Repression of organizer genes in dorsal and ventral Xenopus cells mediated by maternal XTcf3

The T cell factor (TCF) family of DNA binding proteins plays key roles in both embryonic development and in tumor progression. TCFs act as effectors of Wingless/Wnt signaling pathways by binding directly to armadillo/β-catenin (reviewed by Barker et al., 2000; Bienz, 1998; Nusse, 1999). They are known not to act as ‘classical’ transcription factors, but rather as components of multi-protein enhancer complexes (Hecht and Kemler, 2000). In Xenopus embryos, a maternal TCF family member, XTcf3, binds several co-activators and co-repressors including CBP (Hecht et al., 2000), Smad4 (Nishita et al., 2000) and groucho (Roose et al., 1998) as well as β-catenin (Molenaar et al., 1996). At this early stage of development, Wnt signaling is essential for the establishment of dorsoventral pattern. Embryos depleted of maternal β-catenin mRNA develop normally to the gastrula stage but then become radially symmetrical, lacking the dorsal elements of all three germ layers, including neural tubes, somites and notochords (Heasman et al., 1994). The precise role of XTcf3 in this process has not been resolved, but several models have been suggested.

First, XTcf3 could be a transcriptional activator of Wnt-inducible genes, such as siamois and Xnr3, when bound to β-catenin. This model was suggested by cell culture studies, where interaction of XTcf3 with β-catenin was required for the activation of a CAT reporter driven by a multimerized motif for TCF/LEF binding (CTTTGA/TA/T). XTcf3 alone had no activity (Molenaar et al., 1996). This model was also suggested by the fact that a truncated form of XTcf3, lacking the β-catenin-binding domain, repressed axis formation when expressed dorsally, and prevented the dorsalizing activity of ectopically expressed β-catenin (Molenaar et al., 1996). However, this construct could have acted in a dominant-negative fashion either for XTcf3 or for β-catenin itself, since it could block endogenous β-catenin activity. Second, XTcf3 has been suggested to act as a repressor ventrally, and as an activator dorsally in combination with β-catenin. In support of this, expression of a siamois promoter/CAT reporter containing mutated TCF-binding sites resulted in increased CAT expression when it was injected on the ventral side of a Xenopus blastula (Brannon et al., 1997).

No consistent picture has emerged from studies of the role of TCFs in other model systems. Loss-of-function studies in Drosophila support the view that a TCF/β-catenin complex activates wingless target genes, since pangolin (dTCF) mutants phenocopy wingless and armadillo (β-catenin) mutant phenotypes (van de Wetering et al., 1997). However, in C. elegans, the TCF homologue, pop1, may be a transcriptional repressor, since pop1 mutants have the opposite phenotype from β-catenin loss-of-function mutants (Rocheleau et al., 1997; Thorpe et al., 1997). Similarly, Tcf3 is considered to have a repressor function in zebrafish since the headless mutant phenotype, which maps to the tcf3 gene, was rescued by the expression of a Tcf3-Engrailed repressor fusion protein (Kim et al., 2000).

In this situation, loss-of-function studies are essential. In Xenopus, zygotic transcription does not start until the mid-blastula stage, and XTcf3, like other early patterning genes, is expressed in the oocyte and stored as a maternal mRNA (Molenaar et al., 1998; Molenaar et al., 1996). We therefore specifically depleted this maternal store by microinjection of antisense deoxyoligonucleotides (oligos) against XTcf3 into full-grown oocytes in culture, and fertilized them by the host-transfer technique. XTcf3^– embryos developed with a dorso-anteriorized phenotype, and overexpressed the organizer genes siamois, Xnr3, goosecoid and chordin. These effects were specific to the depletion of XTcf3 RNA, since two different oligos complementary to XTcf3 mRNA had this effect and it was rescued by the reintroduction of XTcf3 mRNA into oocytes. Organizer gene expression was increased on both the dorsal and ventral sides of XTcf3^– embryos. Siamois and Xnr3 expression was also activated in isolated animal caps. These results show that in vivo, XTcf3 acts as a repressor of organizer genes throughout the embryo. What then activates these genes in the organizer? The simplest model would be that β-catenin association with XTcf3 blocks the repression of dorsal gene transcription. Alternatively, there could be specific activators of dorsal gene transcription.

In a companion paper, we show that expression of many dorsal genes requires both the vegetally localized maternal transcriptional activator VegT and the maternal Wnt pathway (Xanthos et al., 2002). This suggests that VegT can activate organizer genes when β-catenin de-represses them. We tested this by depleting both VegT and XTcf3 RNAs in the same embryos and confirmed that the dorsal markers goosecoid and Xnr6 are regulated in this way.

Oocytes were manually defolliculated and cultured as described previously (Kofron et al., 1999). Oocytes were injected with oligos in oocyte culture medium (OCM) using two equatorial injections per oocyte for XTcf3 oligos, or one vegetal injection for VegT and Axin oligos, cultured at 18°C and fertilized using the host transfer technique as described previously (Zuck et al., 1998). Rescue experiments were carried out as described in the text either by injecting mRNA equatorially into oocytes, 24 hours after the oligo (Fig. 2F), or by injecting into 4-cell stage embryos (Fig. 2E). Eggs were stripped and fertilized using a sperm suspension and embryos were maintained in 0.2× MMR. For injections of mRNA after fertilization (Fig. 2E), embryos were dejellied, and transferred to 2% Ficoll in 0.5× MMR at the 1-cell stage. mRNAs were diluted with sterile distilled water and injected into blastomeres.

For animal cap assays, mid-blastula embryos were placed on 2% agarose dishes in 1× MMR and the animal caps were dissected using sharp forceps. The caps were then cultured in OCM until sibling embryos reached the mid- to late-gastrula stage. For equatorial explants, embryos were placed on 2% agarose dishes at the mid-blastula stage and the equatorial regions were dissected with tungsten needles (Xanthos et al., 2001) and cultured in OCM until sibling embryos reached the mid-neurula stage. For separation into dorsal and ventral halves at the gastrula stage, the dorsal side of embryos was marked at the four-cell stage using Nile blue crystals. The dorsoventral axis was recognized at the four-cell stage by the pigmentation differences of the dorsal and ventral sides. When wild-type embryos reached stage 10, all the batches were placed on 2% agarose dishes in 1× MMR; pH 7.6 and bisected into dorsal and ventral halves, and frozen in groups of 4 half-embryos at 2-hour intervals through the gastrula stages.

The antisense oligodeoxynucleotides used were HPLC purified phosphorothioate-phosphodiester chimeric oligonucleotides (Sigma/Genosys) with the base composition:

XTcf3 T1: 5′-C*G*A*G*GGATCCCAGTC*T*T*G*G-3′.

XTcf3 T2: 5′-G*A*G*ATAACTCTGA*T*G*G-3′.

VegT (VT9M): 5′-C*A*G*CAGCATGTACTT*G*G*C-3′ (Zhang et al., 1998).

Axin (M7): 5′-T*T*C*C*TCGCCAGGAA*C*T*G*G-3′ (Kofron et al., 2001).

The XTcf3 oligos are completely complementary to all four variants of XTcf3. Asterisks (*) represent phosphorothioate bonds. Oligos were resuspended in sterile, filtered water and injected in doses as described in the text. Full-length XTcf3 in the vector pGlomyc was linearized with XbaI and capped XTcf3 mRNA was synthesized using the T7 mMessage mMachine kit (Ambion). RNAs were phenol extracted, ethanol precipitated and then resuspended in sterile distilled water for injection.

Total RNA was prepared from oocytes, embryos and explants using proteinase K and then treated with RNase-free DNase as described previously (Zhang et al., 1998). Approximately one-sixth embryo equivalent of RNA was used for cDNA synthesis with oligo(dT) primers followed by real-time RT-PCR and quantitation using the LightCycler™ System (Roche) as described by Kofron et al. (Kofron et al., 2001). The primers and cycling conditions used are listed in Table 1. Relative expression values were calculated by comparison to a standard curve generated by serial dilution of uninjected control cDNA. Samples were normalized to levels of ornithine decarboxylase (ODC), which was used as a loading control. Samples of water alone or controls lacking reverse transcriptase in the cDNA synthesis reaction failed to give specific products in all cases.

We injected modified antisense chimeric phosphorothioate/ phosphodiester deoxyoligonucleotides (oligos) into the equatorial regions of oocytes to deplete maternal XTcf3 mRNA. RNA levels were assessed by real time RT-PCR. XTcf3 mRNA was depleted to 30% of control levels by 3.5 ng of oligo T1 (Fig. 1). A second antisense oligo (T2) caused depletions to 10% control levels at 5 ng of oligo (Fig. 1B). Unexpectedly, we found that a second TCF, XTcf4, which had previously been reported to be first expressed at the neurula stage (Konig et al., 2000) was also expressed in the oocyte and early blastula stage embryo (Fig. 1A,B). Therefore we tested whether either of our XTcf3-specific oligos also depleted XTcf4. We found that T1 depleted XTcf4 to 65% of control levels, while T2 was specific for XTcf3. Therefore we confirmed all of our findings with T1 using the T2 oligo. Depletion of maternal XTcf3 did not affect the transcription of zygotic XTcf3, as its expression was equal to that of control embryos at the late gastrula stage (data not shown).

Oocytes were matured in vitro and fertilized 36 hours after oligo injection. XTcf3-depleted (XTcf3^–) embryos developed normally through the cleavage and blastula stages, and became abnormal during gastrulation (Fig. 2A). High doses of oligo T1 (3.5 ng) caused delayed gastrulation, including a failure of blastopore closure, leading to abnormal development, including exogastrulation (Fig. 2B, red). Lower doses also slowed the appearance of the blastopore by 30 minutes, but gastrulation then proceeded (Fig. 2A). At the tailbud and tadpole stages, these embryos had dorso-anteriorized phenotypes (Fig. 2B, blue and mauve; 60/60 cases in 2 experiments). The notochord and anterior endoderm region were enlarged compared to controls, and the hindgut was concomitantly reduced. The most extreme cases of this phenotype at the swimming tadpole stage are shown in Fig. 2C. This experiment was repeated with the T2 oligo. 5 ng of this oligo caused the same effect as 3 ng of oligo T1 (Fig. 2F; 24/26 cases in 2 experiments).

We suspected that XTcf3^– embryos were dorsalized, and that excessive cell movements might be responsible for the gastrulation defects. To test this, we carried out explant assays on sibling embryos to those shown in Fig. 2B. Isolated equatorial regions of wild-type blastulae are known to undergo extensive convergence extension movements concomitant with notochord differentiation (Dale and Slack, 1987). We dissected the equatorial regions of wild-type and XTcf3^– embryos at the mid-blastula stage and cultured them in isolation until the mid-neurula stage. XTcf3^– explants underwent exaggerated and dose-dependent convergence extension movements compared to control explants (Fig. 2B, lower panel).

To confirm that this effect was specifically due to the depletion of XTcf3, we tested whether the dorso-anteriorized phenotype could be rescued by the introduction of synthetic XTcf3 mRNA into XTcf3^– oocytes. This experiment relies on the fact that both the targeted maternal mRNA and injected oligo degrade within 24 hours of injection. RNA injected after this time is not degraded by the oligo. XTcf3^– oocytes were cultured for 36 hours and then half of them were injected equatorially with 100 pg of synthetic XTcf3 mRNA either before (Fig. 2F) or after (Fig. 2E) fertilization. Embryos derived from these RNA-injected embryos were rescued (oligo T1, 12/43 cases dorso-anteriorized; oligo T2, 0/21 cases dorso-anteriorized) compared to those that received oligo only (oligo T1, 60/85 dorso-anteriorized oligo T2, 26/28 dorso-anteriorized) (Fig. 2E,F). In rescued embryos, the notochord was no longer protuberant, and the swollen anterior endoderm was reduced. In most cases the head was reduced in these embryos (Fig. 2E). Similar reduction in head structures was seen when 100-200 pg of XTcf3 mRNA was injected into wild-type oocytes (not shown).

We next analyzed the levels of expression of early zygotic marker genes. XTcf3^– and control embryos were frozen at 2-hourly intervals from the late blastula through the mid-gastrula stages, for real-time PCR analysis of a panel of organizer mRNAs. The organizer genes siamois and Xnr3 have been shown to be direct targets of β-catenin/XTcf3 (Brannon et al., 1997; McKendry et al., 1997). Real-time RT-PCR analysis of XTcf3^– gastrulae showed a dose-dependent increase in the expression of siamois, Xnr3, goosecoid, chordin, cerberus and Xnr6 compared to control levels (Fig. 3A). Importantly, in the same experiment, all these markers, with the exception of Xnr6, were rescued or reduced to below wild-type levels by the introduction of XTcf3 mRNA into XTcf3^– oocytes (Fig. 3A). In comparison, the levels of Bmp4 and Xwnt8 were little affected in XTcf3^– embryos compared to controls. This experiment was repeated using oligo T2 with the same result (data not shown).

In XTcf3^– embryos at the tailbud stage, the ventral mesodermal marker α-T4 globin and general endoderm marker IFABP were expressed at reduced levels compared to controls, while the dorso-anterior endoderm marker Xlhbox8 and heart marker Nkx2.3 were increased (Fig. 3B). These experiments demonstrate that depletion of maternal XTcf3 mRNA causes excessive dorso-anterior development.

XTcf3^– embryos may develop with a dorso-anterior phenotype because dorsal zygotic gene expression is increased by de-repression either in the organizer alone or throughout the embryo. To study this, we dissected explants from wild-type and XTcf3^– embryos and analyzed the expression levels of the organizer genes, siamois, Xnr3, gsc and chordin by real-time RT-PCR. First we isolated animal caps, which in isolation do not normally express dorsal markers, from XTcf3^– and uninjected control embryos at the blastula stage and assayed siamois, Xnr3, gsc and chordin expression during gastrulation. XTcf3^– animal caps showed activation of siamois and Xnr3 expression compared to control caps while gsc and chordin were not expressed (Fig. 4A). Next we examined the expression of siamois, Xnr3, gsc and chordin in dorsal and ventral halves of embryos bisected at three time-points during the gastrula stage (Fig. 4B). In control embryos at all three stages, more than 90% of the expression of chordin, siamois and Xnr3 was restricted to the dorsal side, with the peak of expression of siamois and Xnr3 occurring at stage 10, and the peak of expression of chordin occurring at stage 11. In comparison, XTcf3^– embryos expressed all three organizer markers both dorsally and ventrally. On the dorsal side, the expression level of siamois and Xnr3 was approximately 1.5-fold higher than control levels at stage 10, and for chordin was 3-fold that of control levels at stage 10. On the ventral side the expression of siamois and Xnr3 was increased 15 fold at stage 10, while chordin levels peaked to the same extent at stage 10.5. In stage 10.5 XTcf3^– embryos, the organizer genes were expressed equally highly on the dorsal and ventral sides (Fig. 4B). This experiment was repeated three times with the same result. These data show that the normal function of XTcf3 in vivo is transcriptional repression of the organizer genes siamois, Xnr3 and chordin in the entire embryo throughout gastrulation.

Axin is a major regulator of β-catenin degradation, and we have shown previously that its depletion results in both dorsal and ventral overexpression of β-catenin protein, and the overexpression of organizer genes both dorsally and ventrally (Kofron et al., 2001). However, axin-deficient embryos have much more pronounced dorso-anterior characteristics than XTcf3-depleted embryos, including circumferential cement glands, vertically orientated notochords and lack of tails. To compare these differences directly, we depleted both axin and XTcf3 in the same experiment and compared the effects on the expression of dorsal and ventral marker genes at gastrulation (Fig. 4C). Both depletions caused an increase in dorsal markers siamois and goosecoid, although the extent of overexpression was variable. One significant difference between XTcf3 and axin depletions was on the expression of Xwnt8. In axin^– embryos, Xwnt8 was severely reduced in its expression both dorsally and ventrally, while in XTcf3^– embryos its expression was little affected ventrally and reduced to about 40% control levels dorsally (Fig. 4C) (Kofron et al., 2001). Since it has been shown in zebrafish that a wnt8 mutant has an anteriorized nervous system (Erter et al., 2001), it is likely that in axin^– embryos, loss of Xwnt8 expression enhances the dorsalizing effect of the overexpression of organizer genes. This anteriorization would not be expected to be as extreme in XTcf3^– embryos, where Xwnt8 expression is little affected.

In the absence of XTcf3, the organizer genes were robustly activated on both dorsal and ventral sides of the embryo. One likely candidate that might be responsible for this activation is the maternal transcription factor VegT. We next tested this hypothesis by comparing the expression of organizer genes in wild-type, XTcf3-depleted, VegT-depleted and XTcf3/VegT-depleted embryos. If VegT is required to activate organizer genes, then they should not be expressed in the double depleted embryos, even though they are robustly expressed in XTcf3^– embryos.

A series of mid and late blastula and early gastrula stage embryos was analyzed by real-time RT-PCR. In this experiment, doses of VegT and XTcf3 oligos were sufficient to cause incomplete phenotypes (6 ng VegT and 3 ng T1; Fig. 5A), since the double depleted embryos with higher doses did not survive beyond the gastrula stage. At these doses, VegT^– embryos showed extremely reduced levels of gsc, chordin and Xsox17 compared to controls, while siamois, Xnr3 and Xnr6 were partially reduced (Fig. 5B). XTcf3^– embryos had increased levels of all these markers compared to controls (Fig. 5B). In comparison, embryos depleted of both, VegT^–/XTcf3^– embryos, had reduced levels of all the organizer genes compared to the control levels, indicating that, as well as derepression by XTcf3, these genes need activation by VegT for normal levels of expression. This difference was particularly striking for Gsc and Xnr6, which were almost completely off in double depleted embryos, suggesting that VegT is the only activator of these genes. Xsox17 and chordin were reduced in expression compared to controls. For Xnr3 and siamois, their expression was delayed but then rose to wild-type levels by the gastrula stage, suggesting that another transcriptional activator in addition to VegT modulates their early expression.

XTcf3 mRNA is expressed throughout the early Xenopus embryo (Molenaar et al., 1998). We confirm that it has roles as a transcriptional repressor, in dorsal, ventral and animal cells, suggesting a model for its activity as shown in Fig. 6. In ventral cells, it prevents the ectopic activation of organizer genes (Xnr6, gsc, chordin, cerberus, Xhex, siamois and Xnr3) while dorsally, binding of XTcf3 by β-catenin relieves this inhibition. However, relief of inhibition is not sufficient to activate organizer genes. In cells of the vegetal hemisphere VegT is an essential transcriptional activator of organizer genes including Xnr6 and gsc. In animal cells, XTcf3 also blocks the expression of siamois and Xnr3. The fact that these genes are activated in this region when XTcf3 is depleted suggests that a transcriptional activator of these genes is also present, or that the basal trancriptional machinery is sufficient to maintain this level of expression. However chordin and goosecoid are not activated in animal cells when XTcf3 is depleted, presumably because they have no endogenous animally localized activator. These results confirm and extend previous studies which have shown that siamois and Xnr3 can be induced in animal caps by overexpression of Wnt pathway components whereas chordin and gsc cannot (Brannon and Kimelman, 1996; Carnac et al., 1996; Fagotto et al., 1997; Smith et al., 1995; Steinbeisser et al., 1993; Watabe et al., 1995). Our experiments do not address whether any of these regulatory mechanisms are direct or indirect interactions between the maternal signaling pathways and the zygotic target genes.

In the experiments presented here, we show a depletion rather than a complete loss of XTcf3 function, therefore we cannot discount the formal possibility that the residual amount of XTcf3 remaining may be sufficient to act as an activator of gene expression when it is present in small amounts. However, the dose response experiment (Fig. 3A) does not support a model where XTcf3 acts as a repressor at high concentrations and an activator at low concentrations, since increasing doses of oligo (causing graded depletion of XTcf3 mRNA) causes an increasing expression of dorsal markers, not a biphasic response. Nor do we find evidence of genes that are simply activated by XTcf3 whose expression would be expected to be reduced by XTcf3 depletion.

In the vegetal hemisphere, in XTcf3^–/VegT^– embryos, gsc, Xnr1 and Xnr6 have a very low level of expression, while Xnr3 and siamois are still relatively robustly expressed. This suggests that Xnr3 and siamois have an additional activating mechanism independent of VegT. Since β-catenin– embryos do not express siamois and Xnr3, it is likely that the unknown activator is also regulated by β-catenin. The suggestion that the Wnt pathway may regulate transcription factors other than XTcf3 in early Xenopus embryos was also raised in a recent study of the regulation of zygotic ventrolateral mesodermal genes, where β-catenin was shown to act independently of XTcf3 (Hamilton et al., 2001). While further studies are needed to identify the molecules involved, β-catenin has been shown to interact with other regulatory proteins including TCF4 (Korinek et al., 1997), Xsox17 and Xsox3 (Zorn et al., 1999) and retinoic acid receptor (RAR) (Easwaran et al., 1999).

Although organizer gene expression in XTcf3^– embryos clearly shows that they are dorsalized, the embryos do not resemble Xenopus embryos dorsalized by the depletion of maternal axin (Kofron et al., 2001). Axin is a key component of the β-catenin degradation complex, and in its absence, organizer genes are also over-expressed both dorsally and ventrally (Kofron et al., 2001). However, axin-deficient embryos have much more pronounced dorso-anterior characteristics, including circumferential cement glands, vertically orientated notochords and lack of tails. There are several possible reasons for this. First, it is possible that the XTcf3^– phenotype is very dose sensitive and we have not been able to pinpoint a dose between the dorsalized phenotype described in this work, and the arrest at gastrulation phenotype that would mimic the appearance of axin^– embryos. Secondly, the XTcf3^– phenotype may be qualitatively different from the axin^– phenotype and not simply a less extreme form. We show that while Xwnt8 is downregulated by axin depletion, it is less affected by XTcf3 depletion (Fig. 3A, Fig. 4C). Since zygotic Wnts are known from other studies to be important in posteriorizing the axis (Erter et al., 2001; Kiecker and Niehrs, 2001; Xanthos et al., 2002), it is likely that this accounts for the more extreme anteriorization of axin^– embryos. Additionally, axin depletion causes an increase in soluble β-catenin protein levels, whereas we found that XTcf3 depletion had no effect on soluble β-catenin levels (data not shown). It is possible that β-catenin may activate or de-repress other DNA binding proteins as well as XTcf3, since XTcf3 is clearly not the only factor regulating Xwnt8 in these experiments. Further experiments are needed to understand the regulators of zygotic Xwnt8.

The maternal Wnt pathway has also been implicated in regulating the zygotic expression of BMPs. This was first suggested in experiments where ventralization by ultraviolet irradiation caused an encroachment of Bmp4 mRNA into the organizer region. Conversely, a dorsalizing lithium treatment resulted in a repression of Bmp4 expression at the gastrula stage (Fainsod et al., 1994). Additionally, Baker et al. (Baker et al., 1999) showed that overexpression of Wnt pathway components can inhibit Bmp4 expression during gastrulation. It is difficult to make comparisons between ventralizations and dorsalizations caused by ultraviolet radiation and lithium with those caused by loss-of-function of individual pathway components. However the results that we have from depletion of β-catenin (Xanthos et al., 2002), depletion of axin (Kofron et al., 2001) and depletion of XTcf3 (Fig. 3A, Fig. 4C) are all consistent with the conclusion that there is little regulation of BMP expression by the Wnt pathway, at least until the midgastrula stage. This conclusion is supported by the fact that Schohl and Fagotto (Schohl and Fagotto, 2002) show in a recent paper that regions of Smad1 phosphorylation and nuclear β-catenin do not appear in a simple complementary pattern. An important question that remains to be answered is what does regulate zygotic BMP expression.

The results presented here are consistent with work in C. elegans, where the TCF homologue, pop1, was also found to be a transcriptional repressor (Rocheleau et al., 1997; Thorpe et al., 1997) and in zebrafish, where a Tcf3 homolog is thought to have a repressor function (Kim et al., 2000). However the phenotype of Xenopus XTcf3^– embryos is very different from the zebrafish headless mutant. One possible explanation for this difference might be that the zebrafish mutant may lack zygotic, but not maternal, Tcf3 function. Since zygotic Wnt signaling causes posteriorization of axial structures (Erter et al., 2001; Kiecker and Niehrs, 2001; Xanthos et al., 2002), a mutant that lacks the Wnt inhibitory action of Tcf3 would be expected to have excessive Wnt signaling and therefore headlessness. Preliminary experiments with a morpholino oligo suggest that blocking Xenopus zygotic XTcf3 also causes a headless phenotype.

The conclusion from this work is that maternally encoded XTcf3 normally represses the activation of dorsal (organizer) genes throughout the embryo during the late blastula/early gastrula stages in Xenopus. This is consistent with its global expression at these early stages. In addition, activation of the Wnt pathway, leading to relief of this repression, is not sufficient to activate all organizer genes. Additional specific activators are required; one of these is VegT.

This work was supported by NIH NRSA F32 HD40716-01 (to D. W. H.) and NIH RO1 HD33002 (to J. H.). Technical support was provided by Teresa Demel, Kelly Galey, Helbert Puck and Stephanie Lang.