derrière: a TGF-β family member required for posterior development in Xenopus

Members of the TGFβ gene family are key regulators of mesoderm determination in Xenopus, as suggested by their ability to induce mesoderm in explanted tissue (Harland and Gerhart, 1997) and by the ablation of mesoderm after expression of dominant negative TGF-β receptors in whole embryos (Hemmati-Brivanlou and Melton, 1992). Maternally expressed TGF-βs include Vg1 whose RNA is localized to the vegetal hemisphere and is therefore in the correct position to induce mesoderm (Weeks and Melton, 1987).

In conjunction with maternally expressed β-catenin (Heasman, 1997), TGF-β signaling may also direct dorsoventral (D/V) mesodermal patterning, as suggested by the ability of different concentrations of Activin to induce a graded readout of D/V mesodermal fates in Xenopus explants (Green et al., 1992). Zygotically expressed TGF-β family members that may be involved in D/V mesodermal patterning include the Xnr genes (Xnr1, Xnr2, Xnr3 and Xnr4), whose RNAs are either localized to the entire marginal zone or restricted dorsally to the organizer (Ecochard et al., 1995; Jones et al., 1995; Smith et al., 1995; Lustig et al., 1996a; Joseph and Melton, 1997). Additionally, Bone Morphogenetic Proteins (BMPs) are TGF-β family members that may act both maternally and zygotically to promote ventral fates (Harland, 1994).

Closely linked to D/V mesodermal patterning is formation of the anteroposterior (A/P) axis, which appears in the mesoderm by early gastrula (Zoltewicz and Gerhart, 1997) and, soon after, in the ectoderm (Saha and Grainger, 1992; Sive et al., 1989; Gamse and H. L. S., unpublished data). Inhibition of BMP signaling has been implicated in anteroposterior axis determination, since ventral expression of a dominant negative BMP receptor (Graff et al., 1994; Suzuki et al., 1994: Glinka et al, 1997) or a dominant negative BMP4 ligand (Hawley et al., 1995) leads to formation of a secondary axis that never contains a head. Additionally, non-BMP TGF-βs including Xnrs, activin and an activated form of Vg1 are also able to induce, in normal embryos, a secondary axis lacking a head (Thomsen et al., 1990; Dale et al., 1993; Jones et al., 1995; Lustig et al., 1996a; Joseph and Melton, 1997). Thus, in the context of the normal embryo, posterior fates can be activated either by antagonizing BMP signaling or by promoting non- BMP TGF-β signaling. Despite these intriguing data, however, a normal role for TGF-β family members in posterior patterning has not been defined, since no TGF-β with localized expression or endogenous activity consistent with such a role has been described.

We have begun to address the role TGF-β molecules normally play in A/P axis determination. In a yeast-based selection for signal sequence-containing proteins (Jacobs et al., 1997) expressed during Xenopus A/P patterning, we isolated a novel member of the TGF-β superfamily, Derrière. Mature Derrière protein is very similar to Vg1, however, unlike Vg1, derrière is both zygotically expressed and, by mid-gastrula, its expression is restricted to posterior mesoderm. The derrière expression pattern is identical to the zygotic expression of the T box gene, VegT (Xombi, Brat, Antipodean, hereafter called VegT, Stennard et al., 1996; Zhang and King, 1996; Lustig et al., 1996b; Horb and Thomsen, 1997), and appears to be activated in a regulatory loop with this gene. We demonstrate that derrière activity is required for normal mesodermal patterning and posterior development of the Xenopus embryo. Our findings indicate that TGF-β family members not only play important roles in D/V axis determination, but that one member of this family, derrière, plays a crucial role in posterior axial patterning.

Methods for obtaining embryos are described in Sive et al. (1989). Microdissection was performed as described (Kolm et al., 1997). The starting materials for construction of a random-primed cDNA library that went through the yeast selection (see below) consisted of dorsal mesoderm and dorsal ectoderm of stage 11.5 embryos. The anterior limit was the anterior end of the archenteron and the posterior limit was the blastopore lip. The width of the explants was about 120° centered on the dorsal midline. For animal caps, animal hemisphere ectoderm was isolated within the window of 45 minutes centered around stage 8.5 and subsequently incubated as indicated. Anterior dorsal ectoderm (aDE) was isolated from stage 11 or 11.5 embryos. The anterior limit was the edge of the blastocoel, and the posterior limit was the mid-point between the anterior limit and the blastopore lip. The width of the explants was about 50° to each side of the dorsal midline.

For the random-primed cDNA library for the yeast selection, total RNA (600 μg) was isolated from the dorsal mesoderm/ectoderm explants using Proteinase K method followed by DNase I treatment (Gammill and Sive, 1997), poly(A)+ RNA was isolated using an oligo(dT) (Collaborative Biochemical Products) column. 0.5 μg poly(A)+ RNA was used for first- and second-strand cDNA synthesis (GIBCOBRL SuperScript Choice System). A random nanomer with an internal XhoI site (5′-AAGCTTGGCGGTCTCGAGNNNN- NNNNN-3′) was used at 60 μM for random priming. After second- strand synthesis, double-stranded cDNA was size-selected (300-600 bp) by PAGE and electroeluted before adapter ligation. Two oligos were used to form an EcoRI adapter: BIS1, 5′-AATTCGG- ACTACTACAGGTG-3′, unphosphorylated and BIS2, 5′-CACCT- GTAGTAGTCCG-3′, phosphorylated. cDNA was then separated from free adapters by PAGE, electroeluted and was subjected to PCR using phosphorylated BIS1. T4 DNA polymerase (in the absence of dATP and dTTP) was used to generate a cohesive EcoRI end at the 5′ end of cDNA before XhoI digestion at the 3′ end. The cDNA was directionally cloned into the SST vector (pSUC2T7M13ORI, Jacobs et al., 1997) for the yeast selection. For the full-length cDNA library, 300 whole embryos between stages 11.5 to 12 were harvested for total RNA. Poly(A)+ RNA was subsequently isolated and double-stranded cDNA synthesized using the same methods described above except the following: (1) an oligo(dT) primer with a XhoI site, 5′- CTCGAGTTTTTTTTTTTT-3′, was used along with methylated dCTP for first-strand synthesis; (2) two oligos were used for EcoRI adapter: 5′-AATTCCCATAGCAACAAACAGTA-3′ and 5′-TACT- GTTTGTTGCTATGGG-3′; and (3) no PCR amplification was performed and size selection (0.5 kb and above) and electroelution were not done until after adapter ligation and XhoI digestion. This cDNA was cloned directionally into EcoRI-XhoI double-digested CS2+ vector. The resulting library contains 5×10⁶ independent clones and has an average insert size of 1.3 kb.

The random-primed cDNA library was subjected to a yeast selection as described previously (Jacobs et al., 1997). In brief, the plasmid library was transformed into a yeast strain deleted for its endogenous invertase gene. The plasmid vector used to construct the library carried a modified invertase gene lacking its signal sequence. When a heterologous cDNA encoding a signal sequence was fused appropriately upstream of this defective invertase, the yeast’s ability to grow on restrictive media was restored. The original derrière partial cDNA clone contained 282 base pairs. cDNA sequence was determined and used for database search. The full-length cDNA clone was isolated by standard hybridization with probe synthesized using oligo 5′-GAAAGTGATAGCCACAACTCTGCCATG-3′. The GenBank accession number for derrière is AF065135.

Total RNA was prepared by Proteinase K method and analyzed by northern analysis as described (Kolm and Sive, 1995). Antisense derrière probe was prepared by asymmetric PCR amplification (Sive and Cheng, 1991) of derrière/CS2+ plasmid linearized with EcoRI using the primer 5′-CCCTTAGCATTTCCGTCAGT-3′.

Capped RNAs for microinjection were transcribed in vitro as described (Kolm and Sive, 1995). β-globin and lacZ were as described (Kolm et al., 1997). Other templates were as follows: EcoRI linearized pSP64TBVg1 (for BVg1; Thomsen and Melton, 1993), pβB-64T (for activin; Sokol et al., 1991), pXFD/Xss (for XFD, Amaya et al., 1991); NotI linearized Xnr1/CS2+ (for Xnr1; Lustig et al., 1996a), pCS2+VegTfc (for VegT, Zhang and King, 1996), pdor3 (for Xnr3, Smith et al., 1995), derrière/CS2+ and Cm-derrière/CS2+; SmaI linearized Xbra/pSP64T (for Xbra, Smith et al., 1991); XbaI linearized Xnr4-64TNE (for Xnr4, Joseph and Melton, 1997); KpnI linearized pCS2+Xnr2 (for Xnr2; Jones et al., 1995) followed by Klenow treatment. All were transcribed with SP6 RNA polymerase.

In situ hybridization probes were labeled with digoxygenin-11- UTP (Harland, 1991). Krox20, en-2, otx2, HoxB9 and XCG were as described (Kolm et al., 1997). Other templates were as follows: derrière/CS2+, EcoRI linearized, T7 transcribed; pBSAC100 (for m- actin), EcoRI linearized, T3 transcribed; pSP70-N1 (for NCAM; (Kintner and Melton, 1987), EcoRV linearized, SP6 transcribed.

Microinjections were done as described (Kolm et al., 1997). See Figure Legends for stages and sites of injection. 80 pg lacZ-capped RNA was included where necessary. Within each set of experiments, β-globin-capped RNA was used to make up for differences in amount of test RNA injected such that all embryos received the same total amount of RNA.

β-galactosidase staining was performed as described (Kolm and Sive, 1995). Whole-mount in situ hybridization was performed as described (Harland, 1991), with modifications described in Bradley et al. (1996). For in situ hybridization done on sectioned embryos, whole embryos were fixed for 1 hour first before being sectioned with eyebrow knife and fixed for another hour.

RNA and cDNA samples were prepared as described (Kolm et al., 1997). The optimal cycle numbers (listed below) were determined by titration using whole-embryo cDNA from similar stages. The PCR program used was: 95°C for 30 seconds, 55°C for 40 seconds and 72°C for 40 seconds. NCAM (28 cycles), en-2 (26 cycles) and Krox20 (26 cycles) primers were as described (Hemmati-Brivanlou and Melton, 1994). HoxA7 (26 cycles), m-actin (22 cycles), HoxB9 (27 cycles) and Xcad3 (25 cycles) primers were as described (Kolm and Sive, 1997). XCG (17 cycles) primers were as described (Gammill and Sive, 1997). The sequences of other primers used are as follows (sense primer first): ODC (21 cycles): 5′-CAACGTGTGATGGGCTGGAT- 3′ and 5′-CATAATAAAGGGTTGGTCTCTGA-3′, Xbra (24 cycles): 5′-TTCTGAAGGTGAGCATGTCG-3′ and 5′-GTTTGACTTTGC- TAAAAGAGACAGG-3′, VegT (28 cycles): 5′-TTAGCTTCCCAG- AGACAGAGT-3′ and 5′-CACATATAGCCTTGGGGAAATC-3′, eFGF (27 cycles): 5′-CGGGTTTCATATCCAGGTTTTAC-3′ and 5′-GCGTTATAGTTGTTGGGCAGAAG-3′, gsc (26 cycles): 5′- GGATTTTATAACCGGACTGTGG-3′ and 5′-TGTAAGGGAGCA- TCTGGTGAG-3′, eomes (25 cycles): 5′-GGGCCAACAGCAC- AAGAAATAC-3′ and 5′-TGGAGGCGCATAAGGGAAGAT-3′, Xlim-1 (25 cycles): 5′-GTGTCTGCCTTCTATTCTCCTAA-3′ and 5′-GCACAGCCCGCACACTTGGTA-3′, Pintallavis (25 cycles): 5′-GCAGGCACCCAACAAGATGAT-3′ and 5′-CCAGATTCGGG- GTGCAGAGT-3′, Xnot (26 cycles): 5′-CAGACCTGCCTCCAAA- CTATCC-3′ and 5′-TCTCCCCTGGGCATCCTCATT-3′, siamois (27 cycles): 5′-AGGAACCCCACCAGGATAAAT-3′ and 5′-GTTG- ACTGCAGACTGTTGACTA-3′, Xvent-1 (25 cycles): 5′-GCATC- TCCTTGGCATATTTGG-3′ and 5′-TTCCCTTCAGCATGGTTCAA- C-3′, XK81 (19 cycles): 5′-TCATTCCGTTCCAGCTCTTCTTAC-3′ and 5′-TCCAGGGCTCTTACTTTCTCCAG-3′, Xsox17α (25 cycles): 5′-CAATGGCAGCTACCCTCACC-3′ and 5′-CTTGGCCACATAG- CTCAGATAC-3′, endodermin (21 cycles): 5′-TATTCTGACT- CCTGAAGGTG-3′ and 5′-GAGAACTGCCCATGTGCCTC-3′, derrière (22 cycles): 5′-TGGCAGAGTTGTGGCTATCA-3′ and 5′-CTATGGCTGCTATGGTTCCTT-3′.

Three sets of oligonucleotide duplexes were obtained that encoded the mature region of Derrière (residues 241-354) from after the presumed maturation cleavage site to the stop codon and included NdeI and XbaI restriction sites at the 5′/3′ ends of derrière. This allowed cloning of the ligated duplexes into NdeI-XbaI-restricted E. coli expression vector pAL981 (LaVallie et al., 1993) with an in-frame fusion of derrière to an ATG. The codons of derrière were modified to reflect codon usage found in highly expressed E. coli genes. The derrière expression plasmid was sequenced and used to transform the E. coli strain GI934 (Lu et al., 1996). A fresh overnight culture of GI934 containing the derrière expression plasmid was used to inoculate IMC/Amp medium (M9 media containing 0.2% casamino acids, 0.5% glucose, 1 mM MgSO₄ and 100 μg/ml ampicillin) to an OD₅₅₀ of 0.05. The culture was grown at 30°C until the OD₅₅₀ reached 0.5, then L- tryptophan was added to a concentration of 100 μg/ml and the culture temperature shifted to 37°C. 4 hours later the cells were harvested and stored at −80°C until use. Recombinant Derrière expressed in E. coli was refolded to produce active dimers essentially as described (Schlunegger et al., 1992). In brief, the monomeric denatured protein was isolated from inclusion bodies by acidification and purified by size exclusion chromatography. Derrière protein was refolded at pH 8.5 using 2% CHAPS, 1.0 M NaCl and a glutathione redox couple. Final purification was carried out on a Mono S column (Pharmacia Biotechnology Inc.) and reversed-phase HPLC.

For protein treatments, explants (both animal caps and aDE) were incubated at 20°C until desired stages. BSA (New England Biolabs) was used at 300 ng/ml in 0.5× MBS; bFGF (Promega) was used at 100 ng/ml (plus 200 ng/ml BSA) in 0.5× MBS; Derrière was used at 200 ng/ml (plus 100 ng/ml BSA) in 0.5× MBS. A co-treatment of bFGF and Derrière consisted of 100 ng/ml bFGF and 200 ng/ml Derrière.

The strategy for generating derrière cleavage mutant (Cm-derrière) is as described previously (Hawley et al., 1995). The two primers used to bring in substitutions were: 5′-GCATGGTCGACGGGAGTACT- CATTCATCACCTC-3′ (sense, paired with T7 primer) and 5′-GC- ATGGTCGACGCCTTGAGTTTTGCAATTGGATG-3′ (antisense, paired with SP6 primer) using derrière/CS2+ as template. The resulting 5′ (EcoRI-SalI double-digested) and 3′ (SalI-XhoI double- digested) ends of Cm-derrière were subcloned into CS2+ (EcoRI- XhoI double-digested) by a three-piece ligation and the final construct was confirmed by sequence determination.

We isolated derrière using a yeast-based selection assay for secreted proteins, from a cDNA library prepared from Xenopus mid-gastrula dorsal mesoderm and ectoderm (Jacobs et al., 1997; see Materials and Methods). Based on sequence identity, potentially interesting clones were screened for localized expression by in situ hybridization at late gastrula. derrière showed localized posterior expression in this assay.

The derrière cDNA encoded a protein of 354 amino acids (Fig. 1A) and represented a previously unknown member of the TGF-β superfamily. A putative signal sequence cleavage site was present at the amino terminus (von Heijne, 1986), a maturation processing sequence (RAKR, Rehemtulla and Kaufman, 1992) was present at position 237 and, in the carboxy terminus, seven cysteines were present that are conserved among other TGF-β molecules (Kingsley, 1994). derrière is a member of the Vg-related family of TGF-βs and is most similar to Xenopus Vg1 (56% identity for the full- length protein and 79% within the mature region; Fig. 1B). The percentage identities between the mature region of Derrière and of other Vg1-related proteins shown in Fig. 1B are 79% (xVg1), 76% (zVg1, Helde and Grunwald, 1993) and 73% (cVg1, Seleiro et al., 1996). The next-most-related TGF-βs in the database are substantially less related to Derrière: mouse GDF3 has 62% identity in the mature region, sea urchin Univin has 60% identity in the mature region and Xenopus BMP2 has 59% identity in the mature region.

In order to ask what role derrière might play during embryogenesis, we first examined the distribution of derrière transcripts. Northern blotting (Fig. 2A) indicated that no maternal transcripts were present in unfertilized eggs (lane 1) and at the mid-blastula stage (lane 2). derrière transcripts were first detected at late blastula (stage 9, all stages according to Nieuwkoop and Faber, 1994, lane 3), peaked during gastrula (lanes 4, 5) and were barely detectable at the end of neurula (stage 19, lane 7). The derrière transcript was 1.5-1.6 kb in length, consistent with the size of the cDNA (1554 bp).

By whole-mount in situ hybridization (Harland, 1991) (Fig. 2B), at late blastula (stage 9.5), punctate staining was present in vegetal cells as well as in cells of the entire marginal zone (panel a). By early gastrula (stage 10.5), derrière RNA was present in marginal zone cells with higher intensity on the dorsal side of the embryo (panel b), while by mid-gastrula (stage 11.5), RNA was posteriorly restricted with no expression visible vegetally (panel c). By late gastrula (stage 12.5, panel d) and early neurula (stage 14, panel e), staining was excluded from the dorsal midline. A sagittal section of a mid-gastrula stage embryo (panel f) showed that derrière transcripts were present in both superficial (white arrow) and deep (white arrowhead) mesodermal layers but absent from involuted mesodermal cells.

In summary, these data showed that derrière expression is zygotic, with expression initially throughout the presumptive mesendoderm and subsequent localization to the posterior mesoderm. The expression pattern of derrière appeared identical to that of VegT and suggested that derrière might play a role in mesodermal patterning, particularly in posterior regions of the embryo.

We next asked what genes controlled derrière expression, by testing secreted proteins known to have mesoderm-inducing capacity as well as transcription factors previously shown to be involved in mesoderm determination (Fig. 3A). Mid-blastula (stage 8.5) animal caps were isolated from uninjected embryos and incubated in purified basic fibroblast growth factor (bFGF, Promega) or Derrière protein (see Materials and Methods), with bovine serum albumin (BSA, New England Biolabs) as control. Alternately, 2-cell-stage embryos were injected in one cell with RNAs encoding BVg1 (a fusion between the mature region of Vg1 and the pro region of BMP2; Thomsen et al., 1990), Activin (Sokol et al., 1991), VegT or Xbra (Cunliffe and Smith, 1992). globin RNA (Krieg and Melton, 1984) was injected as a control (Fig. 3A). At mid-blastula, animal caps were removed and cultured until sibling controls reached mid- gastrula (stage 11.5) when caps were harvested using a reverse transcriptase-PCR (RT-PCR based assay). Whole embryos were used as positive controls for RT-PCR, ornithine decarboxylase (ODC) was used as loading control and reactions without addition of reverse transcriptase were included to indicate genomic DNA contamination.

The results of a representative experiment (of three experiments) are shown in Fig. 3B. While BSA did not activate derrière expression (lane 1), both bFGF (lane 2) and Derrière protein itself (lane 3) led to derrière activation. Of RNAs injected, globin as a control did not activate derrière expression (lane 5), BVg1 (lane 6), activin (lane 7), VegT (lane 8) and Xbra (lane 9) all led to accumulation of derrière RNA. These results showed that derrière can be induced by known mesoderm inducers and suggested that an autoregulatory loop might control derrière expression.

Since several TGF-β family members are mesoderm inducers, we asked whether derrière could induce mesodermal markers using an animal cap assay (Fig. 4A). RNA encoding either Derrière or Globin as a control was injected into 2-cell embryos, caps were isolated at mid-blastula and harvested when control embryos reached mid-gastrula (stage 11), early neurula (stage 14) or late neurula (stage 19) to account for maximal expression of the different markers tested.

A representative experiment of at least two for each marker is shown in Fig. 4B. In comparison to globin-injected caps (lanes 1, 4 and 7), derrière was able to activate multiple mesodermal marker genes (lanes 2, 5 and 8). Genes activated included VegT, which has a very similar expression pattern to derrière, the posterior marker eFGF (Isaacs et al., 1992), the prechordal plate marker gsc (Cho et al., 1991), the mesendodermal marker eomesodermin (Ryan et al., 1996), the dorsal mesodermal markers Xlim-1 (Taira et al., 1992) and Xnot (von Dassow et al., 1993). Additionally, derrière induced expression of the posterior mesodermal and ectodermal markers HoxB9 (Sharpe et al., 1987) and an endodermal marker, Xsox17α (Hudson et al., 1997). derrière failed to activate expression of siamois (Lemaire et al., 1995), which is expressed in the dorsal mesendoderm, including the organizer, and did not alter expression of the ventral mesodermal marker Xvent-1 (Gawantka et al., 1995). Additionally, derrière failed to induce expression of the neural-specific marker N-CAM (Kintner and Melton, 1987). Since expression of the epidermal marker XK81 (Jonas et al., 1985) was strongly suppressed in derrière-injected caps, failure to induce neural markers was likely to be due to the almost complete conversion of the cap to mesoderm, eliminating an ectodermal substrate for neural induction.

We asked whether derrière induced mesodermal and endodermal markers in a dose-dependent fashion, as has been demonstrated for other TGF-βs, particularly activin (Green et al., 1992). As shown in Fig. 4C, at the highest concentration of derrière RNA tested (20 pg), HoxA7 (Condie and Harland, 1987), Pintallavis (Ruiz i Altaba and Jessell, 1992), the muscle-specific marker muscle actin (m-actin, Mohun et al., 1984), Xbra (Smith et al., 1991) and Xsox17α were induced (lane 2), at 15 pg HoxA7, Pintallavis, Xbra and Xsox17α were induced (lane 3) while, at 10 pg, only HoxA7 and low levels of Pintallavis and Xsox17α were induced (lane 4). None of these markers was activated at 5 pg derrière (lane 5). At the stage when caps were harvested, Pintallavis is expressed weakly in the dorsal mesoderm and in the tailbud (Ruiz i Altaba and Jessell, 1992), and the expression assayed here may be reflective of both dorsal and posterior fates.

It has previously been shown that an intact FGF signaling pathway is required for signaling by activin (La Bonne and Whitman, 1994). We asked whether FGF signaling is required for derrière function using an animal cap assay. As shown in Fig. 4D, derrière induced elongation of animal caps (panel a; 100%; n=29), and this was reduced to background levels by the presence of XFD, a dominant negative FGF receptor (Amaya et al., 1991, panel b; 6%; n=33, control in panel c; 0%; n=30).

These results showed that derrière could induce mesodermal markers characteristic of dorsal, lateral, anterior and posterior regions, as well as endodermal markers. This induction occurred in a dose-dependent and FGF-dependent manner. A posterior marker, HoxA7, was preferentially activated at lower derrière concentration, while m-actin required the highest does of derrière. Interestingly, an endodermal marker was induced by derrière within the same concentration range as mesodermal markers were induced. This is different from the activity of BVg1 and activin where high concentrations induce endoderm and low concentrations induce mesoderm (Henry et al., 1996).

We next asked whether derrière could alter axial patterning in whole embryos, using a gain-of-function approach. 2-cell embryos were injected either dorsally or ventrally in the marginal zone with 50 pg derrière or globin RNA (Fig. 5A), along with lacZ RNA as lineage tracer. Higher amounts of injected derrière RNA led to abnormal gastrulation and death by the end of gastrulation (not shown). Embryos were harvested at tailbud (stage 24) or hatching (stage 35) for analysis by in situ hybridization or morphological examination.

Representative results are shown in Fig. 5B,C. After injection of derrière RNA dorsally (Fig. 5B), embryos developed with a small head (panel a; 88.4% of injected embryos, n=189), with reduction or absence of eyes and cement gland. Controls injected with globin RNA did not show these defects (panel b; 0%; n=73). Expression of m-actin in the posterior paraxial mesoderm was unperturbed after derrière misexpression (panel c; 0%; n=12) and did not extend more anteriorly than it did in globin-injected controls (panel d; 0%; n=15). derrière reduced expression of the anterior marker otx2 (Blitz and Cho, 1995; Pannese et al., 1995) in forebrain and midbrain (panel e; white arrow), eyes (white arrowhead) and cement gland (100%, n=12) relative to globin-injected controls (panel f; 0%; n=15). derrière did not alter the anterior limit of HoxB9 expression in the spinal cord (panel e; black arrowhead; compare to control in panel f). Although derrière did not alter the expression pattern of the midbrain marker engrailed (panel g; white arrow, en-2, Hemmati-Brivanlou et al., 1991) or the hindbrain marker krox20 (bracket), these markers were expressed much closer to the anteriormost extent of the embryo (100%, n=13) than they were in controls (panel h; 0%; n=16). In accord with morphological data, expression of XCG, a cement gland marker (Sive et al., 1989), was strongly reduced in derrière-injected embryos (panels g and h; white arrowhead). Thus, ectopic expression of derrière dorsally suppressed head formation. This phenotype was very similar to the partial head suppression observed after ectopic dorsal expression of VegT and eFGF (Isaacs et al., 1994; Zhang and King, 1996).

In contrast to the effects of dorsal misexpression, ectopic expression of derrière ventrally (Fig. 5C) led to formation of a partial secondary axis (panel a; 71% of injected embryos, n=123), compared to globin-injected controls (panel b; 0%; n=34). The secondary axis never contained a head and ended posterior to the level of the otic vesicle of the primary axis. m- actin expression in the secondary axis indicated the presence of posterior (paraxial) mesoderm (panel c; 100%, n=12), which was never observed after globin RNA injection (panel d; 0%; n=15), and also expressed the neural-specific marker N-CAM (panel e; 67%, n=12; control in panel f; 0%, n=12). otx2 expression was not observed in the secondary axis (panel g, white arrow; 0%; n=12; control in panel h; 0%; n=15), indicating a lack of anterior tissue, while the lack of HoxB9 expression detected in the secondary axis (panel g), suggested that spinal cord had not formed. en-2 or krox20 expression was not observed in the secondary axis (panel i; 0%, n=14; control in panel j; 0%; n=16), confirming that the secondary axis did not contain tissue anterior to the hindbrain. No neural tube or notochord were apparent in sections through the secondary axis (not shown).

In summary, these data indicated that derrière had the properties of a posterior inducer that suppressed head formation and directed formation of a partial secondary axis, containing paraxial mesoderm and neural tissue.

One explanation for the mechanism by which derrière induces a secondary axis is that the Derrière protein dorsalizes the ventral marginal zone (VMZ), as several Xenopus TGF-β family members are able to do (Sokol et al., 1991; Jones et al., 1995; Kessler and Melton, 1995; Smith et al., 1995; Joseph and Melton, 1997). Since derrière expression is maximal on the dorsal side during early gastrula, it may play a role in normal dorsal mesendoderm (organizer) activity. We tested the ability of derrière to increase dorsal character of the VMZ by injecting RNA into the ventral side of both ventral blastomeres at the 4- cell stage. The VMZ and control dorsal marginal zone (DMZ) were dissected at early gastrula (stage 10.25) and harvested at mid-neurula (stage 17) for RT-PCR analysis (Fig. 6A).

At the time of dissection, misexpressed derrière induced a secondary invagination on the ventral side of the embryo (not shown) similar to that observed after VegT misexpression (Lustig et al., 1996b). At the time of harvest, globin-injected VMZ explants remained round (0% elongation; n=30) while DMZ explants (100%; n=20) and VMZ explants expressing derrière (92%; n=24) elongated. In RT-PCR analyses (Fig. 6B), m-actin was used as an indicator of dorsalization while Xbra was a marker for mesodermal fates. While cultured DMZ explants expressed high levels of m-actin (lane 1), globin- injected control VMZ explants did not (lane 2). In contrast VMZ explants expressing derrière showed strong m-actin expression (lane 3).

These data showed that derrière could activate dorsal- specific fates in the ventral marginal zone and was likely to induce a secondary axis through this activity. From these experiments, is not clear whether derrière acts during gastrula stages to dorsalize the pre-existing VMZ or whether it acts earlier to direct formation of a partial DMZ on the ventral side of the embryo.

The ability of derrière to suppress expression of otx2 in the forebrain and eyes, and to reduce formation of the ectodermal cement gland, suggested that derrière might suppress head formation by directly altering ectodermal patterning. We therefore asked whether Derrière could alter specification of the ectoderm, and compared it to the effects of FGF, a putative direct neural inducer or modifier (Kengaku and Okamoto, 1995; Lamb and Harland, 1995; Kolm and Sive, 1997). Dorsal ectoderm that we had previously shown to be specified as neurectoderm (Kolm et al., 1997) was isolated from mid-gastrula embryos and incubated in Derrière protein, bFGF or BSA as a control (Fig. 7A). In order to determine whether any effects of Derrière were secondary to mesoderm induction, we used two stages of dorsal ectoderm in these assays: one (stage 11) is still competent to respond to Activin to form mesoderm while the other (stage 11.5) is not (Green et al., 1990). Explants were harvested when control embryos reached tailbud (stage 22) for assay by RT-PCR. A representative experiment (of five) is shown in Fig. 7B. Stage 11 dorsal ectoderm failed to express the posterior markers HoxA7 and Xcad3 (Northrop and Kimelman, 1994) which are expressed in both mesoderm and ectoderm, or the paraxial mesodermal marker m-actin (lane 1), but did express the anterior cement gland marker XCG, the midbrain marker en-2 and the hindbrain marker krox20. bFGF induced expression of HoxA7 and Xcad3 (lane 2), but did not induce m-actin, and did not alter expression of XCG, en-2 and krox20. In contrast, Derrière protein activated expression of HoxA7, Xcad3 and m-actin and increased expression of the posterior neural markers en-2 and krox20 (lane 3). The combination of bFGF and Derrière gave results indistinguishable from those of Derrière alone (lane 4). These data indicated that Derrière could induce mesoderm and concomitantly induce expression of posterior neural markers in anterior neurectoderm.

Ectodermal explants of a slightly later stage (stage 11.5) expressed XCG, en-2 and krox20 after culture, but failed to express HoxA7, Xcad3 or m-actin (lane 6). bFGF did not induce any posterior markers, had no effect on XCG expression and slightly decreased both en-2 and krox20 expression (lane 7). At this stage, Derrière could no longer induce expression of m-actin, HoxA7 or Xcad3 (lane 8). Derrière did not alter expression of XCG, and slightly reduced en-2 and krox20 expression. In contrast to the effects of single factor treatments, treatment of stage 11.5 ectoderm with Derrière plus bFGF did not lead to a decrease in en-2 and krox20 expression (lane 9).

We conclude that Derrière can activate posterior neural- specific marker gene expression, but only by inducing mesoderm in competent tissue that is present in stage 11 neurectoderm, but absent from stage 11.5 neurectoderm. In this assay, bFGF could also not directly alter A/P neurectodermal patterning. The data suggest that, in the whole embryo,derrière suppresses head formation by altering mesodermal patterning.

In order to ablate derrière function, we constructed a dominant interfering mutant of derrière (Cm-derrière for cleavage mutant). In this mutant residues, 237-240 are changed from RAKR to GVDG (Fig. 1A and Materials and Methods), which we anticipated would block cleavage of the Derrière pro- protein to the mature form. Since TGF-βs are cleaved after dimerization, a non-cleavable monomer can prevent cleavage of normal monomers and thereby act as a dominant negative protein (Fig. 8A), a strategy that has been successfully used to ablate BMP function (Hawley et al., 1995).

Initially we asked whether Cm-derrière would inhibit induction of gene expression by derrière, using different ratios of derrière:Cm-derrière in an animal cap assay (Fig. 8B). Relative to globin-injected caps (Fig. 8C, lane 1), derrière induced high levels of HoxA7, Pintallavis, m-actin, Xbra, Xsox17α and endodermin (Edd, Sasai et al., 1996, lane 2). At a ratio of 1:4 wild-type:Cm-derrière, expression of m-actin and Edd was ablated (lane 3). At a ratio of 1:10, expression of Pintallavis and Xsox17α was strongly reduced (lane 4). At 1:20, HoxA7 and Xbra expression was reduced (lane 5) and at 1:50, expression of all six genes was abolished (lane 6). These data are consistent with the dose-response to derrière (Fig. 3C) showing that m-actin is the most sensitive and HoxA7 the least sensitive to the level of active Derrière protein. These data showed that Cm-derrière was an effective inhibitor of derrière activity.

We next used the animal cap assay to ask whether the effects of Cm-derrière were specific for derrière or whether it also inhibited other TGF-β family members (Fig. 8B). RNAs encoding various TGF-β family members were injected into 2- cell embryos either alone or with Cm-derrière in ten-fold mass excess. A representative experiment (of at least two) is shown in Fig. 8D. Cm-derrière alone did not activate expression of the markers tested (lane 1), and while derrière strongly activated pintallavis, m-actin and Xbra (lane 2), co-injection of derrière with Cm-derrière suppressed m-actin and Pintallavis gene expression (lane 3). In contrast, Cm-derrière did not attenuate the ability of other TGF-β family members tested to induce Pintallavis, m-actin and Xbra. Inducers tested were BVg1 (lanes 4 and 5), activin (lanes 6 and 7), Xnr1 (lanes 8 and 9), Xnr2 (lanes 10 and 11), Xnr3 (lanes 12 and 13) and Xnr4 (lanes 14 and 15). In contrast to published data (Smith et al., 1995; Hansen et al., 1997), we found that Xnr3 induced Xbra (lane 12). It is also interesting that the expression of some mesodermal markers was increased in the presence of a test TGF-β and Cm-derrière-Xbra in the cases of derrière (lanes 2 and 3) and Xnr2 (lanes 10 and 11), and m-actin in the case of Xnr3 (lanes 12 and 13). We do not presently understand the mechanism of this up-regulation.

We also tested the ability of Cm-derrière to interfere with animal cap elongation which is characteristic of dorsal mesodermal fates (Fig. 8E). Animal caps removed from embryos injected with globin did not elongate (panel a; 0% of injected embryos, n=34) and no change in morphology was observed after co-injection of Cm-derrière and globin (panel b; 0%, n=45). After derrière injection, 100% of caps elongated (panel c; n=58), and elongation was almost completely blocked after co-injection of derrière and Cm-derrière at a 1:10 ratio (panel d; 2%, n=64). The frequency of cap elongation induced by BVg1 (panel e; 97%, n=59) was not altered by Cm-derrière (panel f; 97%, n=59). The frequency of cap elongation induced by activin (panel g; 100%, n=18) was also unaffected by addition of Cm-derrière (panel h; 97%, n=29) although the extent of elongation was slightly reduced. Xnr1-induced cap elongation (panel i; 100%, n=58) was somewhat inhibited by Cm-derrière (panel j; 70%, n=59), while both the frequency and extent of elongation induced by Xnr2 (panel k; 100%, n=59) was only slightly decreased by co-expression of Cm-derrière (panel l; 93%, n=59). The frequency of Xnr3-induced cap elongation was low (panel m; 9%, n=45) and no elongation was observed after co-injection of Cm-derrière (panel n; 0%, n=45). The frequency of cap elongation induced by Xnr4 (panel o; 96%, n=45) was not affected by Cm-derrière co-expression, but the extent of elongation was reduced (panel p; 91%, n=45).

In summary, these data showed that Cm-derrière severely attenuates the activity of derrière in both molecular and morphological assays. The activity of Cm-Derrière was highly preferential for Derrière protein as compared to other members of the TGF-β family, with no interference seen in a molecular assay, and only slight interference for Xnr1 seen in a cap elongation assay. These data indicated that Cm-derrière was a useful reagent with which to analyze the in vivo function of the Derrière protein.

We next used the Cm-derrière construct to analyze the effect of ablating Derrière activity in the whole embryo. Wild-type and albino embryos were injected with Cm-derrière RNA or with control globin RNA, along with lacZ RNA as lineage tracer (Fig. 9A). Embryos were harvested at various stages either for in situ hybridization or for morphological analysis.

Fig. 9B shows representative embryos after Cm-derrière injection. During mid-gastrula stages, a blastopore failed to form where Cm-derrière was expressed (panel a, black arrowhead; 91%; n=56, compared to globin-injected, panel b; 0%; n=38), suggesting that involution had failed in this region. During neurula stages, the blastopore remained open in Cm- derrière-injected embryos (panel c; 100%; n=45, compared to globin-injected, panel d; 0%, n=32). Embryos injected with Cm-derrière failed to develop normal posterior morphology (panel e; 92% of injected embryos, n=131). While the head appeared largely normal even with lineage tracer in the head region, with eyes and cement gland visible, no somites were apparent, the neural tube remained open (as a result of the open blastopore) and no tail developed (compare to globin-injected control of equivalent age; panel f; 0%, n=129). In order to determine the anterior extent of Cm-derrière effects, we examined expression of XCG, en and krox20. As shown in panel g, all three genes were still expressed anterior to the open neural plate, although their patterning was somewhat disrupted (100%; n=12, compared to control in panel h; 0%; n=12). However, m-actin expression was almost completely abolished on the injected side of the embryo (panel i; 100%, n=27) compared to globin-injected controls (panel j; 0%, n=10), consistent with a role for derrière in posterior mesoderm induction. This phenotype looked similar to that observed after injection of a dominant negative FGF receptor (Amaya et al., 1991; Kroll and Amaya, 1996), dominant negative Xbra (Conlon et al., 1996) and Brat (VegT) (Horb and Thomsen, 1997) constructs. In support of the specificity of the Cm- derrière phenotype, we found that a dominant negative Activin ligand, Cm-Activin (Hawley et al., 1995), suppressed head formation but allowed normal posterior development (not shown). This head suppression phenotype was consistent with a results reported after expressing a dominant negative Activin receptor (Dyson and Gurdon, 1997).

In order to control for non-specific effects of Cm-derrière, we asked whether the truncated posterior phenotype could be rescued by native derrière and found that it could, when a ratio of 1:10 derrière:Cm-derrière was co-injected (panel k; 83%, n=77). Interestingly, the rescued embryos looked completely normal and did not display either the microcephaly or secondary axis observed after injection of derrière alone. Additionally, since derrière and VegT have almost identical expression patterns, we asked whether the Cm-derrière phenotype could be rescued by co-injection of VegT and found that it could (panel l; 77%, n=194).

In summary, these data showed that, in whole embryos, Cm- derrière prevents normal posterior development. The phenotype caused by Cm-derrière includes failure of blastopore formation and closure, likely reflecting a failure of normal involution and leading to an open region where the neural tube should have been. No tail formation was observed. Additionally, paraxial mesoderm development was severely reduced as indicated by somite morphology and muscle gene expression. This phenotype could be rescued by co-injection of either wild-type derrière or VegT RNAs, suggesting that a regulatory interaction exists between derrière and VegT.

In this report, we describe derrière, a zygotically expressed TGF-β family member that is closely related to Vg1. derrière appears to be induced as part of a regulatory loop involving the T box gene VegT. Both gain-of-function and loss-of-function assays indicate that derrière plays a pivotal role in posterior development of Xenopus.

derrière is a new member of the Vg family of genes that encode TGF-β molecules (Kingsley, 1994), most similar to Xenopus Vg1. Other members of this family have been found in chick (cVg1) and zebrafish (zVg1). All Vg family members share the ability to induce mesoderm and a secondary axis (Seleiro et al., 1996; Shah et al., 1997). Interestingly, cVg1, like derrière, is expressed posteriorly, suggesting that the function of derrière and cVg1 may be similar. In contrast, zVg1 expression is maternal and ubiquitous, although, since the protein persists through gastrulation, zVg1 may have a later function (Helde and Grunwald, 1993; Dohrmann et al., 1996).

The expression of Xenopus Vg1 RNA suggests a role for this gene in mesoderm and endoderm induction, however, the Vg1 proprotein is inefficiently processed and native Vg1 RNA displays no activity when overexpressed (Dale et al., 1993). The mesoderm-inducing and axis-duplication activity of derrière is similar to that of BVg1 (Dale et al., 1993; Thomsen and Melton, 1993), raising the possibility that BVg1 may partially phenocopy the effects of ectopic derrière. However, several results suggest that derrière and Vg1 have different activities. First, unlike derrière, BVg1 does not suppress head formation (Dale et al., 1993) and, second, lineage tracer co-injected with BVg1 is found exclusively in the endoderm, consistent with the observation that high concentrations of BVg1 induced endoderm and lower concentrations induced mesoderm (Thomsen and Melton, 1993; Henry et al., 1996). We find that lineage tracer co-injected with derrière is found in both mesoderm and endoderm, and consistently, derrière induces mesodermal and endodermal markers at similar concentrations. Third, dominant negative ligands made to the mature region of Vg1 (Joseph and Melton, 1998) ventralize the embryo, distinct from the posterior suppression that we observed with Cm-derrière.

Gain-of-function assays in the whole embryo suggest that derrière is an inducer of posterolateral fate. This suggestion is supported by the suppression of head formation that is likely to be a consequence of ectopic derrière expression in the deep marginal zone cells that constitute the head organizer, from which derrière expression is excluded by midgastrula. Head reduction is accompanied by a decrease in otx2 expression and concomitant repositioning of en-2 expression to the front of the embryo, suggesting a reprogramming of forebrain to more posterior regions. derrière is not able to induce head formation when misexpressed ventrally, further supporting its inability to induce head regions.

The secondary axis induced by overexpression of derrière in the ventral mesoderm also lacked axial tissues, including notochord (not shown) and spinal cord. However, we observed extensive muscle formation, a tissue derived from the posterolateral (paraxial) mesoderm. Neural tissue was also induced, presumably by paraxial mesoderm which is a neural inducer (Jones and Woodland, 1989). Why is only posterolateral tissue induced in the whole embryo, when derrière is able to activate expression of both anterior and posterior axial genes in animal caps? One possibility is that posteriorizing factors present in the embryo, but not in the animal cap, limit derrière activity.

Many of the effects of derrière on the whole embryo are likely to be through its activity as a mesoderm inducer or modifier. Consistent with this proposal, head suppression and secondary axis formation were affected only when derrière was misexpressed in the marginal zone. Further, derrière was able to activate dorsal fates in the ventral marginal zone, presumably directing formation of a second organizer, albeit one that lacked head- and notochord-inducing activity.

The ability of ectopic derrière to reduce expression of the anterior neural marker otx2 and to decrease cement gland formation in whole embryos, as well as to induce expression of the neural marker N-CAM in secondary axes suggested that derrière may directly pattern the neurectoderm. In support of a potential role in neural patterning, derrière is expressed beneath the neurectoderm as A/P patterning is taking place. However, derrière does not act directly on induced neurectoderm. In mid-gastrula neurectodermal explants, derrière strongly increased expression of both a midbrain and hindbrain marker, but only when this tissue could still be induced to form mesoderm. The effects of derrière on neurectodermal patterning are therefore likely to be mediated by modifying the mesoderm responsible for neural patterning. We note, however, that, under similar assay conditions, FGF is unable to posteriorize neurectoderm, whereas under other assay conditions it can (Kengaku and Okamoto, 1995; Lamb and Harland, 1995). The posteriorization activity of derrière may therefore depend on the particular experimental condition.

The expression pattern of derrière appears identical to the zygotic expression of the T box gene, VegT. Several lines of evidence suggest that derrière and VegT form a regulatory loop. VegT is able to induce derrière expression in animal caps and, reciprocally, derrière induces VegT expression. Both derrière and VegT give similar phenotypes in gain-of-function assays, while a dominant negative Derrière protein, Cm-Derrière, gives a phenotype that appears to be similar to that of a VegT dominant negative (Brat-En^R, Horb and Thomsen, 1997). Consistently, VegT is able to rescue the effects of Cm-Derrière protein. Maternal expression of VegT is required for endoderm formation (Zhang et al., 1998), while TGF-β signaling is required for normal endoderm and mesoderm formation (Kimelman and Griffin, 1998). derrière is a good candidate to be an endogenous TGF-β directly activated by maternal VegT. Later, derrière may be required for zygotic VegT expression.

In addition to VegT, the eFGF, FGF9 (Song and Slack, 1996) and part of the Xbra expression domains overlap with that of derrière. Since derrière can induce eFGF in animal caps, and since both VegT and Xbra can also be induced by FGF (Isaacs et al., 1994; Schulte-Merker and Smith, 1995; Lustig et al., 1996b; Horb and Thomsen, 1997), the data suggest that derrière is part of an FGF-dependent regulatory loop, perhaps acting through VegT.

Cm-Derrière preferentially inhibited Derrière activity relative to a panel of other TGF-βs tested. The ability of Cm-derrière to decrease Xnr1 activity in an animal cap elongation assay indicates that Derrière and Xnr1 may form active heterodimers, or that Xnr1 directs elongation by inducing derrière. The phenotype of Cm-derrière-expressing embryos consists of a normal head with a severely reduced trunk and tail. This is entirely different from that of embryos expressing Cm-activin (not shown) or a dominant negative activin receptor (Dyson and Gurdon, 1997) where head suppression is observed. The Cm-derrière phenotype resembles that of embryos expressing a dominant negative FGF receptor (Amaya et al., 1991; Kroll and Amaya, 1996), a dominant negative Xbra protein (Conlon et al., 1996) or a dominant negative Brat protein (Horb and Thomsen, 1997). However, since unlike derrière, FGF cannot direct formation of a secondary axis, derrière must have a unique activity that is not shared with FGF. In addition to FGF signaling, posterior determination in Xenopus appears to involve the wnt and retinoid pathways (Blumberg et al., 1997; Kolm et al., 1997; McGrew et al., 1997). It is not yet clear how the TGF-β pathway and these pathways interact.

Why has the requirement for TGF-β signaling in formation of posterior structures previously been missed? One possibility is that Derrière is the only TGF-β ligand specifically required for posterior determination and this requirement could only have been picked up by specifically inhibiting derrière. Since a general inhibition of TGF-β signaling in Xenopus prevents all mesoderm formation, resulting in anaxial embryos (Hemmati-Brivanlou and Melton, 1992), a later posterior requirement for TGF-β signaling would have been masked. Second, Derrière signaling appears to require an intact FGF signaling pathway as has been demonstrated for Activin (Cornell and Kimelman, 1994; La Bonne and Whitman, 1994). Thus, since FGF signaling is required for posterior tissue formation, in the absence of an active FGF pathway, the Derrière pathway might also not be active.

Our data suggest a model for derrière function (Fig. 10). Maternal VegT activates expression of derrière soon after the mid-blastula transition. Soon after this, a regulatory loop is established between derrière, zygotically expressed VegT, eFGF/FGF9 and possibly Xbra. This loop maintains expression of derrière in the marginal zone, but fails to maintain expression vegetally, perhaps because of a lack of vegetal FGF signaling (Cornell et al., 1995). Dorsally, derrière expression is highest due to the activity of dorsal-specific factors. At early gastrula, Derrière, along with other factors, may induce a dorsoventral array of mesodermal genes in the marginal zone. As derrière expression becomes posteriorly restricted during gastrulation, it may specifically activate genes expressed in the this region of the embryo including those later determining paraxial mesoderm (muscle). Since derrière RNA is not expressed in the somites, it must either activate muscle formation during gastrulation or Derrière protein must persist after its RNA has disappeared. We speculate that derrière acts in conjunction with FGF signaling to promote formation of posterior regions, through downstream transcription factors such as Xbra and VegT, that can be activated by both derrière and FGF. However, derrière must also have a unique activity, likely involving the activation of other factors, to account for its ability to direct secondary axis formation where FGF cannot.

In summary, our results indicate that posterior development in Xenopus requires activity of the TGF-β family member, Derrière. Future directions include analyzing the regulatory loop between FGF, VegT and derrière suggested by these studies, and addressing further the mechanism by which derrière regulates posterior patterning.

We thank Marc Kirschner, Doug Melton, Mary Lou King, Chris Wright, Bill Smith and Jeremy Green for gifts of plasmids. Special thanks to Jen Stover for assisting the yeast selection, to Cheryl Evans and Matt Benasutti for sequencing support, to Chris Dagdigian for help with bioinformatics and to Monique Davies for technical support at Genetics Institute; to Vladimir Apekin for expert frog care, to Jeanne Sweeney-Reis for sections, to Ross Silburn for technical assistance and to Liuda Ziaugra for sequencing support at the Whitehead Institute. We are grateful to Tony Celeste, Peggy Kolm and members of our laboratories for helpful discussions. B. I. S. was supported by a NIH postdoctoral fellowship, and by a fellowship from Genetics Institute. H. L. S. was the Latham Family Career Development Professor at MIT, and a recipient of a NSF Young Investigators Award. This work is supported by the NSF, Genetics Institute and the Human Frontiers Science Foundation Program.