8b, and Axin2 (Muzio et al., 2002; Machon et al., 2007). In contrast, we find no change of wnt7ba, wnt7bb, wnt8b, or axin2 expression after emx knockdown. The fact that the emx3 overexpression phenotypes can be rescued by an extracellular inhibitor of the Wnt/beta catenin pathway (dkk1) is consistent with a model in which emx3 activates expression of an extracellular Wnt ligand. However, rescue by an extracellular inhibitor does not exclude an alternative model in which emx3 activates the Wnt pathway internally in signal receiving cells, for example (but not necessarily) as an activating transcriptional cofactor of the TCF/LEF transcriptional complex. In the latter case, emx3 would be expected to act cell-autonomously, as opposed to a non–cell-autonomous function if emx3 activates a secreted ligand. Our overexpression experiments in chimeric embryos suggest that emx3 acts indeed cell-autonomously, at least with respect to activating expression of emx1 and atoh2b. These results favor a model in which emx3 acts in signal receiving cells, but is not strictly required for Wnt/beta-catenin pathway activation (otherwise rescue with dkk1 would not be possible) but rather plays an enhancing or balancing role that can be easily overridden. Both wnt7ba and wnt7bb duplicates, expressed at the telencephalic-diencephalic border, and wnt8b, expressed at the caudal tip of telencephalon, are possible sources of a Wnt/beta-catenin signal that may function in dorsal telencephalic development.
Early in development, Wnt/beta-catenin signaling blocks forebrain development. Wnt signaling must be inhibited locally for forebrain development to occur properly and for emx3 to be expressed (Wilson and Houart, 2004). A potential role of emx3 in activating the Wnt/beta-catenin pathway may be to support differentiation of dorsal telencephalic neurons at the precise time when inhibiting Wnt/beta-catenin signaling may no longer be necessary for telencephalic specification. In mouse, Emx2 and Wnt/beta-catenin signaling regulate cortical neurogenesis in two ways. First, Emx2-mediated Wnt/beta-catenin pathway activation maintains cell proliferation that generates cells for the caudal–medial cortex (Muzio et al., 2005). Second, down-regulation of Wnt/beta-catenin signaling is necessary for neurogenesis (Hirabayashi et al., 2004; Machon et al., 2007). Establishing the link between emx genes and the Wnt/beta-catenin pathway, or other signaling pathways, will be important to elucidate the precise role of Emx genes in telencephalic neurogenesis.
AB or AB/TL hybrid zebrafish were raised under standard conditions at 28.5°C (Westerfield, 2007). Developmental stages were determined by morphological criteria or hr postfertilization (hr; Kimmel et al., 1995). Plasmid DNA injections for mosaic expression and generation of transgenic lines Tg(emx3: YFP)b1200 and Tg(hsp70l:emx3-myc-GFP)b1202 were performed by coinjection of I-SceI enzyme (Roche, Basel, Switzerland, buffer from NEB, Ipswich, MA; Thermes et al., 2002) using a modified pG1 vector (a gift from Chi-Bin Chien) containing I-SceI sites (introduced by H.-G. Belting).
Alignments of full-length Emx protein sequences and phylogenetic trees were generated with ClustalX 2.0.3 (Larkin et al., 2007) (default parameter settings, positions with gaps not excluded, correction for multiple substitutions), and displayed using njplot 2.0 (Perriere et al., 1996). Genbank accession numbers, Ensembl/Genoscope/IMCB peptide ID, and species are Callorhinchus milii (elephant shark) Emx3 AAVX01298069.1:565-213 (putative exon 1), AAVX01519513.1:481-296 (putative exon 2), AAVX01149546.1: 535-699 (putative exon 3); Danio rerio (zebrafish) Emx1 AF534523, Emx2 D32215, Emx3 NM_131279; Gallus gallus (chicken) Emx1 ENS-GALP00000025888, Emx2 XM_421783; Gasterosteus aculeatus (stickleback) Emx1 ENSGACP00000006428, Emx2 ENSGACP00000004194, Emx3 ENS-GACP00000021610; Homo sapiens (human) Emx1 ENSP00000377670, Emx2 AF301598; Lampetra japonica (lamprey) Emx AB048758; Monodelphis domestica (South American opossum) Emx1 translated putative exon 1 fragment NW_001581849.1, bases 216282-216090, Emx2 ENSMODP0000-0011730, Emx3 XP_001381098; Mus musculus (mouse) Emx1 NM_010131, Emx2 AY117415; Oryzias latipes (medaka) Emx2 ENSORLP0000-0017758; Scyliorhinus canicula (dog-fish) Emx1 AF306637, Emx2 AF306636, Emx3 AF306635; Takifugu rubripes (fugu) Emx1 SINFRUP0-0000179547, Emx2 SINFRUP0-0000139710, Emx3 SINFRUP0000-0144814; Tetraodon nigroviridis (tetraodon) Emx1 GSTENP00019961001, Emx2 GSTENP00004254001, Emx3 GSTENP00024283001 (additional two predicted 5′ exons ignored); Xenopus laevis Emx1 NM_001093430; Xenopus tropicalis Emx1 NM_001005459, Emx2 ENSXETP00000047776, Emx3 CU075359 bases 567-1205 translated (Hubbard et al., 2007; Venkatesh et al., 2007). Conserved synteny was analyzed on http://www.ensembl.org/, using current database releases. Genes were considered orthologs if they were listed in Ortholog Predictions on the Ensembl Gene Report.
Full-length emx1 cDNA in pCS2+ was a gift from the Igor Dawid lab (Kawahara and Dawid, 2002). Full-length emx2 cDNA, a gift from the Mishina lab (Morita et al., 1995), was subcloned into pCS2+ expression vector (Dave Turner and Ralph Rupp, unpublished observations). Full-length emx3 cDNA was amplified by RT-PCR from mixed somite stage total RNA, sequenced for correctness, and ligated into a pG1-derived vector containing a myc tag followed by GFP. The myc tag serves as a spacer between emx3 and GFP to preserve the functions of both proteins. Primers (ggatccaccatgtttcaacataacaaaaaatgcttcacgattgaatctcttgtg, ctcgagagatctgtcttctgaaatgacgtcaatgtcc) introduced six, silent, point mutations that reduced the sequence complementary to emx3MOatg translation blocking morpholino to a block of five bases and six blocks of two bases. For rescue experiments, embryos from crosses of AB and heterozygous Tg(hsp70l:emx3-myc-GFP)b1202 transgenic animals were injected with emx3MO, heat shocked at the desired stage for 10 min at 37°C, selected for presence or absence of Emx3-Myc-GFP fusion protein by fluorescence, raised to the desired stage at 28.5°C, and fixed for in situ hybridization or antibody labeling. A 10-min heat shock produced mRNA levels comparable to endogenous emx3, as determined by in situ hybridization with emx3 probe in embryos fixed directly after heat shock. emx3Homeodomain (emx3hd)-vp16 and emx3hd-engrailed (en) fusions were constructed by amplifying the emx3 homeobox with primers including an artificial translation start consensus sequence (gcggatccaccatgcctttctcccgcaagcccaag, gactcgagctgcgggtctggagactcttcctc), and ligating to pCS2+ vector containing amplified PCR products of the VP16 transactivation domain (primers cgctcgaggcccccccgaccgatgtcagcctgg, gctctagactagtcccaccgtactcgtcaattc) and engrailed repressor domain (primers ctcgaggccctggaggatcgctgcagcccac, tctagatgcatagatcccagagcagatttc). mRNA transcription reactions (mMessage mMachine, Ambion, Austin, TX) were purified directly over RNeasy columns (Quiagen, Valencia, CA), eluted with water, and injected.
We used a combination of two translation blocking morpholinos (Gene Tools, Philomath, OR) against each gene (emx1MOatg, ccgttgccgagaacattgtccgtga; emx1MOupstream, cggtatgaccagaagagtccagct; emx2MOatg, acctcttcggtgtgggttgaaacat; emx2MOupstream, gtttaccagtcaagattccccagtt; emx3MOatg, cacttcttattgtgctggaacattg; emx3MO-upstream, gtcactctgaacccgatgatggag), a combination of two splice blocking morpholinos against emx3 (emx3MOe2i2, gacgtgtctgtctcttacctgcgtc; emx3MOi1e2, tgtcgtcacctgcagagaaacaaac) at 0.05 pmol/embryo, or a splice blocking morpholino against emx2 (emx2MOi1e2, gtaaacacattcttacctgagtttc) at 1 pmol/embryo. All morpholinos except emx2MOi1e2 induced ectopic cell death at concentrations that led to a phenotype and were therefore coinjected with tp53MO morpholino (ZDB-MRPHLNO-070126-7, gcgccattgctttgcaagaattg) (Robu et al., 2007) at 0.5 pmol per embryo in all experiments reported here. Approximately 10% of these embryos still showed low levels of ectopic cell death in the hindbrain, as judged by acridine orange staining (Shepard et al., 2004). Control embryos were injected with tp53MO alone.
For transplantation at blastula stages, donor oocytes were injected with 2% Alexa-555 10-kD dextran (Molecular Probes, D-22910, Eugene, OR). For fluorescent membrane labeling (Köster and Fraser, 2004), embryo medium contained 2 μg/ml Bodipy FL C5-ceramide (Molecular Probes, D-3521, Eugene, OR).
Whole-mount in situ hybridization and antibody labeling were performed as described (Westerfield, 2007). Plasmids for probe synthesis were gifts from the respective authors, prepared by RT-PCR from embryonic total RNA, or purchased from ATCC (Manassas, VA): atoh2b, ZDB-GENE-010608-2 (Liao et al., 1999); axin2, ZDB-GENE-000403-2 (Peng and Westerfield, 2006); dlx2a, ZDB-GENE-980526-212 (Akimenko et al., 1994); elavl3, ZDB-GENE-980526-76 (Good, 1995); elavl4, ZDB-GENE-990415-246 (Thisse and Thisse, 2005); emx1, ZDB-GENE-031007-7 (Kawahara and Dawid, 2002); emx2, ZDB-GENE-990415-54 (Morita et al., 1995); emx3, ZDB-GENE-990415-53 (Morita et al., 1995); eomesa, ZDB-GENE-001228-1 (Mione et al., 2001); egr2b, ZDB-GENE-980526-283 (Oxtoby and Jowett, 1993); fezf2, ZDB-GENE-001103-3 (Levkowitz et al., 2003); fgf8a, ZDB-GENE-990415-72 (Fürthauer et al., 1997); flh, ZDB-GENE-990415-75 (Talbot et al., 1995); foxb1.2, ZDB-GENE-990616-47 (Odenthal and Nüsslein-Volhard, 1998); foxg1, ZDB-GENE-990415-267 (Toresson et al., 1998); foxp2, ZDB-GENE-041203-2 (Bonkowsky and Chien, 2005; Shah et al., 2006); fzd8a, ZDB-GENE-000328-3 (Kim et al., 1998); gad1, ZDB-GENE-030909-3 (Higashijima et al., 2004); gad2, ZDB-GENE-030909-9 (Higashijima et al., 2004); isl1, ZDB-GENE-980526-112 (Inoue et al., 1994); lef1, ZDB-GENE-990714-26 (Dorsky et al., 1999); lhx5, ZDB-GENE-980526-484 (Toyama et al., 1995); neurod, ZDB-GENE-990415-172 (Liao et al., 1999); otx2, ZDB-GENE-980526-406 (Li et al., 1994); pax6a, ZDB-GENE-990415-200 (Nornes et al., 1998); pax6b, ZDB-GENE-001031-1 (Nornes et al., 1998); pea3, ZDB-GENE-990415-71 (Münchberg et al., 1999); reln, ZDB-GENE-020822-1 (Costagli et al., 2002); rx3, ZDB-GENE-990415-238 (Chuang et al., 1999); sfrp5, ZDB-GENE-011108-2 (Peng and Westerfield, 2006); slc17a6, ZDB-GENE-030616-554 (Higashijima et al., 2004); slc17a6l, ZDB-GENE-050105-4 (Higashijima et al., 2004); spry4, ZDB-GENE-010803-2 (Fürthauer et al., 2001); tcf4l, ZDB-GENE-070829-1 (Brockschmidt et al., 2007); wnt7ba, ZDB-GENE-041210-178 (Carl et al., 2007); wnt7bb, ZDB-GENE-081006-1, accession no. FJ356091, cloned by RT-PCR (primers tggtggctctcggtgcgaacatcatc, tttgcaggtaaacacctccgtcctctc) to amplify partial predicted cDNA of wnt7b paralog ENSDARG00000071107; wnt8b, ZDB-GENE-990415-279 (Kelly et al., 1995).
Embryos were incubated with 10 mM BrdU (Roche 10280879001, Basel, Switzerland) for 10 min on ice, washed for 5 min, fixed for 2 hr, and labeled with mouse anti-BrdU antibody (Roche 11170376001, 1:100; Shepard et al., 2004). Antibody labeling was performed as described (Westerfield, 2007) using mouse anti–N-acetylated tubulin (Sigma T-6793, St. Louis, MO) 1:500; anti-pH3 rabbit polyclonal IgG (Upstate 06-570, Billerica, MA) 1:500; mouse anti-GFP (JL-8, BD Biosciences, Franklin Lakes, NJ) 1:500; Alexa 488- and Alexa 546-linked secondary antibodies (Molecular Probes, Eugene, OR) at 1:1,000. pH3 cell counts were performed on consecutive stacks of confocal sections from a Pascal confocal microscope (Zeiss, Wetzlar, Germany) using the Cell-Counter plugin of ImageJ (http://rsbweb.nih.gov/ij/).
Grant sponsor: OEAW; Grant number: APART fellowship; Grant sponsor: DFG; Grant number: VA140/3-1; Grant sponsor: NIH; Grant number: DC04186; Grant number: HD22486.
We thank Hayato Yokoi, Steve Wilson, Christine and Bernard Thisse, Lila Solica-Krezel, Soojin Ryu, Gang Peng, Hitoshi Okamoto, Takao Morita, Randy Moon, Masayoshi Mishina, Marina Mione, Atsuo Kawahara, Corinne Houart, Shin-Ichi Higashijima, Joe Fetcho, Wolfgang Driever, Rich Dorsky, Igor Dawid, Chi-Bin Chien, Cristian Canestro, Josh Bonkowsky, Patrick Blader, and Henry Belting for material and advice. We thank Stephanie Heyl, Sofie Seibel, Jeremy Wegner, Judy Pierce, Jocelyn McAuley, and Emma-Jayne Holderness for their technical assistance. G.A. received an OEAW APART fellowship and Z.M.V. was funded by the DFG.
We used gene nomenclature approved by ZFIN (http://zfin.org/zf_info/nomen.html).