Neural expression of the Xenopus homeobox gene Xhox3: evidence for a patterning neural signal that spreads through the ectoderm

The formation of the basic body plan of the amphibian embryo, including the establishment of A-P polarity, depends primarily on positional differences found in the mesoderm (Spemann, 1938). During gastrulation the mesoderm involutes and induces the overlying dorsal ectoderm towards neural development in a regionspecific manner (Mangold, 1933; Raven and Kloos, 1945; Horst, 1948; see Hamburger, 1988). This regionalization results in the development of different neural structures in different places, e.g. brain, spinal cord and neural crest in, respectively, the anterior, posterior and lateral regions of the neural plate in the early embryo.

The mesoderm plays an essential role in neural induction by signalling a change in fate of dorsal ectoderm from epidermal to neural. However, this mesodermal signal may not be the only requirement for the proper formation of the nervous tissue. Perturbation and grafting experiments in amphibian embryos have suggested that dorsal ectoderm may acquire some neural properties before it is contacted by the underlying axial mesoderm during gastrulation (Goerttler, 1925; Lehmann, 1926; Spemann, 1931). This labile determination of the ectoderm has been proposed to result from a neural influence that spreads through the dorsal ectoderm from the organizer region, a place where neural and mesodermal tissues are continuous (see Spemann, 1938).

The requirement for mesoderm in neural induction has also been demonstrated in exogastrulated amphibian embryos (Holtfreter, 1933). In these embryos, the collapse of the blastocoel forces the mesoderm to evaginate instead of invaginating under the ectoderm. As a result, the prospective neural ectoderm is never in contact with axial mesoderm and does not undergo normal neural development. The mesoderm appears to develop normally in exogastrulated embryos as assessed by the differentiation of tissues such as notochord, somites and pronephros along the dorsal-ventral axis. In addition, gills develop at the anterior end indicating regionally appropriate differentiation along the A-P axis (Holtfreter, 1933). The ectoderm of exogastrulated embryos undergoes expansion by epiboly and becomes an empty sac. The main difference between isolated ectoderm cultured in vitro and exogastrula ectoderm is that the latter keeps a small but important junction area with axial mesoderm. This junction area corresponds to the organizer region. Thus, total exogastrulae provide an opportunity to examine the development of ectoderm under the influence of the organizer region but without underlying axial mesoderm. Histologically, no distinguishable neural tissue was found in the ectoderm of exogastrulae and, thus, the idea of a spreading neural signal emanating from the organizer region was abandoned (Holtfreter, 1933).

Because the determination of neural fates occurs before any histological features are apparent, it has been necessary to find molecular markers that indicate the acquisition of neural properties by ectodermal cells. Recently, the embryonic expression patterns of several genes have provided useful markers of cell identity and differentiation in Xenopus embryos. Some of these genes are general neural markers such as N-CAM (Jacobson and Rutishauser, 1986; Kinter and Melton, 1987; Levi et al. 1987), and a neurofilament gene (Sharpe, 1988). Others show region-specific expression such as some vimentin-like genes (Sharpe, 1988), and homeobox genes (Sharpe et al. 1987; Condie and Harland, 1987, Oliver et al. 1988). Homeobox genes are potentially good markers since they are thought to function as region-specific transcription factors and have been implicated in the control of cell fate determination in vertebrates (Harvey and Melton, 1988; Cho et al. 1988; Wolgemuth et al. 1989; Ruiz i Altaba and Melton, 19896; Balling et al. 1989; Wright et al. 1989; see Scott et al. 1989 for general review).

Although there is no axial mesoderm under the ectoderm of exogastrulae (Holtfreter, 1933), the ectoderm is neuralized (Kintner and Melton, 1987). In addition, it has been possible to show that a spreading signal, perhaps originating from the organizer, the dorsal marginal zone of the early gastrula embryo (see Jacobson and Sater, 1988), is involved in defining the limits (Savage and Phillips, 1989) and extent (Kintner and Melton, 1987; Dixon and Kintner, 1989) of neural induction in the dorsal ectoderm. Moreover, this spreading signal may be responsible for the proposed labile determination of prospective neural ectoderm towards neural development (see Spemann, 1938; Sharpe et al. 1987). However, these experiments do not address the question of whether the spreading signal also has a role in patterning the neural ectoderm along the A-P axis. If this occurs, A-P differences of the ectoderm of exogastrulae should be apparent along the apical-proximal axis of the ectodermal sac, since this axis is equivalent to the A-P axis of normal embryos (see Holtfreter, 1933; Nieuwkoop, 1985). To address this issue, the neural expression of the Xenopus homeo-box gene Xhox3 has been examined in both normal and exogastrulated embryos.

Xhox3 shows two periods of expression during embryogenesis (Fig. 1). During the gastrula and neurula stages, Xhox3 is expressed mainly in the axial mesoderm in a graded fashion with its maximum at the posterior pole (Ruiz i Altaba and Melton, 1989a). Xhox3 is involved in the establishment of A-P mesodermal fates during this early period (Ruiz i Altaba and Melton, 19896). At the late neurula-early tailbud stages, there is a second phase of Xhox3 expression in the nervous system and tail bud. Expression in the tail bud disappears when somitogenesis is completed; the neural expression of Xhox3 persists in a spatially restricted manner in the brain throughout the adult stages (Ruiz i Altaba and Melton, 1989a).

In this study, in situ hybridization analyses have been used to localize the expression of Xhox3 in the developing nervous system. The results show that Xhox3 is expressed in a restricted manner in the anterior neural crest and neural tube. Posterior neural crest and neural tube do not express Xhox3. The expression of Xhox3 as a marker of anterior neural development has been analysed in exogastrulated embryos. Xhox3 is expressed in a restricted region of the apical ectoderm in exogastrulae. These results are consistent with the existence of a neural-inducing signal that spreads through the ectoderm and which contributes to the establishment of A-P pattern.

Xenopus laevis embryos were obtained by fertilizing eggs with testis homogenates in vitro. Embryos developed in O.lxMMR (Newport and Kirschner, 1982) and were staged according to Nieuwkoop and Faber (1967). Exogastrulated embryos were produced by incubation of early blastulae in high salt (1-1.4XMMR) in 2% agarose dishes after removing the vitelline membrane.

RNA extraction, RNase protection assays and probes for the Xhox3, actin and EF-Icr genes were as described in Ruiz i Altaba and Melton (1989a).

In situ hybridization methods were basically as described in Kinter and Melton (1987) and Perry-O’Keefe et al. (1989) with the following modifications. Embryos were fixed with either 0.25% chromium trioxide in acid alcohol or 4% paraformaldehyde for 1 hour at 0—4°C. Sections were cut and treated as described in Perry-O’Keefe et al. (1989) and hybridized to ³⁵S-labelled Xhox3 antisense RNA probes generated by in vitro transcription as described in Ruiz i Altaba and Melton ( 1989a). Because of the low abundance of Xhox3 transcripts and the high background produced by probe sticking to endodermal tissue, the sections were hybridized with a concentration of probe 20- to 30-fold less (∼10ngml∼’xkb length) than that suggested previously (see Perry-O’Keefe et al. 1989). Hybridization was at 50°C. All washes and incubations contained 10 HIM DTT. A final stringent wash at 65 °C in 50% formamide, 0.1 % SSPE, 10 mM DTP was done for 45 minutes. Hybridized sections were exposed to fresh emulsion for one month at 4°C. The sections were photographed with an Axiophot Zeiss microscope and Tech-Pan film (Kodak). As a control for specificity, in situ hybridization with RNA probes derived from a different Xenopus homeobox gene, Xhox99 (N. Hopwood, unpublished), and the histone H4 gene show completely different hybridization patterns (data not shown).

Whole-mount staining of albino embryos was as described in Dent et al. (1989). For visualizing the binding of the anti-/? tubulin (Dent and Klymkowsky, 1988), .12/1.01 (Kintner and Brockes, 1984) or MZ15 (Smith and Watt, 1985) antibodies, secondary antibodies coupled to horseradish peroxidase were used and the peroxidase reaction done in the presence of diaminobenzidine. All antibody incubation were in 1X Trisbuffered saline with 5% fetal calf serum (without azide). Washes contained 0.05 % Tween-20.

Neural expression of Xhox3 is first detected at the early tailbud stage, mainly in the brain (Ruiz i Altaba and Melton, 1989a). In order to define the neural expression of Xhox3, in situ hybridization analyses were performed in tailbud and tadpole stage embryos.

In situ hybridization to serial cross-sections of stage 22-24 (not shown) and stage 28-30 (Fig. 2) embryos reveals that Xhox3 is expressed in both the cranial neural crest and the neural tube. Xhox3 expression in the cranial neural crest is localized to a small region along the A-P axis in the mid/hindbrain area (Fig. 2B,C), which may be hyoid neural crest (see Sadaghiani and Thiebaud, 1987). Moreover, comparison of this hybridization pattern to the distribution of neural crest cells in the mid/hindbrain area (Sadaghiani and Thiebaud, 1987) shows that not all neural crest cells in these areas express Xhox3. Neural crest cells in the forebrain and spinal cord areas do not express Xhox3 (Fig. 2A,D,E). It is not clear when Xhox3 is first transcribed in neural tissue, since there are very low levels of Xhox3 expression at early stages.

Expression of Xhox3 in the central nervous system (CNS) is localized to a small area in the ventral hindbrain (Fig. 2C) and anterior spinal cord (Fig. 2D-E). Cells expressing Xhox3 are found in the mantle layer of the hindbrain and spinal cord (Fig. 2C,D,F) suggesting that Xhox3 is expressed in post-mitotic cells as they begin to differentiate. Expression of Xhox3 in the posterior spinal cord is very low or absent (Fig. 2E).

Examination of horizontal sections shows that the most anterior/ventral neural crest cells that express Xhox3 are found at the level of the midbrain (Fig. 2F). More posteriorly in the hindbrain region, neural crest cells expressing Xhox3 are found flanking only the dorsal neural tube (Fig. 2C) overlapping over a short distance with cells expressing Xhox3 in the hindbrain (Fig. 2G). Some Xhox3 expressing cells of the hindbrain may follow a pattern partly coincident with rhombomeres (Fig. 2F and see below).

In situ hybridization to serial sagittal sections of larvae (stage ∼45) shows that expression of Xhox3 is not detected in any structures derived from the neural crest, but is only detected in small areas of the mid and hindbrain in lateral positions (Fig. 3A).

Examination of in situ hybridization to horizontal sections shows that the pattern of expression of Xhox3 is complex. While the most dorsal horizontal sections show no hybridization to the Xhox3 probe, the pattern of Xhox3 mRNA expression changes in sections prepared from more ventral regions. Figure 3B shows a horizontal section with hybridization to the cerebellum, found in between the midbrain and the hindbrain, and to the lateral area of the hindbrain coincident with the mantle zone of hindbrain neuromeres.

In situ hybridization to serial cross-sections illustrates the pattern of expression of Xhox3 mRNA. Xhox3 is not expressed in forebrain (Fig. 3C) or in the spinal cord (Fig. 3G) but in part of the cerebellum (Fig. 3E) and the ventrolateral area of the midbrain and hindbrain (Fig. 3D,F) perhaps coincident with the motor and second order sensory neuron columns.

The overall expression pattern of Xhox3 is distinct from that of frog (Carrasco and Malacinski, 1987; Oliver et al. 1988) and mouse (Holland and Hogan, 1988; Graham et al. 1989; Duboule and Dolle, 1989) homeobox genes of the Antennapedia-Ultrabithorax class, which are usually expressed in the hindbrain and/or spinal cord with different anterior boundaries. The anterior boundary of expression of Xhox3 in the mid/hindbrain region partly coincides with the region where the engrailed-related EN antigen is expressed (Hemmati Brivanlou and Harland, 1989). Expression of Xhox3 in hindbrain neuromeres, the cerebellum and the neural crest is notable since zinc-finger (Wilkinson et al. 1989a) and other homeobox (Davies et al. 1988; Wilkinson et al. 1989b; Robert et al. 1989; Hill et al. 1989) genes are also expressed in these structures.

The expression of Xhox3 in the ectoderm and mesoderm of exogastrulae was examined by RNase protection. As shown in Fig. 4A, expression of Xhox3 mRNA is detected in the ectodermal sac (lane E) of totally exogastrulated embryos at the early tailbud (stage ∼24) and tadpole stages (stage ∼34). This ectodermal expression is not observed during the early neurula stages (Ruiz i Altaba and Melton, 1989"). Because the junction area where posterior mesoderm is in contact with ectoderm was discarded in these assays (Fig. 4B), the presence of Xhox3 transcripts in the ectodermal sac of stage ∼34 exogastrulae (Fig. 4A) suggests that the expression of an anterior neural marker such as Xhox3 occurs away from the area of contact with posterior mesoderm. Consistent with previous findings (Ruiz i Altaba and Melton, 1989"), the mesodermal expression of Xhox3 (Fig. 4A, lane M, stage ∼24) disappears at the tailbud stage and is not detected in tadpole stages (Fig. 4A, lane M, stage ∼34). Muscle-specific (ms) and cytoskeletal (cyt) actin mRNAs were assayed to control for the presence of mesoderm and ectoderm. EF-In-transcripts were assayed as a control for RNA recovery since this gene seems to be expressed at similar levels throughout the early embryo (Krieg et al. 1989).

The results of RNase protection assays (Fig. 4A) suggest that cells expressing Xhox3 are found away from the boundary between posterior mesoderm and ectoderm. In situ hybridizations of totally exogastrulated embryos at the late tailbud-early tadpole stage (stage ∼28-30) shows that expression of Xhox3 RNA is restricted to a small region in the apical region of the ectodermal sac at or near the distal side of the vesicle in relation to the contact area with posterior mesoderm (Fig. 5A). The Xhox3 signal is detected in the vicinity of pigment cells that appear as white spots under darkfield microscopy (small arrows in Fig. 5A,B). Xhox3 mRNA expression is only detected in the apical region of the ectoderm away from the region that contacts the posterior mesoderm. This region corresponds to the animal pole region that would have become the anterior (head) end of the embryo in the course of normal gastrulation (see above and Nieuwkoop, 1985). Thus, the expression of Xhox3, an anterior neural marker in normal embryos, seems to occur at an equivalent position in the ectoderm of exogastrulae in the absence of contact with anterior axial mesoderm.

To show that mesodermal-ectodermal apposition does not occur at the apical end of the ectodermal sac of total exogastrulae, the distribution of ectodermal and mesodermal tissue in exogastrulated and control embryos of the same age (stage ∼30) as those used for RNase protection (Fig. 4A) and in situ hybridization analyses (Fig. 5) was examined. Three antibodies have been used in whole mounts to detect the presence of ‘neural ectoderm’, somites and notochord (Fig. 6). The first antibody used is directed against -tubulin (Dent and Klymkowsky, 1988), which primarily recognizes the developing CNS (Fig. 6D). However, this antibody labels all nuclei including those found in rows in the somites. The second antibody, 12/10.1, recognizes an epitope exclusively found at these early stages in somites (muscle) (Kintner and Brockes, 1984) (Fig. 6E). The third antibody, MZ15, recognizes keratan sulphate, which is primarily found in the developing notochord at this time (Smith and Watt, 1985) (Fig. 6F).

Figures 6A-C show the result of antibody staining of exogastrulated embryos in whole mounts. Somitic and notochord tissue (mesoderm that has the capacity to induce ectoderm towards neural development) identified by antibody staining, does not penetrate the ectodermal sac at any point. Most cells at the apical end of the ectodermal sac express tubulin at high levels (Fig. 6A) similar to that found in the developing CNS of normal embryos (Fig. 6D). This suggests that the neuralization of these ectodermal cells does not result from the apposition of axial mesoderm but rather from the spreading of a neuralizing signal.

The evidence for the existence of a neural signal spreading through the responding ectoderm derives from several observations. First, neural tissue can induce competent ectoderm towards neural develop-the dark-field photograph is an artefact. Panel B shows a dark-field photograph of ment (Mangold and Spemann, 1927; Mangold, 1933). This homogenetic neural induction exhibits A-P polarity. That is, the newly induced neural tissue has a more anterior character than the inducing tissue (Nieuwkoop, 1952A,b). Second, the formation of neural plate-like structures in the ectoderm of sandwich expiants of dorsal tissue are only observed when the organizer region is included in such expiants (Keller and Danilchick, 1988) suggesting that a neural spreading signal originates from the organizer. Third, the repression of Epil expression in the dorsal ectoderm is dependent on contact with the organizer region before mesoderm starts to invaginate (London et al. 1988; Savage and Phillips, 1989). This observation implies that the spreading neural signal is involved in repressing epidermal differentiation which may be a first step towards neural development, and further points to the organizer region as a source of this signal. Fourth, consistent with these findings, analyses of the posterior neural expression of the Xenopus homeobox gene XlHboxó in response to neural induction by the underlying axial mesoderm suggest that dorsal ectoderm may be somewhat predisposed towards neural development (Sharpe et al. 1987). Finally, a spreading neural signal from the organizer region, while normally acting in concert with the neural signal from the underlying axial mesoderm, is sufficient to induce the expression of neural-specific genes in the dorsal ectoderm in the absence of underlying axial mesoderm (Dixon and Kintner, 1989).

The results mentioned above suggest that dorsal (prospective neural) ectoderm is under the influence of a spreading neural signal that is required for full neural induction. They do not, however, directly address the issue of A-P patterning of the neuralized dorsal ectoderm in the absence of apposing axial mesoderm. To address the question of whether the spreading neural signal has an A-P patterning role, the expression of Xhox3, an anterior neural marker, has been analysed in total exogastrulae. Xhox3 expression in exogastrulae, unlike that of N-CAM (Kintner and Melton, 1987), is not detected at the ectoderm/mesoderm junction area but rather it is expressed by cells near the apical end of the ectodermal sac (Fig. 5). This shows that expression of Xhox3 in the ectoderm of exogastrulae is not the result of neural induction from the adjacent posterior dorsal mesoderm.

Xhox3 is not expressed in non-neural ectoderm, i.e. in isolated animal caps (Ruiz i Altaba and Melton, 1989c) suggesting that the ectodermal expression of Xhox3 in exogastrulae is due to an inductive interaction with other tissues. Because the organizer region is the only dorsal mesodermal tissue in contact with the ectoderm in exogastrulae, the ectodermal expression of Xhox3 supports the idea that cells in or around the organizer region are the source of a neural signal that spreads through the ectoderm. Interestingly, cells expressing Xhox3 are found along the apical-proximal axis of the ectodermal sac of exogastrulae in a position roughly equivalent to that of neural cells expressing Xhox3 along the A-P axis of normal embryos (Figs 2, 5). Thus, the spreading signal may be able to induce the expression of neural genes in an A-P ordered manner. However, the posterior-specific homeobox gene XlHboxó (Sharpe et al. 1987) and the anterior-specific engraf/ed-related EN antigen (Hemmati Brivanlou and Harland, 1989) are only expressed in partial but not total exogastrulae suggesting - that apposition of the underlying axial mesoderm is required for the expression of these two genes. Alternatively, the level of the spreading signal that apical ectoderm receives in total exogastrulae may be sufficient to activate the expression of Xhox3 but not XlHboxó or EN.

This interpretation is based on the idea that, in total exogastrulae, anterior mesoderm does not invaginate under the dorsal ectoderm before evaginating. Indeed, the apical expression of Xhox3 is unlikely to be due to induction by anterior axial mesoderm that transiently invaginated before evaginating during exogastrulation. In such case, Xhox3 expression would be expected to appear at the proximal (or posterior) end but not in the apical (or anterior) end of the ectodermal sac since anterior mesoderm only contacts posterior ectoderm in a transient invagination.

The normal expression of Xhox3 in the CNS and neural crest occurs only anteriorly. In exogastrulae, it is unclear whether the apical expression of Xhox3 in the ectoderm is neural crest or CNS related. Neural-crest-derived melanocytes can migrate into the ectodermal sac of exogastrulae (data not shown). Thus, the presence of melanocytes near the cells that express Xhox3 (Fig. 5A,B) points to the possibility that the latter may be posteriorly induced neural crest cells that migrated towards the apical end of the ectodermal sac. The melanocytes present in exogastrulae are of the trunk neural crest lineage induced by posterior axial mesoderm at the ectoderm-mesoderm junction area. However, Xhox3 is expressed in a subset of cranial but not trunk neural crest cells (Fig. 2). Thus, it is unlikely that the ectodermal cells that express Xhox3 are derived from migrating posterior neural crest. The identity (CNS or neural crest) of cells expressing Xhox3 in the ectoderm of exogastrulae therefore does not affect the implications of the results presented here suggesting the existence of a neural patterning signal that spreads through the ectoderm. However, the possibility exists that a brief and transient invagination of anterior mesoderm before évagination would be sufficient to induce cranial neural crest cells able to migrate to the apical region of the ectodermal sac and express Xhox3. The consistency with which these events would have to occur in otherwise extreme exogastrulae makes this unlikely.

The A-P pattern of the induced CNS is partly the result of a region-specific induction between the axial mesoderm and the overlying neural ectoderm (Mangold, 1933; Spemann, 1938; Hamburger, 1988). Recent findings show that the expression of the Xenopus homeobox gene XlHboxl occurs at similar positions along the A-P axis in both the axial mesoderm and the overlying neural ectoderm (Oliver et al. 1988). Based on this finding, it has been proposed that the congruency of the A-P pattern of gene expression in both the axial mesoderm and the neural ectoderm results from a homogenetic induction between the mesodermal and ectodermal germ layers (De Robertis et al. 1989). However, whereas axial mesoderm induces neural structures with an equivalent A-P character as seen in grafting experiments (Mangold, 1933; Ruiz i Altaba and Melton, 1989c), an alternative mechanism for reliably achieving congruent A-P patterns in both the axial mesoderm and the overlying neural ectoderm involves a patterning influence spreading through both the axial mesoderm and the neural ectoderm. In this case, both the neural inducing signal from the underlying axial mesoderm and the spreading signal from the organizer region would seem to act synergistically to pattern the nervous tissue.

Nothing is known about the molecular nature of the spreading signal. However, based on embryological experiments, it has been proposed that the A-P patterning of the induced neural tissue may be achieved through morphogen gradients (see Saxen and Toivonen, 1962; Hamburger, 1988). In addition to the signal that prevents anterior development by promoting posterior development in the posterior region and which derives from the posterior region of the early embryo (Sive et al. 1989), the results presented here suggest the existence of a positive influence on anterior development in the anterior region. Because the A-P polarity of the axial mesoderm may be established, in part, by gradients of growth factor-like molecules (Ruiz i Altaba and Melton, 1989c), it is possible that similar gradients of diffusible molecules derived from the organizer region contribute to the patterning of the neural ectoderm.

I am very grateful to D. A. Melton for his support of this work in his laboratory and his encouragement of this project. I thank D. A. Melton, C. R. Kintner, L. H. V. Wessendorf, E. Gilland, K. L. Mowry, M. Whitman, H. Perry-O’Keefe and T. Jessell for discussion and comments on the manuscript. I thank C. R. Kintner, F. Watt and M. Klymkowski for making the 12/101, MZ15 and anti-/? tubulin antibodies available, respectively. This work was supported by a grant of the NIH to D. A. Melton.