Activin redux: specification of mesodermal pattern in Xenopus by graded concentrations of endogenous activin B

The mesoderm of the amphibian embryo is formed through an inductive interaction in which cells of the vegetal hemisphere of the embryo act on overlying equatorial cells. The first candidate mesoderm-inducing factor to be identified was activin (Asashima et al.,1990; Smith et al.,1990), a member of the transforming growth factor type βfamily. The significance of activin as an endogenous inducing agent has been emphasised by the facts that (1) use of a dominant-negative activin receptor disrupts mesoderm formation in Xenopus(Dyson and Gurdon, 1997; Hemmati-Brivanlou and Melton,1992; New et al.,1997), (2) activin can exert long-range effects in embryonic tissue (Gurdon et al., 1994; Gurdon et al., 1995; Jones et al., 1996; McDowell et al., 1997) and (3)it can activate different genes at different concentrations(Green et al., 1990; Green et al., 1992; Green and Smith, 1990; Green et al., 1994; Gurdon et al., 1994; Gurdon et al., 1995; Papin and Smith, 2000). Together, these observations suggested that activin might act as a morphogen in the developing embryo.

The role of activin in the early embryo has, however, remained unclear because other candidate inducing factors have been isolated, including Vg1 (Thomsen and Melton,1993; Weeks and Melton,1987), the nodal-related genes(Jones et al., 1995; Joseph and Melton, 1997; Onuma et al., 2002; Takahashi et al., 2000) and Derrière (Sun et al.,1999; White et al.,2002), and because attempts to inhibit its function in a specific manner have produced contradictory results. It is not clear, for example,whether the activin-binding protein follistatin does(Marchant et al., 1998) or does not (Schulte-Merker et al.,1994) inhibit mesoderm formation. The most recent view on the role of activin in early Xenopus development has been articulated by Green, who says `although activin was and still is an excellent model for morphogen action, it may not be important in early vertebrate patterning'(Green, 2002).

In this paper, we first reinvestigate the temporal expression pattern of activin B and demonstrate that zygotic expression of activin Bprecedes that of one of its putative target genes, Xbra(Smith et al., 1991). We then use antisense morpholino oligonucleotides to inhibit the function of activin B in the early Xenopus embryo. Our results indicate that activin B is required for normal mesoderm formation in Xenopus in a concentration-dependent manner. We also demonstrate, serendipitously, that in performing antisense experiments of this sort in Xenopus one must beware of polymorphisms in the 5′ untranslated region of the gene(s) of interest.

Xenopus laevis embryos were obtained by artificial fertilisation and maintained in 10% MMR. They were staged according to Nieuwkoop and Faber(Nieuwkoop and Faber, 1975). Embryos at the one- or two-cell stage were injected with RNA dissolved in water or with morpholino oligonucleotides dissolved in 10% MMR. Embryos were cultured in 10% MMR and isolated animal pole regions were cultured in 100%MMR. Partially purified activin was used at a concentration of 16 U ml^–1 (Cooke et al.,1987) in the presence of 0.1 mg ml^–1 bovine serum albumin.

These were purchased from GeneTools. Sequences were as follows: MO1,5′-GCAGAGGCAGTAACAGGAGAGCCAT-3′; mMO1,5′-GCAGACGCACTAACATGAGAACCAT-3′; MO2,5′-CCCGGCGAGGGTCTCCGAGCGGAAA-3′; MO3;5′-CGAGGGTCTCCAAGCGGAGAGAGAA-3′.

The gross phenotypes obtained with these antisense morpholino oligonucleotides, described in Figs 3A-C and 6A-D, were observed in experiments involving at least 100 embryos for each morpholino.

Expression constructs were generated by subcloning cDNAs between the EcoRI and XbaI sites of pCS2+(Rupp et al., 1994; Turner and Weintraub, 1994). Standard techniques were employed and C-terminal HA tags were introduced into all constructs via the downstream PCR primer. Xenopus activin B was cloned using the primers 5′-AGCTGAATTCCGGACCCAGCGCCGCTCC-3′ and 5′-ACGTTCTAGATCAAGCGTAATCCGGGACATCGTACGGGTATGCACAGCCGCACTCGTCC-3′. Xenopus Follistatin(Hemmati-Brivanlou et al.,1994) was cloned using the primers 5′-AGCTGAATTCCCCACTCCCAGCACTGAGG-3′ and 5′-ACGTTCTAGATCAAGCGTAATCCGGGACATCGTACGGGTACTTACAGTTGCAAGATCCA-3′. Site-directed mutagenesis of the pCS2+.Xenopus activin B construct(see Results) used the primer 5′-CCCAGCGCCGCTCCATGGTTCTCATGTTAGTGCGTCTGCTTCTGGCCGGACTC-3′ and its reverse complement. A construct comprising the Smad-binding domain (SBD)and the FYVE domain of SARA (Tsukazaki et al., 1998), together with an HA tag, was cloned into pCS2+ and used as a loading control in Fig. 2C. This construct was created using the primers 5′-GTGACTAGTTGTAGTTACGTTTGCATTATCTG-3′ and 5′-ACCTCTAGACTAGGGGAAGTGGCTCCAGTCTG-3′.

Transcription of mRNA was carried out as described(Smith, 1993). Plasmids were linearised with NotI.

Whole-mount antibody staining using the monoclonal antibodies MZ15(Smith and Watt, 1985) and 12/101 (Kintner and Brockes,1984), specific for notochord and muscle respectively, was carried out as described (Smith,1993).

In situ hybridisation was carried out essentially as described(Harland, 1994), except that BM purple was used as a substrate. Probes used were for Chordin(Sasai et al., 1994), Derrière (Sun et al.,1999), Goosecoid (Cho et al., 1991), Xbra(Smith et al., 1991), Xnot (von Dassow et al.,1993), Xnr2 (Jones et al., 1995), Xvent-1(Gawantka et al., 1995) and Xwnt-8 (Christian et al.,1991; Smith and Harland,1991). The open reading frame of Xnr2 was amplified using the polymerase chain reaction and the primers 5′-TCTGAATTCATGGCAAGCCTAGGAGTCATC-3′ and 5′-ATTTCTAGAGTTACATCCACACTCATCCAC-3′. It was cloned into pCS2+,linearised with EcoRI and transcribed with T7 RNA polymerase. Each experiment shown was carried at least twice, with at least 20 embryos per treatment.

Total RNA was prepared from five pooled embryos using the TriPure reagent(Roche), followed by DNAseI digestion, proteinaseK treatment,phenol/chloroform extraction and ethanol precipitation. RNA was dissolved in water and used as a template for real-time RT-PCR. Reactions were performed in the LightCycler instrument (Roche) using a SYBR GreenI-based one-step RNA amplification kit (Roche). A standard curve was generated from diluted RNA derived from control embryos and experimental results were quantified accordingly. PCR primers are listed in Table 1. Each experiment was carried out at least twice and usually three times.

The 5′ untranslated region of Xenopus activin B was cloned by reverse transcription of total RNA isolated from five stage 12 sibling embryos followed by PCR amplification. The primers used were 5′-CGACACTGGCAGCACCTTC-3′ and 5′-GGCAGTAACAGGAGAGCCATG-3′.

If activin were to play a role in mesoderm induction in Xenopus,its expression should precede that of genes such as Xbra, which is expressed in presumptive mesoderm and which can be activated in animal pole tissue in response to mesoderm induction(Smith et al., 1991). To investigate this issue, RNA was extracted from Xenopus embryos between stage 7 (early blastula) and stage 23 (by which time the neural tube has closed). Samples were assayed by real-time RT-PCR for expression of activin B and Xbra and, as a reference gene, ornithine decarboxylase. Zygotic expression of activin, which is known to occur shortly after the mid-blastula transition(Clements et al., 1999; Dohrmann et al., 1993; Thomsen et al., 1990),preceded that of Xbra, consistent with the idea that activin B functions as an endogenous mesoderm-inducing agent(Fig. 1).

To investigate the role of activin B in early Xenopus development,we designed an antisense morpholino oligonucleotide that recognises the first 25 nucleotides of the open reading frame (MO1; Fig. 2A). A control morpholino oligonucleotide (mMO1) contained four evenly spaced base changes. The ability of the specific morpholino oligonucleotide to inhibit activin B translation in a dose-dependent manner was confirmed in an in vitro translation reaction, in which the mutated oligonucleotide had no effect(Fig. 2B). The specific morpholino oligonucleotide also inhibited translation in a dose-dependent manner of a tagged form of activin B following injection of mRNA into Xenopus embryos (Fig. 2C), and it inhibited activin-induced expression of Xbrain animal caps (Fig. 2D). The specific morpholino oligonucleotide also inhibited animal cap elongation in response to activin B (Fig. 2E-G) and it prevented the disruption of development that is caused by widespread expression of activin B in the embryo(Fig. 2H,I), in which assay the mutated oligonucleotide mMO1 had no effect(Fig. 2J).

The results that follow indicate that the mutated morpholino oligonucleotide mMO1 has no effect on Xenopus development. To confirm that the oligonucleotide is capable, given the opportunity, of inhibiting translation, we changed four nucleotides in the open reading frame of activin B to match the sequence of the mutated morpholino oligonucleotide. Misexpression of the mutated activin B was still capable of disrupting Xenopus development (Fig. 2K), and this disruption was prevented by the mutated oligonucleotide mMO1 (Fig. 2M)but not by the original version (Fig. 2L). These observations show that the mutated antisense morpholino oligonucleotide is stable and functional.

Xenopus embryos were injected with increasing concentrations of specific or mutated antisense morpholino oligonucleotide and then allowed to develop to tadpole stage 32. The mutated oligonucleotide had no effect on development, but the specific oligonucleotide MO1 caused severe disruption to dorsal axial development, with both head and tail being affected(Fig. 3A-C). Disruption to the dorsal axis was presaged by slow passage through gastrulation (see Fig. 5G-I and Fig. 7H,I), and was confirmed by in situ hybridisation, which showed that expression of Xnot(Fig. 3D,E) persists around the closing blastopore and that expression of chordin is more diffuse than in control embryos (Fig. 3F,G). These observations suggest that disruption of activin B function may cause a disruption of convergent extension (see Discussion).

Use of the monoclonal antibodies MZ15, which is specific for notochord, and 12/101, which is specific for muscle, demonstrated that these tissues are disrupted in embryos in which activin function is inhibited, although notochord and muscle cells are present(Fig. 3H-K).

To examine the concentration-dependent effects of the antisense morpholino oligonucleotide, and to investigate its effects on early development, we assayed the expression of a panel of genes expressed in mesoderm and mesendoderm using RT-PCR (Fig. 3L). Dorsally expressed genes such as Goosecoid(Cho et al., 1991), chordin (Sasai et al.,1994) and Xhex (Jones et al., 1999; Newman et al.,1997) were most severely affected by low concentrations of oligonucleotide, while expression of the pan-mesodermal marker Xbrawas reduced only by the highest concentration. These observations are consistent with the suggestion that patterning of the mesoderm in the Xenopus embryo occurs in response to different concentrations of activin (Green et al.,1992).

In situ hybridisation confirmed that expression of Goosecoid and Xbra is decreased in embryos in which activin B function is inhibited. We note that downregulation of Goosecoid is not accompanied by a significant restriction of its expression domain but by a general decrease in levels of transcription(Fig. 4A,B), whereas downregulation of Xbra tends to occur not throughout the marginal zone but in a more restricted domain (Fig. 4C,D). This may correspond to the dorsal marginal zone, but it is also possible that it reflects the region where the concentration of MO1 is highest (Fig. 4C,D). The expression domains of other genes, including Xwnt8 and Xvent1, which are expressed laterally and ventrally, are, like that of Goosecoid, little affected by the antisense morpholino oligonucleotide (Fig. 4E-H),suggesting that the upregulation of Xvent1 observed in Fig. 6 is due to elevated levels of transcription within its normal expression domain.

The results described above suggest that activin B plays a role in the development of the mesoderm in Xenopus. Experiments using antisense technology, however, require careful controls; we first asked whether the activin B antisense morpholino oligonucleotide functions, as we would predict,by interfering with activin translation or whether it interferes with the activity of some component of the activin signal transduction pathway. To address this point, animal caps were dissected from control embryos or from embryos injected with the specific morpholino oligonucleotide or the mutated version. The caps were treated with activin protein and induction was assessed by observing their elongation (Symes and Smith, 1987). Activin caused animal caps to elongate(Fig. 5A,B), and this elongation was inhibited by the activin-binding protein follistatin(Fig. 5C). No inhibition was observed with the activin B antisense morpholino oligonucleotide MO1(Fig. 5F), indicating that the effects of the oligonucleotide on early development are likely to be due to inhibition of activin B translation and not to interference with the activin signal transduction pathway.

We next attempted to `rescue' the MO1 phenotype by co-injecting the mutated form of activin B RNA that is not affected by the antisense oligonucleotide(see Fig. 2K-M). One difficulty with experiments of this sort is that injected RNA diffuses less well than the morpholino itself, so that the distribution of the two will differ(Nutt et al., 2001; Saka and Smith, 2004). Another is that activin causes a severe phenotype on its own, so that one has to inject enough activin RNA to rescue, but not so much so as to cause defects that are due to overexpression. In our attempts to address these concerns, we injected constructs into one cell of the four cell stage embryo rather than into the newly-fertilised egg, and we varied the concentration of injected RNA in an effort to find a dose that would rescue the MO1 phenotype but not cause defects through overexpression. The best results were obtained in an experiment in which 2 pg of mutated activin BHA RNA was used in conjunction with 20 ng MO1 (Fig. 5G-J). In this experiment, 2 pg activin RNA caused abnormal development in 36% of embryos (n=28; Fig. 5H), and a MO1 phenotype was observed in 64% of embryos(n=28; Fig. 5I). Injection of both mutated activin B-HA RNA and MO1 reduced the incidence of abnormal embryos to 24% (n=25; Fig. 5I). χ²analysis indicates that the observed rescue is significant at P<0.01.

As a final control, we designed an alternative antisense oligonucleotide positioned 5′ of the original target sequence (MO2; Fig. 2A). To our surprise this oligonucleotide proved to have little or no effect on development (see Fig. 6). To explore this observation, we isolated and sequenced the 5′ untranslated region of activin B from members of our own colony of Xenopus laevis (see Materials and methods). The sequence of this region proved to differ from the published sequence, resulting in a two nucleotide mismatch with our second antisense oligonucleotide (Fig. 2A). We note that a four nucleotide mismatch is sufficient to render an oligonucleotide completely ineffective (see Fig. 2A), and indeed a single mismatch can lead to a significant reduction in potency(Khokha et al., 2002), so it is likely that the efficacy of this alternative antisense morpholino oligonucleotide will be significantly compromised.

A third antisense morpholino oligonucleotide, MO3, was therefore designed to hybridise with the 5′ untranslated region of activin B derived from our own colony of Xenopus laevis. This reagent proved to have similar effects to our original activin B morpholino, but to be even more effective(Fig. 6A-D).

These conclusions were confirmed at the molecular level by comparing the expression of Goosecoid (Cho et al., 1991), chordin(Sasai et al., 1994) and Xvent1 (Gawantka et al.,1995) in embryos injected with different antisense morpholino oligonucleotides. In this experiment, embryos were isolated at different stages to investigate any temporal effects of inhibition of activin function. MO1 and MO3 gave similar results, with a significant downregulation of Goosecoid (particularly at mid-gastrula stage 11 for MO3) and of chordin, and a strong upregulation of Xvent1(Fig. 6E-J), indicating that in the absence of activin function embryos acquire more ventral characteristics.

The effects of activin depletion might be exacerbated if other mesoderm-inducing factors, such as the nodal-related proteins or Derrière, are downregulated, or they might be reduced if the expression of such factors is enhanced. To explore this issue, we investigated the expression of these inducing factors in embryos injected with mMO1, MO1 or MO3 at early gastrula stage 10.5. Expression of Xnr1, Xnr4, Xnr5 and Xnr6 was little affected by MO1 or MO3, but expression of Xnr2 was substantially increased and expression of Derrière was reduced. As observed with the expression of genes such as Goosecoid and Xvent1, the change in expression levels of these genes was not accompanied by changes in their expression domains (Fig. 7B-J). This is discussed below.

The work described in this paper shows that activin B plays a role in the early development of the Xenopus embryo, and particularly in the specification of the mesoderm. It thus reinstates activin as a potential inducing or patterning agent in the early amphibian embryo. Xenopusembryos lacking activin B function display defects in both anterior and posterior structures, and axial tissues such as notochord and muscle are present but severely disrupted (Fig. 3A-C,H-K). The expression of dorsally expressed genes such as goosecoid and chordin is downregulated(Fig. 3L), and that of Xvent1 is upregulated (Fig. 6G,J), but as discussed below, the expression domains of these genes are unaffected. We note that embryos in which activin B function is inhibited suffer defects in convergent extension(Fig. 3D-F). This may be a direct consequence of the lack of activin B, or a result of the downregulation of genes such as goosecoid and Xbra(Fig. 3L, Fig. 4A-D).

The requirement for zygotic activin B function in normal mesoderm formation in Xenopus contrasts with a requirement for maternal activin in Medaka (Wittbrodt and Rosa,1994) and with the absence of a requirement for zygotic activin A and activin B expression in mesoderm formation in the mouse (Matzuk et al.,1995). It is clear, however, that in all these species members of the TGFβ family, and particularly the nodal proteins, play significant roles in mesoderm formation (Schier,2003).

Our experiments do not represent the first attempt to investigate the role of activin in Xenopus development, but they may use the most specific tool to inhibit activin function. Dominant-negative activin receptors, for example, disrupt normal development(Hemmati-Brivanlou and Melton,1992; New et al.,1997), but they are as likely to inhibit the functions of other members of the TGFβ family, including BMPs and nodal-related proteins, as they are to inhibit activin (Hawley et al., 1995; Schulte-Merker et al., 1994; Wilson and Hemmati-Brivanlou, 1995; Yamashita et al., 1995). Even a secreted version of the type II activin receptor, which displays significantly greater specificity for activin(Dyson and Gurdon, 1997), may also inhibit other members of the TGFβ family, although we note that the phenotype of embryos expressing such a construct resembles, at least superficially, the phenotypes of embryos injected with MO1 or MO3(Dyson and Gurdon, 1997).

Other attempts to inhibit activin function have employed the activin-binding protein follistatin. Experiments by Schulte-Merker and colleagues were unable to demonstrate a role for activin following injection of RNA encoding rat follistatin(Schulte-Merker et al., 1994),although more recent experiments, using higher concentrations of RNA encoding the Xenopus protein, suggest that follistatin does inhibit mesoderm formation (Marchant et al.,1998). Interpretation of these experiments is further complicated by the observation that the inhibitory effects of follistatin are not restricted to activin, and that it also binds to members of the BMP family(Iemura et al., 1998).

A final approach has involved the use of dominant-negative `cleavage mutants', where expression in the embryo of a TGFβ construct in which the proteolytic cleavage site is mutated prevents the release of active dimers(Lopez et al., 1992). In Xenopus laevis, however, activin cleavage mutants prove to have little effect on development (Hawley et al., 1995; Osada and Wright,1999), and indeed similar results have been obtained in Medaka,although in this species it is maternal activin that appears to be required for proper mesoderm formation (Wittbrodt and Rosa, 1994). Potential pitfalls concerning the use of cleavage mutant constructs have been discussed by Eimon and Harland(Eimon and Harland, 2002), but explanations for inappropriate lack of activity of a construct are few. One possibility is that endogenous and exogenous activin B RNAs are processed in different compartments of the cell. Another is that endogenous activin can employ an alternative cleavage site; although activin is cleaved at a single cleavage site in oocyte expression studies(Hawley et al., 1995), an additional site may be employed after the mid-blastula transition. The inability of activin cleavage mutants to affect early Xenopusdevelopment requires further investigation.

Analysis of the spatial expression pattern of activin B in the early Xenopus embryo is hampered by its low expression level, but dissection of embryos indicates that transcripts are distributed ubiquitously(Dohrmann et al., 1993). This observation suggests that there is translational or post-translational control of activin function, as also occurs with BMP family members and the nodal-related genes (Agius et al.,2000; Bouwmeester et al.,1996; Cheng et al.,2000; Dale and Wardle,1999; Glinka et al.,1997; Jones and Smith,1998; Smith,1999). The spatial control of effective activin concentration is likely to be very complicated; for example, one known activin antagonist,Xantivin (Cheng et al., 2000),is more effective in marginal zone tissue than in the animal cap(Tanegashima et al., 2000),while experiments involving injection of activin into the blastocoels of Xenopus embryos suggest that there is in addition an intrablastocoelic inhibitor of activin function(Cooke et al., 1987).

The complicated regulation of activin function may help explain why it is so difficult to `rescue' MO1- and MO3-injected embryos to normality by introducing activin B RNA. The problem is exacerbated by the facts that endogenous activin B expression levels are so low and that injected RNA diffuses less well in the embryo than does injected morpholino oligonucleotide(Nutt et al., 2001; Saka and Smith, 2004). These problems notwithstanding, we have achieved partial rescue of the phenotype caused by MO1 by injecting a quantity of activin B RNA that is just sufficient, in 36% of embryos, to cause defects through overexpression(Fig. 5G-J). These results confirm the specificity of the observed phenotype and reinforce the conclusion that activin B is required for normal development in Xenopus.

Activin is but one of several mesoderm-inducing factors in the early Xenopus embryo; there are, in addition, five nodal-relatedgenes (Takahashi et al., 2000; Thomas et al., 1997) as well as Derrière (Sun et al.,1999) and Vg1 (Dale et al., 1993; Thomsen and Melton, 1993; Weeks and Melton, 1987). It is remarkable that the abolition of just one of these, activin, should cause such a dramatic phenotype, especially as the embryo seems to make some attempt to compensate for the loss of activin activity; although the inhibition of activin function is accompanied by a downregulation of Derrière, there is a marked upregulation of Xnr2 (Fig. 7). The first of these results is consistent with the observation that a dominant-negative Derrière construct inhibits Xnr2 expression in the Xenopus embryo, indicating that mesoderm-inducing factors might positively regulate their own expression(Eimon and Harland, 2002). The upregulation of Xnr2 in embryos injected with MO1 or MO3 does not accord with this idea, however, and it may be necessary in the future to conduct a systematic analysis of the effects of ablating candidate inducing factors and to ask how they regulate each other's expression. We note that inhibition of all mesoderm-inducing Xenopus nodal-related genes, by expression of the C-terminal region of Cerberus(Bouwmeester et al., 1996),causes severe defects in mesoderm formation. Such embryos form just a small tail-like structure, and expression of α-actin and α-globin is severely reduced (Wessely et al.,2001).

The downregulation of Goosecoid expression in response to MO1 and MO3, and the upregulation of Xvent1, appear to occur without significant changes in the expression domains of these genes (Figs 4, 7). This suggests that during normal development the spatial expression patterns of regionally expressed genes are defined by the combined effects of members of the TGFβ family,including the nodal-related genes and derrière as well as activin B. Loss of just one member of this network, such as activin, may not disrupt spatial expression patterns to a significant extent,but may affect expression levels such that development is severely perturbed.

The final point to be made from the results described is that one should not rely solely on sequences derived from GenBank when designing antisense morpholino oligonucleotides. Polymorphisms, particularly in the 5′untranslated regions of Xenopus mRNAs, are likely to be frequent, and even a single nucleotide difference between oligonucleotide and target RNA may produce a significant reduction in potency(Khokha et al., 2002). It may be impractical to confirm the sequence of the mRNA of interest in every experiment one does, but a failure to obtain a phenotype in one egg batch does not necessarily invalidate the rest of one's results.

This work was funded by a Wellcome Trust Programme grant to J.C.S. We thank members of our laboratory for helpful discussion and the referees of the paper for their insightful comments.