Localized expression of a Xenopus POU gene depends on cell-autonomous transcriptional activation and induction-dependent inactivation

In amphibian embryos, regional differentiation is thought to arise through a series of inductive events. Spemann and Mangold dramatically showed that transplantation of the presumptive dorsal mesoderm region (‘organizer’) of a developing embryo to the ventral side led to induction of a complete secondary dorsal axis (Spemann and Mangold, 1924). In vertebrates, the molecular mechanisms regulating axial specification are poorly understood. However, in Drosophila significant advances have been made in understanding the molecules required for proper anteroposterior and dorsoventral axial development during embryogenesis (reviewed by Manseau and Schup-bach, 1989; Anderson, 1987), and homeobox gene products have been shown to control some of these processes (reviewed by Scott, 1988). Recent experimental results support the assumption that these genes will have analogous roles in vertebrate development. Xenopus homeobox genes have been shown to be expressed in restricted patterns along the anteroposterior axis (Condie and Harland, 1987; Sharpe et al., 1987; Oliver et al., 1988; Ruiz i Altaba and Melton, 1989a), and manipulation of the expression of some of these genes alters normal axial development (Ruiz i Altaba and Melton, 1989b; Wright et al.; 1989, Cho et al., 1991).

POU homeobox genes have been implicated in the commitment of cell types to specific lineages (reviewed by Ruvkun and Finney, 1991; Rosenfeld, 1991). For example, the rat Pit-1 gene is expressed in a highly tissue-specific manner and has been shown to be an important regulator of cell fate in the pituitary gland (Ingraham et al., 1988; Bodner et al., 1988). Mouse dwarf locus mutant strains that lack Pit-1 expression lose cell lineages in the anterior pituitary (Li et al., 1990). In C. elegans, mutations in the unc-86 POU domain gene cause loss of a specific subset of neurons (Chalfie et al., 1981; Finney et al., 1988).

Another possible functional role of POU genes may be to maintain cells in an undetermined state in preparation for a future developmental signal. Both the SCIP and Oct3/4 POU genes may mediate their biological effects in this manner. The SCIP gene is involved in inhibiting myelin-specific gene expression and maintaining á proliferative state during the transition between the premyelinating and myelinating phases of Schwann cell differentiation in rats (Monuki et al., 1990). The Oct3/Oct4 gene is expressed in primordial germ cells, unfertilized oocytes, and the inner cell mass of the mouse blastocyst (Okamato et al., 1990; Rosner et al., 1990; Scholer et al., 1990). During mouse embryogenesis, the gene is turned off during differentiation of primitive endoderm and mesoderm tissues (Rosner et al., 1990). Oct3/4 is also expressed in undifferentiated embryonic carcinoma cells and is down regulated by retinoic acid induced differentiation (Okamato et al., 1990; Rosner et al., 1990). When antisense Oct3/4 oligonucleotides were injected into fertilized mouse oocytes, cell division was arrested at the one-cell stage (Rosner et al., 1991). Oct3/4 might function by maintaining a proliferative totipotent state in embryonic cells.

Since gastrulation is a period when cell fates become specified and determined, we looked for POU genes that could regulate these processes. In this study, we cloned the XLPOU91 gene of Xenopus laevis and analysed its expression pattern and regulation. XLPOU91 expression is enriched in the ventroposterior regions of neurula embryos and may play a role in determining cell fate in this region. Some genes are dependent on cell-cell interactions for transcriptional activation (induction) whereas other genes appear to have an intrinsic clock which signals transcription (cell-autonomous activation) regardless of normal cell-cell interactions. While most locally expressed genes are dependent on induction for transcriptional activation, XLPOU91 is activated throughout the embryo in an autonomous manner. Long-range cell-cell interactions lead to gene shut-off in all but the most ventroposterior regions. In contrast to most locally expressed genes, a negative regulatory pathway appears to control localized expression of XLPOU91.

Degenerate oligonucleotides to the conserved WFC POU homeodomain were synthesized (Burglin et al., 1989). Two pooled 23mer oligonucleotides (300 ng, 0.8 μ M) of 1024-fold and 512-fold degeneracy were end-labelled and hybridized to nylon filters according to Burglin et al. (1989). Filters were washed three times for 10 min each in 6 × SSC and 0.05% sodium pyrophosphate at room temperature. For the more stringent washes, filters were washed according to Burglin et al. (1989), or twice for 15 minutes at 42°C in 2 × SSPE, 0.1% SDS (White et al., 1988). Phage lift filters were exposed on films with intensifying screens at – 80°C for 2 – 6 days. For Southern blot analysis of purified phage clones, the first roomtemperature wash suffices for signal specificity; filters were exposed 1-4 hours to films at – 80°C with intensifying screens.

A Xenopus laevis stage 11 cDNA library in the λ Zap vector was screened. After tertiary screens, purified individual phage were converted into plasmids by in vivo excision. Individual plasmids were analysed by Southern blotting using the degenerate oligonucleotide probes as outlined previously. All cDNAs were sequenced by the chain termination method using the T7 DNA polymerase sequencing kit (Pharmacia).

Ovulation of females, in vitro fertilization, UV-irradiation, embryo culture and dissection, and in situ hybridization were carried out as described previously (Frank and Harland, 1991). Embryos were staged according to Nieuwkoop and Faber (1967). The dorsal-anterior index of UV-irradiated embryos was determined at stage 41 according to Kao and Elinson (1988). UV-treated embryo pools used in these experiments developed an average dorsal-anterior index of <0.5. PIF was prepared from the P388D1 cell line as described by Sokol et al. (1990). PIF-treated explants were grown in 1/2 × modified Ringer’s (MR) from stage 8 until stage 11; explants were then rinsed twice and returned to 1/3 × MR.

For embryonic dissociation in calcium/magnesium-free medium (CMFM: 88 mM NaCl, 1 mM KC1, 7.5 mM Tris-HCl pH 7.6,2.4 mM NaHCO₃ and 0.1 mM EDTA), embryos were grown in CMFM from the stages mentioned in the text until midgastrula (stages 11 – 11.5). Embryos were then rinsed twice and returned for growth in 1/3 × MR. Experiments shown here were carried out with the vitelline membrane intact.

Total RNA was isolated and northern analysis was carried out as described previously (Frank and Harland, 1991). Filters were hybridized with the complementary DNA probes of XLPOU91 (0.5 kb PstI-EcoRI in pSK), XMyoD (0.55 kb BamHl-EcoRI, from pSP73-M24/3, Hopwood et al., 1989), EFlafOA kb Pstl-Sstl, Krieg et al., 1989), the muscle-specific cardiac actin, (0.5 kb PVMII-ECORI from pAClOO, Dworkin-Rastl et al., 1986), epidermal-specific cytokeratin (0.6 kb EcoRl from pGem-2 E13, J. Roberts, Seattle; this gene crosshybridizes to the XK81 cytokeratin transcript, Jonas et al., 1985), GS-17 or fibronectin (0.7 kb HindHI-Pstl, from pl7f; 2.6kb EcoRI, from pXFA; Krieg and Melton, 1985). Although EFIα mRNA increases in abundance during development, it is used as a standard for comparing levels of RNA loaded per well at any given stage. Fibronectin and GS-17 are used as standards for the amount of RNA loaded per well in some experiments.

Filters were exposed to films with intensifying screens at – 80°C (XLPOU91, XMyoD, fibronectin and GS-17, 6 hours -2 days; E13 and EFl a, 3 hours; muscle actin, 5 – 24 hours). For quantitation of transcript levels, preflashed films were exposed and scanned on a Hoeffer scanning densitometer with direct integration using GS370 software. Relative levels of expression in whole embryos were calculated in the following manner: the integrated numerical value for a gene to be examined was divided by the integrated value for EFIα. This number was set at 1 for whole embryos to define a correction factor for an individual experiment. This correction factor could then be applied to other samples analysed on the same filter. Values above 1 are interpreted to indicate that a transcript is enriched in these samples relative to whole embryo RNA.

The cDNA sequence of the XLPOU91 POU domain XLPOU91 is a POU box-containing gene cloned from Xenopus laevis embryos. A stage 11 cDNA library was screened using a degenerate oligonucleotide probe to the highly conserved WFC box of the POU homeobox domain (see Materials and methods). As can be seen in Fig. 1, the open reading frame of this cDNA sequence encodes a POU domain protein. A complete sequence for this transcript has been determined independently by M. Jamrich and M. Perry (personal communication, accession no. M60077). This protein shows the highest degree of similarity to the Oct3/4 (POU5) family of mammalian embryonic transcription factors, 74% in the POU domain and 58% in the homeodomain (Okamato et al., 1989; Rosner et al., 1990; Scholer et al., 1990). In addition, there is a striking eight amino acid match in the usually non-conserved linker region between the POU and homeodomains (Fig. 1). The similarity to the other POU groups is also shown in Fig. 1. At the N terminus of the homeodomain, a histidine residue (starred) is found in the basic region which normally contains only arginine and lysine residues (Fig. 1). This basic region, which lies adjacent to the first homeobox helix, has been shown to be critical for DNA binding by these proteins (Ingraham et al., 1990; Treacy et al., 1991). Within a given POU family (between species), the POU boxes have a >85% similarity, and the homeobox domains have >75% similarity. On the basis of the amino acid similarity in both the POU and homeodomains of XLPOU91 in comparison to the other POU classes, XLPOU91 likely represents a new class of embryonic POU domain proteins.

As seen in the developmental northern blot (Fig. 2A), XLPOU91 is detected as a single 3.5 kb transcript. Long exposures show that the gene is activated at the midblastula stage (Fig. 2B, lanes 1 – 4); peak mRNA levels are reached in the midgastrula-early neurula (Fig. 2A, stages 11 – 14) with levels diminished by late neurula-early tailbud (Fig. 2A, stages 20 – 28). At the peak of expression (stage 12), reconstruction experiments with synthetic XLPOU91 RNA measured 7.5 pg of mRNA per embryo, representing approximately 4 × 10⁶ transcripts. The total amount of mRNA in an oocyte is about 50-100 ng (Rosbash and Ford, 1974; Dolecki and Smith, 1979), and if this value does not change significantly during embryogenesis, XLPOU91 represents about 0.02% of gastrula mRNA.

At stage 9, XLPOU91 is initially detected in all regions of the embryo by northern analysis of mRNA from dissected animal, equatorial and vegetal regions (Fig. 2B, lanes 2-4). However, by early neurula stages, XLPOU91 transcripts are enriched in the ventroposterior regions of the embryo as shown by northern analysis of mRNA from dissected regions (Fig. 3A, B). XLPOU91 mRNA isolated from dissected regions was quantified by comparing its abundance to EF lαmRNA (see Materials and methods). Dissection of early neurulae (stage 14 – 15) into six sections along the anterior-posterior and dorsal-ventral axes revealed highly localized expression. Embryos were dissected through the middle of the blastopore and body cavity into a dorsal and ventral half (Fig. 3B). Each half was then cut into three anterior-to-posterior pieces (Fig. 3B). Relatively low levels of transcript are seen in ventral and dorsal anterior regions as well as in the middle and posterior dorsal regions (Fig. 3A). MyoD levels were assayed as a control for dissection accuracy; in sharp contrast to XLPOH91, MyoD transcripts were mainly expressed in the dorsal-middle and dorsal-posterior regions (Fig. 3A, lanes 3, 4). Northern analysis suggests that XLPOU91 is expressed in a graded manner along the ventroposterior axis with mRNA levels being enriched in the most ventroposterior region by 35-fold in comparison to anterior regions (Fig. 3A, B). This enriched expression in ventroposterior regions was detected as early as late gastrula (not shown). In UV-ventralized embryos, there may be more of the ventroposterior cells fated to express XLPOU91 than in normal embryos (Scharf and Gerhart, 1983). Because XLPOU91 expression is enriched in the ventroposterior regions of the embryo, mRNA levels were examined in UV-ventralized embryos. Ventralized gastrula and early neurula stage embryos all expressed over ten times more transcript than controls (Fig. 3C).

The spatial expression of XLPOU91 was examined by whole-mount in situ hybridization. In late neurulae, XLPOU91 transcripts were detected in the most extreme ventroposterior regions (below the blastopore) of the embryo (Fig. 4). XLPOU91 transcripts were not detected in vegetal pole cells of gastrulae or in the yolky endoderm of neurulae by in situ hybridization, while transcripts were clearly detected by northern analysis (Figs 2B, 4). Perhaps, because vegetal/endoderm cells have such a large cytoplasmic volume, the relative concentration of transcript is too low to detect, or perhaps the yolk is somehow interfering in the detection process. Whereas in situ hybridization shows spatial expression in a clear manner, northern analysis of dissected regions enables a more direct quantitation of mRNA levels in different regions of the embryo.

The XLPOU91 gene was initially activated in all three germ layers (Fig. 2B, lanes 1 – 4) and was expressed at higher levels in the yolky endoderm located in the posterior ventral regions of neurulae than the remainder of the total embryonic meso-ectoderm (Fig. 2B, lanes 5 –7). However, it is expressed equally in yolky endoderm and ventral meso-ectoderm, both of which express higher levels than dorsal meso-ectoderm (not shown).

Because XLPOU91 is initially transcribed ubiquitously throughout blastula-stage embryos but regionally expressed in the ventroposterior regions of the neurula embryos, we were interested in determining what types of signals regulated its transcription. Is the XLPOU91 gene dependent on cell-cell interactions for transcriptional activation or is expression activated in a cell autonomous manner? A rigorous test of cell autonomy is to dissociate cells of the embryo. Under these conditions, genes dependent on cell-cell interactions should not be expressed and genes activated according to an autonomous cellular clock should be expressed. Embryos at the 4-to 8-cell stage were grown in calcium/magnesium-free medium (CMFM) with the vitelline membranes intact until stage 11. In these embryos, the blastocoel collapsed and vegetal and animal pole cells intermingle forming a ‘checkerboard’-like embryo (Sargent et al., 1986). XLPOU91 was expressed at normal levels in these perturbed embryos at stage 11 (Fig. 5A, lanes 2 and 3). XLP0U91 was also expressed at control levels when the vitelline membrane was removed at stage 7, and cells were repeatedly dispersed until stage 11 (not shown). As a control, we examined the expression of the inducible muscle marker MyoD; the CMFM treatment significantly inhibited expression of this marker (Fig. 5A, lanes 2 and 3). Muscle gene expression has been shown previously to be sensitive to embryo dissociation. While muscle cells may be induced in the dissociated state, muscle markers can only be expressed when cells are in contact (Gurdon et al., 1984; Sargent et al., 1986). XLPOU91 expression is cell autonomous in the sense that expression occurs without normal cell-cell contacts. In addition, autonomously expressed genes such as GS-17 and cytokeratin were also expressed in dissociated gastrula-stage embryos (not shown).

Epidermal marker genes are autonomously expressed in animal caps without any type of inductive signal (Sargent et al., 1986; Jamrich et al., 1987). Expression of XLPOU91 in animal caps was examined to determine if XLPOU91 is also autonomously expressed in a similar manner. In animal cap explants dissected at stage 8 and grown until stage 11, XLPOU91 was expressed (Fig. 5B, lanes 1, 2). Genes that are induced in the mesoderm are not normally expressed in animal cap explants at the gastrula stage, though treatment of animal cap explants with peptide growth factors (PGFs) such as TGF-β-like (XTC, PIF-Activin) and FGF-like molecules can induce their expression (Green et al., 1990). Treatment of animal caps with PIF concentrations that induce expression of mesodermal markers (Fig. 5B, lanes 3, 9) does not alter XLPOU91 mRNA levels in animal caps at stage 11 (Fig. 5B, lanes 2, 3).

While XLPOU91 transcriptional activation is not dependent on inductive processes, the proper spatial and temporal down-regulation at late neurula stages could be dependent on cell-cell interactions. An example of such a process is the down-regulation of cytokeratin expression in the neural plate (Jamrich et al., 1987). If indeed a pathway exists which down-regulates XLPOU91 expression, its activity may be dependent on cell-cell interactions at some critical time. To examine whether transient dissociation of embryos could disrupt the temporal down-regulation of XLPOU91 expression at late neurula stages, embryos were grown in CMFM with the vitelline membrane intact from stages 4, 6, 7 or 9 until stage 11.5; at this stage Ca²⁺ and Mg²⁺ were returned to the medium (see Materials and methods) and embryos were grown until stage 20. The embryos grown in CMFM from stages 4, 6 or 7 all displayed the typical morphological abnormalities associated with this treatment (Sargent et al., 1986). CMFM treatment between stages 4 and 7 caused 30-fold more expression of XLPOU91 at stage 20 than was found in control embryos (Fig. 6). The elevated levels of mRNA expression are approximately the same as those observed at peak expression (stage 12) in normal embryos. In contrast, dissociation of embryos between stages 9 and 11.5 caused no morphological abnormalities and XLPOU91 was down-regulated properly (not shown).

To examine whether the failure to down-regulate XLPOU91 expression was correlated with a loss of muscle induction, MyoD expression was also examined in the same stage 20 embryos. Embryos that were put in CMFM between stage 4 and stage 11.5, were assayed for MyoD expression at stage 20. Under these conditions, muscle induction was fully inhibited (over 60-fold; Fig. 6, compare lanes 1 and 2). If dissociation was between stage 6 and stage 11.5, muscle induction was inhibited only six-fold (compare MyoD expression in Fig. 6, lanes 3 and 4), and no effect was observed if dissociation lasted from stage 7 to stage 11.5 (compare MyoD expression in Fig. 6, lanes 5 and 6). Neural markers were also expressed in all of the dissociated embryos (not shown) as has been reported by Sato and Sargent (1989). Autonomously expressed genes such as GS-17 and cytokeratin were not overexpressed like XLPOU91 in CMFM-treated neurula embryos. GS-17 underwent its proper temporal shut down (not shown), whereas cytokeratin expression was extinguished in neurulae as has been reported previously for epidermal markers (Fig. 6; Sargent et al., 1986; Jones and Woodland, 1986). The XLPOU91 down-regulatory pathway is therefore different from pathways regulating other induced or autonomously expressed genes.

Because temporal regulation of XLPOU91 expression at late neurula is impaired by disruption of short-range cell contacts in CMFM, we examined whether the proper temporal down-regulation of gene expression in explants is also controlled by longer-range cell interactions. We tested whether ectodermal animal cap cells removed at stage 8 and grown as explants to late neurula could down-regulate transcription in the same manner as normal embryos. XLPOU91 mRNA levels in explants were compared to levels in whole embryos (after normalizing to EFl α, see Materials and methods). XLPOU91 mRNA was 35-fold more abundant in animal cap explants grown to stage 18 than in whole embryos (Fig. 5B, lanes 7,8; Table 1A). At stage 18 most of the cells of the animal cap region in normal embryos would lie in regions that express low amounts of XLPOU91 (Fig. 4). Thus, like transient embryo dissociation, removing animal cap cells from their normal position in the embryo leads to impaired downregulation of XLPOU91 expression. The failure to down-regulate expression is not simply a property of isolated animal caps; dorsal marginal zone (DMZ) explants (dissected at stage 10.25 and grown to stage 20) also expressed the gene 35-fold more abundantly than control embryos (Table IB). The DMZ tissue is not fated for the ventroposterior regions which normally express high levels of the gene (Keller, 1975), and isolated DMZ differentiates into a variety of dorsal mesodermal tissues (Dale and Slack, 1987). To exclude the possibility that dissection per se prevents downregulation of XLPOU91, whole marginal zone explants were taken at stage 8 to examine if the undisrupted marginal zone region can down-regulate XLPOU91 expression. Marginal zone explants were grown until stage 20, and all exhibited relative levels of XLPOU91 expression comparable to control embryos (Table IB). Down-regulation likely requires coherent cellular contacts within the marginal zone during gastrulation.

Inductive processes in embryos are dependent on cell-cell interactions, so the shut down of XLPOU91 expression at late neurula could be mediated by an inducing factor. The peptide growth factor PIF has been shown to be a strong inducer of mesodermal structures in animal cap explants (Sokol et al., 1990); we therefore wished to determine whether PIF could down-regulate the high XLPOU91 levels observed in isolated animal caps. In contrast to the untreated animal caps, the relative abundance of XLPOU91 transcripts in PIF-treated animal caps resembled that of normal embryos, with the proper temporal down-regulation (by 35-fold) of XLPOU91 expression (Fig. 5B, lanes7-9; Table 1A). In the animal cap assay, only explants differentiating muscle (assayed by cardiac actin expression) down-regulated XLPOU91 expression to control levels; PIF levels ten-fold lower as well as bFGF (both of which failed to induce visible morphological change or muscle-specific gene expression) down-regulated expression by five-fold (Fig. 5B, lanes 10, 11; Table 1A). In contrast to XLPOU91, cytokeratin gene expression is shut down more efficiently by bFGF than by PIF in treated animal caps (Fig. 5; Table 1A).

We wanted to determine whether increased XLPOU91 expression in UV-ventralized embryos (Fig. 3C) could be the result of the lack of a dorsalizing inducer in these embryos. Marginal zone explants from ventralized embryos were treated with dorsalizing concentrations of PIF. Like control ventralized embryos, the ventralized marginal zone explants expressed the gene over ten-fold higher than normal embryos and explants, but PIF treatment lowered XLPOU91 expression to control levels (Table IB). PIF treatment also induced muscle differentiation in these ventralized marginal zone explants (as assayed by MyoD expression; Table IB). While XLPOU91 transcriptional activation is a cell autonomous event, proper downregulation and localized expression correlate with dorsalizing inductive processes.

In this work, we report the cloning of a Xenopus POU domain gene -XLPOU91. This gene has the most similarity to the Oct3/4 family of embryonic mammalian POU genes (Okamato et al., 1990; Rosner et al., 1990; Scholer et al., 1990). In addition, XLPOU91 also has an eight amino acid stretch in the usually non-conserved POU-homeo linker region that matches an identical stretch in Oct3/4 (Fig. 1). We and others have also cloned closely related genes of this same family, whose expression is also restricted to early Xenopus development (Frank and Harland, unpublished; M. Perry, personal communication; M. Jamrich, personal communication). Because of the relatively low level of similarity to the other POU classes, these genes likely belong to a new distinct embryonic family of POU proteins.

XLPOU91 expression is enriched over 30-fold in the most ventroposterior regions when compared to anterior regions (Figs 3,4). The gene is also expressed 10-to 20-fold more abundantly in ventralized embryos than in normal embryos (Fig. 3C). The function of XLPOU91 in this region of the embryo is still unknown, and it is tempting to speculate that this gene may control specification of cell fate in this region, similar to other POU genes in vertebrates and invertebrates. XLP0U91 could also maintain an unspecified totipotent state in this region which would later enable differentiation into tailbud tissue.

XLPOU91 is initially turned on throughout the embryo in a cell-autonomous manner at the midblastula transition (Figs 2, 5), but by early neurula the gene is locally expressed. Explant experiments shed light on the possible regulation of this localized expression. When normal cell-cell or regional interactions are disrupted, expression is not down-regulated at late neurula stages (Figs 5,6). Normal temporal regulation of XLPOU91 expression in late neurula occurred only in whole marginal zone explants or in PIF-treated explants; whereas disruption of normal cell contacts by dissection, dissociation or elimination of dorsal morphogenesis all led to impaired down-regulation of XLPOU91 expression (Figs 5, 6; Table IB). The negative regulatory pathway that is sensitive to embryonic dissociation is not identical to the muscle and neural inducing pathways; muscle and neural markers were expressed in the same CMFM-treated embryos which did not temporally down-regulate XLPOU91 (Fig. 6).

Explants of the whole marginal zone region (taken at stage 8) maintain normal temporal expression, suggesting that the whole marginal zone is specified for XLPOU91 down-regulation by the blastula stage (Table IB). Isolated DMZ explants taken at stage 10.25, however, are still not specified to down regulate XLPOU91 expression, since these regions express the gene at higher levels in explants than when compared to normal levels in the embryo (Table IB). During gastrulation, cell-contact-dependent signalling must occur throughout the entire marginal zone to enable proper shut down of XLPOU91.

These results suggest that localized XLPOU91 expression to the ventroposterior region can be viewed as a default pathway. In this model, a gene can be activated ubiquitously throughout the embryo at the midblastula transition, and dorsalizing inducers turn it off. The limit of a specific inducer’s field of inhibition will leave a locally expressed transcript. Thus even regions thought not to arise by way of dorsal induction may be directed to another fate by default.

Epidermal cytokeratin genes have also been shown to undergo negative regulation during Xenopus gastrulation (Jamrich et al., 1987). When the underlying involuting chordomesoderm comes into contact with the presumptive neural plate and neural induction commences, cytokeratin transcription is extinguished in this region (Kintner and Melton, 1987; Jamrich et al., 1987). In addition, treatment of individual animal cap cells with PGFs such as bFGF and XTC (activin) suppresses epidermal-specific gene expression (Symes et al., 1988; Green and Smith, 1990). The data presented here suggest that while these processes appear similar, the regulation of XLPOU91 expression differs from that of the cytokeratin gene. CMFM treatment allows activation of the cytokeratin gene in gastrulae, but the gene turns off prematurely by neurula stages; XLPOU91 expression, in contrast turns on and remains on in these embryos (Fig. 6). In animal cap explants, bFGF treatment minimally affects XLPOU91 expression in sharp contrast to its inhibitory effect on cytokeratin expression (Table 1A). In addition, XLPOU91 is expressed regionally and not restricted to the epidermal ectoderm like cytokeratin (Fig. 4).

Other zygotically expressed genes have also been shown to be activated in a more general manner than their final pattern of expression. The muscle-specific XMyoD gene is initially activated in animal pole, vegetal pole and ventral marginal zone cells in addition to the dorsolateral marginal zone cells fated to make somites (Rupp and Weintraub, 1991; Frank and Harland, 1991). During gastrulation, XMyoD expression is shut down by a gradual process; first, expression ceases in animal and vegetal pole cells (Rupp and Weintraub, 1991) and by early neurula, transcripts are no longer detected in ventral mesoderm (Frank and Harland, 1991). While the control of XLPOU91 expression is markedly different from XMyoD expression, different regulatory mechanisms may exist in the embryo for initially turning on genes in a general manner, with local expression being the result of a shut down process.

In mammalian cells, TGF-β molecules have been shown to mediate some biological effects by inhibiting gene expression (reviewed by Moses et al., 1990). TGF-β strongly inhibits skin kératinocyte proliferation by shutting down Myc transcription. A TGF-β cis-acting element is responsible for this suppression. Experimental evidence suggests that TGF-β inhibition of Myc expression may occur via modification of the retinoblastoma gene product which likely interacts with this element. Similar negative pathways may exist during amphibian development, and PIF-Activin-me-diated down-regulation of XLPOU91 expression in explants could be an example of an embryonic TGF-β inhibitory pathway.

The XLPOU91 gene is regulated by a novel negative pathway which localizes transcription to the ventropos-terior regions of the neurula embryo. Negative inductive pathways may provide viable regulatory networks for directing morphogenesis in vertebrate embryos and may be more widespread than previously thought. Further experiments will determine if specification of the ventroposterior region in Xenopus embryos is more dependent on a lack of a negative signal rather than a definitive positive inducing signal.

We thank Michael Perry and Milan Jamrich for communicating results prior to publication. Xenopus bFGF was generously provided by the Richard Strohman Laboratory. D.F. was supported by the Rothschild Foundation and University of California Cancer Research Coordinating Committee Fellowships. This work was supported by an NIH grant.