ABSTRACT
In order to study the mechanism of neural patterning in Xenopus, we used subtractive cloning to isolate genes activated early during this process. One gene isolated was opl, (odd-paired-like) that resembles the Drosophila pair-rule gene odd-paired and encodes a zinc finger protein that is a member of the Zic gene family. At the onset of gastrulation, opl is expressed throughout the presumptive neural plate, indicating that neural determination has begun at this stage while, by neurula, opl expression is restricted to the dorsal neural tube and neural crest. opl encodes a transcriptional activator, with a carboxy terminal regulatory domain, which when removed increases opl activity. opl both sensitizes animal cap ectoderm to the neural inducer noggin and alters the spectrum of genes induced by noggin, allowing activation of the midbrain marker engrailed. Consistent with the later dorsal neural expression of opl, the activated form of opl is able to induce neural crest and dorsal neural tube markers both in animal caps and whole embryos. In ventral ectoderm, opl induces formation of loose cell aggregates that may indicate neural crest precursor cells. Aggregates do not express an epidermal marker, indicating that opl suppresses ventral fates. Together, these data suggest that opl may mediate neural competence and may be involved in activation of midbrain, dorsal neural and neural crest fates.
INTRODUCTION
One of the first decisions made during neural determination in Xenopus is a general increase in the competence of the ectoderm to respond to subsequent neuralizing signals. Both transplantation experiments and exposure of explants to pure factors or to natural inducing tissue have shown that neural competence in Xenopus is maximal at early gastrula; by the end of gastrulation, ectoderm displays little competence to form neural tissue in response to mesoderm-derived signals (Nieuwkoop and Koster, 1995; Jones and Woodland, 1987; Sive et al., 1989; Servetnick and Grainger, 1991; Kintner and Dodd, 1991; Lamb et al., 1993; Knecht and Harland, 1997).
Superimposed on an increase in neural competence of the entire ectoderm is an enhanced competence of the dorsal ectoderm (the presumptive neurectoderm) for dorsal fates. The dorsal ectoderm is already biased towards dorsal fates by early blastula stages, since isolated dorsal ectoderm treated with activin goes on to form more dorsal mesoderm than does ventral ectoderm (Sokol and Melton, 1991). By early gastrula, isolated dorsal ectoderm fails to activate expression of the epidermal antigen, Epi-1, whereas this antigen is strongly specified in the ventral ectoderm, which gives rise to epidermis (Savage and Phillips, 1989). Consistently, the dorsal ectoderm is more responsive than ventral ectoderm to neural-inducing signals arising from the dorsal mesendoderm (Sharpe et al., 1987).
By midgastrula, specification assays indicate that neural determination has occurred (Sive et al., 1989; Sharpe and Gurdon, 1990). Initial neural induction is thought to be anterior in character and appears to be mediated by antagonists of Bone Morphogenetic Proteins (BMPs), such as the secreted factors noggin and chordin (Eyal-Giladi, 1954; Sive et al., 1989; reviewed in Sasai and DeRobertis, 1997). Later, the posterior part of the neural plate is converted into posterior neural tissue. Retinoids, fibroblast growth factor and wnt proteins may all be involved in this process, perhaps acting at different anteroposterior levels (reviewed in Kolm and Sive, 1997).
Initiation of dorsoventral neural tube patterning also occurs during gastrulation (reviewed in Placzek and Furley, 1996). In Xenopus ventral parts of the neural tube have been specified by midgastrula (Saha et al., 1997) presumably under direction of ventral midline-derived signals, particularly sonic hedgehog (Roelink et al., 1994). Similarly, at this time, expression of pax3 (a dorsal neural tube marker) is beginning to be restricted to the future dorsal neural plate showing that dorsal neural fates have started to be determined (Bang et al., 1997). Final dorsal neural tube fates require input from adjacent non-neural ectoderm (Liem et al., 1995; Bang et al., 1997), including members of the TGFβ gene family. The neural crest is specified later than dorsal neural tube, by the end of gastrulation, with specification increasing as the neural tube closes (Mayor et al., 1995). Neural crest induction requires signals both from the neural plate and from the juxtaposed non-neural ectoderm, and perhaps also signals from underlying paraxial mesoderm (reviewed in Bronner-Fraser, 1995).
Neural differentiation is regulated by genes such as neurogenin, Xash3 and other members of the bHLH class (Turner and Weintraub, 1994; Zimmerman et al., 1993; Ma et al., 1996). Interfacing with neural differentiation is the correct patterning of the nervous system that must be controlled within the neurectoderm by a distinct set of genes. However, few genes that control early patterning decisions within the neurectoderm have been defined (for examples see Blitz and Cho, 1995; Pannese et al., 1995; Kolm and Sive, 1995a). In order to address this deficit, we have used subtractive cloning (Sagerström et al., 1997) to identify genes expressed specifically in the dorsal ectoderm during gastrula stages, after neural determination has occurred, but before neural differentiation has begun. To date, we have isolated more than 40 genes expressed preferentially in the dorsal ectoderm by midgastrula (Patel and Sive, 1996; Gamse, J., Kuo, J. S., Patel, M. and Sive, H., unpublished data). One of the earliest markers of the neurectoderm that we identified is a zinc finger gene that is similar to Drosophila odd-paired (opa) (Benedyk et al., 1994) that we called opl for odd-paired-like. opl is a member of the Zic gene family, which comprises at least four members (Aruga et al., 1996).
Here we analyse the expression and activity of the opl gene. We show that opl expression demarcates the presumptive neural plate very early, by the onset of gastrulation, and later defines the dorsal neural tube and neural crest. opl can sensitize the neurectoderm for induction, suggesting that opl may regulate neural competence. opl is also able to potentiate expression of a midbrain marker, as well as dorsal neural and neural crest marker genes, delineating a role in neural tube patterning for this gene.
MATERIALS AND METHODS
Growth, dissection and culture of embryos and explants
Xenopus laevis eggs were collected, fertilized and cultured as described in Sive et al. (1989). Embryos were staged according to
Nieuwkoop and Faber (1967). For animal caps, late blastula (stage 9) animal hemisphere ectoderm was isolated and incubated in 0.5× MBS alone, or with other reagents as indicated. Dexamethasone (dex), and hydroxyurea and aphidicolin treatments were as described in Kolm and Sive (1995b) and Harris and Hartenstein (1991), respectively.
Subtractive cloning of opl
PCR-based subtractive cloning modified from Wang and Brown (1991) and described in detail by Patel and Sive (1996) was used to identify opl. RNA was isolated from dissected midgastrula (stage 11.5) dorsal ectoderm and blastula (stage 9-9.5) animal caps aged to midgastrula equivalent (stage 11.5) (Condie and Harland, 1987). cDNA was prepared by using the Pharmacia Timesaver cDNA synthesis kit. cDNA products were digested with AluI and RsaI, then ligated with oligo adaptors. Oligo 1 (5′-TAGTCCGAATTCAAGCAAGAGCACA-3′) was kinased, annealed to oligo 2 (5′-CTCTTGCTTGAATTCGGACTA-3′), and ligated to the dorsal ectoderm (D) cDNA, thereby introducing flanking EcoRI sites (underlined). Oligo 3 (5′-ATGCTGGATATCTTGGTACTCTTCA-3′) was kinased and annealed to oligo 4 (5′-GAGTACCAAGATATCCAGCAT-3′), and ligated to the animal cap (uninduced) ectoderm (U) cDNA, thereby introducing flanking EcoRV sites (underlined). Before each round of subtraction, tracer cDNA (from the D population) was PCR-amplified in the presence of 32P-dCTP, and driver cDNA (from the U population) was PCR-amplified in the presence of biotinylated-11-dUTP (Enzo Diagnostics). Subtractive hybridization was carried out at 68°C, with long hybridizations (40 hours) alternating with short hybridizations (2 hours) to remove rare common sequences and abundant common sequences, respectively. Biotinylated duplexes were removed by treatment with streptavidin and phenol extraction, and enriched tracer and driver cDNAs were amplified for the next round of subtraction. 10 rounds of subtraction were performed to produce D10 and U10 cDNAs. D10 cDNA products with average size of 300 bp (range 150 to 1000 bp) were cut with EcoRI and subcloned into CS2+ vector (Rupp et al., 1994). 800 clones were randomly selected and screened by dot blot analysis with probe prepared from D10 and U10 cDNAs to identify cDNAs expressed more strongly in the dorsal pool. 194 dorsal clones were sequenced and grouped into 75 sequence classes. 44 were of these were detected by in situ hybridization in gastrula stage dorsal ectoderm but not in ventral ectoderm. One of the most abundant clones isolated encoded opl (Fig. 1). A full-length opl cDNA clone was isolated from a midgastrula (stage 11.5) λZap phage library (prepared by C. Nocente-McGrath). Two cDNA clones were identified, the larger one (TG6.11) contained a 2.6 kb insert.
opl protein sequence alignments. Amino acid residues in the compared proteins that are identical to Xenopus opl are indicated by periods. Sequence gaps introduced for optimal protein alignment are indicated by dashes. Numbers at right indicate amino acid position. (A)Alignment of zinc fingers II-V of opl, mouse Zic proteins (Aruga et al., 1996): mZic1, mZic2, mZic3, mZic4, Drosophila Opa (Benedyk et al., 1994) and human Gli3(Ruppert et al., 1988). The first zinc finger is omitted since it is incomplete. This region of the Xenopus opl protein (aa 265-381) is 98% identical to the equivalent regions in mZic1, 97% identical to mZic2, 90% identical to mZic3, 85% identical to Opa and 53% identical to Gli3. (The GenBank accession number for opl is AF028805). (B) Alignment of conceptually translated Xenopus opl, mouse Zic1 and human Zic proteins. The predicted zinc fingers are indicated by a bar above the sequence. Block indicates start of polypeptide deleted in oplΔC constructs. Xenopus opl is 92% similar to both mouse Zic1 and human Zic proteins (Yokota et al., 1996). opl is 60% overall identical to both mouse and Xenopus Zic3 (Nakata et al., 1997).
opl protein sequence alignments. Amino acid residues in the compared proteins that are identical to Xenopus opl are indicated by periods. Sequence gaps introduced for optimal protein alignment are indicated by dashes. Numbers at right indicate amino acid position. (A)Alignment of zinc fingers II-V of opl, mouse Zic proteins (Aruga et al., 1996): mZic1, mZic2, mZic3, mZic4, Drosophila Opa (Benedyk et al., 1994) and human Gli3(Ruppert et al., 1988). The first zinc finger is omitted since it is incomplete. This region of the Xenopus opl protein (aa 265-381) is 98% identical to the equivalent regions in mZic1, 97% identical to mZic2, 90% identical to mZic3, 85% identical to Opa and 53% identical to Gli3. (The GenBank accession number for opl is AF028805). (B) Alignment of conceptually translated Xenopus opl, mouse Zic1 and human Zic proteins. The predicted zinc fingers are indicated by a bar above the sequence. Block indicates start of polypeptide deleted in oplΔC constructs. Xenopus opl is 92% similar to both mouse Zic1 and human Zic proteins (Yokota et al., 1996). opl is 60% overall identical to both mouse and Xenopus Zic3 (Nakata et al., 1997).
opl constructs
Constructs are diagrammed in Fig. 3. All techniques are as described in Sambrook et al. (1989), unless otherwise indicated. Enzymes were obtained from Stratagene or New England Biolabs. All constructs are in pCS2+ or pCS2+MT vectors (Rupp et al., 1994; Turner and Weintraub, 1994). The opl plasmid contained the opl open reading frame (ORF) fragment (699 bp) modified to include an optimal translation initiation site (Kozak, 1987) upstream of the opl initiation ATG and results in two extra amino acids (M-A) at the NH2-end of translated peptide. To create oplΔC, a PstI (end-filled)-XhoI (blunted) opl plasmid backbone was treated with Klenow fragment to remove 16 extra bp distal to the XhoI site (see Fig. 1) and then self-ligated. Myc-tagged opl (MTopl) was made by ligating an end-filled EcoRI-XbaI fragment into the XbaI (end-filled) site of CS2+MT. MToplΔC was made by ligating a NotI-NotI oplΔC fragment into the NotI-digested MTopl vector backbone. To make constructs encoding inducible opl proteins (oplGR), the ligand-binding domain of the human glucocorticoid receptor was PCR-amplified from the MyoDGR plasmid (Hollenberg et al., 1993) with the primers MP-GR1-XhoI (5′-GGCGCCGAGCTCGAGCCTCTGAAAATCCT-3′)and MP-GR2-XhoI (5′-GGCGGGCACTCGAGACTTTTGATGAAAC-3′).Products were then digested with XhoI (underlined) and inserted into an XhoI-digested opl plasmid.
Microinjection
Microinjection techniques were as described (Kolm and Sive, 1995a). 200 to 500 pg of opl, oplΔC or oplGR RNAs were injected per embryo. 0.1 to 10 pg noggin RNA (derived from the CS2+noggin plasmid) per embryo was co-injected in the opl plus noggin synergy experiments. 100 pg lacZ RNA was co-injected as tracer in whole-embryo assays
In vitro transcription
Capped sense RNAs for microinjection were synthesized as in Krieg and Melton (1984) and lacZ sense RNA as in Gammill and Sive (1997). All opl-derived sense RNAs (opl, oplΔC, oplGR, MTopl, MToplΔC) were made by SP6 transcription of a NarI-NarI fragment from CS2+derived vectors. Other templates for sense RNAs are: CAT (p64TCAT; Kuo et al., 1996), SmaI linearized, SP6 transcribed; globin (pXβM), PstI linearized, SP6 transcribed; noggin (pCS2+xnoggin), NotI linearized, SP6 transcribed. Antisense probes for in situ hybridization were transcribed in the presence of digoxigenin-11-UTP or fluorescein-11-UTP as follows: XCG (Sive et al., 1989), otx2 (Pannese et al., 1995), en2 (Hemmati-Brivanlou and Harland, 1989), krox20 (Bradley et al., 1993), slug (Mayor et al., 1995), pax3 (Espeseth et al., 1995) and XK81 (Jonas et al., 1985).
In situ hybridization and sectioning
Whole-mount in situ hybridization was performed on albino embryos essentially as described in Harland (1991) with modifications as in Sagerstrom et al. (1996). β-galactosidase staining was performed as in Kolm and Sive (1995a). For sections, embryos were embedded in JB4 resin (Polysciences 00226) and 5 to 10 μm sections were mounted in Crystal mount (Biomeda M03). Sections were stained with toluidine blue.
Isolation of RNA and northern analysis
RNA was prepared from explants and embryos by proteinase-K treatment and phenol extraction (Condie and Harland, 1987). RNA was analyzed by northern blotting (Sambrook et al., 1989). opl antisense probe was made by asymmetric PCR (Sive and Cheng, 1991) with the opl-D2 primer (5′-GTTGGCCAACTGCTCGG-3′).
Relative quantitative RT-PCR
RNA from pools of 15-30 explants or 2-4 embryos was analysed by RT-PCR as described in Sagerström et al. (1996), using RNA from 5 animal cap equivalents or 0.25 embryo equivalents per reaction. NCAM, en2, krox20 primers are described in Hemmati-Brivanlou and Melton (1994a). EF1α and XCG primers are described in Gammill and Sive (1997). Other primers yielding product sizes as indicated are: otx2 – 485 bp [Xotx2.C (5′-GCAACAGCAGCAGCAGAATG-3′),Xotx2.D(5′-TGTAATCCAGGCAGTCAGTG-3′)] (Pannese et al., 1995); HoxD1 – 245 bp [PJK3 (5′-CAGCCCCGATTAC-GATTATTATGG-3′), PJK4 (5′-CCGGGGAGGCAGGTTTTG-3′)(Kolm and Sive, 1995a); HoxB9 – 201 bp [HLS51 (5′-GCCCCTGCGCAATCTGAAC-3′),HLS52(5′-CAGCAGCGGC-TCAGACTTGAG-3′)] (Wright et al., 1990); slug – 220 bp [(Xslug1 (5′-GGTTCGATTTGCAGTATGG-3′), Xslug2 (5′-GAGCTGCT-TCGTAAAGCAC-3′)] (Mayor et al., 1995); pax3 – 181 bp [pax3A (5′-CAGCCGAATTTTGAGGAGCACAT-3′), pax3B (5′-GGGCAG-GTCTGGTTCGGAGTC-3′)] (Espeseth et al., 1995); NK2 – 406 bp [NK2.1 (5′-AGAAACCGTCGGCTGATGAGTC-3′), NK2.2 (5′-TGGGCGCTATAGGCAGAGAAAG-3′)] (Saha et al., 1993); shh –191 bp [shh1 (5′-GGTTGACCGCGGCCCATCTAC-3′), shh2 (5′-AGGCGCATAAGCTCCAGTGTCC-3′)] (Ekker et al., 1995); Xash3 –227 bp [forward (5′-AAGCGGCTCCTGTCAAACACTAT-3′), reverse (5′-CCCCCGCGCTCCTCTCC-3′)] (Zimmerman et al., 1993); neurogenin – 263 bp [forward (5′-CGCCGCA-ACCCGACTCACCT-3′),reverse (5′-CCTGCATCGCG-GGCTGTTCTC-3′)](Ma et al., 1996); XK81 – 308 bp [forward (5′-TCATTCCGTTCCAGCTCTTCTTAC-3′), reverse (5′-TCCAG-GGCTCTTACTTTCTCCAG-3′)] (Jonas et al., 1985); GATA2 – 415 bp [forward (5′-CTAAACAGAGGAGCAAGAGC-3′),reverse(5′-CTGGAAAGTTCCTCAAAAC-3′)] (Walmsley et al., 1994); m-actin – 222 bp [forward (5′-GCTGACAGAATGCAGAAG-3′), reverse (5′-TTGCTTGGAGGAGTGTGT-3′)] (Stutz and Spohr, 1986).
Western analysis and immunocytochemistry
Western blotting (Kolm and Sive, 1995b) and whole-mount immunostaining (Hemmati-Brivanlou and Harland, 1989) were performed using the c-myc 9E10 antibody (Santa Cruz Biotechnology) diluted 1:1000 (0.1 μg/ml) and the ECL Vector Elite kit from Amersham.
Transient transfection reporter assays
A reporter plasmid (pGL15) containing three Gli3 DNA-binding sites (Vortkamp et al., 1995) and a luciferase reporter was mixed with CS2+ or CS2+opl ORF or CS2+oplΔC plasmids for transfections at the ratio of 5 μg:1 μg. The human pancreatic tumor cell line BXPC (ATCC #CRL 1687) were transfected using the DEAE-dextran/chloroquine method (Aruffo and Seed, 1987), and cell lysates were assayed 48 hours after transfection for luciferase activity using the Promega Luciferase Assay System (E1501) on a Analytical Luminescence Laboratory Monolight 2010 machine. Duplicate transfections were done for each DNA mix and duplicate readings were taken for each lysate sample.
RESULTS
opl encodes a zinc finger protein with similarity to the Zic family
In order to isolate genes activated soon after neural determination has begun in Xenopus, we subtracted cDNAs from microdissected midgastrula dorsal ectoderm with those from uninduced ectoderm aged until midgastrula (Patel and Sive, 1996). One of the genes that we isolated encodes a 443 amino acid protein with five zinc fingers of the C2H2 class (GenBank accession no. AF028805). By comparison with invertebrate zinc finger genes, the zinc fingers of this gene were most similar (85%) to those of the Drosophila odd-paired gene (opa), a pair-rule gene (Benedyk et al., 1994), and we therefore named this gene opl for odd-paired-like. Among the vertebrate zinc finger genes, the zinc fingers of opl were similar to those of the Gli family genes (53% identity to the zinc fingers of human Gli3 (Ruppert et al., 1988)), but even more similar to the family of mouse Zic genes. The Zic genes share intron/exon boundaries with Drosophila odd-paired (Aruga et al., 1996) indicating that these are true homologs of opa (Fig. 1A). Over the entire coding region, opl was very similar to mouse Zic1, with 91% overall amino acid identity (Fig. 1B, (Aruga et al., 1994)). It was much less similar to the proteins encoded by other members of the Zic gene family, with 63% identity to mouse Zic2, 60% identity to Zic3 and 54% identity to Zic4 (Aruga et al., 1996). Recent reports describing isolation of Xenopus Zic members include Zic3 (Nakata et al., 1997), which has 60% identity to opl, while the percentage identity of opl and Zic-r1 (Mizuseki et al., 1998) is not clear. On the basis of these data, opl may be the Xenopus homolog of Zic1. However, since we cannot conclude this unequivocally and in order to reflect the similarity of this gene to Drosophila opa, we chose the name opl.
Temporal and spatial characterization of opl expression
We next examined the time course of opl expression by northern analysis (Fig. 2). opl RNA is activated zygotically by late blastula (stage 9.5, lane 3). Expression was maximal by midgastrula (lanes 4-7), with lower expression persisting into tailbud stages (lanes 8-11).
(A) Northern analysis of Xenopus embryos. One embryo equivalent per lane was analyzed for opl RNA (top row) at various embryonic stages shown. Ethidium-stained 28S rRNA is a loading control (bottom row).
Whole-mount in situ analysis of opl expression. Embryos were analyzed by in situ hybridization as described in Methods. Embryo orientations are indicated by A, anterior; P, posterior; D, dorsal and V, ventral; arrowhead, dorsal lip of blastopore.
By whole-mount in situ hybridization, opl expression was biphasic. By early gastrula (Fig. 3), opl was expressed strongly in the dorsal ectoderm (Fig. 3A,B), with much lower levels in the ventral ectoderm, and trace levels of expression in the marginal zone. Strong expression throughout the presumptive neural plate was observed at midgastrula, with increased opl expression at the lateral edges of the neural plate (Fig. 3C,D). By early neurula, opl expression had entered a second phase, with restriction to the lateral edges of the neural plate, that is the presumptive dorsal neural tube and neural crest (Fig. 3E,F). As the neural tube closed, expression was predominant in the anterior dorsal neural tube and the premigratory presumptive cranial neural crest (Fig. 3G,H). The precise region of presumptive neural crest expressing opl changes during neurulation. At early neurula (stage 15), opl expression was in the most anterior mandibular crest (not shown) while, by stage 18 (Fig. 3G,H), opl was expressed predominantly in the more posterior hyoid and branchial neural crest (Sadaghiani and Thiebaud, 1987). By tailbud, expression was high in two stripes located in the telencephalon and diencephalon of the forebrain, and more posteriorly along the entire length of the neural tube (Fig. 3I,J).
In order to analyse opl expression in more detail, embryos stained for opl RNA were sectioned sagittally. At stage 10+ (the onset of gastrulation), opl expression was observed in both the inner and outer layers of the dorsal ectoderm (Fig. 4A,B) in a region largely without underlying mesendoderm (bracket). By midgastrula (stage 11) (Fig. 4C,D) and late gastrula (Fig. 4E), opl expression was excluded from the extreme posterior neurectoderm. Sequential transverse sections of a tailbud embryo showed that, within the spinal cord, opl expression was restricted to the roof plate of the neural tube (Fig. 4F, black arrow) and the migratory neural crest (Fig. 4F, white arrow). In the hindbrain, opl expression extended into the more ventral alar plate region of the neural tube (Fig. 4G,H). At the midbrain level, opl expression extended through the dorsal half of the neural tube (Fig. 4I), while more anteriorly, in the forebrain, opl expression was lost from roof plate and retained in the alar plate (Fig. 4J).
Sections of embryos analyzed by whole-mount in situ hybridization for opl expression. Embryos were fixed, embedded and sectioned as described in Methods. Sectioning planes are shown in schematic diagrams above panels. Black arrowheads indicate position of dorsal blastopore lip. Black arrows indicate opl expression in roofplate. White arrows indicate spinal crest expression. Brackets indicate alar plate expression in forebrain/midbrain regions. D, dorsal; V, ventral. Scale bar, 25 μm.
Sections of embryos analyzed by whole-mount in situ hybridization for opl expression. Embryos were fixed, embedded and sectioned as described in Methods. Sectioning planes are shown in schematic diagrams above panels. Black arrowheads indicate position of dorsal blastopore lip. Black arrows indicate opl expression in roofplate. White arrows indicate spinal crest expression. Brackets indicate alar plate expression in forebrain/midbrain regions. D, dorsal; V, ventral. Scale bar, 25 μm.
In order to delineate the expression of opl relative to other positions in the neural tube, we performed comparative in situ hybridization with other dorsal ectodermal markers (Fig. 5). At midneurula, expression of the forebrain and cement gland marker otx2 (Blitz and Cho, 1995; Pannese et al., 1995) overlapped with that of opl anteriorly, extending to the anterior boundary of the neural plate, but opl is excluded from the dorsal medial region and the future cement gland (bracket) (Fig. 5A). Consistently, opl was expressed more posteriorly than XCG, a marker of the cement gland (Sive et al., 1989) that lies anterior to the neural plate (Fig. 5B). Expression of slug (Mayor et al., 1995) in the premigratory cranial neural crest overlapped that of opl, confirming that opl expression extended beyond the edges of the neural plate (Fig. 5C). Relative to engrailed 2 (en2) at the midbrain/hindbrain junction (Hemmati-Brivanlou and Harland, 1989) and krox20, demarcating future rhombomeres 3 and 5 of the hindbrain (Bradley et al., 1993), opl expression at early neurula extends posterior to the rhombomere 5 stripe of krox20 (Fig. 5D).
Double whole-mount in situ analysis of opl. In situ analysis was as described in Methods with opl stained light blue and all other markers (otx-2, XCG, slug, en-2, krox20) are stained purple. (A) opl plus otx-2, a marker of forebrain and cement gland (Blitz and Cho, 1995; Pannese et al., 1995), bracket shows otx-2 expression in cement gland primordium; (B) opl plus XCG, a cement gland marker (Sive et al., 1989), bracket shows XCG expression in cement gland; (C)opl plus slug, a neural crest marker (Mayor et al., 1995), bracket shows slug domain contained within opl domain; (D) opl plus en-2, a marker of the midbrain/hindbrain regions (Hemmati-Brivanlou and Harland, 1989) and krox20, a marker of presumptive rhombomeres 3 and 5 in the hindbrain (Bradley et al., 1993). White arrowhead indicates en-2 expression, black arrowheads indicate krox20 expression, white bar is posterior limit of early opl expression.
Double whole-mount in situ analysis of opl. In situ analysis was as described in Methods with opl stained light blue and all other markers (otx-2, XCG, slug, en-2, krox20) are stained purple. (A) opl plus otx-2, a marker of forebrain and cement gland (Blitz and Cho, 1995; Pannese et al., 1995), bracket shows otx-2 expression in cement gland primordium; (B) opl plus XCG, a cement gland marker (Sive et al., 1989), bracket shows XCG expression in cement gland; (C)opl plus slug, a neural crest marker (Mayor et al., 1995), bracket shows slug domain contained within opl domain; (D) opl plus en-2, a marker of the midbrain/hindbrain regions (Hemmati-Brivanlou and Harland, 1989) and krox20, a marker of presumptive rhombomeres 3 and 5 in the hindbrain (Bradley et al., 1993). White arrowhead indicates en-2 expression, black arrowheads indicate krox20 expression, white bar is posterior limit of early opl expression.
In summary, opl expression is an early marker of the presumptive neurectoderm with later expression marking the dorsal neural tube and neural crest. Recent analyses of the Xenopus Zic3 (Nakata et al., 1997) and a Zic1-related gene (Zic-r1) (Mizuseki et al., 1998) indicate that the expression patterns of these genes and opl are similar.
opl is a nuclear protein whose carboxy terminal encodes a regulatory domain
We next characterized the activity and localization of the opl protein. Since mouse Zic1 is a nuclear protein that can bind the DNA target site of the related zinc finger protein Gli3 (Fig. 1A; Aruga et al., 1994), we assumed that opl was a transcription factor and asked whether it had activator or repressor functions. The carboxy terminus of opl is highly serine-rich, suggesting it may be a phosphorylation domain (see Fig. 1B). We removed this region (to create oplΔC) and asked whether it played a regulatory role in opl activity, using a reporter assay. The various constructs used are diagrammed in Fig. 6 and described in Methods. The reporter consisted of luciferase linked to three copies of the Gli3 consensus DNA-binding site (Fig. 7A; (Vortkamp et al., 1995). The activity of intact opl or oplΔC was assayed in transient transfection assays (see Methods). The averaged results of three experiments are shown in Fig. 7A. While opl was able to activate reporter gene expression 2.8-fold above background, oplΔC was able to activate reporter gene expression 14.6-fold above background, indicating that this mutant allele of opl has an enhanced transcriptional activation capacity and designating the carboxy terminus of opl as a domain that is able to suppress opl activity. This transactivation requires Gli3-binding sites, since a reporter lacking these sites (containing an SV40 promoter) was not activated above background levels by either opl or oplΔC (not shown).
opl constructs. The opl construct comprises the coding region of opl; oplΔC deletes a serine-rich carboxy terminal domain (S), oplGR comprises an in-frame fusion of the opl-coding sequence with the hormone-binding domain of the glucocorticoid receptor (Kolm and Sive, 1995b). MT-opl is an in-frame fusion of the opl-coding region with a multimerized myc epitope (MT, black circle) at the amino end of opl (Rupp et al., 1994) and MT-oplΔC is a similar fusion of the multimerized myc epitope with the carboxy deleted opl. See Methods for details.
opl constructs. The opl construct comprises the coding region of opl; oplΔC deletes a serine-rich carboxy terminal domain (S), oplGR comprises an in-frame fusion of the opl-coding sequence with the hormone-binding domain of the glucocorticoid receptor (Kolm and Sive, 1995b). MT-opl is an in-frame fusion of the opl-coding region with a multimerized myc epitope (MT, black circle) at the amino end of opl (Rupp et al., 1994) and MT-oplΔC is a similar fusion of the multimerized myc epitope with the carboxy deleted opl. See Methods for details.
opl is a nuclear protein with a regulatory domain in the COOH-terminal. (A) opl protein transactivates reporter constructs in tissue culture. Transient transfection reporter assays were performed as described in Methods, using the human pancreatic tumor BXPC cell line. The reporter construct is shown schematically: three Gli3 DNA-binding sites are located 5′ to a basal promoter (TATA box) and luciferase reporter gene (Vortkamp et al., 1995) Data shown are the averages of three independent experiments with corresponding error bars. In the absence of the Gli3-binding sites, neither opl nor oplΔC activated a reporter construct (not shown). DNA mixes for transient transfections are as follows, Left panel, reporter plus CS2+ vector; middle panel, reporter plus CS2+opl ORF right panel, reporter plus pCS2-oplΔC. (B) Immunocytochemistry of injected opl and oplΔC proteins. Albino embryos were injected with 500 pg of synthetic MTopl or MToplΔC RNA at 2- to 4-cell stages and analyzed at stage 11 by immunostaining with anti-myc 9E10 antibody as described in Methods. Views of the animal hemispheres are shown, left panel, MTopl-injected embryo, right panel, MToplΔC-injected embryo.
opl is a nuclear protein with a regulatory domain in the COOH-terminal. (A) opl protein transactivates reporter constructs in tissue culture. Transient transfection reporter assays were performed as described in Methods, using the human pancreatic tumor BXPC cell line. The reporter construct is shown schematically: three Gli3 DNA-binding sites are located 5′ to a basal promoter (TATA box) and luciferase reporter gene (Vortkamp et al., 1995) Data shown are the averages of three independent experiments with corresponding error bars. In the absence of the Gli3-binding sites, neither opl nor oplΔC activated a reporter construct (not shown). DNA mixes for transient transfections are as follows, Left panel, reporter plus CS2+ vector; middle panel, reporter plus CS2+opl ORF right panel, reporter plus pCS2-oplΔC. (B) Immunocytochemistry of injected opl and oplΔC proteins. Albino embryos were injected with 500 pg of synthetic MTopl or MToplΔC RNA at 2- to 4-cell stages and analyzed at stage 11 by immunostaining with anti-myc 9E10 antibody as described in Methods. Views of the animal hemispheres are shown, left panel, MTopl-injected embryo, right panel, MToplΔC-injected embryo.
We also asked whether the opl protein was localized to the nucleus and whether C-terminal removal altered its nucleocytoplasmic distribution, using myc-epitope tagged versions of both opl and oplΔC, MT-opl and MT-oplΔC (see Methods). One cell of a 2-cell embryo was injected with RNA encoding either of these constructs and examined at midgastrula for distribution of the epitope-tagged opl using whole-mount immunocytochemistry. As shown in Fig. 7B, both the intact (MT-opl) and carboxy terminal truncated (MT-oplΔC) proteins were nuclear, with no obvious differences in the distribution of either construct. We were also unable to detect any differences in the amount of opl protein produced in embryos from either MT-opl or MT-oplΔC (not shown).
In summary, this analysis suggested that opl is a transcriptional activator and defined a putative regulatory domain in the carboxy region of the opl protein. This region repressed the ability of opl to act as an activator but had no effect on the nucleocytoplasmic distribution of opl.
oplΔC activates neural crest and dorsal neural tube markers
The neurectodermal restriction of opl expression suggested that it may play a role in neural determination. We tested this possibility in an animal cap assay. RNA encoding opl, oplΔC or CAT (as control) was injected into the animal pole region of two cell embryos. At late blastula (stage 9), animal caps were removed, cultured until tailbud (stage 22) and marker gene expression was analysed using a reverse transcriptase-polymerase chain reaction (RT-PCR) assay (Fig. 8A and Methods).
oplΔC can activate neural crest and dorsal neural tube marker genes in animal caps. (A) Experimental scheme. Wild-type embryos were injected in both blastomeres with 200 pg of indicated RNAs at the 2-cell stage. Injection of CAT RNA served as negative control by equalizing the injected RNA dose. Pools of 15-20 animal caps were isolated from injected embryos at late blastula (stage 9) and incubated in saline until harvest at tailbud (stage 22) for RT-PCR analysis. See Methods for details. (B) Expression of marker genes in animal caps. Uninjected embryos serve as control for baseline expression level. Xash3 (Zimmerman et al., 1993), neurogenin (Ma et al., 1996)and NCAM (Kintner and Melton, 1987) are neural-specific markers; slug (Mayor et al., 1995) is a neural crest marker; pax3 (Espeseth et al., 1995) marks the dorsal neural tube, shh marks the floorplate (Ekker et al., 1995); XK81 (Jonas et al., 1985) and GATA2 (Walmsley et al., 1994) are markers of the ventral ectoderm; EF1α (Krieg et al., 1989) served as loading control. Samples processed without RT did not show EF1α signal after PCR (not shown). Data from one representative experiment is shown; similar results were obtained from three experiments. Lane 1, stage 22 embryo; lane 2, CAT-injected animal caps; lane 3, oplΔC-injected caps; lane 4, opl-injected caps.
oplΔC can activate neural crest and dorsal neural tube marker genes in animal caps. (A) Experimental scheme. Wild-type embryos were injected in both blastomeres with 200 pg of indicated RNAs at the 2-cell stage. Injection of CAT RNA served as negative control by equalizing the injected RNA dose. Pools of 15-20 animal caps were isolated from injected embryos at late blastula (stage 9) and incubated in saline until harvest at tailbud (stage 22) for RT-PCR analysis. See Methods for details. (B) Expression of marker genes in animal caps. Uninjected embryos serve as control for baseline expression level. Xash3 (Zimmerman et al., 1993), neurogenin (Ma et al., 1996)and NCAM (Kintner and Melton, 1987) are neural-specific markers; slug (Mayor et al., 1995) is a neural crest marker; pax3 (Espeseth et al., 1995) marks the dorsal neural tube, shh marks the floorplate (Ekker et al., 1995); XK81 (Jonas et al., 1985) and GATA2 (Walmsley et al., 1994) are markers of the ventral ectoderm; EF1α (Krieg et al., 1989) served as loading control. Samples processed without RT did not show EF1α signal after PCR (not shown). Data from one representative experiment is shown; similar results were obtained from three experiments. Lane 1, stage 22 embryo; lane 2, CAT-injected animal caps; lane 3, oplΔC-injected caps; lane 4, opl-injected caps.
One of three representative experiments is shown in Fig. 8B. After injection of CAT RNA, we observed only background levels of the neural-specific markers Xash3 (Zimmerman et al., 1993), neurogenin (Ma et al., 1996), NCAM (Kintner and Melton, 1987), the neural crest marker slug (Mayor et al., 1995), the dorsal neural tube marker pax3 (Espeseth et al., 1995) and the ventral neural tube marker Sonic hedgehog (shh) (Ekker et al., 1995) (Fig. 5B, lane 2). In contrast, high levels of the ventral ectodermal markers XK81 (Jonas et al., 1985) and GATA2 (Walmsley et al., 1994) were observed after this treatment (lane 2). Injection of opl or oplΔC RNAs did not induce expression of NCAM, Xash3 or neurogenin (lanes 3 and 4), however oplΔC (lane 3), but not opl (lane 4) was able to activate expression of slug (4-fold over CAT-injected levels, averaged over three experiments) and pax3 (3.5-fold over CAT-injected levels). shh was not activated by injection of either construct. We also asked whether opl or oplΔC could suppress non-neural ectodermal differentiation, by assaying the expression of XK81 and GATA2 as ventral markers. Both opl and oplΔC were able to suppress expression of XK81 by 2-fold relative to CAT-injected samples. Additionally, while oplΔC was unable to suppress GATA2 expression, opl consistently suppressed expression of GATA2 by 2-fold.
In summary, these data showed that neither opl nor its activated form, oplΔC, could activate expression of neural-specific genes in uninduced ectoderm. However, consistent with the later expression pattern of opl in the dorsal neural tube, oplΔC efficiently induced expression of neural crest and dorsal neural tube markers. These data contrast with those obtained after injection of Zic3 and Zic-r1 RNAs, where multiple neural-specific markers as well as slug were activated (Nakata et al., 1997; Mizuseki et al., 1998). It is not clear whether this represents a difference in the experimental conditions used, or a difference in the activities of opl and these related genes.
oplΔC sensitizes the ectoderm to induction by noggin
The early expression domain of opl in the dorsal ectoderm correlated with the region of the gastrula ectoderm that is most competent to respond to neural induction. We therefore asked whether opl could act as a modulator of neural competence and sensitize ectoderm to the neural-inducing activity of noggin, a factor that acts by antagonizing BMP signaling (Lamb et al., 1993), using the animal cap assay diagrammed in Fig. 8A. Embryos were injected with varying amounts of noggin RNA alone or together with oplΔC RNA or with globin RNA as a control and caps later analysed by RT-PCR for expression of two neural-specific genes, NCAM and Xash3.
A representative experiment (of three) is shown in Fig. 9. Neither globin RNA (lane 1) nor oplΔC alone (lane 2) induced NCAM or Xash3 RNA expression, consistent with results shown in Fig. 8. 0.1 pg noggin RNA did not induce either NCAM or Xash3 (lane 3) and 1 pg noggin RNA did not induce NCAM and activated Xash3 only weakly (lane 4). 10 pg noggin RNA strongly activated both NCAM and Xash3 expression (lane 5). At 0.1 pg noggin RNA plus oplΔC RNA, no NCAM and very weak Xash3 expression was observed (lane 6); however, at 1 pg noggin RNA plus oplΔC, strong expression of both NCAM and Xash3 was observed (lane 7), to levels that were almost as strong as seen with 10 pg of noggin alone. At higher noggin concentration (10 pg) plus oplΔC, no further potentiation of NCAM and Xash3 was observed (lane 8) and, in the experiment shown, Xash3 RNA accumulation declined relative to 1 pg noggin plus oplΔC, perhaps indicating that saturation of the induction machinery had occurred.
oplΔC sensitizes the ectoderm to induction by noggin. Wild-type embryos were injected in both blastomeres with indicated RNAs at the two-cell stage, using 200 pg globin or oplΔC, and 0.1 pg. 1 pg or 10 pg noggin (Smith and Harland, 1992) RNAs, as diagrammed in Fig. 8A, except injection of globin RNA served as negative control by equalizing the injected RNA dose in all lanes. Pools of 15-20 animal caps were isolated from injected embryos at late blastula (stage 9) and incubated in saline until harvest at tailbud (stage 22) for RT-PCR analysis. See Methods for details. Expression of neural markers NCAM (Kintner and Melton, 1987) and Xash3 (Zimmerman et al., 1993) was analysed. EF1α served as loading control (Krieg et al., 1989). Samples processed without RT did not show EF1α signal after PCR (not shown). Data from one representative experiment is shown, similar results were obtained from three experiments. Lane 1, globin-injected caps; lane 2, caps injected with 200 pg oplΔC RNA; lane 3, caps injected with 0.1 pg noggin RNA; lane 4, caps injected with 1 pg noggin RNA; lane 5, caps injected with 10 pg noggin RNA; lane 6, caps injected with 0.1 pg noggin RNA plus 200 pg oplΔC RNA; lane 7, caps injected with 1 pg noggin RNA plus 200 pg oplΔC RNA; lane 8, caps injected with 10 pg noggin RNA plus 200 pg oplΔC RNA.
oplΔC sensitizes the ectoderm to induction by noggin. Wild-type embryos were injected in both blastomeres with indicated RNAs at the two-cell stage, using 200 pg globin or oplΔC, and 0.1 pg. 1 pg or 10 pg noggin (Smith and Harland, 1992) RNAs, as diagrammed in Fig. 8A, except injection of globin RNA served as negative control by equalizing the injected RNA dose in all lanes. Pools of 15-20 animal caps were isolated from injected embryos at late blastula (stage 9) and incubated in saline until harvest at tailbud (stage 22) for RT-PCR analysis. See Methods for details. Expression of neural markers NCAM (Kintner and Melton, 1987) and Xash3 (Zimmerman et al., 1993) was analysed. EF1α served as loading control (Krieg et al., 1989). Samples processed without RT did not show EF1α signal after PCR (not shown). Data from one representative experiment is shown, similar results were obtained from three experiments. Lane 1, globin-injected caps; lane 2, caps injected with 200 pg oplΔC RNA; lane 3, caps injected with 0.1 pg noggin RNA; lane 4, caps injected with 1 pg noggin RNA; lane 5, caps injected with 10 pg noggin RNA; lane 6, caps injected with 0.1 pg noggin RNA plus 200 pg oplΔC RNA; lane 7, caps injected with 1 pg noggin RNA plus 200 pg oplΔC RNA; lane 8, caps injected with 10 pg noggin RNA plus 200 pg oplΔC RNA.
In summary, these data showed that oplΔC could sensitize isolated ectoderm to the neural-inducing effects of noggin, indicating that opl can regulate neural induction, and consistent with a role for opl in mediating neural competence.
In conjunction with noggin, opl activates posterior and dorsoventral neural markers
Having shown that oplΔC could sensitize the ectoderm for responsiveness to noggin, we next asked whether opl or oplΔC could alter the spectrum of neural markers induced by high levels of noggin that were characteristic of particular anteroposterior or dorsoventral axial positions, using the animal cap assay diagrammed in Fig. 8A with CAT RNA as control. Expression levels of the general neural marker NCAM and marker genes differentially expressed along the anteroposterior (A/P) or dorsoventral (D/V) axes of the neural tube were assayed by RT-PCR.
One of four representative experiments is shown in Fig. 10. Uninjected caps (or caps injected with CAT RNAs (not shown)) failed to express all dorsal ectodermal and mesodermal markers examined (lane 2). Injection of noggin plus CAT RNAs (lane 3) led to activation of NCAM, the cement gland marker XCG (Sive et al., 1989) and the forebrain plus cement gland marker otx2 (Blitz and Cho, 1995; Pannese et al., 1995), but activated only low levels of markers that were more posterior. Injection of oplΔC (lane 3) or intact opl (not shown) RNAs alone led to weak activation of the hindbrain markers krox20 (Bradley et al., 1993), HoxD1 (Kolm and Sive, 1995a) and less reproducibly, the spinal cord marker HoxB9 (Wright et al., 1990), but failed to activate either NCAM or the anterior markers XCG, otx2 and en2 (Hemmati-Brivanlou and Harland, 1989). However, expression of oplΔC plus noggin (lane 5) or opl plus noggin (lane 6), reproducibly led to very strong activation of en2, expressed at the midbrain/hindbrain junction (25-fold for oplΔC, and 7-fold for opl, averaged over four experiments). oplΔC (but not opl) was also able to activate expression of krox20 and HoxD1, (lane 5), with the extent of activation varying between experiments, while HoxB9 expression was not reproducibly activated.
opl synergizes with noggin to activate more posterior neural markers in animal caps. Wild-type embryos were injected in both blastomeres with indicated RNAs at the 2-cell stage, using 200 pg CAT, opl or oplΔC, and 10 pg noggin (Smith and Harland, 1992) RNAs, as diagrammed in Fig. 8A. Injection of CAT RNA served as negative control by equalizing the injected RNA dose. Pools of 15-20 animal caps were isolated from injected embryos at late blastula (stage 9) and incubated in saline until harvest at tailbud (stage 22) for RT-PCR analysis. See Methods for details. Expression of a set of neural markers expressed along the anteroposterior (A/P) or dorsoventral (D/V) neural axes was analysed. Uninjected embryos serve as control for baseline expression level. NCAM is a general neural marker (Kintner and Melton, 1987). Expressed in the indicated anterior (A)-to-posterior (P) series: XCG is a cement gland marker, the most anterior ectodermal tissue (Sive et al., 1989); otx-2 is a forebrain and cement gland marker (Blitz and Cho, 1995; Pannese et al., 1995); en2 is a marker of mid/hindbrain boundary (Hemmati-Brivanlou and Harland, 1989); krox20 marks rhombomeres 3 and 5 in the presumptive hindbrain (Bradley et al., 1993); HoxD1 is a hindbrain (posterior to rhombomere 4) and spinal cord marker (Kolm and Sive, 1995a), and HoxB9 is a spinal cord marker (Wright et al., 1990). Expressed in the indicated dorsal (D) to ventral (V) series of markers, slug (marking the neural crest, (Mayor et al., 1995)), pax3 (marking the dorsal neural tube, (Espeseth et al., 1995)), NK2 (marking the ventrolateral neural tube, (Saha et al., 1993)) and shh (marking the floorplate, (Ekker et al., 1995)). M-actin (Mohun et al., 1984) is a mesodermal marker; EF1α (Krieg et al., 1989) served as loading control. Samples processed without RT did not show EF1α signal after PCR (not shown). Data from one experiment are shown, comparable results were obtained in four independent experiments. Lane 1, stage 22 embryo; lane 2, uninjected animal caps; lane 3, noggin plus CAT-injected; lane 4, oplΔC-injected; lane 5, noggin plus oplΔC-injected; lane 6, noggin plus opl-injected.
opl synergizes with noggin to activate more posterior neural markers in animal caps. Wild-type embryos were injected in both blastomeres with indicated RNAs at the 2-cell stage, using 200 pg CAT, opl or oplΔC, and 10 pg noggin (Smith and Harland, 1992) RNAs, as diagrammed in Fig. 8A. Injection of CAT RNA served as negative control by equalizing the injected RNA dose. Pools of 15-20 animal caps were isolated from injected embryos at late blastula (stage 9) and incubated in saline until harvest at tailbud (stage 22) for RT-PCR analysis. See Methods for details. Expression of a set of neural markers expressed along the anteroposterior (A/P) or dorsoventral (D/V) neural axes was analysed. Uninjected embryos serve as control for baseline expression level. NCAM is a general neural marker (Kintner and Melton, 1987). Expressed in the indicated anterior (A)-to-posterior (P) series: XCG is a cement gland marker, the most anterior ectodermal tissue (Sive et al., 1989); otx-2 is a forebrain and cement gland marker (Blitz and Cho, 1995; Pannese et al., 1995); en2 is a marker of mid/hindbrain boundary (Hemmati-Brivanlou and Harland, 1989); krox20 marks rhombomeres 3 and 5 in the presumptive hindbrain (Bradley et al., 1993); HoxD1 is a hindbrain (posterior to rhombomere 4) and spinal cord marker (Kolm and Sive, 1995a), and HoxB9 is a spinal cord marker (Wright et al., 1990). Expressed in the indicated dorsal (D) to ventral (V) series of markers, slug (marking the neural crest, (Mayor et al., 1995)), pax3 (marking the dorsal neural tube, (Espeseth et al., 1995)), NK2 (marking the ventrolateral neural tube, (Saha et al., 1993)) and shh (marking the floorplate, (Ekker et al., 1995)). M-actin (Mohun et al., 1984) is a mesodermal marker; EF1α (Krieg et al., 1989) served as loading control. Samples processed without RT did not show EF1α signal after PCR (not shown). Data from one experiment are shown, comparable results were obtained in four independent experiments. Lane 1, stage 22 embryo; lane 2, uninjected animal caps; lane 3, noggin plus CAT-injected; lane 4, oplΔC-injected; lane 5, noggin plus oplΔC-injected; lane 6, noggin plus opl-injected.
We also examined synergy of opl or oplΔC with noggin with respect to markers expressed along the D/V aspect of the neural tube. Injection of noggin RNA together with control CAT RNA (lane 3) induced low levels of the dorsal marker pax3 (Espesethet al., 1995), the ventrolateral marker NK2 (Saha et al., 1993) and the ventral marker shh. Consistent with the data shown in Fig. 8, oplΔC alone (lane 4) was very effective at activating expression of the neural crest markers slug, twist (Hopwood et al., 1989) (not shown) and pax3. In conjunction with noggin, oplΔC (lane 5) also activated expression of genes normally restricted to the ventrolateral and ventral neural tube including NK2, shh and F-spondin (Ruiz i Altaba et al., 1995) (not shown). Intact opl protein in conjunction with noggin activated high levels of pax3, NK2 and shh, but did not efficiently activate slug or twist (not shown) expression (lane 6). Expression of mesodermal genes, including muscle actin, was not activated by any of these injections.
Overall, this assay demonstrated that opl and oplΔC were able to alter the A/P and D/V spectrum of genes activated by noggin, suggesting that opl may act as a modulator of positional information along the A/P and D/V neural axes. As in the assay for transcriptional activation, oplΔC was more active than native opl protein.
opl induces cellular aggregates in the absence of DNA synthesis
Since opl and oplΔC were able to alter neural marker gene expression in animal cap assays, we wanted to determine whether ectopic opl expression could alter neurogenesis in the intact embryo. To do this, we compared the effects of native opl protein, oplΔC, and a hormone inducible version of opl, oplGR, on the morphology of whole embryos. oplGR comprises the opl-coding region linked in frame to the ligand-binding domain of the glucocorticoid receptor (GR). We have previously shown that such constructs are post-translationally hormone inducible in Xenopus embryos, and can be activated at the time when the endogenous gene is maximally expressed (Kolm and Sive, 1995b). Embryos were injected with RNA encoding either oplGR, along with lacZ RNA as a lineage tracer (Fig. 11A). oplGR constructs were activated by dexamethasone (dex) at early gastrula, and embryos later assayed morphologically.
opl induces cellular aggregates without cell division. Experimental scheme. Wild-type embryos were injected in one cell at the 2-cell stage with 500 pg of oplGR or oplΔC RNA plus 100 pg lacZ RNA as lineage tracer. Injected embryos were incubated in saline alone or with dexamethasone (dex) and/or hydroxyurea plus aphidicolin (HUA) starting at early gastrula (stage 10) until late neurula (stage 19). See Methods for details. (B-H) Morphology and histology of opl-induced cellular aggregates. β-gal staining is obscured by pigment. (B-E) Anterior (A) is to the left, dorsal is to the top. (B) Embryo injected with oplGR RNA without dex addition; (C,D) embryos injected with oplGR RNA and treated with dex from stage 10 until harvest. Arrows point to cellular aggregates in ventrolateral ectoderm. (E) Embryo injected with oplGR RNA and treated with HUA (see Methods) plus dex from stage 10 until harvest. (F-H) Embryos sectioned and stained after oplΔC injection. (F) +dex, transverse section of stage 19 embryo through trunk region. Regions magnified in G and H are boxed and indicate uninjected and injected sides of the embryo respectively, as judged by β-gal staining. (G) Uninjected side; (H) injected side. The width of the ectoderm is bracketed in G and H. D, dorsal; V, ventral.
opl induces cellular aggregates without cell division. Experimental scheme. Wild-type embryos were injected in one cell at the 2-cell stage with 500 pg of oplGR or oplΔC RNA plus 100 pg lacZ RNA as lineage tracer. Injected embryos were incubated in saline alone or with dexamethasone (dex) and/or hydroxyurea plus aphidicolin (HUA) starting at early gastrula (stage 10) until late neurula (stage 19). See Methods for details. (B-H) Morphology and histology of opl-induced cellular aggregates. β-gal staining is obscured by pigment. (B-E) Anterior (A) is to the left, dorsal is to the top. (B) Embryo injected with oplGR RNA without dex addition; (C,D) embryos injected with oplGR RNA and treated with dex from stage 10 until harvest. Arrows point to cellular aggregates in ventrolateral ectoderm. (E) Embryo injected with oplGR RNA and treated with HUA (see Methods) plus dex from stage 10 until harvest. (F-H) Embryos sectioned and stained after oplΔC injection. (F) +dex, transverse section of stage 19 embryo through trunk region. Regions magnified in G and H are boxed and indicate uninjected and injected sides of the embryo respectively, as judged by β-gal staining. (G) Uninjected side; (H) injected side. The width of the ectoderm is bracketed in G and H. D, dorsal; V, ventral.
The results of this analysis are shown in Fig. 11 and collated in Table 1. Without dex treatment embryos appeared normal (Fig. 11B), while after activation of the oplGR protein at early gastrula, large pigmented aggregates were seen in the ventral and lateral ectoderm (Fig. 11C,D) in 50% of all embryos. These aggregates were also observed after ectopic opl misexpression in 15% of embryos, and after ectopic oplΔC expression in 64% of embryos, consistent with the greater activity of oplΔC (Table 1). Aggregates were observed in the ectoderm and adjacent to the neural plate, but not within the neural plate itself. The increased activity of oplGR relative to opl in this assay is similar to that of oplΔC and is likely to be due to the presence of a weak activation domain in the GR, that can overcome the inhibitory domain of intact opl protein (Mattioni et al., 1994).
We next asked whether the aggregates observed after opl misexpression were a result of increased proliferation or cell aggregation. Embryos injected with oplGR were pretreated with hydroxyurea and aphidicolin (HUA) to inhibit DNA synthesis and cell division (Harris and Hartenstein, 1991), before dexamethasone treatment (Fig. 11A). After this treatment, cells remained large, consistent with the absence of cell division (not shown). As shown in Fig. 11E, cell aggregates were observed even in the absence of DNA synthesis in 58% of embryos, similar to the percentage observed without HUA treatment (66%; Table 1). Similar results were obtained after injection of oplΔC in the presence of HUA (Table 1).
A section of an embryo, injected with oplGR RNA and treated with dex at stage 10, formed a ventral aggregate (Fig. 11F). In close up, the uninjected side shows the three germ layers, including a tightly packed ectodermal epithelium (bracket, Fig. 11G). In contrast, on the injected side of the embryos the ectoderm was organized as a loosely packed mass of cells, which included pigmented cells not arranged in the characteristic epithelium seen in control embryos (Fig. 11H). The mesodermal and endodermal layers appeared morphologically normal.
Additionally, we asked when the embryo was competent to form aggregates. After injection of oplGR RNA followed by dex treatment at various times after injection (Fig. 11A), embryos were later scored for appearance of maximally sized aggregates. The results are collated in Table 2. After dex induction at midblastula (stage 8), embryos took 12.5 hours to form aggregates of maximal size, whereas by early gastrula (stage 10) only 8 hours elapsed before aggregate formation. By midgastrula (stage 11.5), only 4 hours elapsed after dex addition before maximal aggregate formation occurred and this was the shortest lag time observed. Older embryos (late gastrula (stage 12), early neurula (stage 15) and late neurula (stage 18) all formed aggregates, but at 8 to 9 hours after dex addition. No aggregates were observed after dex addition at tailbud (stage 24), but it was not clear whether this reflected a lack of competence to form aggregates or degradation of oplGR RNA.
Induction of cellular aggregates depends on time of addition and duration of dex treatment in oplGR-injected embryos

In summary, opl, oplΔC and oplGR proteins were able to activate formation of large pigmented aggregates in the ventral and lateral ectoderm. These appeared to be loose associations of cells and were not caused by an increase in cell proliferation. Embryos were competent to form aggregates from midblastula through late neurula, with the time of maximal competence at midgastrula.
opl activates expression of neural crest and dorsal neural tube markers and represses epidermal gene expression in whole embryos
We extended the analysis in whole embryos to ask whether opl constructs can activate ectopic expression of neural crest and dorsal neural tube markers and suppress ventral ectodermal (epidermal) fates. Embryos were injected with either oplGR or oplΔC RNA, along with lacZ RNA as lineage tracer. For oplGR injections, dexamethasone (dex) was added at stage 11 when response to oplGR was maximal as indicated in the aggregate formation assay (Table 2, and not shown). Neural crest formation was assayed by expression of slug, dorsal neural tube fates were assayed by expression of pax3 and epidermal fate was assayed by expression of XK81.
Representative data is shown in Fig. 12B-E and the results collated in Table 3. Without dex, uniform slug expression was observed on both sides of the embryo (Fig. 12B). After dex treatment, concentration of lacZ lineage tracer was observed reflecting the cell aggregates previously seen in pigmented embryos (see Fig. 11). In dex-treated embryos, slug expression expanded anteriorly on the injected side of the embryo (Fig. 12C) in 42% of injected embryos. Similarly, a large expansion of the pax3 expression domain was observed after dex addition in 75% of injected embryos. In a few cases (4% of injected embryos), patches of pax3 expression were not connected to the neural plate.
opl induces ectopic slug and pax3 and inhibits XK81 expression in embryos. (A) Experimental scheme. Albino embryos were injected in one cell at the 2-cell stage with 500 pg oplGR or oplΔC plus 100 pg lacZ RNA as lineage tracer. Embryos were incubated in saline, or with addition of dexamethasone (dex) at stage 11, until embryos reached midneurula (stage 18), when they were harvested for in situ analysis. See Methods for details. (B-E) Ectopic slug (Mayor et al., 1995) and pax3 (Espeseth et al., 1995) expression in whole embryos. Embryos were injected with oplGR plus lacZ RNAs and incubated without (B,D) or with dex (C,E) before histochemical staining for β-gal (light blue) and in situ hybridization for slug or pax3 (purple). Anterior is to the top in all panels. Arrowheads indicate injected side. Ectopic expression of slug and pax3 is indicated by brackets. (F,G) oplΔC suppressed expression of the epidermal cytokeratin XK81 (Jonas et al., 1985). Embryos were injected with 500 pg globin (F) or oplΔC (G) RNAs mixed with 100 pg lacZ RNA, incubated in saline until stage 18, before histochemical staining for β-gal (light blue) and in situ hybridization for XK81 (purple). Anterior is to the top in both panels. Endogenous XK81 is expressed throughout the ventrolateral epidermis and excluded from the neural plate (white regions). White arrow indicates oplΔC-expressing cells (marked by β-gal staining) do not express XK81. A, anterior; P, posterior.
opl induces ectopic slug and pax3 and inhibits XK81 expression in embryos. (A) Experimental scheme. Albino embryos were injected in one cell at the 2-cell stage with 500 pg oplGR or oplΔC plus 100 pg lacZ RNA as lineage tracer. Embryos were incubated in saline, or with addition of dexamethasone (dex) at stage 11, until embryos reached midneurula (stage 18), when they were harvested for in situ analysis. See Methods for details. (B-E) Ectopic slug (Mayor et al., 1995) and pax3 (Espeseth et al., 1995) expression in whole embryos. Embryos were injected with oplGR plus lacZ RNAs and incubated without (B,D) or with dex (C,E) before histochemical staining for β-gal (light blue) and in situ hybridization for slug or pax3 (purple). Anterior is to the top in all panels. Arrowheads indicate injected side. Ectopic expression of slug and pax3 is indicated by brackets. (F,G) oplΔC suppressed expression of the epidermal cytokeratin XK81 (Jonas et al., 1985). Embryos were injected with 500 pg globin (F) or oplΔC (G) RNAs mixed with 100 pg lacZ RNA, incubated in saline until stage 18, before histochemical staining for β-gal (light blue) and in situ hybridization for XK81 (purple). Anterior is to the top in both panels. Endogenous XK81 is expressed throughout the ventrolateral epidermis and excluded from the neural plate (white regions). White arrow indicates oplΔC-expressing cells (marked by β-gal staining) do not express XK81. A, anterior; P, posterior.
We also asked whether opl altered expression of the ventral epidermal marker, XK81. Since epidermis is specified early, we used the constitutively active construct oplΔC. Embryos injected with globin RNA as a control (Fig. 12F) along with lacZ showed XK81 expression over the entire ventral ectoderm, obscuring the β-gal staining. After injection of oplΔC (Fig. 12G), however, 51% of embryos showed large patches of lacZ expression since XK81 is not expressed in these regions (Table 4). Regions of ventral ectoderm that did not express XK81 always coincided with regions of lacZ expression.
In summary, in whole embryos, opl constructs were able to expand the expression domains of a neural crest and a dorsal neural tube marker, and was able to inhibit expression of an epidermal cytokeratin gene, which is a marker of the ventral ectoderm. The slug expression domain also expands in Zic3- and Zic-r1-injected embryos, consistent with these results.
DISCUSSION
We report isolation of opl, a gene similar to Drosophila odd-paired and a new member of the Zic gene family in Xenopus. opl expression indicates that neural determination begins by the onset of gastrulation, and the opl protein appears to have three activities. First, opl can sensitize the presumptive neurectoderm for induction, suggesting that it may be a neural competence factor. Second, consistent with its later expression in the dorsal neural tube and neural crest, an activated form of opl, oplΔC, can activate neural crest and dorsal neural tube markers. Third, opl can synergize with noggin to induce expression of engrailed. These data indicate that opl may be a key regulator of Xenopus neurogenesis.
opl expression defines an early neurectodermal domain
opl is most similar to the mouse Zic1 gene that is expressed in the embryonic mesoderm and presumptive neurectoderm of mice (Nagai et al., 1997), with later expression persisting in the dorsal neural tube and the dorsomedial region of the somites. Xenopus opl is not expressed at significant levels in the mesoderm. However, the early expression of Xenopus opl, mouse Zic1 and zebrafish opl (Grinblat et al., 1998) during neurogenesis, and their later restriction of expression to the dorsal neural tube suggests that their function may be conserved during vertebrate neural development.
At early gastrula, when opl is first expressed, neural tissue is not yet specified, as judged by isolation and culture of dorsal ectoderm and later analysis for markers of neural differentiation (Sive et al., 1989). However, restricted opl expression in dorsal ectoderm clearly shows that the neurectoderm has begun to be set aside by the onset of gastrulation. A large part of the early expression domain of opl is in dorsal ectoderm not yet underlain by the involuting mesendoderm. This dorsal restriction of opl RNA is dependent on dorsal signals since it is abolished in UV-treated embryos (J. G. and H. S., unpublished data). These data indicate that opl RNA accumulation may be activated by planar signals from the organizer, or by patterning signals present in the dorsal ectoderm prior to the onset of gastrulation (Sokol and Melton, 1991; Wylie et al., 1996). Recent reports concerning other Xenopus Zic family members (Zic3 and Zic-r1) indicate that their expression requires removal of BMP signaling (Nakata et al., 1997; Mizuseki et al., 1998). Consistent with this, opl is induced by noggin (Kuo, J. S., Gamse, J. and Sive, H., unpublished data), however, it is not clear whether such signaling initiates opl expression, or is required for a later maintenance step.
opl can modulate neural competence
The early opl expression domain suggested that it may be a competence factor and, indeed, opl could increase the responsiveness of isolated ectoderm to noggin by 10-fold. Our data indicate that expression of opl may account for some of the enhanced sensitivity of the dorsal ectoderm to neural induction as noted by other investigators (Sharpe et al., 1987; Otte et al., 1991). The translation initiation factor EIF4AII has recently been shown to synergize with noggin to activate expression of some neural crest markers as well as markers of the neurectoderm proper (Morgan and Sargent, 1997). EIF4AII is expressed considerably later than opl, by midgastrula, and may be a target of opl activation. Protein kinase C (PKC) activity has also been implicated in conferring neural competence on the dorsal ectoderm (Otte et al., 1991), while the Notch signaling pathway may also be involved in directing neural competence (Coffman et al., 1993). It is unknown whether opl can modulate or be modulated by either of these pathways.
opl can activate engrailed
In conjunction with noggin, opl and oplΔC strongly potentiated expression of the posterior neural marker engrailed, expressed in the midbrain/hindbrain region. Although opl is expressed in presumptive posterior neurectoderm, we did not expect this result, since opl RNA is present throughout the anteroposterior extent of the future brain at early stages. It is possible, however, that opl protein activity is modulated along the A/P neural axis. One intriguing connection between the ability of opl to potentiate expression of engrailed and Drosophila opa is that opa is necessary for the correct temporal activation of Drosophila wingless (Benedyk et al., 1994), while, in Xenopus, the wingless homologs wnt3A and wnt8 can posteriorize neurectoderm in combination with noggin (McGrew et al., 1995, 1997; Fredieu et al., 1997) This suggests that opl may activate wnt genes or act downstream of wnt proteins. Indeed the expression patterns of opl and wnt3A are very similar from late gastrula (Wolda et al., 1993). However, unlike wnt3A, opl expression did down-regulate anterior neurectodermal markers, suggesting that the activities of these proteins are not entirely equivalent.
opl as an activator of dorsal neural tube and neural crest fates
The later expression of opl in the dorsal neural tube and neural crest correlated with the ability of a truncated form of opl, oplΔC, to activate genes characteristic of these fates. Like opl, the dorsal neural tube marker, pax3, is transiently expressed initially across the entire dorsoventral extent of both the Xenopus and chick neural plates and is later expressed solely in the future dorsal neural tube (Liem et al., 1995; Bang et al., 1997). However, opl is expressed earlier than pax3, consistent with a role for opl in activating pax3. As pax3 expression increases in the future dorsal neural tube, opl expression also increases and later becomes dorsally restricted. The expression of opl in presumptive neural crest from late gastrula again precedes the expression of neural crest-specific markers in this region (Mayor et al., 1995), consistent with a role for opl in regulating neural crest formation. opl expression across both presumptive dorsal neural tube and neural crest regions at late gastrula suggests that these two neural domains are initially defined as a unit that later becomes subdivided into dorsal neural tube and neural crest subdomains. Since recent studies have shown that wnt proteins can also activate neural crest markers in neuralized ectoderm (Saint-Jeannet et al., 1997; Chang and Hemmati-Brivanlou, 1998), another connection between opl and wnt expression is possible.
opl activity is modulated by synergizing factors
Several lines of evidence indicate that opl protein requires synergizing factors. First, oplΔC, but not full-length opl protein could activate dorsal neural tube and neural crest markers in animal caps. In other respects, opl appears to have similar, though weaker activity to oplΔC, as both constructs could repress ventral ectodermal markers, synergize with noggin and catalyse formation of pigmented aggregates. This suggests that oplΔC is an activated form of opl, although we cannot rule out that oplΔC has a novel activity. Second, we showed that, in animal caps, intact opl in conjunction with noggin was able to activate a dorsal neural tube marker, but not neural crest markers. This suggests that factors induced by noggin modulate opl activity; however, since opl could not activate neural crest markers even in conjunction with noggin, some other factor may allow opl to activate neural crest-specific gene expression. Such a factor may act by de-repressing a carboxy terminus inhibitory function in the opl protein, perhaps by altering opl phosphorylation. Third, in whole embryos, activated opl constructs could expand the domain of slug and pax3 expression but only adjacent to the normal expression domains of these genes, and never in the extreme anterior neural plate. This further suggests that opl activity requires cofactors that are present in the neural plate, neural crest and in early gastrula animal caps, but that are absent from or counteracted by inhibitors in the ventral ectoderm of older embryos.
These considerations raise the question of whether the activation of dorsal neural tube fates, the engrailed-inducing capacity and the competence-promoting activity of opl are distinct functions, perhaps requiring different domains of opl or different cofactors. In animal caps, opl in conjunction with noggin was able to promote expression of both dorsal and ventral neural tube markers, indicating that expression of ventral neural tube markers is not incompatible with opl activity. Thus opl may act by a similar mechanism throughout the neural plate, and may play a role in early gastrula to activate both dorsal and ventral neural tube markers. Since, by the end of gastrulation, opl is not expressed ventrally, ventral opl activity is only transient.
Unlike opl, two related genes, Zic3 and Zic-r1 are able to activate both general neuronal markers and a neural crest marker in animal caps without requiring noggin addition (Nakata et al., 1997; Mizuseki et al., 1998). The differences in our results may reflect a difference in assay conditions that were employed, or may indicate different activities for these related genes. Since the neuronal markers that were examined in these studies are expressed in restricted areas of the neural plate, whereas expression of Zic3 and Zic-r1 is much broader, activity of these genes may also be modulated by cofactors. Similarly, Zic3 and Zic-r1 increased slug expression only adjacent to the normal slug expression domain and could not induce it in ventral ectoderm.
opl modulates cell adhesion
One of the most striking phenotypes consistently produced by opl and also produced by Zic3 (Nakata et al., 1997), was the appearance of cellular aggregates in the ventrolateral ectoderm, but not within the neural plate itself. Aggregates did not express neural or neural crest markers unless they directly contacted the neural plate. In both whole embryos and in animal caps injected with opl RNA, aggregates comprised cells that were loosely adherent (not shown). In both the whole embryo and the animal cap, the cells appeared healthy, surviving and dividing after disaggregation in the culture dish. One intriguing hypothesis is that opl suppresses an adhesion pathway responsible for ectodermal-epithelial integrity that is normally down-regulated as neural crest fates are determined (Bronner-Fraser, 1995). Since opl did not induce ectopic slug in the ventral ectoderm, but did induce aggregates there, the aggregates are clearly not committed neural crest cells, but may represent some early step in neural crest commitment. Consistent with this hypothesis, opl is expressed sequentially in the cranial neural crest segments just before the cells mound up, lose adhesivity and begin migrating (Sadaghiani and Thiebaud, 1987).
opl and inhibition of epidermal fates
In Xenopus, formation of ventral fates including epidermis appears to be dependent on Bone Morphogenetic Protein (BMP) activity (Wilson and Hemmati-Brivanlou, 1995). Suppression of this activity by noggin and chordin proteins during gastrulation results in neural determination (Piccolo et al., 1996; Zimmerman et al., 1996). In whole embryos, opl can inhibit expression of the epidermal cytokeratin gene XK81 while, in animal caps, opl also suppresses expression of GATA2, a ventral ectodermal marker that is regulated by BMP4 signaling. Since this suppression is not correlated with an ability of opl or oplΔC alone to activate general neural markers, opl is able to inhibit only part of the dorsoventral ectodermal patterning pathway. In accord with this, opl expression overlaps that of BMP4, until midgastrula when BMP4 RNA begins to be cleared from the presumptive neural plate (J. G. and H. S., unpublished).
A model for opl function
Our data suggest a model for the role of opl in early neural determination (Fig. 13). Early during gastrulation, opl gene expression is induced, either by planar signals from the organizer or by pre-existing signals within the dorsal ectoderm. The early expression domain of opl throughout the neural plate sensitizes the neurectoderm for subsequent neuralizing signals, such as noggin and chordin, derived from the organizer. This sensitization both modulates general neural readout and is also involved in activating relatively posterior neural markers, such as engrailed. By the end of gastrulation, opl expression in the dorsal neural plate and presumptive neural crest directs fates characteristic of these regions. The activity of opl in the presumptive neural crest is modulated by a regulatory domain in the carboxy terminus, suggesting that a laterally, but not anteriorly, localized factor(s) may control opl activity. The ability of opl to suppress cell adhesion may be part of the commitment program for neural crest fates.
Model of the roles opl plays in early neural determination. Schematics of whole embryos during gastrulation are shown (dorsal views, anterior is to the top). Arrow, dorsal lip of blastopore; A, anterior; P, posterior. High level opl expression (spots) is activated in the dorsal ectoderm by early gastrula. By midgastrula opl expression has sensitized dorsal ectodermal cells for later induction and patterning by neural inducers such as noggin (light stripes). opl also acts by suppressing ventral ectodermal fates, such as epidermis (T bar). By late gastrula, opl expression has become restricted to the presumptive dorsal neural tube and neural crest, where it activates expression of neural markers in these regions. A dorsally restricted synergizing activity (dark stripes) is necessary for activation of neural crest fates. See text for further discussion.
Model of the roles opl plays in early neural determination. Schematics of whole embryos during gastrulation are shown (dorsal views, anterior is to the top). Arrow, dorsal lip of blastopore; A, anterior; P, posterior. High level opl expression (spots) is activated in the dorsal ectoderm by early gastrula. By midgastrula opl expression has sensitized dorsal ectodermal cells for later induction and patterning by neural inducers such as noggin (light stripes). opl also acts by suppressing ventral ectodermal fates, such as epidermis (T bar). By late gastrula, opl expression has become restricted to the presumptive dorsal neural tube and neural crest, where it activates expression of neural markers in these regions. A dorsally restricted synergizing activity (dark stripes) is necessary for activation of neural crest fates. See text for further discussion.
Future challenges are to analyse the consequence of opl ablation, to isolate the downstream direct targets of opl, and to further understand how opl fits into the hierarchy of neural regulatory genes.
Acknowledgement
We thank Igor Dawid, Richard Harland, Paul Krieg, Nancy Papalopulu, Michael Sargent, Dave Turner and Andrea Vortkamp for kind gifts of plasmids. We thank W. David Willison for help with the subtractive screen, Cathy Nocente-McGrath for the midgastrula λZap phage library, Melissa Rones for helping isolate the full-length opl cDNA, Harvey Lodish for tissue culture reagents and facilities, Liuda Ziaugra for expert DNA sequencing, and Jeanne Reis for histological expertise. We greatly appreciate the helpful comments and discussions with members of the Sive laboratory during the interminably long course of this work, and thank Andrew Lassar for criticism of the manuscript. We also thank Jun Aruga for interesting discussion concerning Xenopus Zic3 experiments. This work was supported by grants from the NSF and NIH to H. L. S. We appreciate support from Genetics Institute, Cambridge MA. J. S. K. was the recipient of a Medical Scientist Training Program Fellowship (PHS grant 2-T32-GM07753) from the Harvard MD-PhD Program. M. P. was the recipient of a fellowship from the Human Frontiers Science Program. X.L. was the recipient of an NIH postdoctoral fellowship. H. L. S. was the recipient of a NSF Young Investigators Award and the Latham Family Development Professorship at MIT.