A Maternal Smad Protein Regulates Early Embryonic Apoptosis in Xenopus laevis

Abstract

We identified cDNAs encoding the Xenopus Smad proteins most closely related to mammalian Smad8, and we present a functional analysis of this activity (also referred to recently as xSmad11). Misexpression experiments indicate that xSmad8(11) regulates pathways distinct from those regulated by the closely related xSmad1. Embryos that develop from eggs depleted of xSmad8(11) mRNA fail to gastrulate; instead, at the time of gastrulation, they initiate a widespread program of apoptosis, via a CPP32/caspase 3 pathway. Embryos that avoid this fate display gastrulation defects. Activation of apoptosis is rescued by expression of xSmad8(11) but not xSmad1. Our results demonstrate an embryonic requirement for Smad8(11) activity and show that a maternally derived Smad signaling pathway is required for gastrulation and for mediating a cell survival program during early embryogenesis. We suggest that xSmad8(11) functions as part of a maternally derived mechanism shown previously by others to monitor Xenopus early embryonic cell cycles.

The specification and patterning of embryonic regions corresponding to primary germ layers (ectoderm, mesoderm, and endoderm) rely heavily on maternal cues that are deposited and sometimes localized during oogenesis (31, 47, 56, 59). Some maternal transcripts persist even after the midblastula transition (MBT), when zygotic transcription, asynchronous cell-cycles, and acquisition of cell motility all begin (21). After MBT, development remains dependent on maternally derived cues, but now in association with zygotic factors that regulate induction and cell specification (5, 29). A less well defined shift occurs when maternal factors are no longer required; this is sometimes referred to as the early gastrula transition (EGT), or maternal-to-zygotic transition (discussed in reference 23). Until EGT, maternal components provide poorly understood survival cues that inhibit an apopototic pathway that is also maternally derived (19, 41).

Bone morphogenetic proteins (BMPs) expressed after MBT influence the development of all three germ layers, and antagonism of BMP signaling is now recognized as a dominant feature of mesoderm and ectoderm patterning (12, 16, 20). Experiments inhibiting zygotic BMP function by ectopic expression of chordin, noggin, gremlin, or Wnt8 (2, 12, 30, 38) have demonstrated the critical role of BMPs in the development of ventral mesoderm and ectoderm during gastrulation. In presumptive ectoderm, BMPs activate expression of an epidermal differentiation program; in their absence, neural-system-specific transcription factors are expressed (49, 50). The phenotype of the mouse null mutation confirms that BMP4 is required for mesoderm development (51), although early embryonic lethality and potential compensation by related family members limit interpretation of the phenotype. How maternal components associated with the BMP pathway contribute to organizing the body plan and regulating cell survival remains entirely unclear, although it is expected that Smad proteins play a role in mediating the BMP signal.

The Smad proteins are intermediate signaling factors that function downstream of the transforming growth factor beta (TGF-β), activin, and BMP receptors (8, 24). Smad activity is intricately regulated through further interactions with a large number of embryonic signaling pathways (52). Studies with Xenopus laevis and zebrafish show nonlocalized transcripts and proteins encoding Smad1, -2, -3, and -5 in unfertilized eggs, and expression patterns remain ubiquitous during early stages of development (9, 32). However, Smad activity is thought to be controlled primarily by posttranslational mechanisms. Smads of the receptor-associated subclass are phosphorylated; Smad2 and -3 are phosphorylated following activin stimulation, and Smad1, -5, and -8 are phosphorylated in response to BMPs (4, 17).

The normal function of Smad8 has not been reported previously. However, the mammalian Smad8 proteins are phosphorylated by constitutively active forms of ALK-2, ALK-3, and ALK-6 (7, 26), which are all type I serine/threonine kinase receptors associated with BMP signaling. Like Smad1 and Smad5, Smad8 associates with Smad4 following BMP-stimulated phosphorylation and activates transcription of BMP-responsive reporter genes (26, 57). Due to a lack of further information, the early embryonic functions of Smad1, -5, and -8 have been considered to be largely redundant with respect to ventral mesoderm patterning. We isolated the Xenopus cDNAs encoding the proteins most closely related to mammalian Smad8, one of which has since been submitted to GenBank as Xenopus Smad11 (14). Here we present a functional analysis of the endogenous activity. We show that depletion of maternal transcripts encoding xSmad8(11) activates programmed cell death at EGT, in addition to affecting ectoderm development and gastrulation. Our data indicate that Smad8(11) is a mediator of the maternally derived regulatory pathway that inhibits apoptosis prior to establishment of normal cellular checkpoints.

MATERIALS AND METHODS

Isolation of Xenopus Smad8-related cDNAs.

Degenerate oligomers were designed consistent with sequences within the highly conserved Mad homology domains MH1 (Cys-Ile-Agn-Pro-Tyr-His) and MH2 (Thr-Pro-Cys-Trp-Ile-Glu). The sequences of the primers were 5"-TGC ATC AAC CCI TAY CA (TE576) and 5"-CTC GAT CCA RCA IGG IGT (TE577), respectively, where I stands for dI, Y stands for dC or dT, and R stands for dG or dA. Individual cDNA samples were prepared by random priming using RNA isolated from stage 10 ventral marginal zone explants, stage 28 ventral blood island tissues, or circulating blood cells from 1-week-old tadpoles. PCR was performed with the degenerate primers, and the expected 950-bp product was cloned into a TA vector (Invitrogen). Southern blotting experiments with stringent washing conditions were used to identify cDNAs distinct from xSmad1. One such cDNA was used to screen a stage 13 embryo cDNA library (from T. Sargaent, National Institutes of Health) by hybridization, and a single cDNA that extended the sequence was isolated. Both 5" rapid amplification of cDNA ends (5" RACE) and 3" RACE were used to complete the sequence beyond the ATG initiation and TAA stop codons. New PCR primers were then designed to amplify the entire open reading frame. A single reverse transcription-PCR (RT-PCR) product was cloned, and random isolates were sequenced. Three nearly identical cDNA sequences were identified; we designated them xSmad8A, xSmad8B, and xSmad8C. These cDNAs differ by only 6 (xSmad8A versus -8B), 8 (xSmad8A versus -8C), and 4 (xSmad8B versus -8C) of 466 amino acids. The xSmad8B sequence is nearly identical to a GenBank sequence referred to as xSmad11 (14). For the purposes of this study we refer to the cDNAs collectively as xSmad8(11), since each of the antisense oligomers and RT-PCR experiments would detect all three gene products.

Misexpression and antisense techniques.

The dominant-negative xSmad1 and xSmad8A constructs were generated by PCR, based on the same mutation characterized for mSmad8 (26), and the mutations were confirmed by sequencing. Full-length or mutant cDNAs were subcloned into the pCS2+ expression vector. The xSmad1 cDNA was kindly provided by G. Thomsen. Each of the cDNA constructs was linearized and used to synthesize capped mRNAs in vitro by using the mMessage mMachine kit (Ambion). RNA integrity and concentration were confirmed by gel electrophoresis and by in vitro translation. Fertilized eggs were injected with 2.3 ng of purified mRNA at the 1-cell stage; control embryos were injected either with equivalent volumes of water, with an irrelevant control RNA that does not encode functional protein, or with an mRNA encoding β-galactosidase (each control yielded normal embryos).

The host transfer technique was used essentially as described previously (61) to eliminate maternal xSmad8(11) transcripts. Each oligomer was diluted in water, and 4.6 nl was injected into the cytoplasm of each oocyte above the grayish equatorial line. Several oocytes were always removed to confirm successful and specific depletion of xSmad8(11) mRNA by RT-PCR analysis. Embryos were staged as described previously (35). For rescue experiments, 1 ng of xSmad8A or xSmad1 mRNA was injected into fertilized eggs derived from the oligomer-injected oocytes.

The sequences of the antisense oligomers (5" to 3") are as follows: oligomer 1, GGGGACGAAGAAGAAAAATGGGCAGAGAAAG (TE643); oligomer 2, GGCGTGCATTGGATTTGCTGTGTCTACC (TE658); oligomer 3, AGTCCACCGCTTTCTCTGCCCATTT (TE637); oligomer 4, GGTTGCCCTGGACAACTTAAAGCC (TE629); oligomer 5, CAAACCCGGCAATAGATGACATGGGAG (TE659); oligomer 6, GGCATTAGTGGTTCATTATTCAGCGAGGTG (TE644); oligomer 7, TGCATTGTGTGGCATTAGTGGTT (TE625); oligomer 8, CGTCTCTGTGATCTGGTAC (TE680); oligomer 9, GACAGAACACCAATGCAA (TE687); oligomer 10, GCGGGTGTTTTCAATAGTG (TE649); oligomer 11, CTGGGCAAATAGTTGGT (TE682); and oligomer 12, AGAGTTTACGATACGGAAGAGATTGGAT (TE650).

Gene expression analysis by semiquantitive RT-PCR and whole-mount in situ hybridization.

Oocytes and embryos were homogenized in RNA extraction buffer (50 mM Tris-HCl [pH 7.4], 100 mM NaCl, 30 mM EDTA, 1% sodium dodecyl sulfate, and 1 mg of proteinase K/ml) and were incubated at 37°C for 1 h. After phenol-chloroform extraction, nucleic acids were digested with 1 U of RQ1 DNase (Promega) at 37°C for 3 h. Typically, 0.25, 0.50, or 1.00 μg was used to synthesize cDNA at 37°C for 3 h in the presence of Moloney murine leukemia virus reverse transcriptase (Gibco BRL) and p(dN)6 random primers. PCR was performed by using Z-Taq (Panvia/Takara Shuzo Co.), and products were analyzed to confirm the semiquantitative nature of the assay (dependent on cycle number). For xSmad8(11) PCR, denaturing was carried out at 94°C for 30 s, with annealing at 68°C for 30 s and extension at 72°C for 90 s. Under these conditions, PCR products were generated with a linear response to both cycle number and input RNA (29 cycles). The primers used for this study were designed to flank the positions of antisense oligomers; for xSmad8(11), the forward primer (TE643) was GGGACGAAGAAGAAAAATGGGCAGAGAAAG and the reverse primer (TE645) was TCCAATGTGTCTGCGGGTGTTTTCAATAGT. As a control for specificity, samples were also monitored for levels of xSmad1 with annealing at 59°C (28 cycles). For xSmad1, the forward primer (TE718) was GTCTTGCCACCTGTCCTTGTTCCAC and the reverse primer (TE719) was CGAGACCGAGGAGATGGGATTATGG. Other primer sequences are described in the Xenopus Molecular Marker Resource (http://vize222.zo.utexas.edu/Marker_pages/primers.html) or in the reference given in parentheses after the primer name, as follows: xGATA-2 (58), xVent-1 (15), xVent-2 (37), xChordin (39), and xSox17-β (22).

For analysis by whole-mount in situ hybridization, albino embryos were fixed as described previously (3) at 4°C overnight and were rehydrated in 100% ethanol. An antisense probe for xSmad8A incorporating digoxigenin-UTP was prepared by in vitro transcription and used to analyze transcript patterns as described previously (3). Based on the pattern in whole-mount and Northern blotting experiments, the probe does not cross-react with transcripts encoding related Smads including xSmad1. BM-purple alkaline phosphatase substrate (Boehringer Mannheim) replaced nitroblue tetrazolium (NBT)/5-bromo-4-chloro-3-indolylphosphate (BCIP) as a substrate for color development. Control sense strand probes did not generate detectable signals.

Histology, terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end labeling (TUNEL) staining, and caspase assays.

Embryos were fixed in MEMFA buffer (0.1 M MOPS [pH 7.4], 2 mM EGTA, 1 mM MgSO₄, and 3.7% formaldehyde) containing 0.025% glutaraldehyde for 30 min at room temperature. The samples were rehydrated through a graded series of ethanol, cleared in xylene for 1 h, and embedded in Paraplast (Oxford) at 60°C. Sections were cut at 10 μm and placed on glass slides. After being deparaffinized in xylene twice for 2 min, they were rehydrated in a graded ethanol series, stained in Harris modified hematoxylin (Fisher Scientific) for 1 min, and counterstained in 0.5% eosin Y for 1 min.

Whole-mount TUNEL staining of Xenopus embryos was carried out as described previously (18), except that BM-purple AP substrate (Boehringer Mannheim) replaced NBT/BCIP. As a positive control, 60 pg of α-amanitin was injected into both blastomeres of 2-cell-stage embryos (18, 41). After the TUNEL reaction, 14-μm sections were cut and placed on slides. Caspase assays were performed by using the fluorometric-based CaspACE Assay System (Promega) according to the manufacturer's directions. Embryos derived from control or antisense oligonucleotide-injected oocytes were collected following host transfer and fertilization and were lysed at stage 10.5 in hypotonic buffer by freeze-thawing. Protein concentrations were determined (Bio-Rad), and equal amounts of total protein were used, with a 96-well format, to measure interleukin-1β-converting enzyme/caspase 1 or CPP32/caspase 3 activity in the absence or presence of specific inhibitors. Specific activities for each caspase subclass were calculated by comparison to standard calibration curves.

Nucleotide sequence accession number

The sequences for xSmad8A, xSmad8B, and xSmad8C have been submitted to GenBank under accession no. AF464927, AF464928, and AF464929, respectively.

RESULTS

Identification of Xenopus cDNAs highly related to mammalian Smad8, with conserved sequences associated with the BMP-regulated subfamily.

In a search for components of the Xenopus TGF-β signaling pathway, we performed RT-PCRs using degenerate oligomers consistent with the MH1 and MH2 coding sequences (Mad homology domains), which are highly conserved among Smad proteins (see Materials and Methods). Three nearly identical but distinct sequences of xSmad cDNAs were isolated, with an overall amino acid sequence identity of at least 98%, and very closely related to mammalian Smad8. During the course of this study, an xSmad sequence essentially identical to those we isolated was submitted to GenBank (AF329832) and referred to by Fortuno et al. (14) as xSmad11. We note that no Xenopus cDNA more closely related to mammalian Smad8 is known, nor has a mammalian cDNA more closely related to xSmad11 been described. However, it is possible that a larger subfamily of related Smads exists (Smad1, -5, -8, -11, etc)., so that lacking further information, we refer to the activity we describe here as xSmad8(11).

The predicted protein encoded by the xSmad8(11) cDNA is shown, in comparison to related Smad proteins, in Fig. 1A (only the Smad8A variant is shown; the GenBank sequence for Smad11 is essentially identical to the Smad8B sequence). The sequence is 466 amino acids, including a carboxyl SSVS motif, identical to that in other BMP-associated Smad family members. Within the linker region between the conserved MH1 and MH2 domains, there is a 34-amino-acid stretch that is missing in the reported rodent Smad8 proteins, although a related sequence is present in Smad1 and Smad5 from other species. Indeed, two human Smad8 cDNAs have been reported, one of which (hSmad8a) contains this linker sequence (48), apparently from differential splicing. The Xenopus protein is 88% identical to human Smad8a and clearly falls into the BMP subclass of receptor-regulated Smad proteins (Fig. 1B).

FIG. 1.

Sequence comparison of Smad8-related proteins. (A) The sequence of xSmad8(11) is aligned with those of other members of the BMP receptor-regulated Smad subfamily. Shown is the sequence of xSmad8A, which is one representative clone of three nearly identical isolates. Smad8B (not shown) is essentially identical to the cDNA referred to as Smad11 by Fortuno et al. (14). For comparison, xSmad2 is also included. A 34-amino-acid stretch that is missing in the reported sequences for rodent Smad8 is indicated by a horizontal line. The sequences used to design antisense oligonucleotides to deplete the maternal message are indicated by arrows and numbered. Note that sequence 1 is for a control sense strand oligomer. (B) The sequences were used to generate a phylogenetic tree that places xSmad8(11) in the BMP-associated subclass of regulatory Smads.

Xenopus Smad8(11) is expressed as a maternal mRNA and throughout development.

Because xSmad8(11) is transcribed at low levels, we used a semiquantitative RT-PCR assay to analyze mRNA levels during embryogenesis (Fig. 2A). The quantitative nature of the assay is demonstrated in the far right panel of Fig. 2A. The xSmad8(11) mRNA is present in unfertilized eggs, and transcripts are detected throughout development to swimming-tadpole stages. Transcript levels in the maternal mRNA pool are stable through the time of zygotic transcription at MBT (stage 8). During tail bud stages (21-23) the levels decline, but they increase again through tadpole stages.

FIG. 2.

The xSmad8(11) transcript is present as a maternal mRNA, and the gene is expressed throughout development. (A) Transcript levels for xSmad8(11) were determined by semiquantitive RT-PCR. Transcripts are present in unfertilized eggs (U) and throughout development to swimming tadpoles. Stages are indicated above the left panels. EF1-α expression is shown as a loading control and to indicate the timing of initiation for zygotic transcription after MBT. The RT(−) reactions demonstrate that the product is entirely dependent on the RNA template. The samples in the right panel (from stage 2 embryos) confirm that the analysis is semiquantitative. mRNA levels were quantified following phosphorimager analysis and are graphed below the panel. (B) Whole-mount in situ hybridization was used to determine the spatial pattern of xSmad8(11) transcripts during embryogenesis. Embryos analyzed using the antisense probe are arranged as indicated by stage. Low levels were detected throughout the embryos. Arrows for stages 34 to 38 indicate the relatively high spatial expression in the branchial arch region. In the lower right corner, two control embryos (top, stage 22; bottom, stage 36) that were hybridized to a sense strand probe are shown.

Spatial analysis performed by using whole-mount in situ hybridization demonstrates that the mRNA is widely expressed (Fig. 2B). For these experiments we used the xSmad8A probe, although it is expected that this will cross-hybridize with the others. We found that maternal xSmad8(11) transcripts are not substantially localized prior to MBT, which was confirmed in RT-PCR experiments using RNA from isolated animal and vegetal blastomeres of stage 6 embryos (data not shown). There is also no significant difference in transcript levels between the dorsal and ventral halves. At later stages of development, some patterns emerge. For example, during neurulation the mRNA is present in a continuous sheet of the archenteron roof (Fig. 2B). At prehatching stages, transcript levels are enriched in specific embryonic derivatives including eye and ear vesicles, the oral evagination, a small area of endoderm encircling the foregut cavity, the tail blastema, olfactory placodes, and the adjacent frontal epithelial tissues. After hatching, xSmad8(11) mRNA is enriched within cells of the branchial arch and the cephalic sense organs (Fig. 2B, stages 34, 36, and 38). In adult frogs, low transcript levels were detected in all tissues analyzed (data not shown).

Smad8(11) does not have the ventralizing activity of Smad1.

We used misexpression experiments in an attempt to gain insight into potential functions for Smad8(11). For this purpose, mRNA was transcribed in vitro and injected into fertilized eggs, and the embryos were allowed to develop to tadpole stages. The forced expression of xSmad8(11) failed to generate any obvious developmental abnormality, even when embryos were injected with relatively high levels of mRNA (1 to 5 ng) or when the mRNA was targeted to specific blastomeres at the 2- to 4-cell stages. The injected embryos appeared identical to control embryos injected with H₂O or the same amount of an irrelevant mRNA (Fig. 3; compare Control and Smad8). This was somewhat surprising given the strong ventralized phenotype that is elicited by injection of mRNA encoding the closely related xSmad1 (Fig. 3). This negative result suggests that Smad8(11) does not contribute to regulating the specification of the ventral axis and is distinct in activity from Smad1 (noted also as “data not shown” by Fortuno et al. [14]).

FIG. 3.

Smad8(11) regulates embryonic pathways distinct from those regulated by the closely related Smad1. Embryos were injected at the 1-cell stage either with 2.3 ng of control (nonsense) RNA or with RNA encoding either xSmad1, xSmad8A (labeled Smad8), or dominant-negative mutant proteins for xSmad1 or xSmad8A, as indicated (Δ designates a mutant protein). Embryos were photographed at stage 24 and are arranged with the anterior to the left. Views are lateral except for Smad1Δ and Smad8Δ, for which the views are from the dorsal side. In a typical experiment, the phenotypes shown represent (from left to right) 100, 96, 66, 92, and 42% of the injected embryos, respectively (n = 50 for each sample). Black arrows, ventralization; white arrows, secondary axes; grey arrows, gastrulation defect.

To strengthen this conclusion, we generated putative dominant-negative variants of both xSmad1 and xSmad8(11), based on the previous structure-function analysis of mSmad8 (26). The mutant proteins are truncated at the identical location in the C terminus within the MH2 domain, so that they lack the SSVS sequence required for activation by the BMP receptor complex; this mutation inhibits a response to BMP signaling in cell culture experiments (26). We constructed the analogous mutation in xSmad1 for direct comparison. Not surprisingly, forced expression of the mutant xSmad1 protein resulted in defects in dorsal-ventral axis development. Most of the embryos developed either with a duplicated axis (Fig. 3, Smad1Δ) or with more subtle axis defects. In striking contrast, the mutant xSmad8(11) protein never caused overt axis defects, although it did generate a defect in gastrulation instead. In repeated experiments, approximately half the embryos expressing mutant xSmad8(11) failed to complete gastrulation, resulting in embryos with open dorsal structures at the posterior end, often associated with a bifid tail structure. Although we did not investigate the phenotype further, we believe it is specific, since it was not seen in cohorts of embryos injected with control, xSmad1, xSmad8(11), or mutant xSmad1 mRNAs. We conclude that, at least under conditions of forced expression, xSmad8(11) is not associated with the Smad1-related embryonic pathways that control ventral axis development.

Depletion of maternal xSmad8 mRNA disrupts gastrulation and activates cell death.

To analyze the normal function of Smad8(11) during early embryogenesis, we used the host transfer technique (61) to eliminate maternal xSmad8(11) mRNA. Briefly, isolated oocytes were injected with antisense oligomers and cultured to facilitate degradation of targeted mRNA. The oocytes were then matured in vitro, labeled for identification with vital dyes, and transferred into an egg-laying host female. Mature eggs were collected and fertilized in vitro. The sequences used for designing antisense oligonucleotides to target the maternal message are indicated in Fig. 1A. Of an initial set of 12 oligomers (including 1 control sense strand oligomer, oligomer 1), several effectively depleted maternal xSmad8 mRNA in cultured oocytes (Fig. 4A). As a control for specificity, transcript levels were also measured for xODC (a housekeeping gene) and xSmad1 [the xSmad gene that is most closely related to xSmad8(11)]. Oligomers that targeted xSmad8(11) were then synthesized with phosphorothioate modifications (PM). These oligomers were also effective at specifically depleting maternal xSmad8(11) mRNA (Fig. 4B). As shown in Fig. 4C, injection of PM oligomer 3 leads to degradation of xSmad8(11) mRNA, but not of xODC or xSmad1 mRNA, in a dose-dependent manner.

FIG. 4.

Antisense oligomers are effective at specifically depleting maternal xSmad8(11) transcripts in cultured oocytes. (A) An RT-PCR assay was used to determine the effectiveness of mRNA depletion. The oligomers used for oocyte injection are given above the gel (see also Fig. 1A). Injected oocytes were collected after culturing (24 h plus 12 h of progesterone treatment), and RNA was harvested, typically from a set of 3 to 10 oocytes. Oligomer 1 is a sense strand sequence used as a control. As controls for specificity we confirmed that these oligomers had no effect on levels of the closely related xSmad1 mRNA, or on the xODC gene, which also served as a loading control. The RT(−) samples show the dependency of the product on the RT reaction. RT(−) reactions were carried out for all primer pairs, but only representative examples for ODC are shown here. (B) Four of the oligomers (oligomers 3, 6, 11, and 12) and the control oligomer 1 were synthesized using modified phosphorothioates to increase stability and decrease toxicity. Each of the antisense oligomers was effective and specific. (C) Oligomer 3 specifically depletes xSmad8(11) with a linear dose response to injected DNA. Increasing amounts of the oligomer were injected (0, 0.9, 1.8, 3.5, and 7 ng) as indicated above the gel. In this experiment, minor traces of xSmad8(11) transcripts could still be detected after RT-PCR when oocytes were injected with 7 ng of oligomer (shown to demonstrate the dose response), although in most experiments this dose was sufficient to deplete transcripts below the level of detection.

Following host transfer and in vitro fertilization, embryos derived from either control or antisense oligomer-injected oocytes appear completely normal until the start of gastrulation (data not shown). However, most embryos derived from eggs injected with the highest dose (7 ng) of the antisense oligomer fail to form a dorsal blastopore lip, and by stage 11⁻ white mottled cells, which often disaggregate from the embryo, appear (Fig. 5A, top left panel, embryo on the right). The first visible sign of apparent cell death may occur at one or several places anywhere around the periphery of the embryo. The effect then proceeds rapidly, and whitish cells occasionally spill into the space between the embryo and the vitelline membrane. We refer to this phenotype as type I. In sections of a typical type I embryo (Fig. 5B), interspersed, disorganized cells from all three presumptive germ layers are found loosely associated beneath the vitelline envelope in the region normally defined by a blastocoel space. In contrast, embryos injected with the control oligomer develop normally.

FIG. 5.

Depletion of maternal Smad8(11) disrupts gastrulation. Embryos injected with antisense oligomers develop normally up to stage 10 (not shown). (A) The upper left panel shows a typical type I mutant embryo derived from an oocyte injected with 7 ng of an antisense oligomer (on the right) and compared to a control uninjected embryo (on the left). The upper right panel shows a type II embryo with delayed gastrulation derived from an oocyte injected with 3.5 ng of an antisense oligomer (on the right), compared to a control embryo (on the left). Typical embryos at stage 17 are shown below for each. (B) Histological analysis of control and type I mutant embryos at stage 10.5. Arrowhead indicates the position of the dorsal lip in the control embryo. (C) A representative type II mutant embryo at stage 12 maintains a large blastopore (indicated by lines). The blastocoel space is displaced, and the embryo fails to develop an archenteron compared to a control embryo on the left (ar).

Embryos derived from eggs injected with less antisense oligomer often form a blastopore lip, but gastrulation is delayed and incomplete (Fig. 5A, right panel); we refer to this phenotype as type II. In sections of type II embryos, cells are attached, but they appear loosely associated compared to the control embryos (Fig. 5C). In addition, large, yolky presumptive endoderm cells are present at the vegetal surface. To test if this phenotype is due to a partial loss of xSmad8(11), different amounts of antisense oligonucleotides were tested for a dose response. Embryos were classified at the equivalent of stage 11 according to phenotype: normal, no blastopore lip and apparent cell death (type I), or an abnormal delay in blastopore lip formation resulting in a gastrulation defect (type II).

The results, compiled and presented graphically in Fig. 6, are consistent with a dose-dependent relationship of the phenotype. When either 7 or 3.5 ng of an antisense oligomer was injected, 100% of the embryos developed as either type I or type II mutants. At lower doses, the percentage of type I embryos decreased while the number of normal class embryos increased. Most importantly, subsequent coinjection of the mRNA-depleted embryos with in vitro-generated xSmad8(11) mRNA rescued the phenotype, while injection of xSmad1 mRNA, encoding the most closely related subfamily member, failed to rescue. Rescue by xSmad8(11) was progressively more effective when embryos were derived from oocytes injected with smaller amounts of antisense oligomers. In some experiments we found 100% rescue to normal development, although this was not always achieved with respect to the type II phenotype, which limits our ability to interpret this as a Smad8(11)-specific phenotype. In order to test if the C-terminal sequence that is a target for phosphorylation is required for xSmad8 function, we attempted to rescue xSmad8-depleted embryos by subsequent injection of xSmad8Δ mRNA. The mutant mRNA failed to rescue the cell death phenotype (see Fig. 6). When 1.8 ng of antisense oligomer 3 was used to deplete xSmad8 mRNA, the cell death phenotype occurred in 24% of the injected embryos, 0% of the embryos subsequently injected with xSmad8 mRNA, and 32% of the embryos subsequently injected with xSmad8Δ RNA.

FIG. 6.

Summary of the phenotypes in xSmad8(11)-depleted and rescued embryos. Data were obtained by combining results from 12 individual experiments and are expressed as the total numbers (n) and percentages of normal (N), type I, and type II embryos. Results obtained with different doses of antisense oligomer 3, with or without subsequent injection of the embryos with either in vitro-transcribed xSmad8A mRNA, Smad8Δ mRNA, or Smad1 mRNA, are shown. Note that xSmad8 mRNA rescues the cell death phenotype, while neither Smad8Δ nor Smad1 mRNA is able to do so. Results of similar experiments using the control sense strand oligomer 1 and a different antisense oligomer (oligomer 12) are also shown. The percentages of each phenotype are also given in the bar graph at the bottom. When oocytes were not depleted but fertilized eggs were subsequently injected with xSmad8Δ (∗) or xSmad1 (∗∗), the resulting embryos are not normal. Consistent with the results presented in Fig. 3, xSmad8Δ-injected embryos had a gastrulation defect, while 90% of the xSmad1 mRNA-injected embryos were ventralized. The remaining 10% of the xSmad1-injected embryos failed in gastrulation, but there is no evidence that this is related to the type II mutant phenotype caused by xSmad8(11) depletion.

In several additional experiments, essentially identical phenotypes were obtained by using other antisense oligomers. For example, the phenotype of embryos injected with 3.5 ng of oligomer 12, which targets a distinct region of the transcript, yielded no normal embryos, 33% type I embryos, and 67% type II embryos. Furthermore, subsequent xSmad8 mRNA injection resulted in a strong rescue of the cell death phenotype (Fig. 6).

Maternal xSmad8(11) regulates expression of genes in early ectoderm.

Semiquantitative RT-PCR assays were used to analyze the expression of differentiation markers in xSmad8(11)-depleted embryos compared to normal embryos (Fig. 7A). Because type I mutant embryos die, RNA was isolated from normal or type II mutant embryos at stages 9 to 12. The only consistent difference we found was for the transcription of ectoderm markers. Both neural (NCAM) and epidermal (keratin) transcript levels were decreased in xSmad8(11)-depleted embryos. Levels for xGATA-2, normally expressed in ectoderm and ventral mesoderm, were also reduced significantly. In contrast, mesoderm-specific genes were expressed, although the post-MBT increase in zygotic transcript levels was uniformly delayed relative to that for normal embryos (seen also for EF-1α and xBra levels, which are activated only beginning at MBT). No differences were found in expression of the endoderm marker xSox17-β. Similar decreases in expression of ectoderm markers were found when type II embryos were analyzed at stage 17, and this was rescued by subsequent xSmad8 injection (Fig. 7B). Consistent with the gene expression analysis, histological analysis of surviving type II embryos showed differentiated mesoderm tissues and normal-appearing endoderm but not obvious differentiated neural structures (data not shown).

FIG. 7.

Ectoderm differentiation genes fail to be transcribed normally, while mesoderm induction occurs but is delayed, in embryos depleted of maternal Smad8(11). (A) Representative semiquantitative RT-PCR assays to compare expression of differentiation markers among control and xSmad8-depleted embryos at the indicated stages. Samples in the panels on the right demonstrate the linearity of product accumulation relative to input RNA in the RT reactions. In mutant embryos there is a delay in accumulation of EF-1α and xBra that is consistent with the delay in gastrulation, since these zygotic levels increase only after MBT. These levels do eventually recover, in contrast to those for NCAM, keratin, or GATA-2 (compare lanes marked 11−). Although some lanes are labeled “type II,” it is impossible to distinguish phenotypes until apoptosis occurs in type I embryos at stage 10.5. Therefore, samples taken at stages 9 and 10 may be a mixture of type I and type II mutant embryos. (B) Similar analysis of ectoderm markers from embryos allowed to develop to stage 17 before harvesting of RNA. Embryos either were or were not injected with antisense oligomers, and after host transfer they were injected with control or xSmad8 mRNA.

Maternal xSmad8(11) depletion leads to apoptosis at EGT.

The appearance of type I embryos suggests activation of a cell death program. Indeed, whole-mount TUNEL staining of these embryos (Fig. 8A) revealed apoptotic nuclei at the equivalent of stage 10.5 (left panel) or stage 17 (right panel). Control embryos showed no TUNEL-positive cells at stage 10.5, or at the surfaces of stage 17 embryos, as expected, since programmed cell death at neurula stages is restricted to inner-layer cells (19). Injection of α-amanitin has been used to demonstrate an embryonic program of cell survival that requires zygotic transcription (18, 41). TUNEL-positive embryos derived from α-amanitin injection began to collapse during gastrulation (stage 10.5 through 11⁻), like xSmad8(11)-depleted embryos. However, the sites that initiate apoptosis are different, since apoptotic cells were seen only below the surface monolayer of ectoderm in α-amanitin-treated embryos (Fig. 8Bi). In contrast, xSmad8-depleted embryos had TUNEL-positive cells in both inner and outer layers (Fig. 8Bii). To determine if apoptosis initiates prior to cell disaggregation, xSmad8(11)-depleted embryos were analyzed immediately after control embryos reached stage 10, before any overt phenotype manifested. The depleted embryos maintained apparently normal structure and cell layer associations at this stage. Nevertheless, TUNEL-positive cells were located in a wide distribution around these embryos, and apoptotic cells could be identified within presumptive ectoderm comprising three to four cell layers (Fig. 8Cii). In addition, TUNEL-positive cells were found in the marginal zone prior to blastocoel collapse or the appearance of dissociated cells. The results are consistent with a primary defect in cell survival rather than a cell adhesion defect that promotes apoptosis.

FIG. 8.

Depletion of maternal Smad8(11) activates programmed cell death. (A) Whole-mount TUNEL staining detects apoptotic cells around the surfaces of type I embryos. (Left) At stage 10.5, control (con) pigmented and albino embryos (far left) are negative in this assay. In contrast, the depleted embryos fail in gastrulation and the disaggregating cells are TUNEL positive (oligo). As a positive control for apoptosis, α-amanitin was injected into fertilized eggs as described elsewhere (18, 41). (Right) Type II and type I embryos, along with a control embryo, at neurula stage 17. (B) Control α-amanitin-injected (i) or Smad8(11)-depleted (ii) embryos were analyzed at the equivalent of stage 10.5 by whole-mount TUNEL staining and then embedded in paraffin and sectioned. TUNEL-positive cells are found within the entire space of the collapsed blastocoel, or within the loosened presumptive ectoderm layer as indicated by arrows. In some cases the apoptotic cells emerge through the vitelline envelope. Upper panels are higher-power views of the same samples shown in the lower panels. (C) Smad8(11)-depleted embryos were harvested for analysis immediately at the equivalent of stage 10.0, prior to manifestation of an overt phenotype. (i) Control from a stage 10 uninjected embryo. (ii) While the depleted embryos fail to form a blastopore lip, apoptotic cells are detected even prior to blastocoel collapse or cell disaggregation. TUNEL-positive cells are evident both in marginal-zone presumptive mesoderm (arrow) and within layers of presumptive embryonic ectoderm.

Finally, we measured caspase activities following xSmad8(11) depletion relative to those in control embryos. The relative activities of both caspase subfamilies (ICE/caspase 1 and CPP32/caspase 3) were quantitatively determined in lysates prepared from host-transferred embryos, following injection of oocytes with either control or antisense oligomers. The specificity of the activity was monitored by using specific inhibitors of ICE and CPP32 (Ac-YVAD-CHO and Ac-DEVD-CHO, respectively). As shown in Fig. 9, xSmad8(11) depletion resulted in a specific activation of approximately threefold for CPP32 activity, while ICE caspase activity was unchanged. These data are consistent with xSmad8(11) depletion leading to activation at EGT of a caspase 3-dependent program of apoptosis.

FIG. 9.

Activation of CPP32 caspase by depletion of maternal xSmad8(11). Caspase assays were performed on embryonic lysates derived from host-transferred oocytes, which had been injected with control (0) or 7 ng of antisense oligomer 3. Assays were performed either in the absence or in the presence of specific caspase inhibitors, and specific activity was determined relative to standard curves generated using purified substrates. Results presented are means from three independent experiments.

DISCUSSION

xSmad8(11) has activity distinct from those of xSmad1 and -5.

Comparison of the primary structure of all Smads shows that there are conserved Smad8-specific sequences including a serine/threonine-rich stretch (S/TSPISSLFSFTS) at the N terminus. Although the cDNAs we isolated are most closely related to mammalian Smad8, they include a 34-amino-acid stretch encoded in the central linker region that is missing in the reported rat and mouse Smad8 sequences. Fortuno et al. (14) also identified one of the cDNAs we describe (Smad8B) and considered it to be a novel family member named Smad11. The nomenclature is problematic, since neither a Smad5 homologue nor a gene more closely related to mammalian Smad8 has been described for Xenopus. Regardless, a related linker stretch is present in other Smads, and the human Smad8 gene derives both types of cDNAs by differential splicing (48). It seems reasonable to consider that all species may be capable of generating either isoform, perhaps in a cell-specific manner. Interestingly, contained precisely in this sequence is the PPXY binding site of the E3 ubiquitin ligase Smurf-1, which has been shown to target Smad degradation via the 26S proteosome (60). Differential Smad splicing could therefore regulate competence for BMP signaling. A third human Smad8 isoform, lacking the C-terminal phosphorylation site, has been described (36). The conserved-sequence differences between Smad8(11), on the one hand, and Smad1 and -5, on the other, are significant, since overexpression of Smad1 or Smad5, but not of Smad8(11), causes ventralization in Xenopus embryos (44, 45).

Maternal xSmad8(11) activity is essential for normal embryogenesis.

Maternal xSmad8(11) is required to mediate cell survival and morphological movements through early gastrula stages. This function is specific to xSmad8(11), as evidenced by the facts that it is rescued only by subsequent injection of this mRNA and that embryos containing closely related and relatively abundant BMP-associated Smads (including Smad1 and perhaps Smad5) fail to compensate for its loss. The total amounts of the Smad1 (5), Smad8, and Smad11 (13) proteins remain nearly constant across all three presumptive germ layers and the dorsal-ventral axis from stage 9 to 10⁺ (determined by use of an antiserum raised against a fully conserved epitope). Notably, the initial phosphorylation at stages 9 to 10 does not require zygotic transcription (13). Therefore, protein derived from maternal xSmad8(11) mRNA would be activated (phosphorylated) prior to MBT.

Smad8 is believed to mediate BMP signaling, yet the phenotype caused by loss of maternal xSmad8(11) mRNA has not been observed in contexts of disrupted BMP signaling, for example, by using BMP antagonists that promote either a dorsal mesoderm or neural fate. We note that the Smad5 (55) and BMP7 (11) null mutations result in patterned apoptosis of mesenchymal cells later in development, while in some cases, including BMP2 and BMP4 knockouts, abnormal apoptosis may not be ruled out. The BMP4 null mutation (51) did not reveal a function for cell survival, and this may be due in part to difficulties in evaluating maternal functions by mouse knockout approaches. Indeed, the lack of penetrance for the BMP4 null phenotype suggested a developmental function for maternal BMP signaling (51). Likewise, experiments with Xenopus have defined a zygotic function for BMP signaling in ventral mesoderm patterning (25, 43), but these studies did not consider the maternal contribution. It is also possible that currently used inhibitors and antagonists do not specifically target the ligand/receptor combinations that are responsible for regulating xSmad8(11). Given that our limited knowledge is so far restricted mostly to in vitro or ectopic expression experiments, it is also possible that xSmad8(11) has BMP-independent functions. It is clear from this and previous studies that Smads of the 1/5/8/11 subfamily have specific nonredundant embryonic functions at the earliest stages of embryogenesis (6, 10, 46, 55).

Smad8(11) regulates a maternal program that inhibits apoptosis.

The existence of maternal cell survival cues that play out their function at EGT has been suggested from previous experiments (1, 18, 41, 42). This maternal program may be necessary because of the lack during early development of the typical cell cycle-regulated checkpoints. In amphibian development, a fertilized embryo carries out rapid cycles of DNA replication and mitosis (33, 34), controlled by the degradation and synthesis of cyclins A and B (28). The apoptotic response is never elicited prior to stage 10.5, either due to an intrinsic inability to activate a cell death pathway or due to an inhibitor of programmed cell death (18). At MBT the cell cycle lengthens to accommodate checkpoints and zygotic transcription begins. After MBT and until EGT, this maternally encoded program persists and inhibits apoptosis. If embryos are developmentally arrested at EGT by prior treatment with hydroxyurea or cycloheximide, ICE-like caspases are activated and degrade cyclins A1 and B1, leading to apoptosis (42). The early events of apoptosis at the beginning of gastrulation are described as randomly distributed in cells of the animal hemisphere; after EGT, apoptosis is patterned in the developing neural plate (19). Our results indicate that xSmad8(11) is part of a maternal program that regulates cell survival prior to the EGT. This may not be the same pathway that is triggered by treating pre-MBT embryos with DNA synthesis inhibitors, since this induces ICE-like caspases (23, 42). Instead, maternal xSmad8(11) depletion causes apoptosis by activation of CPP32/caspase 3-related enzymes at EGT.

Is this program regulated by BMP signaling? Certainly there is a large body of literature linking BMP signaling with regulation of apoptosis in several contexts, and Smad8 is capable of being activated by BMP signaling (26). The C-terminal phosphorylation target of xSmad8 is required for regulating cell survival, but this only demonstrates the importance of the sequence and is not direct evidence of BMP-mediated regulation. A possible link between BMPs and cell survival comes from studies of a mitogen-activated protein kinase kinase kinase called xTAK1 (54). A kinase-negative xTAK1 mutant protein inhibits ventral mesoderm induction caused by ectopic BMP or Smad1 or -5 (40), and in tissue culture cells, the negative regulatory protein Smad6 can inhibit BMP2-induced activation of TAK1 (27). Forced expression of xTAK1 or of its activator, xTAB1, causes cell death and the dissociation of embryonic cells at the beginning of gastrulation (40); this phenotype is remarkably similar to our type I Smad8(11)-depleted embryos. The ventralizing activity of TAK1 is revealed only if the apoptotic pathway is inhibited, and an inhibitor of apoptosis (XIAP) has been linked to the TAK1/TAB1 complex (53). We were unable to rescue xTAK1-induced apoptosis by coinjection of xSmad8(11) mRNA (data not shown). This does not rule out a connection to BMP signaling, since TAK1 could act downstream of xSmad8, but our data are also consistent with a BMP-independent function for xSmad8. Future experiments will seek to determine if the function of xSmad8 in regulating early embryonic cell survival is controlled by the BMP pathway or represents a BMP-independent function of Smad8.