Abstract

B1-type SOXs (SOXs 1, 2, and 3) are the most evolutionarily conserved subgroup of the SOX transcription factor family. To study their maternal functions, we used the affinity-purified antibody antiSOX3c, which inhibits the binding of Xenopus SOX3 to target DNA sequences [Development. 130(2003)5609]. The antibody also cross-reacts with zebrafish embryos. When injected into fertilized Xenopus or zebrafish eggs, antiSOX3c caused a profound gastrulation defect; this defect could be rescued by the injection of RNA encoding SOX3ΔC-EnR, a SOX3-engrailed repression domain chimera. In antiSOX3c-injected Xenopus embryos, normal animal–vegetal patterning of mesodermal and endodermal markers was disrupted, expression domains were shifted toward the animal pole, and the levels of the endodermal markers SOX17 and endodermin increased. In Xenopus, SOX3 acts as a negative regulator of Xnr5, which encodes a nodal-related TGFβ-family protein. Two nodal-related proteins are expressed in the early zebrafish embryo, squint and cyclops; antiSOX3c-injection leads to an increase in the level of cyclops expression. In both Xenopus and zebrafish, the antiSOX3c phenotype was rescued by the injection of RNA encoding the nodal inhibitor Cerberus-short (CerS). In Xenopus, antiSOX3c's effects on endodermin expression were suppressed by injection of RNA encoding a dominant negative version of Mixer or a morpholino against SOX17α2, both of which act downstream of nodal signaling in the endoderm specification pathway. Based on these data, it appears that maternal B1-type SOX functions together with the VegT/β-catenin system to regulate nodal expression and to establish the normal pattern of germ layer formation in Xenopus. A mechanistically conserved system appears to act in a similar manner in the zebrafish.

Author Keywords: SOX3; B1-SOXs; Nodals; Germ layer specification; Mesoendoderm; Embryonic axis specification; Xenopus; Zebrafish

Article Outline

• Introduction

• Materials and methods

• Embryos and their manipulation

• Immunocytochemistry

• Plasmids and morpholinos

• In situ hybridization and quantitative RT–PCR analyses

• Zebrafish EST isolation and analysis

• Results

• Characterization of the antiSOX3c phenotype

• Studies in the zebrafish

• Discussion

• SOX3 in the early embryo

• Other targets of SOX3

• A conserved role for B1-type SOXs in endoderm formation?

• Acknowledgements

• References

Introduction

The patterning of the metazoan embryo depends upon the interplay between cellular asymmetries and inductive interactions. In many nonamniotic organisms, cellular asymmetry is initially predominant; these asymmetries are established during oogenesis, modified during meiotic maturation, and activated by fertilization (Slack, 1991). In placental mammals, the initial asymmetries are subtler and their relationship to the embryonic axes less direct; that said, there is clear evidence for asymmetries early in mouse and human development (Zernicka-Goetz, 2002). The clawed frog Xenopus laevis has been a particularly fruitful model system for studying the role of maternal components in axis determination and cellular differentiation. As laid, the egg has an obvious animal–vegetal axis defined overtly by the distribution of pigment and yolk platelets; that axis is modified by sperm entry to define the dorsal/ventral and anterior/posterior axes of the later embryo (Kumano and Smith, 2002; Lane and Sheets, 2002 and Pandur et al., 2002). A number of maternal RNAs are asymmetrically distributed in the oocyte and mature egg (Chan and Etkin, 2001). Of these, the vegetally localized mRNA encoding the T-box containing transcription factor VegT plays a critical role in subsequent embryonic patterning. Depletion of maternal VegT mRNA by antisense oligonucleotide injection into late stage oocytes disrupts the localization of other maternal RNAs (Heasman et al., 2001) and leads to defects in endomesodermal differentiation (Houston et al., 2002; Xanthos et al., 2001 and Zhang et al., 1998).

Zygotic gene expression begins in earnest in Xenopus at the midblastula transition (MBT). Recent studies, however, indicate that developmentally critical zygotic gene expression begins earlier. Yang et al. (2002) found that zygotic expression of Xnr5 and Xnr6 can be detected at the 256-cell stage. Xnr5 and Xnr6 encode nodal-related proteins, members of the TGFβ family of secreted proteins (Chang et al., 2002 and Schier, 2003). They have been implicated in both endodermal differentiation and in the establishment of the Nieuwkoop center (Takahashi et al., 2000), a dorsal–vegetal group of cells specified by the 32-cell stage (Gimlich, 1986). Xnr5 expression is regulated by VegT and the β-catenin/TCF system (Hilton et al., 2003; Rex et al., 2002 and Yang et al., 2002). Maternal VegT binds directly to sites within the Xnr5 promoter and acts as a transcriptional activator while TCFs bind to and suppresses Xnr5 expression. Activation of cytoplasmic β-catenin following sperm entry and cortical rotation relieves TCF-mediated repression and allows Xnr5/6 transcription.

Another maternal factor, SOX3, regulates Xnr5 expression and binds directly to sites within the Xnr5 promoter (Zhang et al., 2003). SOX3 is one of a subset of SOX proteins that antagonize the β-catenin/TCF system of the early Xenopus embryo (Zorn et al., 1999a). SOX3 is a member of the B group of SOX proteins, defined by the presence of a HMG box with approximately 50% sequence identity to the SRY (sex-related on the Y) protein. The B-type SOXs are the most phylogenetically conserved of the SOX proteins (Bowles et al., 2000). B-type SOXs are also conserved with respect to embryonic expression. In ascidians (Miya and Nishida, 2003), hemichordates (Lowe et al., 2003), echinoderms (Kenny et al., 1999), and vertebrates (Avilion et al., 2003; Girard et al., 2001; Penzel et al., 1997 and Zhang et al., 2003), B-type SOXs are supplied maternally, distributed asymmetrically, and later expressed within the developing nervous system (Sasai, 2001 and St Amand and Klymkowsky, 2001).

SOX proteins bind to DNA in a sequence-specific manner through interactions with the minor groove. Binding leads to the introduction of a sharp bend in the DNA (Bewley et al., 1998 and Love et al., 1995) that can facilitate or suppress interactions between other regulatory factors (Scaffidi and Bianchi, 2001). The SOXs examined to date bind to variants of a common core consensus sequence AACAAT, although their optimal binding sites are likely to be longer (Mertin et al., 1999 and van Beest et al., 2000). The β-catenin-binding LEF/TCF proteins are also HMG-box transcription factors, and the SOX binding consensus sequence defines a subset of Wnt/β-catenin-regulated LEF/TCF sites (Klymkowsky, 2004). B-type SOXs have been divided into transcriptional activator B1 and transcriptional repressor B2 subgroups (Uchikawa et al., 1999). This is clearly an oversimplification, however, given recent evidence that the B1-type SOX, SOX3, acts as transcriptional repressor in the early Xenopus embryo (Zhang et al., 2003; see below).

In the context of early developmental events, zygotic expression of B1-type SOXs plays a critical role. In the sea urchin, down-regulation of B1-type SOX proteins by a β-catenin-dependent process is essential for normal endodermal differentiation and morphogenesis (Kenny et al., 2003). In the mouse, the B1-type SOX, SOX2, collaborates with Oct3/4 to regulate FGF4 expression (Yuan et al., 1995). FGF4 acts in an autocrine–paracrine manner to maintain the pluripotency of embryonic and extraembryonic cells (Avilion et al., 2003 and Goldin and Papaioannou, 2003). All of these studies, however, focus on zygotic B1 SOX functions. The role, if any, of maternally expressed genes is more difficult to study given the perdurance of maternally supplied proteins. In the mouse, such studies require the generation of conditional mutations, typically via a cre-lox strategy. The gene must be mutated in the oocyte lineage without disrupting oogenesis. One obvious way to circumvent this problem is to introduce a reagent that directly inactivates or inhibits the maternal protein. In the case of SOX3, we have generated such a reagent, an affinity-purified antibody directed against the C-terminal domain of the Xenopus SOX3 polypeptide. The antibody blocks SOX3 binding to DNA (Zhang et al., 2003) and cross-reacts with zebrafish. Using this antibody, we examined the functions of maternal B1 type SOXs in Xenopus and zebrafish. The results indicate a phylogenetically conserved role for B1-type SOX proteins in establishing the primary animal–vegetal axis of the early embryo through the regulation of nodallike protein expression.

Materials and methods

Embryos and their manipulation

Xenopus embryos were prepared, injected, and maintained according to standard laboratory methods (Dent et al., 1989 and Sive et al., 2000). Zebrafish embryos were maintained according to standard protocols (Westerfield, 1994) and pressure injected at the 1- to 8-cell stage with antiSOX3c antibody together with fluorescein dextran (LFD) as a lineage tracer. For rescue experiments, RNA encoding SOX3ΔC-EnR was injected in combination with rhodamine dextran (LRD) as a lineage tracer. The antiSOX3c and antiXTCF3c (antiTCF3c) antibodies have been described previously (Zhang et al., 2003). Antibody concentrations were determined by Lowry assay, and all dilutions were performed using antibody injection buffer. Zebrafish embryonic extracts (approximately 16–20 postfertilization) were prepared following the Zfin web protocol (http://zfin.org/zf_info/zfbook/cont.html); immunoblot analysis was carried out as described previously (Zhang et al., 2003). For RT–PCR analyses, RNA was extracted from manipulated embryos at the 50% epiboly stage using Trizol.

Immunocytochemistry

Whole-mount immunocytochemistry of Xenopus was carried out as described in Dent et al. (1989) with minor modifications. Zebrafish embryos were fixed in 4% paraformaldehyde for 1–2 h at room temperature, followed by 100% methanol. Embryos were incubated in primary antibody overnight, washed, and incubated with secondary antibodies overnight. The antiSOX3c antibody was used, diluted 1:2000. Antibody staining was visualized using alkaline phosphatase-conjugated secondary antibodies (BioRad) and the BCIP/NBT reaction.

Plasmids and morpholinos

pCS2 plasmids encoding Xenopus SOXD-V5H, SOX2-V5H, SOX3-V5H, SOX3ΔC-EnR-myc, and SOX3ΔC-VP16-myc are described in Zhang et al. (2003). The pCS2.CerS plasmid was supplied by E. DeRobertis (UCLA) (Piccolo et al., 1999); mixer and dominant negative mixer-EnR plasmids (Henry and Melton, 1998) were supplied by D. Melton (Harvard). The ability of various RNAs to rescue the effects of antiSOX3c was determined by injecting RNA into fertilized eggs followed by antibody injection. A morpholino that specifically blocks translation of the SOX17α2 RNA (Hudson et al., 1997) was generously supplied by Aaron Zorn (U. Cincinnati). A control morpholino was purchased from GeneTools, Inc.

In situ hybridization and quantitative RT–PCR analyses

Whole-mount in situ hybridization of Xenopus was carried following standard methods (Sive et al., 2000) using digoxigenin-labeled probes against brachyury (J.C. Smith, Cambridge, UK, and Wolfgang Driever, U. Frieberg, Germany), wnt11 (K. Mowry, Brown U., USA), endodermin (supplied by Aaron Zorn, U. Cincinnati), goosecoid (Eddy DeRobertis, UCLA), squint and cyclops (David Kimmelman, U. Washington). Quantitative RT–PCR was carried as described in detail in Zhang et al. (2003). The primers for endodermin were [upstream: 5′-AGC AGA AAA TGG CAA ACA CAC-3′—downstream: 5′-GGT CTT TTA ATG GCA ACA GCT-3′] (Sasai et al., 1996) and those for XSOX17β were [upstream: 5′-CAG GTG AAG AGG ATG AAG AG-3′—downstream: 5′-CAT TGA GTT GTG GCC CTC AA-3′(Hudson et al., 1997). The primers for Xnr5 and Xnr5 are as described in Zhang et al. (2003). Primers for zebrafish EF1α were [upstream 5′-GGC CAC GTC GAC TCC GGA AAG TTC-3′—downstream 5′ CTC AAA ACG AGC CTG GCT GTA AGG-3']; squint [upstream 5′-CCC ACA CCA GTA GAT GAA ACC TTC-3′—downstream 5′-CCA GGT GCC TCA GTG CAG GAA ACC′3′]; and cyclops [upstream 5′-GCT GCT GTT TTC AGA GCA GCA GGA C-3′—downstream 5′-CGG TAT GCA TTG TAC TTC TTA GGG-3′].

Zebrafish EST isolation and analysis

Predicted zebrafish B1-type SOX protein sequences, with the exception of DrSOX19 (Vris and Lovell-Badge, 1995), DrSOX31 (Girard et al., 2001), and DrSOX21 (Rimini et al., 1999), were taken from the Ensembl Danio rerio whole genome assembly Version 15.2.1 (http://www.ensembl.org/Danio_rerio/). The protein names and corresponding Ensembl/NCBI identifiers are listed in Table 3. Protein sequence alignment was performed using the Web-based ClustalW1.8 program (http://searchlauncher.bcm.tmc.edu/multi-align/multi-align.html), and its graphical representation was generated using Boxshade (http://www.ch.embnet.org/software/BOX_form.html).

Results

To study the role of maternal SOX3 in Xenopus laevis, we first tested the effects of injecting a morpholino directed against the 5′ UTR and coding sequence of SOX3 RNA. While the morpholino induced a decrease in the level of SOX3 polypeptide and some changes in gene expression (Zhang et al., 2003), it produced no overt embryonic phenotype (data not shown). While disappointing, this result was not completely unexpected given the high level of maternal SOX3 polypeptide present (Fig. 1C) (Zhang et al., 2003) and initiation of SOX2 expression in the late blastula (Kishi et al., 2000). SOX2 has been found to be functionally similar to SOX3 in the mouse (Avilion et al., 2003) and in Xenopus (Kishi et al., 2000; see below). Given the lack of an overt SOX3 morpholino phenotype, we turned to injection of the antiSOX3c antibody. antiSOX3c, generated against the C-terminal 20-amino acids of the Xenopus SOX3 polypeptide, blocks the binding of endogenous Xenopus SOX3 to specific DNA target sequences (Zhang et al., 2003). When RNAs encoding epitope-tagged forms of Xenopus SOX2 and SOX3 were injected into fertilized eggs, it was apparent that antiSOX3c was more efficient at immunoprecipitating SOX3 than SOX2 (Fig. 1A). The C-terminal region of SOX3 detected by antiSOX3c is conserved between B1-type SOX proteins; it differs at 7 out of 20 positions from Xenopus SOX2 (Fig. 1B).

Injection of fertilized Xenopus eggs with 50 ng per embryo of antiSOX3c led to a profound gastrulation defect (Figs. 2B, D; Table 1). At lower amounts of antibody, gastrulation was more often complete, but subsequent development was often aberrant (Fig. 2C). Often a distinctive ‘taillike’ protrusion was observed (Figs. 2B, C). Four controls were used to confirm the specificity of the antiSOX3c phenotype. Incubation of the antiSOX3c antibody with the KLH-peptide antigen against which it was raised completely abolished all effects of the antibody on development (data not shown). Fertilized eggs injected with equal or higher concentrations of a similarly generated, affinity-purified rabbit antibody directed against the C-terminal 20 amino acids of XTCF3 developed completely normally ( Fig. 2A; Table 1). Immunostaining revealed the persistence of both antiSOX3c and antiTCF3c antibodies through stage 30 of development (see below). Finally, and most importantly, we were able to rescue the phenotypic effect of antiSOX3c injection in a significant percentage of embryos by injecting either SOX3ΔC-EnR or SOX2V5H RNAs. SOX3ΔC-EnR is a chimeric version of Xenopus SOX3 in which the C-terminal epitope recognized by antiSOX3c has been replaced by the engrailed transcriptional repressor domain. SOX3ΔC-EnR mimics the activity of wild-type SOX3 in the early Xenopus embryo (Zhang et al., 2003). XSOX2V5H is an epitope-tagged form of Xenopus SOX2 which, like SOX3, inhibits β-catenin-dependent axis formation in the early embryo (data not shown). When both blastomeres of 2-cell embryos, derived from antiSOX3c-injected eggs, were injected with RNAs encoding either SOX3ΔC-EnR or SOX2V5H, we observed rescue of the antiSOX3c phenotype in approximately 30% of embryos (Fig. 2E). When fertilized eggs were injected first with SOX3ΔC-EnR or SOX2V5H RNAs and then with antiSOX3c, we observed an increase in the extent of phenotypic rescue to approximately 40% ( Fig. 2F; Table 1). No rescue was observed when antiSOX3c-injected embryos were injected with RNAs encoding SOX3ΔC-VP16, an activating version of SOX3 or the divergent, I-type SOX, SOXD ( Fig. 2F; Table 1). Injection of SOX3ΔC-VP16 produces a gastrulation defect on its own (data not shown).

Characterization of the antiSOX3c phenotype

Dorsal injection of SOX3 RNA ventralizes Xenopus embryos (Zhang et al., 2003 and Zorn et al., 1999a). In embryos that complete gastrulation, such as those shown in Fig. 2C, antiSOX3c injection appears to produce a limited dorsalization, with a dorsal–anterior index (Kao and Elinson, 1988) of between 7 and 8 (with 5 being normal). Given that a known target of SOX3 regulation, Xnr5 (Zhang et al., 2003), is expressed very early during development (Yang et al., 2002), the late stage defects found in antiSOX3c-injected embryos are likely to be indirect. We therefore concentrated our characterization of antiSOX3c's effect on markers of differentiation expressed in late blastula/gastrula stage embryos (stages 9–12).

The T-box transcription factor brachyury, an immediate-early marker of mesodermal differentiation (Smith et al., 1991), is regulated in part by the Wnt–β-catenin/TCF system (Vonica and Gumbiner, 2002) and plays key roles in both mesodermal differentiation and the morphogenic movements of gastrulation (Cunliffe and Smith, 1992 and Kwan and Kirschner, 2003). In uninjected and antiTCF3c-injected embryos, brachyury expression occurs in a uniform ring around the blastopore and within the nascent notochord (Figs. 3A, C). In antiSOX3c-injected embryos staining for brachyury, RNA was highly irregular—stronger in some regions and missing in others (Figs. 3B, D). The Xenopus ortholog of wnt11 is expressed in a pattern quite similar to, although weaker than, that of brachyury (Figs. 3E, F). Expression of wnt11 is directly regulated by brachyury (Tada and Smith, 2000). antiSOX3c injection disrupted the pattern of wnt11 expression (Fig. 3G). For both brachyury and wnt11, there was often a dramatic animal-poleward displacement of the expression domain away from the blastopore (Figs. 3B, D, G—arrows), a pattern never observed in uninjected or antiTCF3c-injected embryos. The blastopore itself, however, was still formed in an at least superficially normal manner.

Goosecoid encodes a homeobox-containing protein expressed in dorsal (head) mesoderm (Blumberg et al., 1991). In uninjected (Fig. 3H) or antiTCF3c (Fig. 3I) injected gastrula stage embryos, goosecoid RNA was restricted to a domain corresponding to the Spemann organizer (Cho et al., 1991). In antiSOX3c-injected embryos, the region of goosecoid expression is larger and more diffuse (Fig. 3J). While expression of goosecoid inhibits brachyury expression, overexpression of brachyury does not alter goosecoid expression (Artinger et al., 1997). It is possible that the increase in goosecoid expression is responsible in part for the irregularity in brachyury expression observed in antiSOX3c-injected embryos.

Endodermin is a pan-endodermal marker that encodes a secreted α2-macroglobulin-like protein (Sasai et al., 1996). In the intact embryo, endodermin expression is initially restricted to the region of the Spemann organizer (Fig. 3K) and later extends around the blastopore (Sasai et al., 1996). Endodermin expression is unaffected by antiTCF3c injection (Figs. 3L, M) but was expanded animally in response to antiSOX3c (Fig. 3N). Antibody injected at the 1-cell stage is not uniformly distributed within later stage embryos (data not shown). To examine the effects of a more uniform distribution of injected antibody, both blastomeres of the 2-cell embryo were injected with antiSOX3c (25 ng per blastomere, 50 ng per embryo total). When controls had reached stage 12/13, embryos were fixed and examined by in situ hybridization for brachyury (Figs. 3O–Q), wnt11 (Fig. 3R), and endodermin (data not shown). In each case, the domain of marker expression was greatly expanded, particularly toward the animal pole and away from the blastopore. In the case of brachyury, the effect was particularly dramatic, with expression often occurring within a distinct bandlike domain of the animal cap in exogastrulating embryos (Figs. 3P, Q). Taken together, the aberrations in brachyury, endodermin, goosecoid, and wnt11 expression appear sufficient to account for the gastrulation defects observed in antiSOX3c-injected embryos.

SOX3 binds to sites within the Xnr5 promoter, and injection of antiSOX3c leads to an approximately two- to threefold increase in Xnr5 RNA levels (Zhang et al., 2003; see below). Seeing as Xnr5 is the earliest known zygotic target of the VegT/β-catenin system (Yang et al., 2002), we examined whether the phenotypic effects of antiSOX3c injection could be rescued by the injection of RNA encoding Cerberus-short (CerS). Cerberus, a secreted protein that antagonizes Wnt, BMP, and nodal signaling (Piccolo et al., 1999), is required for head formation in Xenopus (Silva et al., 2003). CerS is an engineered form of the Cerberus protein that specifically inhibits nodal signaling (Piccolo et al., 1999). While high levels of CerS RNA injection produced an aberrant phenotype on their own, it was possible to rescue the antiSOX3c phenotype is a significant percentage of embryo by using lower levels of CerS RNA. When antiSOX3c-injected embryos were injected with 300 pg per embryo CerS RNA, approximately 40% of embryos appeared to develop normally ( Figs. 4A–C, I; Table 2).

Analysis of CerS-rescued, antiSOX3c-injected embryos by whole-mount immunocytochemistry (Figs. 4D–G) or immunoblot for injected antibody (Fig. 4H) revealed no obvious difference in the amount or general distribution of antiSOX3c antibody between aberrant and rescued embryos. CerS RNA injection reduced the levels of Xnr5 and Xnr6 RNAs in antiSOX3c-injected embryos to levels similar to that found control embryos (Fig. 5A), suggesting that nodal activity is required to maintain Xnr5/6 expression (Onuma et al., 2002). SOX17 is a marker of endoderm differentiation and induces expression of endodermin in animal cap experiments (Hudson et al., 1997). The initiation of SOX17 expression appears to be directly controlled by VegT, but the maintenance of expression is dependent upon nodal signaling (Clements and Woodland, 2003 and Xanthos et al., 2001). Injection of antiSOX3c leads to an approximately twofold increase in SOX17β RNA levels, as measured by quantitative RT–PCR, and this increase was blocked by the injection of CerS RNA (Fig. 5A).

The dramatic effect of CerS RNA on the antiSOX3c phenotype leads us to examine whether antiSOX3c acted on genes downstream of nodal-related proteins in the endoderm specification pathway. In Xenopus, both the homeobox-containing protein Mixer (Henry and Melton, 1998) and SOX17 (Hudson et al., 1997) are involved in the expression of endodermin. We compared uninjected embryos stained for endodermin (Fig. 5B) with embryos injected with antiSOX3c alone (Fig. 5C), antiSOX3c, and an RNA encoding a dominant negative form of Mixer, Mixer-EnR (