Expression of the LIM class homeobox gene Xlim-1 in pronephros and CNS cell lineages of Xenopus embryos is affected by retinoic acid and exogastrulation

Homeobox genes are involved in the control of various steps in embryogenesis such as body axis determination, regional specification, and tissue or cell type specification. In addition, some homeobox genes are constitutively expressed in adult tissues where they may have primarily maintenance functions. By amino acid sequence similarities in the homeodomain, homeobox genes are divided into several subfamilies, each of which may have distinct functions during development or in the adult. For example, in vertebrates, members of Hox gene clusters are expressed in a spatiotemporal manner implying that they specify identities along the anteroposterior axis posterior to the midbrain (McGinnis and Krumlauf, 1992; Scott, 1993), whereas Otx, Emx, and Dlx/dll genes may be involved in the regional specification in the midbrain and forebrain (Simeone et al., 1992, 1993; Price et al., 1991; Papalopulu and Kintner, 1993; Dirksen et al., 1993). Certain homeobox gene subfamilies are identified by conserved domains in addition to the homeodomain, as exemplified by Pax (Gruss and Walther, 1992), POU (Schöler, 1991), and LIM class genes.

The LIM class of homeobox genes, initially identified by sequence homology between the C. elegans genes mec-3 and lin-11 and the rat DNA binding factor Isl-1, is defined by the association of two copies of a cysteine-rich domain with a homeodomain (Way and Chalfie, 1988; Freyd et al., 1990; Karlsson et al., 1990). The LIM family now includes the Xenopus genes, Xlim-1, Xlim-2, Xlim-3 and the mouse homolog lim-1 (Taira et al., 1992, 1993; Fujii et al., 1994), the Drosophila gene apterous (Cohen et al., 1992; Bourgouin et al., 1992), and the rodent genes, lmx-1 (German et al., 1992) and LH-2 (Xu et al., 1993). The C. elegans LIM genes are required for the development of specific cell lineages, mec-3 for mechanosensory neurons (Chalfie and Sulston, 1981; Way and Chalfie, 1988) and lin-11 for vulval cells (Ferguson et al., 1987; Freyd et al., 1990). The apterous gene functions in wing differentiation and development of a subset of embryonic muscles (Cohen et al., 1992; Bourgouin et al., 1992). Taken together these observations suggest that LIM class homeobox genes in invertebrates are concerned with the specification and differentiation of specialized cell types. Isl-1 and Xlim-3 are expressed in certain subsets of neurons in the central nervous system as well as the pituitary and pineal glands, implying a similar function for LIM genes in vertebrate development (Thor et al., 1991; Taira et al., 1993).

While several LIM genes arise in cell type-specific patterns, Xlim-1 is expressed at the gastrula stage in the dorsal blastopore lip and the dorsal mesoderm, a region known as the Spemann organizer. Since this region is responsible for establishing the dorsal axis, a global role in patterning rather than in cell type specification may be suggested as the function of Xlim-1 during gastrulation. A notable feature of the Xlim-1 gene is that retinoic acid (RA) enhances its expression in animal explants (Taira et al., 1992), as distinct from the response of other dorsally expressed genes such as goosecoid (gsc; Cho et al., 1991) and XFKH1/XFD-1/Pintallavis (Dirksen and Jamrich, 1992; Knöchel et al., 1992; Ruiz i Altaba and Jessell, 1992).

As reported previously, Xlim-1 continues to be expressed from the early gastrula to later stage embryos in a biphasic manner: the level of expression peaks at late gastrula, then decreases, and increases again at tailbud stage; Xlim-1 RNA was also detected in several adult tissues including the brain, eye, kidney, and testis (Taira et al., 1992). To examine the role of Xlim-1 in development, we analyzed its expression pattern by whole-mount in situ hybridization and found that Xlim-1 is expressed in three distinct cell lineages, those leading to (i) notochord, (ii) pronephros, and (iii) certain subsets of cells in the central nervous system (CNS). The modification of the Xlim-1 expression pattern under the influence of RA suggested a relationship between RA-induced malformations and enhanced Xlim-1 expression. Since Xlim-1 constitutes a useful marker gene for the lineages in which it occurs we examined whether its expression in these lineages depends on ectoderm-mesoderm interactions during and after gastrulation. Using exogastrulae, we found that Xlim-1 expression in the pronephric and neural lineages is strongly inhibited by the suppression of normal germ layer interactions.

Blastula embryos (stages 8 – 9) were treated with 10⁻⁵ M all-trans RA in 0.25× MMR (Peng, 1991) for 30 minutes, washed three times with 25 ml of 0.25× MMR, and transferrsed into 5 ml of 0.25× MMR; usually 30-40 embryos were incubated in this volume. Embryos were fixed at stage 40 and phenotypes were analyzed using the dorsoanterior index (DAI, Kao and Elinson, 1988). We examined the effect of the number of washing steps (0– 3) and the volume (5 or 25 ml) of solution in which embryos were incubated after RA treatment. The number of washes did not greatly affect the DAI, estimated here as the arithmetic mean of the values scored in a group of embryos: we obtained an average DAI of 1.3 with no wash versus a 1.8 after three washes. In contrast, the volume of culture medium had great influence on the outcome: the average DAI was 1.3 after incubation in 5 ml versus 3.1 in 25 ml of medium. This observation suggests that RA continues to act after the 30 minute treatment, with the degree of sub-sequent dilution in the culture medium a critical variable.

Histochemical analysis was done by standard procedures.

RNA was extracted and analyzed as described by Taira et al. (1992). DNAs for probes are XFKH1 (Dirksen and Jamrich, 1992), gsc (Cho et al., 1991), and Xbra (Smith et al., 1991).

The vitelline membrane was manually removed from stage 6.5–7 embryos and they were incubated in 1× MMR on 1% agarose made up in 1× MMR. In some protocols, 1.1-1.4× MMR or NAM (Peng, 1991) is used. These solutions caused dissociation of embryos when used on top of 1% agarose/1× MMR, suggesting that the above protocols may be designed to be used with an agarose base made up in low-salt solution. If so, the volume ratio of solution to agarose would be critical, which is not the case in our version.

Whole-mount in situ hybridization was performed according to Harland (1991) with albino embryos and Xlim-1 digoxigenin RNA probe, as described (Taira et al., 1992).

For immunostaining, embryos were fixed as above and preserved at −20°C in methanol. After rehydration, embryos were incubated in PBT (PBS, 2 mg/ml BSA, 0.1% Triton X-100) plus 10% sheep serum for more than 1 hour at room temperature, and incubated with diluted primary antibody in PBT plus 10% sheep serum overnight at 4°C. After four washes with PBT for 2 hours each at room temperature, samples were incubated with secondary antibody in PBT plus 10% sheep serum overnight at 4°C. Embryos were washed in the same way, and chromogenic reaction was performed either with DAB (3,3′-diaminobenzidine tetrahydrochloride; Sigma) for peroxidase-conjugated secondary antibodies (Amersham), or NBT (nitro blue tetrazolium; Promega) and BCIP (5-bromo-4-chloro-indolyl-phosphate; Promega) for alkaline phosphatase-conjugated secondary antibodies (Boehringer Mannheim) according to manufacturer’s recommendations. Primary antibodies and dilutions used are Tor 70 (1:500) for notochord (Bolce et al., 1992), anti-NCAM 4d (1:20) and anti-HNK-1 (Sigma, 1:500) for neural tissue.

In a previous paper, we showed that the zygotic expression of Xlim-1 starts at initial gastrula, peaks at late gastrula, then decreases and increases again at tailbud stages; early expression is predominantly in the dorsal mesoderm, i.e., the organizer region of the gastrula (Taira et al., 1992). To examine the localization of Xlim-1 mRNA during later development we performed further whole-mount in situ hybridization, and found that Xlim-1 is expressed in three lineages, leading to (i) the notochord, (ii) the pronephros, and (iii) certain regions in the CNS (Figs 1, 3). We describe the three lineages separately below, concluding that the first phase of Xlim-1 expression (Taira et al., 1992) is associated with the involution of the dorsal mesoderm and formation of the notochord, while the second phase is associated with the development of the pronephros and the CNS.

As dorsal mesoderm involutes, it organizes itself into prechordal plate and notochord; Xlim-1 RNA accumulates in both of these regions (Fig. 1A,B). During the subsequent formation of the neural tube, Xlim-1 expression disappears gradually from prechordal (head) mesoderm and then from the notochord in an anterior-to-posterior gradient (Fig. 1A,B,D-F). While most of the signal has disappeared from the notochord by mid neurula stages (Fig. 1E,F), Xlim-1 RNA persists in the most caudal region (indicated by arrows in Fig. 1K,L) where the notochord is still developing, implying that Xlim-1 expression is closely associated with notochord formation from the gastrula stage onward.

During gastrulation Xlim-1 is expressed primarily in the dorsal mesoderm, but signal gradually appears in lateral mesoderm where it is clearly visible by late gastrula (stage 13; Fig. 1B, black arrow), evolving into a stripe that encircles the embryo (except in the neural plate) at early neurula (stage 14; Fig. 1C). That this staining is primarily mesodermal rather than ectodermal is illustrated in the left panel of Fig. 3G. During subsequent development, Xlim-1 RNA appears to condense into a dorsolateral domain that eventually becomes distinct as the developing pronephros and pronephric duct (Fig. 1E,F,H,K). Sections of embryos stained by whole-mount in situ hybridization confirmed the intense labeling of the pronephros at tailbud stages (Fig. 1J). While the intensity of staining decreases later, it persists through embryogenesis (unpublished observations), and adult kidney still contains Xlim-1 mRNA (Taira et al., 1992).

In the early neurula (stage 14), two rows of Xlim-1-expressing cells are visible in the neural plate about midway along the anteroposterior axis (Fig. 1D, white arrow). These rows subsequently expand anteriorly and posteriorly, concomitant with the disappearance of the signal from the notochord (Fig. 1E,F). Xlim-1 expression in the neuroectoderm adjacent to the stained mesoderm can be traced back to earlier stages; staining could be detected at stage 12 (not shown), and is apparent from stage 12.5 onward (indicated by white arrows in Fig. 1B and open triangles in Fig. 3G, left panel). Thus the lateral staining in the early neurula consists of expression in a coordinate domain of both mesoderm and ectoderm. Regional rather than germ layer-specific expression of a homeobox gene has been shown previously in Xenopus embryos for XlHbox1/HoxC6 (Oliver et al., 1988). Fig. 1G, showing cross-sectional views obtained by placing the focal plane at the middle of the embryo, illustrates Xlim-1 expression in the forming neural tube, where it appears limited to the deep cell layer.

At tailbud stages, neural expression of Xlim-1 is seen in the hindbrain and spinal cord (Fig. 1H-K). A cross section through an embryo that had first been hybridized with Xlim-1 shows signal in the lateral region of the spinal cord (Fig. 1J), corresponding to a similar region of expression in the folding neural plate (Fig. 1G). In the hindbrain, there is a gap in the signal (white arrows in Fig. 1H,I,K) which probably corresponds to rhombomere 4, as judged by our in situ hybridization results (not shown) with a mixed probe of Xlim-1 and Krox-20, which labels rhombomeres 3 and 5 (Bradley et al., 1992). This interpretation is consistent with the fact that the gap at later stages corresponds to the position of the otic vesicle, which arises next to rhombomere 4 in Xenopus (Ruiz i Altaba and Jessell, 1991b). In the midbrain Xlim-1 RNA is first detected by stage 22 (open arrow in Fig. 1I), and increases in intensity subsequently (open arrow in Fig. 1K). In later development, Xlim-1 expression also becomes apparent in the forebrain (see Fig. 3J), leading to a complex final pattern, as will be discussed elsewhere.

As reported previously, RA induces Xlim-1 in animal explants and has a synergistic effect with activin A, and also enhances Xlim-1 expression in intact gastrula embryos (Taira et al., 1992). Treatment of embryos with RA, under the conditions described in Materials and Methods, led to the anterior and posterior malformations described previously (Durston et al., 1989; Sive et al., 1990; Ruiz i Altaba and Jessell, 1991a,b; Papalopulu et al., 1991a; Sharpe, 1991), as illustrated in Fig. 2A. The transverse section in Fig. 2B shows that the major axial system, including the notochord, formed in RA-treated embryos. The expression of Xlim-1 in gastrula embryos treated with RA (Taira et al., 1992) was compared by northern blotting to the expression of three other transcription factor-encoding genes. As shown in Fig. 2C, Xlim-1 expression is enhanced by RA, gsc is strongly inhibited, and XFKH1/XFD-1/pintallavis and Xbra are largely unchanged. Thus, the expression of Xlim-1 is affected by RA in a different way as that of the other three genes in spite of the overlapping expression domains of these genes at the gastrula stage.

A possible relationship between RA-induced phenotypes and changes in Xlim-1 mRNA localization was studied by in situ hybridization. RA treatment leads to enhanced and expanded expression in all three Xlim-1-positive lineages. Although Xlim-1 mRNA could be detected initially in late blastula embryos, enhanced expression in RA-treated embryos was first apparent shortly after the beginning of gastrulation (stage 10.25; Fig. 3A). This enhancement is more pronounced at stage 11 (Fig. 3B,C); in the normal gastrula, Xlim-1 mRNA is abundant in the dorsal mesoderm and present at a low level laterally, while the RA-treated gastrula shows signal throughout the mesodermal mantle with a dorsoventral gradient of intensity. At this stage, some signal was also detected in the dorsal (preneural) ectoderm of the RA-treated embryo, but not in the control (Fig. 3C). Thus, by midgastrula, RA-treated embryos have enhanced expression in all three Xlim-1-positive lineages. However, there is no generalized ectopic expression, as illustrated by the absence of Xlim-1 mRNA-positive cells in the somitic domains of treated embryos (Fig. 3D,E).

By focusing on a sagittal plane of late gastrula embryos (Fig.3F) it is apparent that, in the RA-treated embryo, the Xlim-1 mRNA-positive region has expanded anteriorly into what is most likely prechordal mesoderm (black triangle), and into ventroposterior mesoderm, which is stained less intensely. In later stages (Fig. 3H-J), notochordal expression largely disappears from the RA-treated embryos as it does from the controls. Xlim-1 expression in lateral mesoderm is also enhanced and expanded, as seen in cross-sectional views at late gastrula (Fig. 3G) where staining extends all the way to the ventral edge of the embryo. As the Xlim-1-positive domain in the lateral mesoderm condenses into the pronephros in the control, it does also in the RA-treated embryo, yielding a substantially larger structure (Fig. 3H). In some cases, the normally separated right and left pronephric regions are joined by an Xlim-1-positive thread of tissue (Fig. 3I). These effects on pronephros development are consistent with the view that RA treatment results in posteriorization of anterior regions (Durston et al., 1989; Sive et al., 1990; Ruiz i Altaba and Jessell, 1991a,b; Papalop-ulu et al., 1991a). The extreme case, in which both sides of the pronephros are connected, may result from the recruitment of cells into the pronephric lineage, consistent with the observation that RA, in conjunction with activin, induces pronephric tubules in animal explants (Moriya et al., 1993). Since Xlim-1 is also expressed in the ventroposterior mesoderm in RA-treated embryos (Fig. 3F), caudal regions might also be transformed to lateral mesoderm, contributing to the enlargement of the pronephros in the caudal direction (Fig. 3H), and to the truncation of the tail region (Papalopulu et al., 1991a; Fig. 2A). As in lateral mesoderm, RA-treated embryos display earlier, enhanced, and expanded Xlim-1 expression in the neuroectoderm. In Fig. 3C the control embryo at early gastrula shows no ectodermal signal (note clear area overlying the stained dorsal mesoderm), while the RA-treated embryo shows some ectodermal staining. At late gastrula (Fig. 3G), ectodermal staining has arisen in the control (arrowheads), but is much expanded and more intense in the treated example. The dorsal view of an early neurula (Fig. 3E) also indicates that the area of neuroectodermal expression (white arrow) is more anterior in the RA-treated embryo than in the control (Fig. 1B,D). This may be the earliest indication of the situation at later stages where Xlim-1 is expressed in the anteriormost tip of the CNS (see below). In the tailbud stage (Fig. 3H), the width between the two rows of stained neural cells is greater than in the control, probably related to the thickened neural fold in RA-treated embryos (Ruiz i Altaba and Jessell, 1991b; Papalopulu et al., 1991a; our observation). In the late tailbud stage (Fig. 3J), the RA-treated embryo has an enlarged hindbrain-like region constituting the anteriormost portion of its CNS. This region contains a relatively high concentration of Xlim-1 RNA, as does the hindbrain in the normal embryo. Thus, the expression patterns of Xlim-1 in both lateral mesoderm and neuroectoderm of RA-treated embryos are consistent with the phenotypic effects that are observed.

Cell interactions in tissue specification are continuous processes. While the early events of mesoderm and neural induction specify these regions in a global sense, it is clear that subsequent complex intercellular signaling is required to generate individual tissues and cell types. Since Xlim-1 appears to be a useful marker for cells of the notochord, pronephros and parts of the CNS, we examined the dependence of its expression in these lineages on cellular interactions through the use of exogastrulation experiments. After treatment of embryos with high salt solution (see Materials and Methods), we selected exogastrulae with good separation of ectoderm from endomesoderm for whole-mount in situ hybridization with Xlim-1, and for staining with the notochord-specific antibody Tor 70 or with NCAM and HNK-1 antibodies as neural markers. Stages mentioned in this section refer to the stage of normal sibling embryos. During processing some of the exogastrulae separated at the junction of the ectoderm and the endomesoderm, e.g. in Fig. 4A-D.

At stage 13, Xlim-1 mRNA distinctly stained the notochord of the endomesoderm in most of the exogastrulae examined (Fig. 4A). Staining was most intense at the distal end, i.e. the anterior, as it is in normal embryos (Figs 1A, 4A). In stage 11exogastrulae, Xlim-1 expression was seen on one side, presumably the dorsal side (data not shown). Together these results suggest that the expression of Xlim-1 in dorsal mesoderm is not greatly affected by exogastrulation.

In contrast, the lateral mesoderm and neuroectoderm lineages expressed Xlim-1 at much lower levels in exogastrulae than in controls. At stage 17 (Fig. 4B), the endomesodermal halves still showed signal in the posterior notochord (thin arrows), as does the control (N), but lateral mesodermal staining (thick arrows) was quite weak. This situation is more apparent at stage 22/23 (Fig. 4C). Endomesodermal halves show traces of remaining notochord staining, but only one of the five individuals shown has a faint area of staining that may represent lateral mesoderm/pronephros (thick arrow). In the ectodermal halves, two of five show faint staining that may correspond to neural expression (open triangles). The ectodermal fragments c and d actually stain at the junction points, but these regions correspond to posterior notochord, as seen in the panel doubly stained with Tor 70 (thin arrows, Fig. 4C, lower panel). In contrast to the exogastrulae, the control embryo hybridized in the same vial (N), showed clear staining in the pronephros and nervous system.

Fig. 4D shows the results of a double staining experiment in which exogastrulation appears to have been less complete than in the previous experiment and where, as a consequence, ectoderm and endomesoderm did not separate during preparation. Xlim-1 staining was seen at the junction, corresponding to the posterior tip of the notochord as witnessed by Tor 70 staining (thin arrows). Only a trace of staining in lateral mesoderm was present in one of six exogastrulae (thick arrow), in contrast to the intense staining of the pronephros in the control (N). Ectodermal staining was seen in all six of these exogastrulae (triangles); however, its intensity is much weaker than in the control and appears to be proportional to the proximity of the notochord to the ectoderm in each individual case. For example, compare the middle embryo in the top row, which has comparatively high ectodermal Xlim-1 expression and in which the notochord appears to enter the ectoderm, with the embryo at the lower left where the notochord ends within the endomesoderm and Xlim-1 staining is faint.

Exogastrulae express certain neural markers including NCAM and HNK-1 (Kintner and Melton, 1987; Ruiz i Altaba, 1992). In agreement with these earlier observations, we detected NCAM by immunocytochemistry in all of the exogastrulae that were examined, forming a discrete tube-like structure (Fig. 4E). The expression of the HNK-1 antigen was more variable in this material, with about half the exogastrulae being positive for this marker (Fig. 4F). Thus the exogastrulae used in these experiments express certain neural markers in the same way as reported previously (Kintner and Melton, 1987; Ruiz i Altaba, 1992), and the reduction in the expression of Xlim-1 in the neuroectoderm points to a distinct behavior of this gene compared to the NCAM-encoding gene.

Since RA has been reported to enhance the expression of the neural markers HoxB9/XlHbox6 and XIF6 in neuroectoderm-mesoderm explants (Sharpe, 1991) and enhances Xlim-1 expression in the embryos, we tested the effect of RA on this gene in exogastrulae. We observed no increase in Xlim-1 staining in ectoderm and lateral mesoderm, although increased expression arose in the dorsal mesoderm (data not shown). From these results we conclude that Xlim-1 expression in neural tissue and pronephros requires interactions between ectoderm and mesoderm. As further considered in the Discussion, planar neural induction alone may not be sufficient to induce Xlim-1 in neural tissue.

In whole-embryo RNA blots, Xlim-1 exhibits a biphasic temporal expression pattern (Taira et al., 1992); we show here that this pattern arises as a composite of the localization of Xlim-1 mRNA in three lineages in which it is regulated separately. In the first phase, Xlim-1 mRNA accumulates mostly in dorsal mesoderm evolving into prechordal mesoderm and notochord. In the second phase, this RNA appears in the lateral mesoderm and the neural plate, eventually concentrating in the pronephros and the CNS.

A possible role of Xlim-1 in axis establishment is suggested by its initial expression in the dorsal mesoderm and notochord, the organizer region of the gastrula embryo (for review see Gilbert and Saxén, 1993). During neurula stages, Xlim-1 RNA disappears from the notochord in a cranio-caudal gradient, but persists in the caudalmost tip where the notochord continues to develop as part of tail elongation. Thus Xlim-1 expression is associated with notochord formation and growth, and possibly, its function. Disappearance of Xlim-1 mRNA seems to correlate with the functional degeneration of the notochord, such as loss of its potency in motor neuron induction (Yamada et al., 1993). Interestingly, the Xlim-1 expression pattern in the notochord resembles that of integrin α3: this gene is expressed first in dorsal mesoderm, then in the forming notochord, and later becomes restricted to its caudal region (Whittaker and DeSimone, 1993).

Xlim-1 expression leading to the pronephros lineage, begins in the lateral mesoderm of the gastrula, condenses to the pronephric region during neurulation, and continues at least through tailbud stages. According to Nieuwkoop and Faber (1967), the first morphological indication of the pronephros and Wolffian duct anlagen is observed at stages 21 and 23, respectively. However, the condensation of Xlim-1 mRNA-containing cells in the presumptive pronephric region occurs at stage 15 to 18 (Fig. 1E,F), preceding morphological differentiation. In the adult the pronephros has been replaced by the mesonephros, which forms the definitive kidney in amphibians, which also expresses Xlim-1 (Taira et al., 1992). In overall pattern but in a different context, Xlim-1 expression in the nephric lineage is similar to that of the related gene Xlim-3 in the pituitary lineage: Xlim-3 mRNA is first detectable at the anterior edge of the neural plate prior to the formation of the stomodeal-hypophyseal (pituitary) anlage, becomes associated with the forming pituitary, and then persists into the adult (Taira et al., 1993).

Xlim-1 expression may be compared to the expression of Pax-2 which has been studied in kidney development of the mouse (Dressler et al., 1990) and zebrafish (Püschel et al., 1992). The Pax-2 gene product is first detected as the pronephros forms but, unlike Xlim-1, it disappears after completion of morphogenesis (Dressler et al., 1990; Püschel et al., 1992), and its expression is associated with and required for epithelium-mesenchyme interactions (Dressler and Douglas, 1992; Rothenpieler and Dressler, 1993).

Neural expression of Xlim-1 is first detectable at late gastrula (stage 12) in the ectoderm adjacent to the region of Xlim-1-positive lateral mesoderm. At this stage the neural plate is not morphologically distinguished from the epidermis, but neuron precursors are already present, undergo their final round of DNA synthesis between stages 11 and 15, and form the primary neurons after one cell division at stage 13 to 16 (Hartenstein, 1989). The neural plate is composed of two layers of cells: the superficial, or epithelial, and the deep, or sensorial layer. Xlim-1 mRNA is first expressed in both layers (Fig. 3G), but during neural tube formation appears localized in the sensorial layer (Fig. 1G) where primary neurons form. If Xlim-1 expression is associated with the differentiation of certain neurons, waves of differentiation would sweep from the middle towards the anterior and then the posterior as the range of Xlim-1-positive cells expands (Fig. 1E,F). Interestingly, the expansion of neural expression is concomitant with the disappearance of Xlim-1 mRNA from the notochord (Fig. 1E), possibly implying that neural expression of Xlim-1 correlates with a functional change in the notochord that is visualized by the disappearance of this mRNA. At tailbud stages, additional Xlim-1 expression appears in the midbrain and later in the forebrain, becomes important in the diencephalon and telencephalon at tadpole stages (unpublished results), and is found in the adult brain (Taira et al., 1992).

It is notable that Xlim-1 expression in each of these three lineages starts before morphogenesis appears, suggesting that Xlim-1 may play a role in cell type specification in these tissues.

Dorsal mesoderm generates anteroposterior polarity while forming the dorsal axis during its involution in gastrulation. Therefore, anterior deletions elicited by UV irradiation can be explained by the ventralization of the dorsal mesoderm (Gerhart et al., 1989). Likewise, anterior malformations caused by RA are due, at least in part, to the ventroposteriorization of the dorsal mesoderm (Durston et al., 1989; Sive et al., 1990; Ruiz i Altaba and Jessell, 1991a; Papalopulu et al., 1991a). This view is consistent with the observation that the expression of the dorsal marker gsc is reduced by RA (Cho et al., 1991; Fig. 2). Xlim-1 also behaves as a dorsal gene in terms of its expression pattern, inhibition by UV ventralization (unpublished results), and enhancement by LiCl dorsalization (Tadano et al., 1993), but its response to RA is distinct. This difference may be explained by noting that the RA effect on the expression of Xlim-1, as well as Pintallavis, is abrogated by the protein synthesis inhibitor cycloheximide (CHX) (Ruiz i Altaba and Jessell, 1992; Tadano et al., 1993), suggesting indirect action. In contrast, RA has been shown to induce directly the HoxA1/Hox1.6 and HoxD4/Hox4.2 genes (Langston and Gudas, 1992; Pöpperl and Featherstone, 1993; see Scott, 1993, for Hox gene nomenclature). In Xenopus embryos, RA enhances the expression of many Hox genes (Papalopulu et al., 1991b; Sive and Cheng, 1991; Leroy and De Robertis, 1992; Dekker et al., 1992; López and Carrasco, 1992). It may be considered whether Xlim-1 is regulated by a combination of RA-inducible Hox genes (see Boncinelli et al., 1991), that is, through the so-called Hox code. In comparing the expression of Hox genes (Dekker et al., 1992) and of Xlim-1 (Fig. 3) we note that (i) RA induction of HoxB3/Xhox2.7, HoxB4/Xhox-1A, HoxB5/Xhox-1B, and HoxB7/XlHbox2 appears to precede the first appearance of RA-enhanced expression of Xlim-1; (ii) the expression domains of HoxB4 and HoxB5 overlap with that of Xlim-1 in the area of hindbrain and pronephros; (iii) RA induces these Hox genes and Xlim-1 in regions anterior and posterior of their normal range, which correspond to the areas of anterior and posterior defects and to the enlarged pronephros and hindbrain (Figs 2A, 3; Durston et al., 1989; Papalopulu et al., 1991a).

Xlim-1 expression can be divided into two types, (1) activin inducible and CHX insensitive, and (2) RA inducible and CHX sensitive (Tadano et al., 1993). Type (1) probably corresponds to Xlim-1 expression in dorsal mesoderm and notochord and is not mediated by RA-inducible genes. In contrast, Xlim-1 expression in lateral mesoderm and neuroectoderm correlates temporally with the delayed induction of this gene by activin plus RA in animal explants (Taira et al., 1992), and therefore may require RA-inducible genes. This speculation is in good agreement with the fact that RA increases the incidence of pronephros formation in activin-treated animal caps (Moriya et al., 1993).

Several neural tissue-specific genes are activated in the ectodermal region of exogastrulae and in ‘Keller sandwich’ explants, including the general neural makers NCAM, NF-3, Xen1, Xen2, Xen3, Xit, and HNK-1 (Kintner and Melton, 1987; Dixon and Kintner, 1989; Doniach et al., 1992; Ruiz i Altaba, 1992), and the regionally specific genes en-2 and Krox-20 (Doniach et al., 1992). Induction through the plane of the ectoderm is thought to be responsible for activation of these genes. In contrast, XFKH1/XFD-1/Pintallavis, as Xlim-1, is not expressed in the ectoderm of exogastrulae (Dirksen and Jamrich, 1992; Ruiz i Altaba and Jessell, 1992). Since XFKH1/XFD-1/Pintallavis is expressed in the floor plate its expression in ectoderm may require contact with the notochord which induces the floor plate (Yamada et al., 1993). Xlim-1 is expressed parallel to the anterior-posterior axis, and it is possible that its expression in the neural tube requires a signal from the notochord or floor plate. Such a signal might be mediated by RA since RA was detected in the floor plate (Wagner et al., 1992), but exogenous RA did not induce Xlim-1 in exogastrulae. While RA enhanced the expression of HoxB9/XlHbox6 and XIF6 in neuroectoderm explants (Sharpe, 1991), it did not generate neurons positive for the Xit monoclonal antibody in exogastrulae (Ruiz i Altaba, 1992). Thus, a mesoderm-derived signal appears to be required even in the presence of RA for the expression of certain neural marker genes including Xlim-1.

The absence or severe reduction of Xlim-1 expression in the lateral mesoderm of exogastrulae is somewhat unexpected, since such embryos differentiate mesodermal structures including the notochord (Fig. 4C,D). It should be noted that the cell interactions implicated by this suppression of Xlim-1 are much earlier and thus distinct from the extensive epithelial-mesenchymal interactions that are required for kidney differentiation (reviewed by Saxén, 1987). It appears that signals exchanged between ectoderm and mesoderm are required in providing identity to the central region of the lateral plate (Fig. 1C); this region includes in its progeny the cell groups that can be visualized as Xlim-1-positive elongated condensations at stage 17/18 (Fig. 1F), prior to the appearance of the classically recognized pronephros anlage at stage 21 (Nieuwkoop and Faber, 1967). The Xlim-1 gene appears to be a target of this postulated signal.

The dependence of the expression of Xlim-1 in lateral mesoderm, and the possible dependence of the formation of pronephric precursor cells, upon a signal from ectoderm may resemble the requirement for a neural tube-derived signal for muscle cell differentiation in the somite. Dissection and explant culture experiments in avian embryos have shown that early somites, when separated from the neural tube, do not differentiate vertebral muscle, but gain this ability by contact with the neural tube (Rong et al., 1992; Christ et al., 1992). Since the pronephros, unlike the somite, is not in direct contact with the neural tube, it is unlikely that the ectodermal signal for pronephric Xlim-1 expression arises from the neural tube.

Genetic analysis has shown that the C. elegans LIM class genes mec-3 and lin-11, and the Drosophila gene apterous, are involved in cell type specification (Way and Chalfie, 1988; Freyd et al., 1990; Cohen et al., 1992; Bourgouin et al., 1992); mec-3 may also be needed to maintain cell differentiation (Chalfie and Au, 1989). The Xenopus gene, Xlim-3, is expressed entirely in a lineage-specific manner in the pituitary, pineal, and in certain subsets of cells in the CNS and the retina in the embryo and adult (Taira et al., 1993). Cell type specificity has been shown for the expression of other vertebrate LIM class genes (Korzh et al., 1993; Xu et al., 1993). Together with the present results on Xlim-1 these observations suggest that LIM class homeobox genes may be involved in both cell type specification and maintenance of differentiation in vertebrate as in invertebrate animals.

We thank Richard Harland for providing Tor 70 antibodies, David Wilkinson for the Xenopus Krox-20 plasmid, Eddy De Robertis for the gsc plasmid, Jim Smith for the Xbra plasmid, and Michael Rebagliati for critical reading of the manuscript. The anti-NCAM monoclonal antibody 4d was obtained from the Developmental Studies Hybridoma Bank maintained by the Department of Pharmacology and Molecular Sciences at Johns Hopkins University School of Medicine, Baltimore, MD, and the Department of Biology at the University of Iowa, Iowa City, IA, under contract number NO1-HD-2-3144 from the NICHD.