Bimodal and graded expression of the Xenopus homeobox gene Xhox3 during embryonic development

Homeobox-containing genes have been found in a variety of vertebrate species (see McGinnis et al. 1984). One reason for the interest in these vertebrate genes is that Drosophila homeobox-containing genes orchestrate decisions about cell fates and embryonic pattern, and it is therefore possible that their vertebrate counterparts have similar functions (see Gehring, 1985). Consistent with this idea, transcriptional studies on Xenopus and mouse homeobox genes have shown that they are expressed during early embryogenesis in a region-specific manner (Muller et al. 1984; Carrasco et al. 1984; Harvey et al. 1986; Carrasco & Malacinski, 1987; Condie & Harland, 1987; Sharpe et al. 1987; Gaunt et al. 1986; Gaunt, 1987; Utset et al. 1987; Toth et al. 1987; Dony & Gruss, 1987; Holland & Hogan, 1988; Breier et al. 1988; Oliver et al. 1988; Graham et al. 1988).

All the Xenopus homeobox genes previously reported contain a homeobox of the Drosophila antennapedia-ultrabithorax class (see above references). We have screened the frog genome with the Drosophila even skipped homeobox which is significantly different from the antp-ubx homeobox class. In this report we describe the characterization of Xhox3, a new frog even skipped-c\ass homeobox gene that was isolated in this screening. Xhox3 encodes a message of about 2-4 kb that is first transcribed at the midblastula transition (MBT) and maximally expressed at the late gastrula-early neurula stage. RNase protection assays of dissected embryos show that Xhox3 is expressed in two distinct periods. The early period is characterized by an anteroposterior (A-P) graded expression in the mesoderm with the highest concentration at the posterior end. The late period begins at the tailbud stage and is characterized by new expression of Xhox3 in the central nervous system (CNS), predominantly in the brain. In addition, Xhox3 expression is detected in the tail bud at this later stage. We have analyzed Xhox3 expression in ultraviolet light (UV) and lithium ion (Li) treated embryos, i.e. in ‘posteriorized’ and ‘anteriorized’ embryos respectively (see Kao & Elinson, 1988). These tests show that anterior cell fates are correlated with low levels of Xhox3 expression and posterior fates with higher levels. We discuss these findings in light of other homeobox gene studies, e.g. Condie & Harland (1987) and propose a role for frog homeobox genes in pattern formation along the A-P axis.

Female Xenopus laevis frogs were primed with 200 units of mare serum (Sigma G4877) and induced to lay eggs 1 or 2 days later by injection of 1000 units of human chorionic gonadotropin (Sigma CG-10). Embryos were fertilized in vitro with testis homogenates and reared in 0T x MBSH (Gurdon, 1977). Developmental stages are according to Nieuwkoop & Faber (1967). Embryos were induced to exogastrulate by stripping the vitelline membrane at the late blastula stage and incubation in high salt (1−1·2 ×MBSH) thereafter. Ultraviolet light irradiation of fertilized eggs and classification of the resulting embryos was as described in Malacinski et al. (1977) and Scharf & Gerhart (1983). Lithium-treated embryos were obtained as described by Kao et al. (1986). In all cases, some embryos were allowed to develop to the tadpole stages in order to assess the effects of the treatment.

Restriction enzymes, SP6 and T7 RNA polymerases, RNa-sin, RQ1 RNase-free DNase, E. coli DNA polymerase I and T4 DNA ligase were obtained from Promega Biotech. RNase A and T1 were from Sigma. ³²P radionucleotides were from Amersham. Random primers were purchased from BRL.

Xenopus genomic (Krieg & Melton, 1985) and neurula stage cDNA (Kintner & Melton, 1987) libraries were screened at low and high stringencies as described by Ruiz i Altaba et al. (1987). The Drosophila even skipped gene probe was a 240 bp SacII-AccI fragment containing the homeobox (Macdonald et al. 1986). All probes were synthesized by random priming protocols as described by Feinberg & Vogelstein (1983). The Xhox3 genomic probe was an EcoRI to Aval fragment 0·9 kb long (see Fig. 1A).

Genomic DNA was extracted from late tadpoles (stage 35). RNA was obtained by proteinase K digestion and phenolchloroform extractions followed by 4M-LIC1 and ethanol precipitations. Poly (A)⁺ RNA was selected by oligo-dT cellulose (type HI, Collaborative Research) chromatography. To analyse the spatial distribution of Xhox3 transcripts, embryos were fixed in 95 % ethanol-5 % glacial acetic acid at 0−4°C for 5 to 10 min and dissected while still in the fixative. The dissected pieces were directly homogenized in proteinase K, and RNA was extracted as described above. Northern and Southern blots and RNA quantification experiments were done as described by Ruiz i Altaba et al. (1987). DNA sequencing was done by the chain termination method of Sanger et al. (1977) using M13 vectors (Messing & Vieira, 1982).

RNase protection assays were performed as described by Melton et al. (1984) and Krieg & Melton (1987) using 10 embryo or part equivalents of total RNA unless otherwise stated. The MBT analysis shown in Fig. 2D required 30 embryo equivalents of poly (A)⁺ RNA per sample. The RNase digestion step was usually done at room temperature for 30−60 min using only RNase A. The Xhox3 probe was synthesized by transcribing AluI cut pcXhox3BA with T7 RNA polymerase. pcXhox3BA is a BglII deletion subclone of pH8, a pSP70 plasmid subclone of the large 1·8 kb Eco RI piece of the Xhox3 cDNA (see Fig. 1A). The EF-Ia probe (Krieg & Melton, 1989) was synthesized by transcribing Pvull cut pSP72-EF-Icr (Kintner & Melton, 1987) with T7 RNA polymerase. The muscle-specific actin probe was synthesized by transcribing Hmdlll cut pSP65-AC100 (Dworkin-Rastl et al. 1986; Kintner & Melton, 1987) with SP6 RNA polymerase. The N-CAM probe was synthesized by transcribing Pvull cut pSP72-N-CAM (Kintner & Melton, 1987) with T7 RNA polymerase.

In situ hybridization experiments were performed as described in Melton (1987) and Kintner & Melton (1987). Antisense RNA probes were synthesized by in vitro transcription in the presence of [³⁵S]UTP. The Xhox3 probe was synthesized by transcribing H/ndlll cut pH8 with T7 RNA polymerase (see above).

The homeobox portion of the Drosophila gene even skipped (eve) (Macdonald et al. 1986) was used to screen a Xenopus genomic library (Krieg & Melton, 1985) under low-stringency conditions. Analysis of a positive recombinant clone showed that the similarity resides in a 7 kb EcoRl fragment. This fragment was inserted into pSP70 (Krieg & Melton, 1987) to produce the plasmid pgXhox3. Sequence analysis of a 300 bp Xenopus restriction fragment that hybridizes to the eve probe reveals the C-terminal end of a homeobox with similarity to that of eve. Moreover, the Xhox3 homeobox is interrupted in the helix 3 region by an intron suggesting that the Xhox3 gene is split (see Fig. 1A).

Northern blots using a Xhox3 genomic probe at high stringency demonstrate the presence of a predominant 2-4kb transcript. This transcript is first detected by Northern blots at the late blastula stage, peaks at the late gastrula stage and persists through the tailbud stage (Fig. 2B). The smaller band seen in the Northern blot could be a weak cross-hybridization signal or the product of the other Xhox3 allele (see below). Because Xhox3 transcripts are relatively abundant at the neurula stage, a Xenopus neuruia stage cDNA library (Kintner & Melton, 1987) was screened with a Xhox3 genomic probe (see Fig. 1A) under high-stringency conditions. One of the cDNA clones obtained, named pcXhox3, appears to contain a full-length cDNA as judged by the size of the insert (∼2·4kb).

Partial sequence analysis of the cDNA clone reveals that the homeobox domain is very similar (88-3%) at the amino acid level to that of eve and less so to that of XhoxlA (Harvey et al. 1986), a Xenopus homeobox gene of the antp-ubx type (see Fig. 1B). Preliminary sequence data show that the similarity between Xhox3 and eve disappears outside the homeobox. Nevertheless, the position of the homeobox is, in both cases, near the Â-terminal end (see Fig. 1 A, Macdonald et al. 1986 and Frasch et al. 1987) unlike most other homeobox genes in flies and vertebrates.

Two sets of bands are detected with different intensities in genomic Southern blots (Fig. 2A). It is therefore possible that there are two different copies of the Xhox3 gene in the Xenopus laevis genome and these may be the only genes with homeobox similarity to eve. Moreover, neither the genomic nor the cDNA clones contain repeated sequences as judged by high (Fig. 2A)

To confirm and extend the Northern blot findings shown in Fig. 2B, a more sensitive method was used to determine the time course of expression of Xhox3 mRNA. Fig. 2C shows the expression profile of Xhox3 as assayed by RNase protection. Xhox3 mRNA is first detected at the late blastula stage (stage 8−9), peaks in early neurula (stage 13−14) and decreases slightly through the tailbud and tadpole stages (stages 26−36). The amount of Xhox3 mRNA present at various stages was determined by Northern blots and RNase protection experiments similar to those shown in Fig. 2. A value of about 3pg or 2·3×10⁶ molecules per embryo was obtained for stage 13, when Xhox3 expression peaks (see Fig. 7).

To determine whether Xhox3 is first transcribed precisely at the MBT, pools of embryos were collected at timed intervals between stages 7 and 10. The expression of the gene coding for the translational factor EF-lα was used to mark the MBT as transcripts of this gene accumulate to very high levels shortly after the MBT (Kintner & Melton, 1987; Krieg & Melton, 1989). As shown in Fig. 2D, these data show that Xhox3 is first expressed at the MBT, when the zygotic genome is transcriptionally activated (Newport & Kirschner, 1982).

The early expression of Xhox3 shows a graded distribution in the mesoderm along the A-P axis Dissections of gastrula and neurula embryos reveal that Xhox3 transcripts are distributed in a graded fashion along the prospective A-P axis. Fig. 3 shows that Xhox3 RNAs are more abundant at the posterior than at the anterior pole during the mid and late gastrula and neurula stages. This regional concentration begins at the early-mid gastrula stage and becomes more pronounced as development proceeds (see Fig. 3, stage 19). Dissection experiments similar to those in Fig. 3 show that expression of Xhox3 in the posterior third of the axis is 5-10 fold higher than in the anterior third (data not shown).

Incubation of embryos in high salt prevents invagination of the mesoderm during gastrulation. Exogastrulated embryos obtained in this way show a clear separation of the three germ layers (Holtfreter, 1933) and this allows for the facile dissection of ectoderm, mesoderm and endoderm (Kintner & Melton, 1987). Dissection of the three germ layers from exogastrulated embryos demonstrates that during this early period Xhox3 transcripts are predominantly if not exclusively found in the mesoderm as shown in Fig. 4. Note that Xhox3 expression coincides with that of muscle-specific actin, a mesodermal marker. Because normal neural development is impaired in exogastrulated embryos (Holtfreter, 1933), it is formally possible that there is some Xhox3 expression in neural ectoderm at these early stages.

In all cases, RNase protection assays were also performed with an EF-lα probe to control for RNA recovery, since EF-I a transcripts are found in all cells at similar levels during early embryogenesis (Krieg & Melton, 1989). Localization of Xhox3 transcripts during the gastrula and neurula stages by in situ hybridization resulted in inconsistent results due to the low abundance of Xhox3 mRNA, the fact that Xhox3 expression is not localized to just a few cells and the large amounts of yolk in the embryo which caused a high background. Nevertheless, preliminary in situ hybridization results are consistent with the data on Xhox3 localization that were obtained by embryo dissections and RNase protections. We have concluded that a more detailed examination of Xhox3 expression at these early stages will require the use of Xhox3 antibodies.

After neurulation, the location of Xhox3 expression changes. By the early tailbud stage (stage 26, see Fig. 5) Xhox3 expression appears at the anterior end. Embryo dissections at this stage reveal a bipolar distribution of Xhox3 transcripts in the A-P axis. This unusual bipolar distribution is maintained through the early tadpole stages (stage 36, Fig. 5).

Dissections along the dorsoventral axis show that Xhox3 mRNA is more concentrated in dorsal regions (Fig. 5). Xhox3 expression at stage 36 is more closely correlated with expression of a neural marker (N-CAM) than with a mesodermal marker (muscle-specific actin) (see Fig. 5). For example, both Xhox3 and N-CAM, but not muscle-specific actin, are expressed in the anteriormost region (head) of stage 36 tadpoles (fraction 1, Fig. 5). The expression detected in the posterior pole and in ventral regions is likely to be due to Xhox3 expression in the growing tail bud (Fig. 5 and see below).

In situ hybridization of tadpole stage embryos shows Xhox3 expression in some areas of the mid and hind brain. A sagittal section of the head area of a tadpole stage embryo hybridized with a labelled Xhox3 antisense RNA probe is shown in Fig. 6. Detection of Xhox3 transcripts by in situ hybridization at this late stage is possible because Xhox3 expression is concentrated in a small area in the CNS where there is little yolk.

In feeding tadpoles (stage 45), a stage when somitogenesis is completed and the gut has differentiated, the posterior pole of Xhox3 expression disappears (Fig. 7 and compare with stage 36 in Fig. 5). For this reason we believe that the posterior expression observed at stage 36 (Fig. 5) is in the growing tail bud. At stage 45, Xhox3 is expressed mainly in the CNS. The dissection results of stage 45 tadpoles shown in Fig. 7 reveal Xhox3 expression primarily in brain (fractions 2 and 3, Fig. 7) and possibly some in the spinal cord (fraction 4, Fig. 7). This is corroborated by the finding that Xhox3 is expressed in dissected tadpole (Fig. 7, lane B) and adult frog (data not shown) brains.

The normal distribution of Xhox3 transcripts along the A-P axis during the gastrula and neurula stages shows that mesodermal cells along the A-P axis express Xhox3 at different levels. Anterior cells express Xhox3 at the lowest levels and the highest concentration of Xhox3 RNA is found in posterior cells. To see whether the level of Xhox3 expression is correlated with different A-P cell fates, the levels of Xhox3 mRNA were quantified in embryos irradiated with u.v. light or treated with lithium ions (Condie & Harland, 1987). These treatments have opposite effects on positional fates of cells in early embryos, u.v. irradiation causes cells to assume more posterior and ventral fates (Grant & Wacaster, 1972; Malacinski et al. 1977; Scharf & Gerhart, 1980; Cooke & Smith, 1987) whereas lithium treatment drives cells towards more anterior and dorsal fates (Masui, 1961; Kao et al. 1987; Breckenridge et al. 1987; Kao & Elinson, 1988). As shown in Fig. 8, Xhox3 mRNA levels dramatically increase in posteriorized u.v.-treated embryos. In contrast, the levels of Xhox3 RNA sharply decrease in anteriorized lithium-treated embryos. The amount of Xhox3 RNA in u.v. and lithium embryos is about 5-fold above and 5-fold below the amount in untreated control embryos (see Fig. 9).

Comparison of the profiles of Xhox3 expression during development in these experimentally perturbed and control embryos shows that the enhanced (u.v.) and diminished (Li) levels of Xhox3 mRNA persist until the tailbud stage (Fig. 9). This change in Xhox3 RNA levels during development coincides with the transition of Xhox3 expression from one period to the next, that is, from a posterior graded distribution in the mesoderm to a bipolar distribution mainly in the anterior CNS and posterior tail bud. These data suggest that only the early expression of Xhox3 in the mesoderm can be directly affected by u.v. or Li.

The frog Xhox3 gene shares some structural features with the Drosophila eve gene. The homeoboxes of Xhox3 and eve are 88 % similar at the amino acid level, and in both cases these are located close to the N-terminus. Nevertheless, we have no reason to believe that the Xhox3 gene is the functional homologue of the Drosophila gene even skipped. To provide a background for functional studies of Xhox3 we have looked at its temporal and spatial expression in developing embryos.

Early expression o/Xhox3

The pattern of Xhox3 expression can be divided into two distinct periods. The first period of expression commences at the MBT and levels of Xhox3 RNA peak at the early neurula stage. At this time, Xhox3 is expressed predominantly in the posterior mesoderm. The results of dissection experiments suggest that Xhox3 expression is not localized at the beginning of gastrulation (not shown). However, these dissections are not sufficiently exact to be certain that the relatively small region on the dorsal side that forms most of the anterior mesoderm has the same levels of Xhox3 mRNA as does the posterior region. In contrast, by the end of gastrulation the transcripts are clearly most concentrated at the posterior pole (near the blastopore). These observations (Fig. 3) suggest that Xhox3 expression gradually declines in cells at the anterior end and at the same time increases in cells at the posterior end. This is an unusual expression pattern because the posterior pole (the circumblastoporal region) contains a different group of cells during each stage of gastrulation. Thus, the posterior expression pattern of Xhox3 is presumably achieved by activating Xhox3 transcription in cells at or near the posterior pole of the gastrulating embryo. Any increase in Xhox3 expression that occurs in prospective anterior mesodermal cells, before they migrate away from the posterior circumblastoporal region, must be followed by a decrease in Xhox3 expression later on. The observed A-P differences in Xhox3 expression could be achieved by a combination of changes in the distribution of cells transcribing Xhox3, changes in the rate of transcription and/or Xhox3 mRNA stability. Interestingly, the posterior specific expression pattern of Xhox36 in the mesoderm of early neurula stages may be achieved in a similar fashion (Condie & Harland, 1987). The posterior specific expression of Xhox36 and the graded posterior expression of Xhox3 in the mesoderm seem to be different from the neural posterior expression of the frog homeobox gene XlHboxó, since the latter one arises as a result of neural induction (Sharpe et al. 1987).

The unipolar expression of Xhox3 observed during gastrulation and neurulation changes to a bipolar pattern in postneurula development. At this time (tailbud stages, see Fig. 5) the two ends of the embryo show the highest levels of Xhox3 mRNA. The data are consistent with the idea that the posterior expression of Xhox3 is due to expression in the tail bud of tailbud and tadpole embryos (Figs 5 and 7), while late in development Xhox3 is transcribed anteriorly only in the CNS (Figs 5, 6 and 7).

The tail bud tip has been postulated to be a special region of active cell division where new mesoderm is induced by the posteriormost part of the organizer and eventually segmented into tail somites in an A-P order (see Spofford, 1948; Bijtel, 1958 and Elsdale & Davidson, 1983). Although it is not yet known which cells in the tail bud express Xhox3, its expression in this special zone suggests a role in patterning and possibly segmentation along the A-P axis (see below) since tail bud mesoderm is progressively patterned during tail growth. Consistent with this idea, the posterior pole of expression of Xhox3 disappears when tail somitogenesis is completed (compare Fig. 7 with stage 36, Fig. 5).

Several Drosophila homeotic genes (for example ubx, antp) have two periods of expression, one starting around the cellular blastoderm stage and a later one in the larval central nervous system (see for example Teugels & Ghysen, 1985 and Akam, 1987 for review). Expression of homeobox genes in the CNS has also been observed in vertebrates including frogs (Carrasco & Malacinski, 1987; Sharpe et al. 1987; Oliver et al. 1988), humans (Mavilio et al. 1986; Simeone et al. 1986; Simeone et al. 1987) and mice (Jackson et al. 1985; Awgulewitsch et al. 1986; Joyner & Martin, 1987; Utset et al. 1987; Fainsod et al. 1987; Wolgemuth et al. 1987; Toth et al. 1987; Dony & Gruss, 1987; Holland & Hogan, 1988; Breier et al. 1988; Graham et al. 1988). In this context it is not surprising that Xhox3 transcripts are found in the brain of tadpoles (Fig. 6) and frogs. Unlike Xhox3, other frog homeobox genes such as XlHBox 1 (or AC I or Xeb 1) (Carrasco & Malacinski, 1987; Oliver et al. 1988), Xhox36 (Condie & Harland, 1987), XlHBox 6 (Sharpe et al. 1987) and most mouse homeobox (Hox) genes are expressed in different regions of the spinal cord and not in the brain with the exception of En-1 and En-2 (Joyner & Martin, 1987; Davis et al. 1988).

The A-P axis of Xenopus is established during gastrulation when axial or dorsal mesoderm moves under the overlying ectoderm from prospective posterior to more anterior positions. During gastrulation, the mesoderm induces overlying ectoderm to form nervous tissue (see Slack, 1983 for review). The A-P differences in the nervous tissue (e.g. brain versus caudal spinal cord) are believed to be a consequence of the differences or regionalization of the underlying inducing mesoderm (Mangold, 1933; Suzuki et al. 1984; see Slack, 1983 for review). For example, at the end of gastrulation the mesoderm in anterior regions will differentiate into head mesoderm and will induce anterior but not posterior neural structures.

What is not known is how axial mesodermal cells attain different A-P properties, that is, how positional information is established and/or interpreted by mesodermal cells in the A-P axis. Whatever mechanism is responsible for establishing the differences in positional information, it is believed to operate during gastrulation (see Gerhart & Keller, 1986 for review and references). Gastrulation-arrested embryos illustrate this point clearly. In these experiments mesodermal cells that would normally form head structures and induce brain stop migrating at trunk levels and then proceed to differentiate into trunk structures and induce spinal cord. This result indicates that mesodermal cells do not have their A-P differences determined until gastrulation movements are complete (Gerhart et al. 1984). Additional evidence is provided by the transplantation experiments pioneered by Mangold and Spemann. The region dorsal to the blastopore, the organizer, is capable of inducing a secondary axis when transplanted to the ventral region of a host embryo (Spemann & Mangold, 1924; Spemann, 1938). The extent or completeness of the secondary A-P axis depends on the stage of the donor embryo (Spemann, 1938). In such grafts, the transplanted tissue contributes to notochord and somites but also influences surrounding mesodermal cells to change their fates, for example from prospective blood to axial muscle (Smith & Slack, 1983; Gimlich & Cooke, 1983; Jacobson, 1984). This suggests that the organizer is a source of an axis-forming property or signal that is not cell-autonomous (Smith & Slack, 1983). Since mesodermal cells have different A-P values which are directly related to the extent and/or timing of their migration it is possible that the organizer acts as a source of a signal that establishes positional values in the A-P axis.

If the posterior pole of the embryo or circumblastoporal region (where the organizer region is found operationally), acts as a source of a morphogenetic signal that establishes different A-P cell fates one might expect to find genes whose expression respond to this signal. Interestingly, the expression of the Xenopus homeobox genes Xhox3 (this work) and Xhox36 (Condie & Harland, 1987) is first detected during the late blastula-early gastrula stages and is region-specific along the A-P axis with the highest concentration in the posterior mesoderm. The expression of Xhox3 in the tail bud further suggests a relationship between the organizer region and Xhox3 expression since the posteriormost part of the organizer induces posterior neural plate to become tail mesoderm (Spofford, 1948; Bijtel, 1958; Woodland & Jones, 1989).

Some treatments of the early Xenopus embryo, including u.v. irradiation and Li ion uptake, are known to cause a respecification of the mesodermal pattern which can be interpreted as changes in the specification of A-P fates. For example, in severely affected embryos the character of the invaginating mesoderm changes so that it behaves as fully anterior after exposure to lithium (Kao et al. 1987; Cooke & Smith, 1988) or fully posterior after u.v. irradiation (Cooke & Smith, 1987). Interestingly, the expression of Xhox3 (Fig. 7) and Xhox36 (Condie & Harland, 1987) changes dramatically after u.v. or Li treatment. Since A-P mesodermal cell fates are changed in u.v. and Li treated embryos, the correlation of a graded series of different A-P patterns, from fully anterior (extreme Li) to fully posterior (extreme u.v.) with increasing levels of Xhox3 expression is intriguing. This points to the idea that the A-P graded expression of Xhox3 in the mesoderm is causally related to different A-P cell fates, e.g. low expression with anterior cell fates and high expression with posterior cell fates (see Figs 3 and 7). Refinement of specific cell fates may require the local expression of other homeobox genes such as Xhox36 (Condie & Harland, 1987) and XlHBoxl (Oliver et al. 1988).

A possible role for homeobox genes in the establishment of mesodermal A-P cell fates follows from the results summarized above and the following considerations. Suppose that the posterior pole or circumblastoporal region, where the organizer region is found, acts as a source of a morphogenetic signal that controls the expression of different Xhox genes in different A-P mesodermal regions. Different concentrations of this signal along the axial mesoderm during gastrulation may be responsible for the activation and/or spatial restriction of homeobox gene expression. Thus, we propose that the posterior pole (the organizer) directs the establishment of A-P cell fates by generating a signal(s) with a graded distribution that regulates Xhox expression in the mesoderm. Homeobox genes may act early in lineage precursor cells to restrict the available fates to its daughters, thereby creating regional differences. This proposal is similar to Wolpert’s theoretical model of positional information (1969) involving a diffusible property (the signal of the posterior pole) and local activators/repressors (homeobox gene products). The proposal of a graded distribution of the signal of the organizer is similar to previously proposed theories (see Toivonen & Saxen, 1962).

In fact, the data suggest that homeobox gene expression in the mesoderm is initially uniform in the A-P axis and is spatially restricted during gastrulation. Interestingly, some Drosophila pattern forming genes are first expressed evenly and then restricted along the A-P axis by the action of polar organizer centers (see Akam, 1987; Nusslein-Volhard et al. 1987; Gaul & Jackie, 1987; Tautz, 1988). Studies on the expression of some mouse homeobox genes indicates that they are also expressed in a restricted manner in the A-P axis during gastrulation, notably in mesodermal regions (Gaunt et al. 1986; Gaunt, 1987; Dony & Gruss, 1987; Holland & Hogan, 1988). It is therefore possible that the proposed role for frog homeobox genes may be valid for vertebrates in general. The early function of establishing cell identities along the A-P axis may be common to some homeobox genes in all species, as homeobox genes have been isolated from totally unsegmented organisms (Dolecki et al. 1986).

We are grateful to P. Macdonald and G. Struhl for the generous gift of the Drosophila eve homeobox probe. We thank P. Macdonald, G. Struhl, P. Krieg, C. Kintner, J. C. Smith, R. Harvey, D. Tannahill, J. Yisraeli, K. Mowry and other members of this laboratory for discussion or comments on the manuscript. This work was supported by a grant from the NIH.