Concerted action of two dlx paralogs in sensory placode formation

The vertebrate ear and nose arise from sensory placodes, which are morphologically visible thickenings of embryonic ectoderm. Early grafting studies in salamander (Jacobson, 1966; Yntema, 1933; Yntema, 1950) and chick (Waddington, 1937) have led to a model for placodal development (Torres and Giraldez, 1998). Sensory placodes develop from a stripe of ectoderm that lies laterally to the neural plate during gastrulation. Early in development, this stripe of tissue shows multiplacodal competence, such that any region of this stripe can develop into any of the sensory placodes, i.e. otic, olfactory or lens placode, given the proper inductive environment. As development proceeds, the domains within this stripe that are competent to form placodes become progressively restricted, ultimately producing localized regions committed to their respective placodal fates (Jacobson, 1963).

Several spontaneous and gene-targeted mutations have been described in mouse that produce ear defects (reviewed by Baker and Bronner-Fraser, 2001; Torres and Giraldez, 1998). In addition, genetic studies in zebrafish have identified over 20 different mutations that affect various aspects of otic placode development, including induction of the placode and morphogenesis and differentiation of the inner ear (Malicki et al., 1996; Whitfield et al., 1996). Nevertheless, the molecular mechanisms underlying the initial steps of sensory placode formation remain to be elucidated.

Recently, a large number of genes expressed in the placodes during development have been identified, in particular for the otic placode (reviewed by Baker and Bronner-Fraser, 2001; Torres and Giraldez, 1998). dlx3 and dlx7 are two homeobox-containing genes that are among some of the earliest genes expressed in the otic and olfactory primordia in zebrafish (Akimenko et al., 1994; Ekker et al., 1992; Ellies et al., 1997). The expression pattern of these two genes during development is strikingly coincident with the model of progressive restriction for placodal development. In mid-to-late gastrulation, dlx3 and dlx7 are expressed in a continuous stripe around the lateral edge of the neural plate that appears to be coincident with the region of multiplacodal competence. Expression then becomes downregulated everywhere in the stripe, except in the presumptive otic and olfactory primordia. dlx3 and dlx7 expression is restricted to the otic and olfactory primordia by the end of gastrulation when these regions have become committed to their placodal fates.

The vertebrate Dlx family consists of genes with homeoboxes similar in sequence to that of the Drosophila Distal-less gene (Cohen et al., 1989). Dlx genes have been identified and cloned in several vertebrate species. Dlx genes are organized as linked pairs that lie in a convergently transcribed configuration and are separated by a few kilobases (kb) of intergenic sequence (Ellies et al., 1997; McGuinness et al., 1996; Nakamura et al., 1996; Simeone et al., 1994), and are thought to have originated from the duplication of an ancestral gene pair. Six Dlx genes constituting three linked pairs have been identified in mice (Nakamura et al., 1996; Porteus et al., 1991; Price et al., 1991; Robinson and Mahon, 1994; Robinson et al., 1991; Simeone et al., 1994; Stock et al., 1996; Weiss et al., 1995) and humans (Nakamura et al., 1996; Scherer et al., 1995; Simeone et al., 1994), and eight Dlx genes have been identified in zebrafish (Akimenko et al., 1994; Ekker et al., 1992; Stock et al., 1996), six of which constitute three pairs, and dlx5 and dlx8, which are not linked to one another (Ellies et al., 1997; Stock et al., 1996). Expression of the different Dlx genes overlaps in many structures during early development, including the forebrain, migrating neural crest, branchial arches, otic and olfactory placodes, and fins/limbs (Akimenko et al., 1994; Ekker et al., 1992; Ellies et al., 1997; Morasso et al., 1995; Acampora et al., 1999; Bulfone et al., 1993; Depew et al., 1999; Dolle et al., 1992; Eisenstat et al., 1999; Fernandez et al., 1998; Neidert et al., 2001; Papalopulu and Kintner, 1993; Porteus et al., 1991; Price et al., 1991; Qiu et al., 1997; Robinson and Mahon, 1994; Robinson et al., 1991; Simeone et al., 1994; Yang et al., 1998). However, there are exceptions to this pattern, such as the differential expression of linked Dlx genes during mouse tooth (Zhao et al., 2000) and placental development (Morasso et al., 1999).

In vitro studies have demonstrated that Dlx proteins share similar DNA-binding properties, and they are thought to act as transcriptional regulators (Feledy et al., 1999; Liu et al., 1997; Zerucha et al., 1997; Zerucha et al., 2000; Zhang et al., 1997). Much of the proposed function of Dlx genes is based on studies of their expression patterns. The idea that a functional redundancy may exist among these genes is largely based on the overlapping nature of their expression patterns and on analyses of Dlx-targeted gene disruptions in mice. Single knockouts for mouse Dlx1 (Qiu et al., 1997) and Dlx2 (Qiu et al., 1995), which constitute a linked pair, have very similar craniofacial abnormalities. However, the double Dlx1/Dlx2 mouse knockout has dentition abnormalities not seen in either of the single knockouts (Qiu et al., 1997), providing evidence for functional redundancy among these genes. Dlx5 mouse knockouts also show craniofacial malformations and inner ear defects (Acampora et al., 1999; Depew et al., 1999). Dlx3 mouse knockouts die of early placental failure (Morasso et al., 1999), and no other defects have been reported. In humans, a four base pair deletion in the DLX3 gene is linked to Tricho-dento-osseus (TDO) syndrome, which is a hereditary autosomal dominant disorder characterized by abnormalities in the hair, teeth and cranial bones (Wright et al., 1997; Price et al., 1998). No known mutations affecting Dlx genes have been identified to date in zebrafish.

Based on their expression pattern, we test here the hypothesis that dlx3 and/or dlx7 may play a crucial role in determination and development of the otic and olfactory placodes. We describe a large, γ-ray induced deletion in zebrafish, b380, which removes both dlx3 and dlx7. Homozygotes for the deletion show no evidence of otic and olfactory development. We also demonstrate that injection of dlx3 and dlx7 antisense morpholino oligonucleotides into zebrafish embryos produces an otic and olfactory phenotype similar to that of b380 mutants. Together, these data indicate that dlx3 and dlx7 share an essential and partially redundant function in otic and olfactory placode development.

Both the AB line (Eugene, OR) and a partially inbred line of store-bought fish were used as wild-type lines. The b380 mutation was induced in the AB line with γ-rays (Fritz et al., 1996), and these mutants were hybridized with the line of store-bought fish.

Zebrafish (Danio rerio) embryos were raised at 28.5°C and staged as previously described (Kimmel et al., 1995). Embryos were fixed in phosphate-buffered solution containing 4% paraformaldehyde. Whole-mount in situ hybridization was performed as described (Thisse, 1998) using single-stranded RNA probes labeled with digoxigenin-UTP or fluorescein-UTP (Boehringer Mannheim). The zebrafish dlx3 (Ekker et al., 1992), dlx7 (Ellies et al., 1997), eya1 (Sahly et al., 1999), pax2.1 (Krauss et al., 1991) and pax8 (Pfeffer et al., 1998) probes have been previously described.

Morpholinos were obtained from Gene Tools, LLC. The antisense oligonucleotide sequences were designed to bind to the 5′ UTR or flanking sequences including the initiation methionine. To minimize the possibility of nonspecific effects, we generated two morpholinos complementary to non-overlapping sequence for each gene. Sequences were as follows: dlx3-1 MO, 5′-ATATGTCGGTCCACTCATCCTTAAT-3′; dlx3-2 MO, 5′-TTTCCAAGGCAGACCGAAGCAAGTC-3′; dlx7-1 MO, 5′-TAACCGTCAAGTCCATAAAGCCCGA-3′; and dlx7-2 MO, 5′-TCAGACATGAAACTCATAGACATCA-3′. The standard control morpholino (5′-CCTCTTACCTCAGTTACAATTTATA-3′) from Gene Tools, LLC was also injected in some experiments. Morpholino injections were essentially performed as described (Nasevicius and Ekker, 2000). One- to eight-cell embryos were injected with 8-22 ng of morpholino. No phenotypic differences were observed when injecting between 8-20 ng morpholino per embryo. Above 20 ng, some embryos showed nonspecific necrosis. In all subsequent experiments, embryos were injected with a total amount of 16 ng of morpholino oligonucleotide(s), i.e. 16 ng dlx3 MO, 16 ng dlx7 MO or (8 ng + 8 ng) dlx3 MO and dlx7 MO. Double injections were made with all possible combinations of dlx3 MOs with dlx7 MOs.

We amplified a 1 kb dlx7 fragment, including poly A sequence, from 17 h zebrafish cDNA. The following primer was used to amplify the 5′ end of the fragment: 5′-CGGTTAATGATGTCTATGAGTTTC-3′. This generated a fragment that begins with nucleotide position –6 with respect to the ATG start codon (A is +1) and does not contain the sequence complementary to dlx7-1 MO. The PCR fragment was cloned into pCRII-TOPO vector (Invitrogen). A 925 bp dlx3 fragment, beginning with nucleotide position –26 (with respect to the ATG start codon), was cloned into pT3TS vector (provided by Dr Steve Ekker). The vector contains a poly A sequence following the 3′ region of the insert fragment. The dlx3 fragment did not contain the sequence complementary to dlx3-2 MO. Capped sense RNA for both of these constructs was transcribed using Ambion mMessage mMachine in vitro transcription kits. dlx3 and dlx7 RNA were injected into one-cell stage zebrafish embryos at several different concentrations to determine a nontoxic concentration level for each RNA. The RNA was then used in morpholino rescue experiments. dlx7 RNA (6 pg) was co-injected with (9 ng + 9 ng) dlx3-1 MO and dlx7-1 MO, and 5 pg of dlx3 RNA was co-injected with 18 ng of dlx3-2 MO.

We examined dlx3 and dlx7 expression patterns in zebrafish embryos using single gene-specific probes in whole-mount RNA in situ hybridization. We find the expression patterns of these genes to be highly overlapping based on comparisons between these different in situ hybridization. At the late gastrula stage, dlx3 and dlx7 are expressed in a continuous stripe of presumptive ectoderm around the lateral edge of the neural plate (not shown). After initiation of somitogenesis, this expression becomes progressively restricted to the prospective otic and olfactory primordia. At the 18-somite stage, dlx3 and dlx7 expression is detectable in the olfactory placodes, otic vesicles, visceral arches (Fig. 1A,B) and the median fin fold (not shown). Our data confirms the patterns originally described by others using these single gene-specific probes (Ekker et al., 1992; Akimenko et al., 1994; Ellies et al., 1997).

To definitively determine whether these genes are co-expressed by the same cells, we examined the expression overlap in double in situ hybridization. We find that dlx3 and dlx7 are expressed in the same cells in the otic vesicles (Fig. 1D), olfactory placodes (Fig. 1C), visceral arches and median fin fold (not shown) at 18 somites. These results conclusively demonstrate that these two Dlx genes are expressed in many of the same cells during zebrafish embryonic development, and that the domains of co-expression include the otic and olfactory primordia. This high degree of expression overlap indicates a potential for functional redundancy among these genes.

In a γ-ray mutagenesis screen for deletions that remove specific cloned genes in zebrafish, we have identified a mutation, b380, that removes both dlx3 and dlx7 (Fritz et al., 1996). The deletion is approximately 6 cM in size and, in addition to dlx3 and dlx7, removes at least six other known genes and seven expressed sequence tags (ESTs). b380 heterozygotes develop normally to adulthood and show no obvious defects. From early somitogenesis up to approximately 20 somites, embryos homozygous for the b380 deletion show two predominant phenotypes; they appear to completely lack otic and olfactory placodes (Fig. 2E, olfactory phenotype not shown), and they also fail to form somites. The embryos begin to develop a severe compound phenotype during later stages of somitogenesis, presumably because of the removal of so many genes, and they typically die at ∼56 hours of development (56 h).

The expression patterns of dlx3 and dlx7 suggested their specific removal in the large b380 deletion was the cause of the placodal defects in this mutant. We therefore wanted to test whether the disruption of dlx3 and dlx7 could mimic the b380 placodal phenotype. We also wanted to examine the individual functions of these genes to determine whether a functional redundancy existed, which was suggested by the overlapping nature of the expression patterns. To address these issues, morpholino oligonucleotides (dlx3 MO and dlx7 MO) were generated. Morpholinos are antisense, chemically modified oligonucleotides that have been shown to bind and block translation of mRNA in vitro, in tissue cell culture (Summerton et al., 1997; Summerton et al., 1999), and in some in vivo applications (Arora et al., 2000; Qin et al., 2000; Heasman et al., 2000). Recently, it has been demonstrated that morpholinos function as highly effective and specific translational inhibitors in zebrafish (Nasevicius and Ekker, 2000). We injected these morpholinos both individually and together (dlx3+7 MO) into one-cell stage wild-type embryos and then examined these embryos for morphological defects. To address the issue of morpholino specificity for the target genes, we generated two morpholinos complementary to non-overlapping sequence for each gene, and we injected these in all possible combinations in the co-injection analyses. In addition, we also injected a control morpholino.

Embryos injected with dlx7 MO do not exhibit any morphologically detectable defects and are indistinguishable from wild type (Fig. 2A,B). By contrast, otic vesicles are smaller than wild-type in ∼98% (79/81) of dlx3 MO-injected embryos. By 24 h, the otic vesicles of these embryos exhibit smaller lumens and typically only one otolith (Fig. 2C, Fig. 3). This effect is very consistent over a wide range of concentrations of injected MO (see Materials and Methods) and is observed for both dlx3 MOs. We did not detect any other visible defects in the injected embryos up to 7 days of development, and no defects were observed in embryos injected with the control morpholino (not shown). We conclude from these results that dlx3 is necessary for proper otic development, whereas dlx7 is dispensable in this process.

Embryos co-injected with dlx3+7 MO displayed even more severe otic defects than observed in embryos injected with dlx3 MO alone (Fig. 2D). Whereas 89% (72/81) of embryos injected with dlx3 MO developed at least one otolith, 100% (57/57) of embryos injected with dlx3+7 MO completely lacked otoliths and formed only very small lumens (Fig. 3). Because the dlx7 MO does not produce any defects when injected singly, it appears that dlx3 and dlx7 are partially functionally redundant in the process of otic development. This also suggests that the otic phenotype of b380 mutants is mediated by the loss of dlx3 and dlx7.

To address the redundancy of dlx3 and dlx7 further, and to test the specificity of the dlx3 MO and dlx7 MO, we attempted to rescue the otic defects of MO-injected embryos by co-injection of either dlx3 or dlx7 in vitro synthesized mRNA. Forty percent (21/53) of embryos co-injected with dlx3 MO and dlx3 mRNA had larger lumens and two otoliths in at least one otic vesicle, compared with only 2% (2/81) of embryos injected with dlx3 MO alone (Fig. 2G,H; Fig. 3). In addition, we find that 53% (21/40) of embryos co-injected with dlx7 mRNA and dlx3+7 MO had larger lumens than those injected with dlx3+7 MO and at least one otolith (Fig. 2I,J; Fig. 3). Increasing amounts of either dlx3 or dlx7 mRNA led to toxicity and developmental disruption of the injected embryos. We have not been able to elicit otolith formation in b380 mutant embryos by injecting dlx3 and/or dlx7 mRNA, although we can occasionally observe formation of a small otic vesicle (data not shown). This could be due to the limiting amounts of mRNA that can be injected into embryos.

To determine at what stage placode development fails when dlx3 and/or dlx7 function is lost, we examined the expression of otic and olfactory markers that are known to be transcribed during different stages of placode formation and differentiation. eya1 is a member of the eyes absent-like gene family, which encodes a group of transcriptional coactivators, and it is transcribed in both the otic and olfactory structures as they undergo differentiation (Sahly et al., 1999). At 18 somites, eya1 transcripts are detectable in the otic vesicles, olfactory placodes, anterior pituitary, anterior and posterior lateral line placodes, and in the developing somites in wild-type embryos (Fig. 4A,G), embryos injected with a control MO (Fig. 4B,H), and in embryos injected with dlx7 MO (Fig. 4C,I). However, the number of eya1-expressing otic and olfactory cells is reduced in embryos injected with dlx3 MO (Fig. 4D,J). eya1 expression is reduced even further in otic domains of embryos injected with dlx3+7 MO (Fig. 4E) and is reduced to a level comparable with that observed for dlx3 MO in the olfactory domain (Fig. 4K). Expression is undetectable in the otic domain of b380 mutants (Fig. 4F), and olfactory expression is reduced to almost a single cell (Fig. 4L). Similar results were obtained for expression of anxV (Fig. 4M-R), a later marker of olfactory placodes (S. Farber and M. E. Halpern, personal communication). These results show that otic and olfactory placode development are already blocked at the time when patterning and differentiation normally take place, suggesting an earlier role for dlx3 and dlx7 function in the development of these structures.

To determine whether dlx3 and dlx7 are required earlier for placode formation, we examined the expression of two early markers of sensory placodes, pax8 and pax2.1. pax8 is the earliest known marker for otic development in both fish and mammals (Pfeffer et al., 1998). In zebrafish, pax8 expression is detectable in the presumptive otic primordia as early as 80% epiboly, around the same time that dlx3 expression becomes detectable and several hours before otic placodes become morphologically visible. We find that otic expression of pax8 in b380 and dlx3 MO-, dlx7 MO- and dlx3+7 MO-injected embryos is indistinguishable from wild type at 3 somites (Fig. 5A-E). All of these embryos show strong pax8 expression in the presumptive otic primordia and the pronephros at this stage.

pax2.1 expression begins in the otic primordia later than pax8, at ∼3 somites, and expression in this domain lasts through the otic vesicle stage (Krauss et al., 1991; Pfeffer et al., 1998). In addition to the otic expression domain, pax2.1 is also expressed strongly in the optic stalk, midbrain-hindbrain boundary (mhb) and pronephros at this stage (Fig. 5F). Expression of this gene appears normal at 3 somites in embryos injected with dlx7 MO (Fig. 5G). By contrast, embryos injected with dlx3 MO show reduced expression of pax2.1 in the presumptive otic primordia (Fig. 5H). This reduction in expression is specific; expression levels in the optic stalk, mhb and pronephros are comparable with those seen in wild-type and dlx7 MO-injected embryos. We are unable to detect any pax2.1 expression in the otic domain of embryos co-injected with dlx3+7 MO or in b380 mutants (Fig. 5I,J). This loss of expression is also specific to the otic domain, similar to the pattern seen in dlx3 MO-injected embryos. Thus, in the absence of dlx3 and dlx7 function, otic placode formation appears to be blocked after initiation of pax8 transcription in the otic domain but before pax2.1 otic expression. It is during this period from late gastrulation to early somitogenesis that dlx3 and dlx7 are expressed in a stripe surrounding the neural plate.

An additional level of complexity in understanding the function of dlx3 and dlx7 comes from potential cross-regulatory interactions between these two genes, which have been shown to exist among some Dlx genes (Zerucha et al., 1997; Zerucha et al., 2000). Therefore, we wanted to determine whether any similar cross-regulation exists between dlx3 and dlx7. Because morpholinos target gene expression at the level of translation, we might expect dlx3 and dlx7 mRNA expression patterns to be unaffected in injected embryos in the absence of cross-regulation. However, while dlx3 expression in embryos injected with dlx7 MO is comparable with wild type at 6 somites (Fig. 6A,B), expression is noticeably upregulated in embryos injected with either dlx3 MO alone (Fig. 6C) or co-injected with dlx7 MO (Fig. 6D). At this same stage, dlx7 expression is also upregulated in embryos injected with dlx3 MO (Fig. 6G). By contrast, dlx7 expression is downregulated in embryos injected with dlx7 MO alone (Fig. 6F), and appears similar to wild-type in co-injected embryos (Fig. 6E,H) (see Table 1 for summary of data). Except for the noted up- and downregulation, the patterns of dlx3 and dlx7 expression appear to otherwise be normal in all of these MO-injected embryos. We also examined expression of dlx3 and dlx7 in b380 homozygotes and, as expected, no transcripts for these genes were detected (not shown).

To exclude artifacts of injection or in situ hybridization, we performed these experiments three separate times with the same results. In addition, dlx3 expression in dlx7 MO embryos serves as in internal control for this experiment; we never observed an upregulation of expression in these embryos, making it unlikely that the injection process itself increases the permeability of the embryos during the hybridization procedure, leading to a perceived upregulation. In addition, we never observed an up- or downregulation of other markers we examined in embryos injected with the control morpholino (see Fig. 4B,H,N, for example). We independently confirmed aspects of this cross-regulation in a collaboration with K. Wuennenberg-Stapleton in J. Ngai’s laboratory (UC Berkeley, personal communication), who performed a microarray gene chip analysis comparing expression levels between RNA isolated from 16 h wild-type, b380 mutant and dlx3+7 MO injected embryos. In this analysis, a dlx3 clone was spotted twice to serve as a control in the b380 to wild-type RNA comparison. As expected, dlx3 was strongly downregulated in RNA from b380 embryos (see Table 2). These experiments also showed a statistically significant and reproducible upregulation of dlx3 (approximately twofold) in the RNA from dlx3+7 MO injected embryos.

We find that this overexpression of dlx3 begins to dissipate by mid-to-late somitogenesis. At 21 somites, dlx3 expression in dlx7 MO injected embryos (Fig. 6J,N) is still comparable with wild type (Fig. 6I,M). Expression still appears to be upregulated in dlx3 MO embryos, most notably in the visceral arches (Fig. 6K,O). However, the total number of cells expressing dlx3 within the otic region is reduced, similar to the trend observed with other otic markers expressed during differentiation, such as eya1 (Fig. 4A-E). Upregulation is not nearly as notable in dlx3+7 embryos at the 21 somite stage (Fig. 6L,P). Expression may still be upregulated in otic cells, but the total number of expressing cells in this region is reduced even more than in dlx3 MO embryos. Both intensity and size of the regions expressing dlx3 in the olfactory organ and the visceral arches appear to be reduced in these embryos.

The zebrafish dlx3 and dlx7 genes have been identified as genes expressed during the initial and formative stages of sensory placode development (Akimenko et al., 1994; Ekker et al., 1992; Ellies et al., 1997; Torres and Giraldez, 1998). We provide functional evidence that dlx3 and dlx7 act in concert to promote otic and olfactory placode formation. Our data demonstrate that the expression domains of dlx3 and dlx7 are completely overlapping through the first 24 hours of development, creating the potential for redundant functions. The removal of both dlx3 and dlx7 activity by the b380 deletion mutation leads to a failure of otic and olfactory placode development, both by morphological and gene expression criteria. Furthermore, the morpholino mediated knockdown experiments strongly support the view that the sensory placode phenotype elicited by b380 mutant embryos is due to the lack of dlx3 and dlx7 gene activity. Specifically, loss of dlx3 and dlx7 function results in a severe reduction of the otic and olfactory placodes, mimicking the phenotype of b380 mutant embryos. Interestingly, our findings suggest that while the functions of dlx3 and dlx7 are redundant, they are not equivalent. The loss of dlx7 activity appears not to result in any detectable phenotype. However, dlx3 MO injections lead to a clearly visible reduction in the size of the otic placode and inner ear. This suggests that dlx3 alone is fully capable of supporting otic placode and inner ear formation, while dlx7 can only partially compensate for the loss of dlx3.

Transplantation and extirpation experiments in a number of organisms have shown that competence to respond to placode-inducing signals is initially broadly distributed throughout ectoderm (Baker and Bronner-Fraser, 2001; Torres and Giraldez, 1998). As development proceeds, competent ectoderm becomes progressively restricted to a region lateral to the neural plate and subsequently commits to the appropriate placodal fates. Expression of dlx3 and dlx7 recapitulates these steps, beginning by late gastrulation (90% epiboly), and appears to coincide with the region competent to respond to placode inducing signals. Shortly after onset of somitogenesis, dlx3 and dlx7 expression is downregulated along the margin of the neural plate but upregulated in the prospective otic and olfactory placodes, possibly in concert with commitment to those fates.

Given our observations for dlx3 and dlx7 inactivation, we can envision two possible models for how these genes may function in the process of otic and olfactory placode development. One possibility is that dlx3 and dlx7 function prior to somitogenesis to maintain competence in ectodermal cells to respond to signals required for placode induction. Alternatively, these genes could be required later, immediately after their upregulation during early somitogenesis, for the proper development and differentiation of otic and olfactory placodal cells that have already been induced and committed to placodal fates. In support of the latter model, we find that at least some induction of otic characteristics occurs in the absence of dlx3/7. Expression of pax8 is the earliest known manifestation of otic induction (80% epiboly) (Pfeffer et al., 1998), and pax8 appears to be induced and expressed normally in dlx3/dlx7-deficient embryos. However, our pax2.1 expression data better support the first model. pax2.1 otic expression is initiated shortly after the onset of somitogenesis, and coincides with the time when dlx3 and dlx7 expression is upregulated in the otic anlagen (Krauss et al., 1991; Pfeffer et al., 1998). If dlx3 and dlx7 function is required only during this later stage for placodal differentiation after all inductive events have occurred, we would expect that at least initiation of pax2.1 expression would occur normally. However, pax2.1 is not expressed in b380 mutant embryos or dlx3/7 MO embryos, implying an earlier function for dlx3 and dlx7 in placodal development.

It has recently been shown that induction of the otic placode begins earlier than previously thought. Phillips et al. have demonstrated that zebrafish fgf3 and fgf8 cooperate during mid-gastrulation to induce the otic placode and pax8 expression (Phillips et al., 2001). However, it has also been shown that induction of the otic placode is required and proceeds over a period of time in zebrafish (Mendonsa and Riley, 1999) and in other species [see Groves and Bronner-Fraser (Groves and Bronner-Fraser, 2000), for recent molecular data on chick]. The importance of continued inductive influences is further supported by mutations in genes that presumably disrupt hindbrain-derived signals and lead to inner ear defects (Chisaka et al., 1992; Cordes and Barsh, 1994; Epstein et al., 1991; Lufkin et al., 1991; Moens et al., 1998). For example, mutant alleles of zebrafish valentino gene result in inner ear abnormalities, and hindbrain expression of this gene does not begin until early somitogenesis (Moens et al., 1998; Moens et al., 1996). Thus, even after induction of pax8, the placodal ectoderm needs to respond to further signals, and it is possible that dlx3 and dlx7 are necessary to maintain this competence.

It should be noted that these two models are not mutually exclusive. For example, dlx3 and dlx7 could be required early for competence in preplacodal ectoderm, and later for proper patterning or differentiation of the inner ear and olfactory organ. We will be better able to pinpoint the time of action of these genes as more early markers for otic and olfactory primordia become available. However, we might have to await the availability of conditional mutations in zebrafish to fully determine the role of these genes.

Our results indicate that pax8 expression is not sufficient to promote otic placode formation, or even pax2.1 expression, which is highly overlapping with pax8 during early somite stages. Targeted inactivation of the mouse Pax8 gene also suggests that it is not required for otic placode development (Mansouri et al., 1998). Disruption of mouse Pax2 (Favor et al., 1996; Torres et al., 1996) or zebrafish pax2.1 (Brand et al., 1996) results in later perturbations in otic vesicle development but does not prevent placode formation. Of other genes expressed in the otic placode, targeted disruption of mouse Eya1 leads to the most severe inner ear defect described so far (Xu et al., 1999). However, initial otic placode induction and Pax2/8 expression occur normally in these Eya1-deficient embryos. Inner ear development arrests later at the otic vesicle stage, and this is followed by apoptotic regression of the otic primordia. Our observations provide a possible explanation for the resilience of sensory placodes to disruption by mutations. First, the signals required for placode induction are redundant [fgf3/8 for otic placode (Phillips et al., 2001)]. Second, the responding ectoderm itself expresses functionally redundant molecules that support the initial steps of sensory placode formation.

Possible crossregulatory interactions between Dlx genes have previously been proposed based on the overlapping expression domains observed for both paralogous and orthologous Dlx genes (Zerucha et al., 1997). In zebrafish, dlx3 has been shown to act as a transcriptional activator of dlx4 through a 1.7 kb fragment flanking the 5′ region of dlx4 (Zerucha et al., 1997). More recently, Zerucha et al. (Zerucha et al., 2000) have shown that a highly conserved DNA element found in the intergenic region between mouse Dlx5/Dlx6 and zebrafish dlx4/dlx6 is sufficient to direct the forebrain specific expression seen with the endogenous genes. Furthermore, Dlx proteins bind to this intergenic enhancer in vitro and can upregulate transcription of a reporter gene in a co-transfection assay.

Our results suggest that both cross- and autoregulatory interactions exist among dlx3 and dlx7 that modulate the levels of their activities. These interactions are evident in early somitogenesis and, at least in the case of dlx3, are still functioning by mid-to-late somitogenesis stages (summarized in Table 1). In particular, morpholino-mediated knockdown of Dlx3 protein consistently leads to an upregulation of both dlx3 and dlx7 message, suggesting that Dlx3 might act through a shared enhancer element. This upregulation is observed in both RNA in situ hybridization as well as with microarray analysis. However, further experiments will be required to address whether the observed feedback mechanisms involve direct interactions between Dlx3/Dlx7 proteins and the regulatory regions of their genes.

Similar to the Hox genes, the Dlx gene family has arisen through several rounds of duplication, which is thought to have facilitated the development of more complex and novel morphological structures in vertebrates (Ohno, 1970). In support of this, we have shown that two Dlx genes act in concert to promote the development of sensory structures that are absent in primitive chordates that only possess one Dlx gene, such as amphioxus (Holland et al., 1996). However, it remains unclear whether the roles of Dlx3 and Dlx7 are conserved during vertebrate evolution. In lamprey, Dlx genes are expressed in the otic and olfactory placodes, suggesting an ancient role for these genes in sensory organ development (Neidert et al., 2001). Support for a conserved role of the dlx3 gene in sensory placode development comes from expression analysis in chick and axolotl. Axolotl Dlx3-3 is expressed in the lateral line placodes and the otic vesicle (Metscher, 1997), and chick DLX3 is expressed in the otic and olfactory placodes (Pera and Kessel, 1999). Although in chick, DLX5 transcription surrounding the anterior neural plate, including the prospective olfactory placode, starts earlier than DLX3 transcription (Pera et al., 1999), and is more reminiscent of the zebrafish dlx3 expression pattern.

Functional studies of DLX3 in Xenopus (Beanan and Sargent, 2000; Dirksen et al., 1994), mouse (Morasso et al., 1999) and humans (Price et al., 1998) do not provide evidence for a conserved role in sensory placode development, and expression of these orthologs in mouse and Xenopus differs from zebrafish. Mouse embryos homozygous for a Dlx3 targeted gene disruption die around day 10 because of placental patterning defects, and no otic or olfactory defect has been reported for these embryos (Morasso et al., 1999). A four base pair deletion in the human DLX3 gene, resulting in a truncation of the protein close to the C terminus, leads to Tricho-Dento-Osseus (TDO) syndrome, which is characterized by craniofacial abnormalities, kinky hair and tooth defects (Price et al., 1998). However, the dominant nature of this mutation complicates the interpretation of the role of DLX3 in human development. The Xenopus dlx3 homolog, X-dll2, appears to promote the development of epidermis and may suppress the development of mesoderm and neural tissue (Beanan and Sargent, 2000; Dirksen et al., 1994). Developmental function of the dlx7 ortholog remains to be addressed in other vertebrates. Whether the discrepancies of dlx3 expression and function observed among different vertebrate species reflect an actual functional difference or rather a problem in the assignment of orthology is unclear.

However, at least in mouse, the linked Dlx5/Dlx6 genes are expressed in a very similar pattern to zebrafish dlx3/dlx7 during gastrulation, and therefore it has been suggested that these paralogous genes are more likely to be functionally equivalent (Quint et al., 2000). Notably, mouse embryos homozygous for a targeted Dlx5 gene disruption display inner ear patterning defects and olfactory epithelium abnormalities (Acampora et al., 1999; Depew et al., 1999). This may correspond to the phenotype that we observe in dlx3 MO-injected embryos, although that remains to be addressed. No Dlx5/Dlx6 double mutant has been reported in mouse, although it would be interesting to examine whether the phenotype of such a mutant would reveal redundant functions for these genes. Functional redundancy for Dlx1 and Dlx2 has been observed in mice with targeted disruptions of these genes. Mice that lack only Dlx1 or Dlx2 display craniofacial malformations (Qiu et al., 1997; Qiu et al., 1995). In addition to craniofacial abnormalities, mice that lack both proteins have severe subcortical telencephalon abnormalities and lack maxillary molars, phenotypes not observed in either of the single mutants (Qiu et al., 1997). This suggests that mouse Dlx1 and Dlx2 posses both redundant and nonredundant functions. Although we observe partially redundant functions for zebrafish dlx3 and dlx7, our results indicate that dlx3 function alone is sufficient for sensory placode formation. Functional redundancy of dlx3 and dlx7 is further supported by the observation that both dlx3 mRNA and dlx7 mRNA are able to partially rescue the dlx3+7 MO phenotype. Thus, it remains unclear why both of these seemingly redundant genes have been preserved through evolutionary time. Several theoretical models predict that one copy of a redundant gene pair will eventually become silenced by drift if the two genes are entirely equivalent (Kimura, 1983; Li, 1980; Watterson, 1983). However, one model proposes that selection will maintain both copies of a redundant pair if one member is slightly less efficacious and also mutates at a slightly lower rate than the other gene (Krakauer and Nowak, 1999). Our data fits this model in that dlx7 appears to be a slightly less efficacious member of a redundant gene pair. It will be interesting to determine whether dlx7 also possesses a lower mutation rate than dlx3, providing further support for this evolutionary model.

We thank S. Farber and M. E. Halpern for providing the anxV probe prior to publication, and K. Wuennenberg-Stapleton for microarray analysis. We also thank W. Kelly, M. E. Halpern and K. Moses for helpful discussions, and R. Russo for technical assistance. This work was supported by a grant from the NIDCD (R01-DC04701), and K. S. was partially supported by a NIH Predoctoral Training Program in Genetics grant (NIGMS, T32 GM08490).