|Home | About | Journals | Submit | Contact Us | Français|
The Ovo gene family encodes a group of evolutionarily conserved transcription factors and includes members that reside downstream of key developmental signaling pathways such as Wg/Wnt and BMP/TGF-β. In the current study, we explore the function of Ovol2, one of three Ovo paralogues in mice. We report that Ovol2 is expressed during early–mid embryogenesis, particularly in the inner cell mass at E3.5, in epiblast at E6.5, and at later stages in ectodermally derived tissues such as the rostral surface (epidermal) ectoderm. Embryos in which Ovol2 is ablated exhibit lethality by E10.5, prior to which they display severe defects including an open cranial neural tube. The neural defects are associated with improper Shh expression in the underlying rostral axial mesoderm and localized changes of neural marker expression along the dorsoventral axis, as well as with expanded cranial neural tissue and reduced cranial surface ectoderm culminating in a lateral shift of the neuroectoderm/surface ectoderm border. We propose that these defects reflect the involvement of Ovol2 in independent processes such as regionalized gene expression and neural/non-neural ectodermal patterning. Additionally, we present evidence that Ovol2 is required for efficient migration and survival of neural crest cells that arise at the neuroectoderm/surface ectoderm border, but not for their initial formation. Collectively, our studies indicate that Ovol2 is a key regulator of neural development and reveal a previously unexplored role for Ovo genes in mammalian embryogenesis.
The embryonic brain develops from a region of specialized dorsal ectoderm known as the neural plate, within which the specification of the anterior neuroectoderm is initiated by the anterior visceral endoderm and is subsequently maintained by signals from the underlying anterior axial mesendoderm derived from the node (Brennan et al., 2001; Wilson and Houart, 2004). Ectoderm also gives rise to the surface ectoderm, precursor of the epidermis and its appendages. Studies in Xenopus, zebrafish, and chick suggest that the induction of neural plate and patterning of neural and non-neural ectoderm are governed by a complex interplay between BMP, Wnt, and FGF signaling pathways (Altmann and Brivanlou, 2001; Bally-Cuif and Hammerschmidt, 2003). Specifically, a BMP signaling gradient, set forth by BMPs and their antagonists, is proposed to differentially pattern the ectoderm; inhibition of BMP activity induces a neural fate, high BMP activity specifies an epidermal fate, and intermediate BMP activity induces the formation of neural crest cells at the neural/non-neural border (Hemmati-Brivanlou and Melton, 1997; Marchant et al., 1998; Nguyen et al., 1998; Wilson et al., 1997). Additionally, FGF signaling promotes a neural fate, whereas active Wnt signaling inhibits the ectoderm’s response to FGF signaling, thereby permitting epidermal specification (Wilson et al., 2001;Wittler and Kessel, 2004). Also implicated in this process, largely by studies in Xenopus, are transcription factors that either promote (e.g., the Zic family members) or inhibit (e.g., Msx1, Dlx3) neural plate differentiation (Feledy et al., 1999; Kuo et al., 1998; Luo et al., 2001; Mizuseki et al., 1998; Nakata et al., 1997; Suzuki et al., 1997). To date, little is known about the genetic pathways underlying neural/epidermal patterning in mammals, as studies have been hampered by the in utero development of embryos and by functional redundancy between multiple members of particular gene families.
A number of important signaling centers are formed to elaborate anterior–posterior (A–P) and dorsal–ventral (D–V) patterning of the prospective brain (reviewed in Joyner, 2002; Rubenstein and Beachy, 1998). The rostral domain of axial mesendoderm including the prechordal plate specifies the ventral aspect of the cranial neuroectoderm, while the surface ectoderm adjacent to the neural plate emanates BMP signals that specify a dorsal neural fate. The balance between these opposing activities determines the final D–V specification of the cranial neural tube (Joyner, 2002). Shh is an important ventralizing signal from the prechordal plate and the notochord. It induces the expression of itself and of Ptch1, encoding an Shh receptor and inhibitor, in the ventral neuroectoderm, as well as the expression of ventral forebrain markers such as the transcription factor Nkx2.1 (Briscoe and Ericson, 2001; Dale et al., 1997; Goodrich et al., 1997; Jessell, 2000; Shimamura and Ruben-stein, 1997). In the dorsal neuroectoderm, transcription factors Pax3 and Msx1/2 are expressed in response to dorsalizing BMP signals and are required for dorsal neural differentiation (Goulding et al., 1993; Liem et al., 1995). Understanding the function of known and yet uncharacterized transcription factors and how they interact with the morphogenetic signaling events is key to understanding the genetic networks that underlie the formation of complex tissues such as the brain.
Cranial neurulation is an integral component of brain morphogenesis. This process requires morphogen-induced cell shape changes and movements, a delicate balance between proliferation and apoptosis in the neuroepithelium, expansion of the cranial mesoderm, and efficient migration of neural crest cells away from the neural plate (Copp et al., 2003). Disruption of any of these cellular and morphological events leads to exencephaly, ultimately resulting in defective brain formation. The medial expansion of the surface ectoderm provides a major driving force for folding and subsequent closure of the neural plate (Smith and Schoenwolf, 1997). Genetic studies in mice have elucidated a critical involvement of a number of genes in cranial neurulation (Copp et al., 2003; Smith and Schoenwolf, 1997), including twist, which encodes a basic helix–loop–helix protein expressed in head mesenchyme, and the transcription factor AP-2, which is expressed in cranial surface ectoderm. Thus, factors present outside of the neuroepithelium can also affect the morphogenesis of the cranial neural tube.
The Ovo gene family encodes evolutionarily conserved zinc-finger transcription factors whose function in embryogenesis is under-explored. The founding member of this family, Drosophila ovo, is required for epidermal denticle formation and oogenesis and has been shown to be genetically downstream of canonical Wg/Wnt signaling (Mevel-Ninio et al., 1995; Oliver et al., 1994; Payre et al., 1999). lin-48, the ovo homologue found in C. elegans, is required for cell fate specification during hindgut development (Chamberlin et al., 1999). There are three ovo homologues in mammals, designated Ovol1, Ovol2, and Ovol3. In mice, Ovol1 is required for hair follicle differentiation, where it is regulated by nuclear effectors of Wnt signaling, and also for kidney and male germ cell development (Dai et al., 1998; Li et al., 2002b, 2005). Recently, human OVOL1 has been identified as a gene that is responsive to TGF-β1/BMP7 treatment via a Smad4-dependent pathway (Kowanetz et al., 2004). That members of the ovo gene family act downstream of signaling pathways required for diverse processes during both early and late stages of embryonic development raises the possibility that ovo genes might be important for embryogenesis.
Previous studies revealed Ovol2 expression in brain, testis, and epithelial tissues such as skin and intestine of adult mice (Li et al., 2002a). In this study, we show that it is also expressed during early–mid embryogenesis, particularly in the epiblast and its ectodermal derivatives in the developing head. Using a gene targeting approach, we ablated Ovol2 expression and observed that Ovol2 is required for the development of the embryonic brain. We further demonstrate that Ovol2 is required for maintaining the proper expression of signaling molecules such as Shh and Wnt1 along the cranial D–V axis, for positioning the neuroectoderm/surface ectoderm border in the head region, for maintenance of migrating neural crest cells, and for closure of the cranial neural tube.
A 120 kb BAC clone containing the Ovol2 locus was identified by screening a 129/Sv genomic library (Incyte Genomics) and was used to generate a targeting vector designed to delete exons 1a, 1b, and 2 (E1A, E1B, E2) that contain the start codon of the Ovol2 open reading frame (Fig. 1A; Li et al., 2002a). A 2 kb BamHI fragment corresponding to the genomic region upstream of E1A was cloned as the 5′ arm into the pPGKneobpA-lox2PGKDTA vector at the HindII site by blunt-end ligation. Primers corresponding to sequences in E2 and E3 were used to amplify a 14 kb genomic sequence which was subsequently cloned into the TOPO-TA vector (Invitrogen). A 6.8 kb PstI fragment was released from the resulting plasmid and blunt-end ligated into the NotI site of the targeting vector as the 3′ arm. A 1 kb SmaI–AflII fragment encoding the EGFP protein (from pEGFP-N1, Clontech) was inserted downstream of the 5′arm at the NheI site by blunt-end ligation.
The targeting vector was linearized with SacII and electroporated into E14 embryonic stem (ES) cells. After selection in the presence of G418, clones were screened by PCR and Southern hybridization (Figs. 1B–D), two of which were injected into blastocysts. The resulting chimeras were mated with wild-type C57BL/6J females to produce F1 progeny of a C57BL6/J×129Ola (B6×129) mixed genetic background, which were further intercrossed to produce homozygous mutant F2 progeny for study. Subsequent breeding of F1 heterozygous males with wild-type CD1 outbred females generated heterozygous offspring enriched in a CD1 background that were also used to produce homozygous mutant mice for study. Mutant phenotypes in mice derived from both ES clones were identical.
Genomic DNA or lysate isolated from ES cells, tail clippings, or yolk sacs was used for PCR genotyping using the following primer sequences (designated as arrows in Fig. 1A): PCRi: 5′-TGCTTCGTGGGTTGGCCTGAGAAC-3′ and 5′-CAGATGTCCATCGGTGCCTTGGGC-3′; PCRii: 5′-GTTTCGCTTGGTGGTCGAATGGGCAG-3′ and 5′-CACCACAGAGGCTGGGAGTGACATC-3′; PCRiii: 5′-CGCTCCTTCTTTCTAGCAAGTCTCCCG-3′ and 5′-AAGTCGTGCTGCTTCATGTG-3′; PCRiv: 5′-GCTGACCCTGAAGTTCATCTGACCA-3′ and 5′-CGCTTCTCGTTGGGGTCTTTGCTCA-3′. Southern blot analysis was performed essentially as described (Sambrook and Russell, 2001) using the probe indicated in Fig. 1A. RT-PCR was performed using RNA isolated from either single embryos (Superscript One-Step RT-PCR with Platinum Taq kit, Invitrogen) or pooled embryos (Superscript III Reverse Transcriptase, Invitrogen) according to the manufacturer’s recommendations. The following primer sequences were used to detect Ovol2 transcripts: 5′-CCCACCATGCCCAAAGTCTTTCTGGTA-3′ and 5′-GGCGTCGTGAAGCTCTGGAGTTTCAG-3′ or 5′-GAGCCTGGGCTGTCTGCTCCGCG-3′ and 5′-CCCAAGCTTCGCTGCCAGATGTCCATCGGTGC-3′. Primers specific for Gapdh were used as a control.
For morphological analysis, embryos were dissected at the appropriate stages, removed from the yolk sac, fixed for 1–2 h at 4°C in 4% paraformaldehyde/PBS, and washed in PBS. Pictures were taken using a Spot-RT camera mounted on a Leica MZFLIII dissecting microscope equipped with a GFP filter. GFP images were similarly acquired, except that the embryos were unfixed at the time of analysis. For histological analysis, whole deciduae at the appropriate stages were removed from the uterus, fixed for 2 h at room temperature in Bouin’s fixative, and subsequently washed in 70% ethanol until embryos were clear. Following dehydration and embedding in paraffin, 5 µm sections were cut and stained with hematoxylin and eosin.
Embryos were dissected at the indicated stages, fixed for 2 h at 4°C in 4% paraformaldehyde/PBS, dehydrated through a series of increasing methanol concentrations, and stored at −20°C. Mutant embryos in both B6×129 and CD1-enriched backgrounds that showed severe neural fold rounding at E8.5 were used for final analysis along with stage-matched controls. Whole mount in situ hybridization was performed essentially as described (Nagy et al., 2003). For select probes, OCT sections (14 µm) were prepared after hybridization for further analysis. The following probes were transcribed using either T7, T3, or Sp6 RNA polymerase (T7 and Sp6, New England Biolabs; T3, Stratagene): Ovol2 (GenBankAY0990537); Shh (Echelard et al., 1993); Otx2 (Ang et al., 1996); Fgf8 (Mahmood et al., 1995); Krox20 (Wilkinson et al., 1989); Ptch1 (Goodrich et al., 1996); Wnt1 (Parr et al., 1993); Nkx2.1 (Camus et al., 2000); Pax3 (Goulding et al., 1991); Msx1 (Catron et al., 1996); Msx2 (Catron et al., 1996); Wnt6 (Parr et al., 1993); Dlx3 (Morasso et al., 1996); and Sox10 (Pusch et al., 1998).
Proliferating cells in the embryo were labeled by injecting pregnant female mice with 50 µg BrdU/g body weight 30 min before dissection. Embryos were dissected in PBS and fixed for 1–2 h in 4% paraformaldehyde/PBS, equilibrated in 30% sucrose, and frozen in OCT. Frozen sections (12–14 µm) were stained with an anti-BrdU monoclonal antibody (Roche) as previously described (Li et al., 2005). Antibodies against Ki67 (Vector Laboratories) and phosphorylated histone-H3 (Upstate Biotechnology) were used to stain frozen sections using standard immunostaining protocols. The number of phospho-H3-positive nuclei in the neuroectoderm was counted from serial sections and compared to the total number of cells as determined by DAPI-positive nuclei.
Nile blue sulfate staining was performed essentially as described (Anderson et al., 2002), except that embryos were stained for 1 h on ice. The TUNEL assay was performed on whole embryos fixed in 4% paraformaldehyde using the In Situ Cell Death Detection Kit (Roche), according to the manufacturer’s recommendations.
To understand the biological role of Ovol2, we used homologous recombination in embryonic stem (ES) cells to ablate the murine Ovol2 gene. We replaced the initial three exons of Ovol2 with a cassette containing the EGFP and neor genes, which would result in EGFP expression being controlled by the Ovol2 promoter region (Fig. 1A). The targeting construct was electroporated into ES cells followed by selection with G418, yielding two independently targeted clones that were injected into blastocysts and gave germline transmission (Figs. 1B–D). No homozygous Ovol2 mutant pups were recovered from heterozygote intercrosses in either the B6×129 or the CD1-enriched genetic backgrounds examined (Table 1). Further analysis of embryonic litters at various stages indicated that Ovol2 mutant embryos died around embryonic day 10.5 (E10.5; Table 1), and χ2 analysis of genotype distributions at earlier stages of development revealed no significant deviation from the expected Mendelian ratios (data not shown). Thus, these data indicate that Ovol2 is required for embryonic development before E10.5.
To identify the tissues whose development might be affected by loss of Ovol2, we analyzed its spatio-temporal expression during early–mid embryonic development. RT-PCR analysis revealed that Ovol2 is expressed in ES cells and E7.5, E8.5, E9.5, and E10.5 wild-type embryos (Fig. 2A). EGFP fluorescence in heterozygous Ovol2 embryos was examined as a read-out of Ovol2 promoter activity. EGFP expression was detected in the inner cell mass of blastocysts (Fig. 2B) and later in the epiblast but not the extraembryonic region of E6.5 embryos (Fig. 2C). As development progressed, EGFP protein continued to be present broadly in epiblast-derived tissues, with enhanced fluorescence in what appear to be anterior definitive endoderm and axial mesoderm (arrowhead and arrow in Fig. 2F, respectively) and no expression in the visceral endoderm (Figs. 2D–F). By ~E8.5, EGFP fluorescence became more restricted and was observed in the cephalic neural folds (arrow), the surface ectoderm bordering the neural plate (arrowheads), and the gut tube (open arrowhead), but not in the developing heart (*; Figs. 2G–H).
Whole mount in situ hybridization was then performed using Ovol2 antisense probes. Ovol2 transcripts were detected in E8.75 embryos in the gut tube endoderm and in the surface ectodermal component of the head and the newly formed first branchial arch, but not in heart (Figs. 2I–K). At E9.5, Ovol2 expression was observed in the forebrain area, as well as in the branchial arches, otic vesicles, developing gut, and at low levels throughout the entire embryo (Figs. 2L, N–P). No hybridization signal was observed in stage-matched Ovol2 mutant embryos at E8.75 or at E9.5 (Fig. 2M and data not shown), indicating that the signals observed in the wild-type embryos are specific to Ovol2 and that the mutant allele generated in the present study likely represents a null allele. Moreover, the similarity in the distribution of Ovol2 transcripts and EGFP fluorescence confirms that the presence of EGFP protein is a good indicator of the endogenous pattern of Ovol2 expression.
We next performed a detailed morphological analysis of Ovol2 mutant embryos at different stages prior to death. Initial analyses were performed on F2 homozygous mutants from F1 heterozygote intercrosses in a mixed B6×129 genetic background. Given the expression of Ovol2 in the presumptive forebrain area and in the surface ectoderm adjacent to the cranial neuroectoderm, we first focused on the developing brain. The earliest, most consistent signs of morphological defects were detected in mutant embryos at ~E8.5–8.75, when they displayed larger, more rounded neural folds (white arrowheads; Figs. 3A–B, D–E). Optic invaginations, which normally occur at this stage, were not observed in the mutant (black arrowheads in Figs. 3A, D). These defects were not due to a general delay in development as the mutant and wild-type embryos analyzed possessed the same number of somites. The rounded neural folds in the E8.5 mutant embryos were more evident at a histological level. In wild-type embryos, a sharp transition from neuroectoderm to surface ectoderm was observed, and there was a clear morphological distinction between these two adjacent epithelia (Fig. 3G). In the mutant, however, the neuroectoderm was expanded and folded laterally towards the surface ectoderm, as identified by its thin appearance, resulting in an apparent lateral shift of the neuroectoderm/surface ectoderm junction (Fig. 3H). The transition from the columnar epithelial morphology of the neuroectoderm to the cuboidal epithelium of the surface ectoderm was not as distinct in the mutant and appeared to be more gradual (see insets in Figs. 3G–H). By E9.5, all of the mutant embryos displayed a failure of cranial neural tube closure, as well as a failure of the turning process (Fig. 3F; compare to to3C).3C). There was no evidence of formation of the optic eminence or otic pit in the mutant, and the branchial arches were often underdeveloped. By E10.5, most mutant embryos were fully resorbed, leaving empty yolk sacs that appeared deflated and seemed to lack proper vascularization (data not shown).
A similar cranial phenotype was observed when the mutant allele was bred into a CD1-enriched genetic background (Figs. 3I–N). Sectioning through these mutant embryos revealed that, at E9.5, their forebrain developed as two separate hollow tubes instead of forming a single vesicle (Fig. 3O). There was a severe paucity of cranial mesenchyme (*) and no sign of surface ectoderm covering the forebrain region. We should note that the apparent expressivity of the neural phenotype was lower in this background than that observed in the B6×129 background. For example, not all mutant embryos in the CD1-enriched background showed a dramatic rounding of their neural folds at E8.5, and rudimentary optic and otic structures were often seen at E9.5 (data not shown). Furthermore, most mutant embryos were able to turn by E9.5, although they displayed a twisted body axis with multiple kinks (Figs. 3F, N; also see Fig. 5G). Surprisingly, we recovered a single mutant at E9.5 showing a rather normal external appearance, with near-normal head morphology, body shape, and sensory structures. However, this “escaper” still displayed areas of open cranial neural tube as well as open regions in trunk neural tube (spina bifida; Fig. 3P).
Consistent with Ovol2 expression in the developing gut tube, histological analysis revealed abnormal gut morphology in the Ovol2 mutants, the extent of which was not significantly affected by strain background (Fig. 3H, data not shown). Overall, the mutant gut epithelium appeared less developed than the wild-type, and foregut often lacked the eosin-positive acellular debris normally found within the wild-type lumen. Marker expression studies (Mackay and Dai, unpublished) further suggested that the differentiation of the hindgut endoderm is delayed in the Ovol2 mutant.
While Ovol2 expression was not detected in the developing heart tube at E8.5, all mutant embryos at this stage displayed severely defective heart development (Figs. 3D–F). Similar to neural development, the heart phenotype was slightly more severe in the B6×129 background than that in CD1. Compared to wild-type E8.5 embryos that had a single linear or already looped heart tube, the somite-matched mutant embryos either displayed a failure in the fusion of the left and right heart primordia or lacked any sign of heart primordium altogether. This cardia bifida phenotype was confirmed at a histological level, where it was clear that in some mutants the two heart fields failed to fuse properly (arrows in Figs. 4B–C). As development progressed, mutant hearts were able to form a single heart tube but were much smaller than their wild-type counterparts and showed abnormal patterns of looping morphogenesis and septation, resulting in a single ventricular chamber abutted by two laterally positioned atrium-like structures (Figs. 4D–M). Moreover, the mutant hearts displayed an apparently dilated indistinct atriosinus region (*; compare Figs. 4M to I) and an indistinct outflow tract/aortic sac region (arrows; compare Figs. 4J to F). These severe heart defects in the mutants would be expected to result in an embryonic heart failure, which was most likely the cause of the subsequent embryonic death.
Correct A–P and D–V patterning re essential for neural development and neural tube closure (Joyner, 2002). We first examined A–P brain patterning by performing whole mount in situ hybridization using a panel of probes marking particular regions of the developing brain along the A–P axis. Otx2 is normally expressed in the neuroectoderm and anterior mesendoderm of the developing forebrain and midbrain and is essential for the proper patterning of anterior neural tissues (Ang et al., 1996). Wild-type embryos showed an expected pattern of Otx2 expression, and Otx2 expression in Ovol2 mutant embryos was not significantly different, indicating that the forebrain and midbrain domains were specified properly in the absence of Ovol2 (Fig. 5A). Further experiments using probes for Fgf8, expressed in the anterior neural ridge located at the rostral margin of the neural plate and the isthmic organizer located at the midbrain/hindbrain boundary (Mahmood et al., 1995), and Krox20, expressed in the hindbrain (Wilkinson et al., 1989), confirmed that A–P brain patterning was not significantly affected in Ovol2 mutants (Figs. 5B–C).
To examine D–V patterning of the developing mutant brain, we performed in situ hybridization using a probe for Shh. At E8.5, Shh is normally expressed in the axial mesoderm of the notochord and prechordal plate (arrow in Fig 5D, top panel) and later in the ventral aspect of the brain (arrowhead in Fig. 5E, top panel). In Ovol2 mutant embryos, Shh expression in the notochord appeared normal; however, the prechordal plate region marked by Shh was not as rostrally extended as seen in the wild type and lacked the characteristic hook-shaped morphology (arrows in Figs. 5D–E). In addition, Shh expression appeared more diffuse in the truncated prechordal plate region and was weaker in the ventral forebrain of the mutants (Figs. 5E–F). As the embryos developed further, the defects in prechordal plate and ventral forebrain became more pronounced as the domain of Shh expression was not properly maintained (Fig. 5G; arrowhead indicates a gap within the Shh expressing domain of mutant ventral forebrain). Additionally, the notochord in the mutants displayed multiple kinks (* in Fig. 5G), which further indicated that the body axis was twisted or improperly extended. Expression of Ptch1 appeared diffuse in the mutant ventral brain (Fig. 5H), likely due to the improper distance from the source of Shh signal. We next examined the expression of Wnt1, which is normally localized to the dorsal aspect of the cranial neural folds of the midbrain region (Parr et al., 1993). As expected, wild-type embryos expressed Wnt1 in a spatially restricted area adjacent to the neuroectoderm/surface ectoderm junction (Figs. 5I–J). In the mutants, however, the domain of Wnt1 expression was expanded ventrally, although a dorsal concentration was still evident (Figs. 5I–J). Additionally, the dorsal boundary of Wnt1 expression was shifted laterally, consistent with the apparent shift of the neuroectoderm/surface ectoderm border observed at a morphological level (see arrowheads in Figs. 3H and and5J).5J). Taken together, these results indicate that the loss of Ovol2 does not affect A–P patterning of the brain but leads to slightly altered expression of signaling molecules that are known to be involved in D–V patterning of the neural tube.
To examine whether D–V specification is disrupted in the Ovol2 mutant brain, we investigated the expression of several well-characterized D–V neural markers. The expected ventrally or dorsally localized expression of Nkx2.1, Pax3, Msx1, and Msx2 in the cranial neural tube suggests that the overall D–V domains were largely maintained in the Ovol2 mutant (Figs. 6A–B and data not shown). Remarkably, Msx2 and Pax3 expression showed a similar lateral shift towards the presumptive surface ectoderm (Figs. 6A–B), as observed for Wnt1 expression.
The lateral shift of dorsal neural markers in the mutant, together with the apparent increase in the contour of the neuroectoderm, suggests that mutant neuroectoderm is expanded. At least three possibilities may account for such an expansion: overspecification of a neuroectodermal fate at the cost of a surface ectodermal fate, increased neuroectodermal proliferation, and/or decreased neuroectodermal apoptosis. To address whether the surface ectoderm is specified properly in the mutant, we investigated the expression of a surface ectoderm marker, Wnt6. Wnt6 expression was observed in the mutant in the presumptive surface ectoderm that juxtaposes the dorsal border of the neural plate, indicating that Ovol2-deficient ectoderm was able to differentiate into surface ectoderm (Fig. 6C). However, the ratio of the cranial surface ectoderm to neuroectoderm was reduced two-fold in the mutants compared to that observed in the wild type (Fig. 6D; P < 0.0001). Therefore, a concomitant expansion of neuroectoderm and reduction of surface ectoderm in the mutant contributes to the observed lateral shift of the cranial neuroectoderm/surface ectoderm border.
This phenotype is reminiscent of that observed in Xenopus and zebrafish when Dlx3 function was disrupted (Woda et al., 2003). To address whether Dlx3 expression is affected in Ovol2 mutant, we performed whole mount in situ hybridizations using a Dlx3 probe on E8.5 embryos. Previous studies reported Dlx3 expression in the first and second branchial arches in mouse embryos at E9.5 (Robinson and Mahon, 1994) and weakly in the rostral ectoderm at E8.0 (Quint et al., 2000). Consistent with these reports, we observed strong Dlx3 expression in the first branchial arch and weak expression in the cranial surface ectoderm in wild-type E8.5 embryos. Compared to their wild-type littermates, all seven mutant embryos examined displayed much weaker expression in comparable regions, with the difference most apparent when color development for whole mount in situ hybridization was prolonged to enhance the surface ectoderm signal (Figs. 6E–G). These results indicate that a functional Ovol2 is required for Dlx3 expression in these areas.
To assess a possible contribution of proliferation to neuroectoderm expansion, we next analyzed the extent of proliferation in the neuroectoderm at E8.5. Widespread proliferation was detected using BrdU labeling and immunological staining for Ki67 in both wild-type and mutant embryos, consistent with rapid growth at this stage of development (data not shown). We instead determined the mitotic index using an antibody that recognizes phosphorylated histone H3, a marker for cells in mitosis (Gurley et al., 1978). The mitotic index was calculated as the ratio of the number of cells in the neuroectoderm that stained positive for phosphorylated histone H3 over the total number of cells in the neuroectoderm as determined by DAPI staining (Figs. 7A–B). We observed regional variability in both the wild-type and mutant embryos, so we averaged the mitotic index over multiple serial sections. Analysis of either the cranial region alone or both cranial and caudal regions combined revealed no increase in mitotic index (Fig. 7B and data not shown). Instead, we observed a slight but statistically significant (P = 0.037) decrease in the rate of proliferation in the mutant neuroectoderm.
We also investigated the status of apoptosis in the neuroectoderm at E8.5. Using both Nile blue sulfate staining (Anderson et al., 2002) and TUNEL assays, we did not detect any significant difference in the number of apoptotic cells in wild-type and mutant neuroectoderm at this stage (Fig. 7C and data not shown). Collectively, our data exclude overproliferation or decreased apoptosis as likely causes of cranial neural expansion in the mutant but instead suggest that the lateral shift of the neuroectoderm/surface ectoderm border is a consequence of the expansion of neuroectoderm at the cost of surface ectoderm in the cranial region.
The mispositioning of the neuroectoderm/surface ectoderm border in Ovol2 mutants raises the possibility that border cell fates might be affected. The best characterized cell type arising at the border is neural crest cells that undergo an epithelial–mesenchymal transition, delaminate from the neuroepithelium, and migrate away from the neural plate border to various predetermined target locations (Gammill and Bronner-Fraser, 2003). Upon morphological examination, we noticed that cells with neural crest morphology were present in mutant E8.5 embryos, except that they clustered next to the neuroectoderm/surface ectoderm border (see arrow in Fig. 3H). Furthermore, we observed a similar characteristic pattern of apoptosis along the cranial neuroectoderm/surface ectoderm border of both wild-type and mutant E8.5 embryos, previously implicated as from apoptotic neural crest cells (Graham et al., 1994; Trainor et al., 2002). Therefore, it appears that neural crest formation is initiated in the mutant (see below). However, when Nile blue staining was performed on E9.5 embryos, a dramatic increase in the number of apoptotic cells was observed in the mutant, particularly in forebrain, branchial arch, and trunk regions (Figs. 7D–F, H–J). TUNEL assays, while not as sensitive as the Nile blue staining in our hands, confirmed that apoptosis was indeed elevated in the mutant (Figs. 7G, K). The distribution of these apoptotic cells in Ovol2 mutant embryos is reminiscent of that of migrating neural crest cells.
To directly assess the neural crest population in Ovol2 mutants, we next examined the expression of Sox10, which marks migrating neural crest cells (Cheng et al., 2000; Pusch et al., 1998). At E8.5–E8.75, largely normal but slightly enhanced Sox10 expression was observed in the mutant embryos, confirming that neural crest cells are specified and formed (Fig. 8A). A detailed analysis of sections revealed that the Sox10-expressing cells were displaced laterally towards the surface ectoderm, consistent with the lateral shift of neuroectoderm/surface ectoderm boundary (Fig. 8D). Furthermore, Sox10-positive cells still appeared to be an integral part of the neuroectoderm in the mutant, suggesting that their emigration was not efficient. This notion is consistent with the morphological finding of clustered neural crest cells next to the neuroectoderm/surface ectoderm border (see above). By E9.5, the mutants displayed regions of Sox10 expression that appeared smaller, and either more (compare i in Figs. 8B–C) or less (compare ii in Figs. 8B–C) intense than the wild type. Additionally, the mutants displayed either missing or ectopic regions of Sox10-expressing neural crest cells (arrows in Figs. 8B–C). Sectioning through the embryos indeed revealed the presence of condensed clusters of Sox10-expressing cells in the mutant, some of which failed to leave the region immediately adjacent to the laterally shifted neuroectoderm/surface ectoderm border (arrowheads in Fig. 8E). Taken together, these results suggest that neural crest maintenance and migration are disrupted in the absence of Ovol2.
In this study, we uncovered a previously unsuspected function of Ovol2 in early–mid embryogenesis. Our data demonstrate that a functional Ovol2 gene is required for the proper development of multiple tissues including brain, neural crest, gut tube, and heart. Presently, our studies do not unequivocally distinguish primary defects from those that arise as secondary consequences of the initial tissue defect, particularly because Ovol2 expression in early embryos is rather ubiquitous and only gradually becomes restricted to specific tissues or sites. While several independent cellular and molecular mechanisms might underlie the cranial neural tube defect, the major focus of this work (see below), the gut defects might arise as a direct effect of loss of Ovol2 expression in the developing gut endoderm. Little or no Ovol2 expression was detected in the developing heart at the time when it was affected, raising the possibility that this defect might be secondary. In this context, we note that over-emphasized anterior neural development, which was observed in these mutants, has been shown to interfere with heart development (Fishman and Chien, 1997; Jacobson, 1960). Alternatively, the heart defect may be a result of improper neural crest migration and maintenance as it is known that a specific population of neural crest cells, known as the cardiac neural crest, contributes to the development of the outflow tract and possibly other parts of the developing heart (Stoller and Epstein, 2005). Finally, it is possible that Ovol2 is expressed in heart precursor cells at earlier stages of development and is required for heart tube formation and looping morphogenesis.
The ventral midline emanates important signals that pattern the developing neural plate along the A–P and D–V axis. The loss of Ovol2 led to a defective pattern of Shh expression in the rostral domain of the axial mesoderm. Curiously, while EGFP fluorescence suggested Ovol2 expression in the ventral midline tissues at the headfold stage, Ovol2 transcripts were not detected in the prechordal plate or notochord after E8.5, yet the extent of disruption of Shh expression in these axial structures was more severe at E9.5 than at E8.5. Therefore, it remains to be addressed whether the improper rostral extension of Shh expression is a direct effect of loss of Ovol2 in the axial structures.
Consistent with the prevailing notion that Shh signaling from the prechordal plate induces Shh expression in the neuroectoderm of the ventral brain (Shimamura and Rubenstein, 1997), mutant embryos displayed a reduced level of Shh expression in the ventral forebrain, likely due to improper Shh signaling from the prechordal plate. However, the expression of several well-known D–V markers appeared grossly normal in Ovol2 mutant neuroectoderm. Specifically, the expression of ventral markers such as Ptch1 and Nkx2.1 was induced in a ventrally restricted fashion. Furthermore, dorsal markers such as Pax3 and Msx1/2 were restricted to the dorsal domain of the E8.5 mutant cranial neural tube, confirming that the ventralizing signals were sufficient to repress the expression of these genes in the ventral domain (Liem et al., 1995). Instead, we observed rather specific and local changes in gene expression. For example, we consistently observed ectopic sites of Ptch1 and Nkx2.1 expression in the mutant brain, too weak and diffuse to clearly document. Additionally, Wnt1, a gene whose expression is normally confined to the dorsal aspect of the neuroectoderm, was abnormally expanded into the ventral domain of the mutant neuroectoderm. Since we observed no significant increase in the expression levels of BMP2 and BMP4 – candidate dorsalizing signals – in the mutant (data not shown), we surmise that the ventral expansion of Wnt1 expression is likely due to compromised repression from sub-optimum ventral Shh signaling but not over-emphasized dorsalizing signals (Basler et al., 1993; Dickinson et al., 1995; Goulding et al., 1993; Kanzler et al., 2000). This notion is also more in line with the observed reduction (not increase) in the area of surface ectoderm, thought to be the source of dorsalizing signals (Dickinson et al., 1995; Lee and Jessell, 1999; Liem et al., 1995). This said, we cannot rule out the possibility that the expression of other BMPs (e.g., BMP7) or that the downstream components of BMP signaling are somehow affected, leading to a change in BMP activity gradient across the neuroectoderm D–V axis. Collectively, our studies demonstrate that the loss of Ovol2 leads to local changes in the expression of specific signaling molecules and transcription factors but not a general disruption of D–V patterning in the developing brain. The mid-gestation lethality of Ovol2 mutants precludes a detailed analysis of specification of ventral neural fates to evaluate the consequence of the observed gene expression defects. Future studies using a conditional Ovol2 allele to bypass the lethality issue will likely be informative.
Another interesting and likely independent consequence of Ovol2 ablation is the lateral shift of the neuroectoderm/surface ectoderm border in the dorsal aspect of the rostral neural tube. This is apparent at a histological level and is confirmed by expression studies of neuroectoderm and surface ectoderm markers. Quantitative analysis of the relative linear dimensions of these ectodermal derivatives indicates that the neuroectoderm is expanded, while surface ectoderm is reduced in the cranial region of the mutant embryos. These results suggest that Ovol2 is involved in patterning the neural and non-neural ectoderm, a role that is consistent with its expression in the ectoderm at early stages of embryogenesis. Specifically, Ovol2 appears to function as an anti-neural gene (see below), where its absence in the ectoderm leads to over-specification of a neural fate. To our knowledge, this work represents the first example elucidating the functional involvement of a transcription factor in patterning the neural and epidermal ectoderm in mice.
The lateral and ventral expansion of Wnt1 expression in the mutant neuroectoderm deserves more attention. Genetic studies in mice suggested that Wnt signaling, via redundant functions of Wnt1 and Wnt3a in the dorsal neural midline, is required for proliferation or survival of dorsal neural precursor cells, including the neural crest, and/or for maintaining a stable dorsal neural fate (Dickinson et al., 1994; Ikeya et al., 1997; Saint-Jeannet et al., 1997). It may be possible that this expanded Wnt1 expression causes the lateral shift of the neuroectoderm/surface ectoderm border by promoting dorsal neuroectodermal proliferation (Dickinson et al., 1994); however, we consider this possibility highly unlikely due to the following reasons: (1) we observed decreased proliferation in the mutant neuroectoderm, contradictory to the expected effect of Wnt1 in dorsal ectoderm, and (2) it has been shown that contact with the surface ectoderm can induce ectopic Wnt1 expression in the neuroectoderm (Liem et al., 1995), suggesting that formation of the neuroectoderm/surface ectoderm border precedes border-specific expression of Wnt1. Furthermore, induction of neural crest, a border-specific cell type, can occur without Wnt1 and Wnt3a (Basler et al., 1993; Dickinson et al., 1995; Goulding et al., 1993). Therefore, we favor the hypothesis that the lateral shift of the mutant neuroectoderm/surface ectoderm border reflects a primary defect in neural/non-neural patterning and that the lateral shift of Wnt1 expression is simply a consequence of an expanded neural plate, which pushes the border and its associated gene expression laterally.
The anti-neural function of Ovol2 resembles that of Xenopus/ zebrafish Dlx3, which encodes a homeodomain transcription factor. Loss of Dlx3 leads to an expansion of neuroectoderm at the expense of surface ectoderm, resulting in mispositioning of the neural/non-neural border (Woda et al., 2003). Dlx3, like Ovol2, is first expressed broadly throughout embryonic ectoderm and then becomes restricted to the non-neural ectoderm (Woda et al., 2003). The current model states that Dlx3 functions by repressing the expression of anterior proneural genes, whereby directing or permitting ectoderm to adopt a non-neural fate in response to developmental signals such as Wnt and BMP (Beanan et al., 2000; Feledy et al., 1999). Unlike another anti-neural gene, Msx1, which is an immediate-early target of BMP signaling, activation of Dlx3 expression requires additional factors that are synthesized following BMP signaling (Feledy et al., 1999). It is tempting to speculate that the Ovol2 protein may be one such factor that is required to activate Dlx3 expression in surface ectoderm and its precursor cells, a notion supported by the significant reduction of Dlx3 expression in Ovol2 mutant embryos.
It is important to note that a similar function has not yet been described for Dlx3 in mice. Dlx3-deficient mice die between E9.5 and E10, within a similar time-frame as Ovol2 mutants; however, the initial study only focused on the placental phenotype of these mice (Morasso et al., 1999). In light of our findings, it should be worthwhile to further explore the Dlx3 mutant mice to address a likely conserved involvement of Dlx3 in mammalian neural induction and patterning. The hypothesis that Ovol2 and Dlx3 function in a common pathway to position the neuroectoderm/surface ectoderm border and that Dlx3 is genetically downstream of Ovol2 should be systematically tested. Furthermore, given that its ovo relatives in flies, mice, and human are regulated by BMP and Wnt signaling pathways, Ovol2 itself might be subject to regulation by the anti-neural BMP and Wnt signaling in the ectoderm, a notion worth addressing in the future.
Neural crest cells emerge from the neuroectoderm/surface ectoderm border, where complex interactions between signaling pathways are thought to be key to neural crest induction and emigration (Gammill and Bronner-Fraser, 2003). It is conceivable that a mispositioned border might result in aberrant tissue and molecular interactions, which could in turn lead to defective neural crest development. Our studies demonstrate that, despite the lateral shift of neuroectoderm/surface ectoderm border in Ovol2 mutants, neural crest formation is largely unaffected. This result may not be too surprising considering the largely normal levels of expression of Msx1/2, presumptive targets of BMP signals from the surface ectoderm, and Wnt6, recently proposed to be a major neural crest-inducing signal from the surface ectoderm (Trainor, 2005), in the Ovol2 embryos. Instead, the emigration of nascent neural crest cells from the neural plate and the survival of migrating neural crest cells were impaired, demonstrating a role for Ovol2 in these later processes.
How Ovol2 might regulate neural crest cell migration and survival is another interesting question. Besides its role in promoting neural crest precursor proliferation, Wnt signaling is also implicated in neural crest migration as treatment of cultured neural crest cells with reagents that mimic an activated Wnt signal results in inhibition of migration (de Melker et al., 2004). Therefore, the ventrally expanded Wnt1 expression of the mutant neuroepithelium may create a stronger source of Wnt signal that inhibits migration of the neural crest cells. Genetic perturbation experiments suggest that neural crest migration and survival is more than just an intrinsic property but rather is controlled non-cell-autonomously by signals from multiple tissues including the surface ectoderm (Trainor, 2005). Since an abundant site of Ovol2 expression at ~E8.5 is the surface ectoderm, it is likely that Ovol2 regulates signals originating from the surface ectoderm, which in turn regulate neural crest migration. As the actual signals that regulate cranial neural crest migration in mice (independently of a function in neural crest induction) have not been unequivocally identified (Trainor, 2005), it is difficult to postulate at this time the specific nature of these signals. Furthermore, we cannot rule out the possibility that Ovol2 is expressed in neural crest cells at a level below our methods of detection and is required autonomously in these cells for their migration and maintenance.
The first two cranial neural crest streams are destined to migrate to the first and second branchial arches, respectively (Kulesa et al., 2004). The branchial arches in the Ovol2 mutants were consistently less developed than those in the wild type. Although some neural crest cells were able to migrate to and populate the branchial arches in the mutant, the Sox10-positive neural crest domain was smaller (Fig. 8 and data not shown), suggesting that branchial arch underdevelopment is at least in part due to neural crest defects. Alternatively, since Ovol2 is expressed in the surface ectoderm of branchial arches, it is possible that mutant branchial arches fail to emit the necessary signals (e.g., FGFs; Trumpp et al., 1999) to attract or maintain migrating neural crest cells, thus leading to their apoptotic death along the migrating paths and within the branchial arches. Regardless of the underlying mechanism, our findings uncover an important role for Ovol2 in neural crest cell migration and maintenance, an involvement that is apparently independent of the initial induction event. Unfortunately, the death of Ovol2 mutant embryos shortly afterwards prevents us from analyzing the neural crest derivatives at later developmental time points. Again, conditional ablation of Ovol2 will help circumvent this problem.
At a morphological level, the cranial neurulation defect in Ovol2 mutant embryos mimics those observed in mice deficient in Ptch1 (Goodrich et al., 1997) or Atrophin 2 (Zoltewicz et al., 2004), both associated with abnormal Shh expression and/or Shh signaling. While uncontrolled Shh signaling due to the loss of a negative regulator underlies the morphological defect in Ptch1 mutants, Atrophin 2-deficient embryos display a complete loss of Shh expression in the anterior midline. The defect in Ovol2 mutants, however, is milder and is characterized by an abnormal distribution of Shh gene products and signaling activity in the cranial region, creating local domains with either slightly diminished or enhanced Shh signaling activity. The apparent discrepancy between the severe morphological anomaly and mild Shh signaling defect implies the existence of additional factors contributing to a failure of cranial neural closure in the absence of Ovol2. The lateral shift of the neuroectoderm/surface ectoderm border in Ovol2 mutant likely represents one such contributing factor. Morphological defects of cranial neural folds and a disrupted neuroectoderm/surface ectoderm border similar to what we observed in the Ovol2 mutants have been previously associated with a high incidence of neural tube closure defects in mice (Gunn et al., 1993), providing correlative support for this hypothesis. Another probable cause is the reduced proliferation in Ovol2 mutant neuroectoderm at E8.5, which might lead to an improper shape of the developing neural plate, thereby creating a physical barrier for its dorsal closure at the anterior end. Finally, the untimely migration of neural crest cells away from the neural plate border and the reduced amount of head mesenchyme likely also contribute to the open brain defects. Collectively, our results demonstrate that Ovol2 is involved in multiple cellular, molecular, and morphological events required for brain morphogenesis and implicate mice containing Ovol2 mutations as useful tools to study this important biological process and to model after human diseases with similar brain defects.
We would like to thank the UCI Transgenic Mouse Facility and the UCI Optical Biology Core for technical assistance; Bill Skarnes for the generous gift of E14 ES cells; Grant MacGregor for getting us started with embryological techniques; Grant MacGregor and Patrick Tam for helpful suggestions and critical reading of the manuscript; and Marianne Bronner-Fraser, Siew-Lan Ang, Ian Gallicano, Maike Sander, Andy McMahon, Maria Morasso, Alex Joyner, Phil Soriano, Jonathan Epstein, and David Wilkinson for in situ hybridization probes; and reviewers of the manuscript for helpful insights. D.R.M. was supported by an NIH institutional training grant on Structure and Function of Macromolecules. This work was supported by the NIH Research Grant R01 AR47320 and Independent Scientist Award K02 AR51482 awarded to X.D.