Xmsx-1 modifies mesodermal tissue pattern along dorsoventral axis in Xenopus laevis embryo

Mesoderm patterning of Xenopus laevis embryos is governed by at least two activities. Dorsalizing activity was first defined by Slack et al. (1987), and many genes have been identified in this category; factors include goosecoid (Cho et al., 1991), noggin (Smith and Harland, 1992), chordin (Sasai et al., 1994), follistatin (Hemmati-Brivanlou et al., 1994) and siamois (Lemaire et al., 1995). Among these factors, noggin, chordin and follistatin also possess neural inducing activity (Smith et al., 1993; Sasai et al., 1995; Hemmati-Brivanlou et al., 1994). In addition, more recently, extensive studies have identified ventralizing factors, namely bone morphogenetic protein-4 (BMP-4) (Dale et al., 1992; Jones et al., 1992), Xwnt-8 (Christian and Moon, 1993) and genes regulated by BMP-4 signaling (Gawantka et al., 1995; Onichtchouk et al., 1996; Schmidt et al., 1996; Ault et al., 1996; Ladher et al., 1996; Mead et al., 1996). It is suggested that these dorsalizing and ventralizing activities make a gradient along the dorsoventral axis and determine the fate of mesoderm.

BMP-4, a secretory protein belonging to the TGF-β super family, is a key factor in the formation of ventral mesoderm, as shown by its expression pattern (Fainsod et al., 1994; Schmidt et al., 1995), and functional studies utilizing dominant negative receptor (Graff et al., 1994; Suzuki et al., 1994; Maéno et al., 1994a) and antisense RNA (Steinbeisser et al., 1995). We have recently investigated downstream factors of BMP-4 signaling. Activation of ras pathway is essential for the BMP-4-mediated erythropoietic differentiation (Xu et al., 1996), as also shown in activin-mediated and FGF-mediated pathways (Whitman and Melton, 1992; MacNicol et al., 1993). BMP-4 also regulates XGATA-2in ventral mesoderm and animal pole tissue. XGATA-2 is expressed in both prospective epidermis and ventral areas (Walmsley et al., 1994; Kelley et al., 1994), and this factor functions as a stimulator of epidermis-dependent erythropoietic differentiation (Maéno et al., 1996).

To identify nuclear factors involved in ventral pattern formation, we amplified the DNA fragments with homeobox motif from the cDNA of UV-treated gastrula embryos by poly-merase chain reaction. We obtained msx-1 gene fragments in a certain frequency. Msx-1 is expressed in various places in a specific time course in developing vertebrate embryo (Robert et al., 1989, 1991; Coelho et al., 1991; Vainio et al., 1993; Liem et al., 1995; Graham et al., 1994; Tureckova et al., 1995). In Drosophila, msh, which is a homolog gene of msx-1 and msx-2, is expressed in muscle progenitor cells and neuronal cells (Lord et al., 1995). The above studies suggest that the temporal expression of msx controls the subsequent growth and differentiation of immature cells and the patterning of tissue structure and that there is a relationship between msx-1 expression and BMP-4 signaling (Vainio et al., 1993; Liem et al., 1995; Graham et al., 1994; Chen et al., 1996). Thus, we have attempted to elucidate the function and regulation of Xenopus msx-1 (Xmsx-1) in dorsoventral specification in mesoderm formation, in relation to ventralizing activity of BMP-4 signaling. The present study shows that the expression of Xmsx-1 is regulated at least in part by BMP-4 signaling in Xenopus embryonic cells, but the role of Xmsx-1in mesoderm patterning is distinct from that of BMP-4.

Xenopus laevis embryos were obtained by artificial insemination after induction of females with 250 i.u. of human chorionic gonadotropin. Developmental stages were designated according to Nieuwkoop and Faber (1967). To obtain ventralized embryos, fertilized eggs were freed of the jelly layers with 2.5% thioglycolic acid (pH 8.1), placed in a quartz-bottomed dish filled with 5% MMR and irradiated with UV light for the designated lengths of times using a UV illuminator (Vilber Lourmat Co. Ltd., 254 nm, 29mW/cm²). These embryos were incubated in 5% MMR at 18°C to stage 14. To obtain the dorsalized embryos, 32-cell-stage embryos were treated with 0.3 M LiCl in 30% MMR for 30 or 40 minutes, thoroughly washed in 30% MMR and then incubated at 23°C to stage 14.

Xenopus Msx-1 plasmid was generous gift from Dr Ramirez (Mount Sinai Medical School, NY). pSP72 containing whole insert (1.8 kb) was linearized with SalI and capped RNA was synthesized using T7 polymerase according to manufacture’s protocol (Ambion). Synthetic RNA of BMP-4 and a dominant negative form BMP-4 receptor (DN-TFR11) were made as described previously (Maéno et al., 1994a; Suzuki et al., 1994). The two blastomeres of dejellied embryos were injected with these RNAs at 2-cell or 4-cell stage (18.4 nl/embryo) in 3% Ficoll/Steinberg’s solution. The embryos were developed in Steinberg’s solution up to stage 10+, when animal cap (AP), dorsal marginal zone (DMZ) or ventral marginal zone (VMZ) tissue was excised and further cultured in sterilized Steinberg’s solution with 30 μg/ml kanamycin.

Total RNA from whole embryos or explants was extracted by AGPC method (RNA isolation kit, Stratagene), loaded in denatured 1% agarose gel and transferred to N-bond membrane (Amersham). Probes used in this study were as follow; Xmsx-1, 1.8 kb EcoRI fragment (Su et al., 1991); α-actin, 1.2 kb BamHI/HindIII fragment (Mohun et al., 1984); α-globin, 0.8 kb PstI fragment (Sandmeier et al., 1988); EF-1α, 0.4 kb PstI/SacI fragment (Krieg et al., 1989). Hybridization was performed in Hybrisol I (Oncor) and the membrane was exposed to Kodak XAR-5 film after washing. The same blot was sequentially hybridized with the different probes to detect each message in the same explant samples.

Complementary DNA was synthesized from 500 ng total RNA extracted from 10-15 explants. Polymerase chain reaction (PCR) was performed in 10 μl solution containing 1 μl of cDNA (10% of the obtained cDNA), 1× Ex-Taq buffer, 0.2 mM each of dNTP, 5 μCi of α-³²P -dCTP (3000 Ci/mmol), 350 ng of each primer and 0.25 U of Ex-Taq DNA polymerase (Takara, Tokyo). The PCR program consisted of 94°C for 1 minutes, 55°C for 1.5 minutes and 72°C for 1 minutes (22 cycles). After extenstion step (72°C for 10 minutes), 4 μl of the sample was loaded on 5% polyacrylamide gel. Primers used for PCR were as follows: Xmsx-1, 5′-GCA-GGA-ACA-TCA-CAC-AGT-CC-3′ and 5′-GGG-TGG-GCT-CAT-CCT-TCT-3′, EF-1α, 5′-CCT-GAA-TCA-CCC-AGG-CCA-GAT-TGG-TG-3′ and 5′-GAG-GGT-AGT-CTG-AGA-AGC-TCT-CCA-CG-3′ (Suzuki et al., 1993).

Whole-mount in situ hybridization was performed as described by Harland (1991). N-terminal fragment (EcoRI/HindIII, 440 bp) of Xmsx-1 was subcloned into pSP72 and digoxigenin-labelled ribo-probes were synthesized using Sp6 RNA polymerase. The positive signals were visualized using BM purple (Moors Jr. et al., 1995). Whole-mount immunostaining was performed as described (Klymkowsky and Hanken, 1991) using monoclonal antibodies against N-CAM (4d, from Developmental Studies Hybridoma Bank) and keratan sulphate (MZ-15, gift from Dr F. M. Watt) to stain neural tissues and notochord, respectively. Embryos and explants were fixed in Dent solution (20% dimethyl sulfoxide in methanol), incubated in diluted first antibodies and followed by 1:400 diluted peroxidase-conjugated secondary antibodies. Staining was visualized with diaminobenzidine and H₂O₂.

The embryos and explants cultured for 2 days (stage 35/36 of control embryo) were fixed in 2% paraformaldehyde/50% MMR solution, dehydrated, embedded in paraffin, sectioned at 7 μm, and stained with hematoxylin and eosin. The appearance of each tissue phenotype in explants was determined morphologically and scored as described by Ariizumi and Asashima (1994).

In an attempt to find a homeobox gene with ventralizing activity, we amplified DNA fragments from unidirectional cDNA library made from stage 10+ UV-treated embryos, with homeobox core sequence (TTT-T(G/T)G-AAC-CAG-AT(C/T)-TTC-AC(T/C)-TG) as a primer and T7 sequence (GCG-CGT-AAT-ACG-ACT-CAC-TAT-A) in plasmid as another primer. From the rescued plasmids derived from the amplified DNA fragments, we found a sequence identical to Xmsx-1 (formerly called as Xhox7.1). Thus, we examined the expression of Xmsx-1in the UV- and LiCl-treated embryos in order to assess whether dorsoventral patterning affected Xmsx-1 expression. As shown in Fig. 1, the Xmsx-1 expression in whole embryos at stage 14 was largely increased in UV-treated embryos and inhibited in LiCl-treated embryos (compared to stage-matched wild-type embryos).

Whole-mount in situ hybridization analysis showed that Xmsx-1was just detectable in animal pole area at early gastrula (stage 10^1/4) (Fig. 2A). At mid-gastrula stage (stage 11) (Fig. 2B,C), the positive staining was detected in marginal zone and animal pole area, laterally and ventrally, but not dorsally. The localized expression was much more obvious at late gastrula (stage 13); the message was strongest in dorsal-lateral region, but was completely absent in the presumptive neural area (Fig. 2D,E). At the neurula stage (stage 15), the message was detected in the ridges of neural fold and anterior lateral edges of the neural region, where the neural crest precursors exist (Fig. 2F). These expression patterns suggested that Xmsx-1 may have a role in dorsoventral specification of the whole embryo at the gastrula stage, in addition to dorsal determination of the neural tube at the neurula stage.

Previous studies have shown that, in mammalian and avian species, BMP-4 signaling governs msx-1 expression (Vainio et al., 1993; Liem et al., 1995; Graham et al., 1994). Since BMP-4 is a key factor regulating ventral development in Xenopus embryo, we examined whether the expression of Xmsx-1 is controlled by BMP-4 signaling. BMP-4 RNA was injected, at the 4-cell stage, into the dorsal marginal zone (DMZ) and expression of Xmsx-1 was determined in DMZ at stage 10+. As shown in Fig. 3, Xmsx-1 was prematurely induced by the BMP-4 RNA injection in DMZ region. The obvious elevation of Xmsx-1was only observed when 5 ng of BMP-4 RNA was introduced. It is concluded that the activation is a direct effect of BMP-4 signaling, since activin RNA did not affect the expression of Xmsx-1(Fig. 3).

We and others have shown previously that a truncated mutant BMP-4 receptor RNA (DN-TFR11), which exhibits dominant-negative effect on BMP-4 signaling, can alter the fate of ventral tissues to dorsal tissues (Graff et al., 1994; Suzuki et al., 1994; Maéno et al., 1994a). Thus we examined Xmsx-1 expression when endogenous BMP-4 signaling was blocked in the embryonic tissues by the injection of DN-TFR11 RNA into animal pole (AP) or VMZ region at 2- or 4-cell stages. The explants of each region excised at stage 10+ early gastrula were cultured until stage 14 neurula. Xmsx-1 was abundantly expressed at stage 14 (Fig. 4A), and also detected in VMZ and AP explants at this stage (Fig. 4B). In contrast, the expression of Xmsx-1was drastically reduced both in AP and VMZ explants injected with DN-TFR11 RNA (Fig. 4B). From these results, it was confirmed that Xmsx-1expression is regulated and maintained by BMP-4 signaling in embryonic cells.

To assess the activity of Xmsx-1 on ventralization, we injected Xmsx-1RNA in DMZ area and embryos were allowed to develop until stage 35/36. The effect on dorsal development was dose dependent and no significant effect was observed when 0.2 ng RNA was injected per embryo. As shown in Fig. 5B, embryos injected with 5 ng Xmsx-1RNA exhibited microcephaly with smaller eye capsules. The average D. A. I. of Xmsx-1-injected embryos was 3.55 (n=33). Whole-mount immunostaining with anti-N-CAM antibody (Fig. 5C,D) and notochord-specific antibody (Fig. 5E,F) showed that dorsal tissues were abortive in injected embryos. Most embryos had a trace of notochord in head region (see arrowhead in Fig. 5F). Histological examination clearly showed that the embryos injected with Xmsx-1RNA in dorsal blastomeres have a large block of muscle tissue in the dorsal mesodermal area and notochord is absent at the trunk level (Fig. 6).

We also examined the phenotype of DMZ explants isolated at stage 10+. Injected or control explants were subsequently cultured for 2 days, and the explants were examined by histological and immunohistochemical analyses. The control DMZ explants exhibited a typical dorsal phenotype as shown by the appearance of the cement gland (Fig. 7A). In contrast, those explants injected with 5 ng Xmsx-1showed a swollen, clear tissue type (Fig. 7B). Histological examination of these explants showed that, while the control DMZ explants contained neural tissue, cement gland, notochord and muscle in most cases, injected explants had mesenchyme instead of dorsal tissues (Figs 7D-F, 8). Although it is clear from the tissue pattern in the injected explants that Xmsx-1 possesses a ventralizing activity (inhibiting notochord and neural tissue development), we found that muscle, a dorsolateral type tissue, was abundant in the explants (Figs 7D, 8). Staining of these explants with antibodies also demonstrated that neural tissue (Fig. 7C) and notochord (data not shown) were absent after injection of Xmsx-1.

We compared the biological activity on mesoderm specification of Xmsx-1with that of BMP-4. The blastomeres of DMZ, LMZ or VMZ regions were injected with either BMP-4 RNA or Xmsx-1 RNA, and each region was explanted and cultured for 2 days. BMP-4 completely converted dorsal phenotype of the DMZ explants to a ventral phenotype. In both DMZ and LMZ explants, muscle α-actin mRNA was diminished and α-globin mRNA was induced. In contrast, Xmsx-1-injected explants preserved the production of abundant muscle α–actin mRNA and did not express any trace of α-globin mRNA (Fig. 9). Furthermore, surprisingly, Xmsx-1 can strongly induce α-actin mRNA and completely inhibit α-globin mRNA expression in VMZ explants (Fig. 9). These results revealed that, even though expression is regulated by BMP-4 signaling, Xmsx-1functions in determination and differentiation of dorsolateral region of mesodermal derivatives including muscle tissue.

We also investigated whether Xmsx-1could rescue the dorsalizing effect of DN-TFR11 on VMZ explants. As shown in previously, injection of high doses of DN-TFR11 RNA into the ventral blastomeres and subsequent culture of VMZ from stage 10+ abrogated the ventral phenotypes in the explants and led to the appearance of dorsal phenotypes such as notochord, muscle and neural tissue. As expected from the results of Fig. 9, coexpression of DN-TFR11 and Xmsx-1RNAs in the VMZ explants did not recover the α-globin expression (data not shown). This indicates that, although Xmsx-1is a downstream target gene of BMP-4 signaling that is involved in ventral specification, other factors independent from BMP-4 signaling might be involved in the specific expression and function of Xmsx-1during embryonic axis formation.

A number of studies have shown that BMP-4, a growth peptide belonging to TGF-β family, is a key factor in the regulation of the pattern of mesoderm formation in Xenopus embryo (Dale et al., 1992; Jones et al., 1992; Fainsod et al., 1994; Schmidt et al., 1995; Graff et al., 1994; Suzuki et al., 1994; Maéno et al., 1994a; Stein-beisser et al., 1995). The action of BMP-4 in relation to the dorsalizing factors (Sasai et al., 1995; Wilson and Hemmati-Brivanlou, 1995), FGF signaling (Northrop et al., 1995), and activin signaling (Dale et al., 1992; Jones et al., 1992; Suzuki et al., 1995) has been analyzed, and also recent studies have investigated the target factors located downstream of the BMP-4 signal (Gawantka et al., 1995; Onichtchouk et al., 1996; Ladher et al., 1996; Schmidt et al., 1996; Ault et al., 1996; Mead et al., 1996; Xu et al., 1996; Graff et al., 1996; Liu et al., 1996). Our previous study focused on the role of XGATA-2 in ventral mesoderm formation. XGATA-2 expression can be induced by the BMP-4 overexpression in the animal pole area and XGATA-2 can activate the epidermis-dependent stimulation on the ventral mesoderm to form blood cells. However, the knockout of BMP-4 signal by a dominant-negative receptor (DN-TFR11) did not inhibit completely XGATA-2 expression in animal pole tissue. Thus we concluded that XGATA-2 is not solely maintained and regulated by BMP-4 signal (Maéno et al., 1996). In contrast to XGATA-2, the expression of Xmsx-1 is drastically decreased by the DN-TFR11 injection (Fig. 4). In this respect, Xmsx-1expression (and possibly function) is dependent on BMP-4 signaling that leads to ventralization of the frog embryo. The correlation of BMP-4 signaling with the expression and the action of msx-1has also been suggested by previous studies using different experimental models (Graham et al., 1994; Liem et al., 1995; Tureckova et al., 1995).

In agreement with the previous work (Su et al., 1991), whole-mount in situ hybridization showed that Xmsx-1 is highly expressed in the dorsal region of neural tube and in neural crest cells adjacent to the forebrain and midbrain at the neurula stage. Although in situ analysis indicated that the Xmsx-1message is localized in neural cells at neurula stage, our northern blot analysis showed that VMZ explants at the neurula stage also express a significant amount of message (Fig. 4B). Thus, it would be pertinent to conclude that Xmsx-1 is not solely expressed in neural cells but also in lateral and ventral portions at this stage. Furthermore, our data emphasized that Xmsx-1 message was detected in lateral and ventral areas of mid to late gastrula embryos (Fig. 2). This expression pattern suggests a role of Xmsx-1in determining dorsoventral pattern of mesoderm derivatives. Especially, the strong expression in dorsolateral areas (shown in Fig. 2D) might reflect an importance of this gene in differentiation of muscle tissue, as is also shown in studies using RNA injection (Figs 6, 9). Further studies using a cytogenetic marker system will be important to prove whether muscle progenitor cells temporally express Xmsx-1or not.

While the present study showed that Xmsx-1disturbs the differentiation of notochord and neural tissues, Chen and Solursh (1995) reported recently that the injection of Xmsx-1 into uncleaved fertilized egg caused axis duplication of dorsal structures including notochord, neural tube and somites. The reason for this discrepancy is unclear. As suggested in our expression study, taken together with other studies on msx-1 expression (Su et al., 1991; Liem et al., 1995; Tureckova et al., 1995), the function of Xmsx-1may be complex and distinct in different cell lineages. In the present study, the ventral cells, if they are injected with msx-1, can be dorsalized (blood cells were converted to muscle tissue), and likewise the dorsal cells can be ventralized (notochord was converted to muscle tissue). Therefore, it is speculated that the injection of Xmsx-1 RNA into fertilized cells reorganized the dorsoventral axis in some injected embryos and secondary axis was formed. The differentiation of muscle tissue from ventral mesoderm in this study might be coincident with this observation.

The most important finding in this study is a difference in biological activities on mesoderm formation between BMP-4 and Xmsx-1. (1) Animal cap explants from a stage 10+ embryo, which was previously injected with BMP-4 RNA, showed a swollen structure with slight elevation of α-globin mRNA (Dale et al., 1992; Maéno et al., 1994a). Thus BMP-4 is a mesoderm inducer. In contrast, animal cap injected with Xmsx-1RNA gave rise to atypical epidermis (data not shown). This indicates that Xmsx-1 has no activity to induce any mesoderm and may regulate the mesoderm patterning after primary induction of mesoderm. (2) Xmsx-1ventralized the dorsal mesodermal structures and mesenchymal cells were obviously induced in DMZ explants in terms of morphological examination. However, it did not induce any trace of α-globin mRNA. BMP-4, in contrast, could turn on α-globin mRNA (Fig. 9), suggesting that Xmsx-1 does not function in the development of blood cells in the embryo. This phenomenon was also supported by the following result. We have shown previously that BMP-4 can stimulate the epidermis-dependent erythropoietic activation in combination explants of ventral mesoderm with animal pole tissue (Maéno et al., 1994b; Xu et al., 1996). Xmsx-1failed to activate the blood program in this system. Thus, activity stimulated by BMP-4 can not be replaced by Xmsx-1in ectodermal cells. (3) BMP-4 completely abrogated muscle formation in the DMZ explants based on muscle α-actin expression, but Xmsx-1did not. Furthermore, in VMZ explants, α-actin message was turned on and α-globin message was turned off by the injection of Xmsx-1 RNA. These observations suggest that, although the expression of Xmsx-1is regulated by BMP-4 signaling, other factors could be involved in the positive and negative regulation of Xmsx-1expression.

Finally, the inducible effect of Xmsx-1 on muscle differentiation is inconsistent with previous reports, in which human msx-1 suppresses myoD expression and differentiation of myoblasts to muscle cells (Song et al., 1992; Woloshin et al., 1995). However, we propose that myogenic differentiation in fibroblasts and embryonic mesodermal cells involve distinct mechanisms. As also supposed in myogenic regulation by twist (Spicer et al., 1996; Baylies and Bate, 1996), msx-1 may be necessary in early phases of muscle precursor cells, but not in differentiating myoblasts. Thus, the undifferentiated mesodermal cells may need a transient msx-1 expression for the determination of these cells to myogenic lineage, but at a later stage, the cells do not need msx-1 for further differentiation. Experiments to elucidate distinct regulation mechanisms in these two systems are currently being undertaken.

We thank Dr L. I. Zon for critical review on this manuscript, Dr F. Ramirez for Xmsx-1 plasmid, Dr N. Ueno for BMP-4, DN-TFR-11 and α-actin plasmids, Dr R. Weber for α-globin probe, Dr T. Sargent for EF1-α probe, Dr F. M. Watt for MZ-15 monoclonal antibody, and Dr Y. Watanabe for support and encouragement. This work was partly supported by The Asahi Glass Foundation and The Naito Foundation.