sizzled function and secreted factor network dynamics

A large number of secreted signaling agonists are expressed in the early Xenopus embryo, including Wnts, BMPs, Nodals, and FGFs together with their antagonists (De Robertis, 2009; De Robertis and Kuroda, 2004; Schier, 2009; Smith, 2009). One component of the early embryo's extracellular signaling system is Sizzled, a secreted protein with homology to the extracellular domain of the Wnt receptor Frizzled (Collavin and Kirschner, 2003; Salic et al., 1997). Previous studies reported that sizzled expression is regulated by, and regulates BMP signaling (Lee et al., 2006; Muraoka et al., 2006) and that Sizzled “functions in a negative feedback loop that limits allocation of mesodermal cells to the extreme ventral fate” (Collavin and Kirschner, 2003). Our interest in sizzled was spurred by the observation that sizzled RNA levels in the dorsolateral mesoderm increased in response to blocking the expression of the transcription factor Snail2/Slug, an important regulator of both mesoderm and neural crest differentiation in Xenopus (Shi et al., 2011).

Sizzled was originally identified by Salic et al., as a Wnt inhibitor. Its expression begins in the ectodermal (animal cap) region at the midblastula transition (stage 8.5), subsequently its expression becomes restricted to the ventral marginal zone (VLZ) and ventral animal cap (Salic et al., 1997). In retrospect, the complexity of Sizzled's function is not surprising. Sizzled is a member of the family of “secreted frizzled receptor-like proteins” (SFRPs) (Bovolenta et al., 2008; Esteve and Bovolenta, 2010). In addition to binding to and inhibiting Wnt signaling, SFRPs have been found to bind to Frizzleds directly, or in a complex with Wnts, thereby activating Wnt signaling. They can enhance Wnt diffusion (Mii and Taira, 2009) and have been found to bind other proteins, such as RANKL, blocking its ability to activate NF-κB signaling through the RANK receptor (Hausler et al., 2004). NF-κB is involved in the patterning of the early Xenopus embryo (Beck et al., 1998; Kennedy et al., 2007; Kennedy and Kao, 2011; Lake et al., 2001; Tannahill and Wardle, 1995; Zhang et al., 2006). Perhaps more surprisingly, first Sizzled and then the related protein Crescent were found to act indirectly as BMP inhibitors by binding to, and inhibiting the activity of secreted Tolloid-like/BMP1-like metalloproteinases, which in turn inhibit the BMP inhibitor Chordin (Lee et al., 2006; Misra and Matise, 2010; Muraoka et al., 2006; Ploper et al., 2011; Yabe et al., 2003). Subsequently it was found that the related protein, SFRP2, binds to and activates Tolloid-like proteins, such as procollagen C proteinase (Kobayashi et al., 2009), raising the prospect of complex positive and negative interactions between SFRPs and secreted enzymes.

We originally set out to test the hypothesis that blocking the increase in sizzled RNA found in Snail2/Slug morphant embryos would rescue the Snail2/Slug morphant phenotype. This experiment was complicated by the fact that blocking sizzled expression itself produce both mesodermal and neural crest phenotypes, which we show can be rescued by the injection of RNAs encoding BMP and Wnt antagonists, suggesting that Sizzled normally acts, either directly or indirectly, as both a BMP and a Wnt inhibitor.

Expression of sizzled RNA begins following the midblastula transition (stage 8.5) in the animal (ectodermal) region, before becoming restricted to the ventral regions of gastrula stage embryos (Salic et al., 1997). In our hands whole-mount in situ hybridization (Fig. 1A–E), standard (Fig. 1F) and quantitative RT-PCR (qPCR) studies (Fig. 1G) revealed detectable levels of sizzled RNA in both ventral and dorsal axial marginal zones of early gastrula stage embryos, as well as in the dorsal region of early neurula stage embryos. qPCR analysis of whole embryos (Fig. 1H) confirmed that sizzled RNA levels increased in snail2, snail1, and twist1 morphant gastrula (stage 11.5) embryos (both cells of two cell stage embryos injected). In RNA SEQ of whole snail2 morphant Xenopus tropicalis embryos (Table 1), sizzled RNA levels were increased ∼3 fold compared to control morpholino injected embryos (analyzed at stage 11.)

Given the increase in sizzled RNA levels in snail2 C2/C3 morphant DLMZs found previously (Shi et al., 2011), we were interested in answering the question, does blocking the increase in sizzled expression (through morpholino injection) rescue the snail2 morphant phenotype? As describe below, this goal was complicated by the fact that i) loss of snail2 activity also leads to an increase in RNA levels for two other secreted signaling antagonists, cerberus and chordin (Shi et al., 2011; Zhang and Klymkowsky, 2009), and ii) the observation that sizzled has its own role within the DLMZ with respect to mesodermal and neural crest differentiation. Since we have yet to identify a useful anti-Xenopus Sizzled antibody, we tested the efficacy of the sizzled morpholino by co-injecting embryos with RNAs (200 pg/embryo) encoding GFP and Sizzled-HA or a similarly epitope-tagged form of a mutated and inactive form of Sizzled, SizzledD92W-HA (Lee et al., 2006). Both RNAs contain the morpholino's target sequence; in both cases, the morpholino (injected at 7 ngs/embryo) significantly reduced, but did not eliminate the accumulation of the exogenous polypeptide (Fig. 2A). Similar results were observed when an RNA encoding a GFP-tagged form of Sizzled was injected; the sizzled morpholino dramatically reduced fluorescence, while injection of control morpholino had no effect on Sizzled-HA/GFP accumulation (data not shown). Since the level of injected RNA is estimated to be greater than the level of endogenous sizzled RNA, we expect that the sizzled morpholino will produce a fairly robust hypomorphic (reduction of function) phenotype. We also note that the amount of morpholino used in these and all other studies (7 ngs/embryo) is at the low end of that commonly used, which has been reported to be as high as 40 to 60 ngs/embryo.

We first characterized sizzled morpholino “half-embryo” effects; in such studies the morpholino was injected equatorially into one cell of a two-cell embryo (Table 2). In gastrula stage embryos, we found loss of expression of the mesodermal marker brachyury (Fig. 2B,C) and increased expression of the endodermal marker endodermin (Fig. 2E,F). Both could be rescued by injection of 200 pgs sizzled-HA RNA (Fig. 2D,G); higher levels of RNA produced their own effects (see below.) In later stage embryos, we found disruption of pronephric development (Fig. 2H,H*) and myotomal organization (Fig. 2I–N) on the morphant sides of embryos.

Previously we found that snail2 morpholino injection of the C2/C3 blastomeres of the 32-cell embryo led to a loss of mesoderm and neural crest, while snail1 and twist1 morphant, C2/C3 injected embryos lost mesoderm but not neural crest (Shi et al., 2011). When the sizzled morpholino was injected into C2/C3 blastomeres we found the loss of xbra expression in early gastrula stage embryos (Fig. 3A–C), loss of the neural crest markers snail2 (Fig. 3D–F), twist1 (Fig. 3G–I) and sox9 (Fig. 3J–L) as well as the loss of myoD expression (Fig. 3M,O) in later stage embryos (Table 3). The sizzled morpholino effect on sox9 expression was efficiently rescued by the injection of the sizzled RNA but not by injection of sizzled D92N RNA (Fig. 3O). An interesting observation was that while sizzled morpholino injection led to a “simple” reduction of the snail2 and sox9 expression domains, its effects on twist1 expression were more complex, with both reduced and apparently ectopic ectodermal expression observed (Fig. 3I).

Because the C2/C3 lineage gives rise to a range of tissues (Dale and Slack, 1987; Moody, 1987; Nakamura et al., 1978), we examined whether the effects on ectodermal neural crest marker expression were due to inductive (secreted factor-mediated) effects by using a standard explant sandwich approach (Bonstein et al., 1998; Shi et al., 2011). When wild type ectoderm (animal cap) is cultured alone it remains undifferentiated epidermis (Fig. 4A); when cultured with wild type DLMZ, the animal cap expresses neural crest markers, such as sox9 (Fig. 4B). We found that DLMZ from sizzled morpholino-C2/C3 injected embryos produced a much weaker inductive effect in terms of sox9 expression (Fig. 4C). Quantitative RT-PCR analysis of explants (Fig. 4D) supported this conclusion; the increases in sox9, snail2, and twist1 RNA levels observed in wild type animal cap/wild type DLMZ explants were reduced to below the levels observed in wild type animal caps alone in wild type animal cap/sizzled morphant DLMZ explants. This indicates that loss of sizzled function leads to a loss of inductive activity, as well as other negative effects on sox9, snail2, and twist1 expression in this system.

BMP and Wnt signaling have been implicated in mesodermal induction of neural crest (see Shi et al., 2011 and references therein). To examine the signaling system influenced by sizzled loss of function, we took three approaches. First, we isolated DLMZ from sizzled morphant, C2/C3 injected embryos and subjected them to qPCR analysis. The result shows a small but reproducible increase in the levels of wnt8 and bmp4 RNAs and more dramatic decreases in the levels of noggin and sfrp2 RNAs (Fig. 5A). Levels of dkk, frzb1, and chordin RNAs appeared unchanged. We then asked whether the C2/C3 sizzled morphant phenotype could be rescued by injection of noggin or sfrp2 RNAs, either alone or together. Alone, neither noggin or sfrp2 RNAs rescued the C2C3 sizzled morphant phenotype, but together rescue was robust (Fig. 5B). Since SFRP2 interacts with tolloid-like proteins (Kobayashi et al., 2009), it is possible that it has both anti-Wnt and (indirect) anti-BMP activities. We therefore repeated these studies using RNAs encoding Noggin and the Wnt antagonist Dickkopf (Dkk), which acts in a distinctly different mechanism from SFRPs (Bafico et al., 2001; Semenov et al., 2001). Again, we found that only the combination of the BMP inhibitor Noggin and the Wnt antagonist Dkk efficiently rescued the sizzled morphant phenotype (Fig. 5C), suggesting that both signaling pathways are involved, and that the Wnt/BMP signaling balance is important in neural crest induction by DLMZ-derived factors. This supports our previous conclusion (Shi et al., 2011), that both signaling systems are involved in mesodermal induction of neural crest.

In the course of experiments to rescue the sizzled morphant phenotype using sizzled RNA injection, we noted a dramatic dose response. At low RNA levels (∼200 pgs/embryo), the levels used to rescue sizzled morphant phenotypes, the effects of sizzled RNA injection were relatively subtle–as an example, we found increases in levels of sox9 (Fig. 6A,B) and twist1 (Fig. 6D,E) expression on the injected side of neurula (stage 18) embryos. In contrast moderately higher levels (∼600 pgs/embryo) led to dramatic effects on both injected and contralateral sides of the embryo in terms of sox9 (Fig. 6C), twist1 (Fig. 6F), snail2 (Fig. 6G,H), chd7 (Fig. 6I), and c-myc (Fig. 6J–L), reinforcing the general caution associated with interpreting over-expression studies, particularly with regards to secreted proteins and in systems that display dramatic adaptive behavior (De Robertis, 2009), as illustrated by the changes in noggin and sfrp2 RNA levels seen in sizzled morphant embryos.

Because the X. tropicalis genome sequence is available (Hellsten et al., 2010), while that of X. laevis is not, we carried out an RNA SEQ analysis of sizzled morphant X. tropicalis embryos. Both blastomeres of two cell embryos were injected with 7 ngs/embryo control or sizzled morpholino. RNA was isolated when control embryos reached stage 11. Our analysis identified ∼700 RNAs whose levels were changed in sizzled morphant embryos compared to controls (p values<0.002) (supplementary material Sizzled morphant, X. tropicalis RNA SEQ data – Excel file). While it is clear that targeted regional and perturbation studies, which we plan to carry out in X. laevis once its genome sequence is available, are required to make sense of these data, we do note that a number of genes encoding transcription factors associated with mesodermal and muscle differentiation, as well as genes encoding secreted factors are altered (Table 4).

The role of Sizzled in dorsal-ventral axis formation is well established (Collavin and Kirschner, 2003). Here we demonstrate that sizzled also plays a role within the DLMZ, where its loss of function leads to a range of down-stream effects, including disrupting neural crest, as well as pronephros and somatic muscle formation, both mesenchymal tissues (Kielbowna, 1981; Muntz, 1975; Wessely and Tran, 2011). While there has been some confusion about whether Sizzled acts as a Wnt inhibitor (Salic et al., 1997) or an indirect BMP inhibitor (through the sizzled tolloid ⊣ chordin ⊣ BMP ⊣ pathway) (Lee et al., 2006; Muraoka et al., 2006), our studies suggest that it plays both roles in the DLMZ; that said, it is not clear whether its ability to inhibit Wnt signaling is direct (e.g. by binding to Wnts) or indirect (in analogy to its ability to inhibit BMP signaling.) Moreover, it is likely that the phenotypic effects observed in sizzled C2/C3 morphant embryos represent the end result of a cascade of effects, in which changes in noggin and sfrp2, along with other genes (altered in response to sizzled loss of function) play an important role, linked in part to more general effects on BMP and Wnt signaling.

What is clear is that SFRP family proteins, like Sizzled, have more complex functions than originally expected. Our RNA SEQ studies in X. tropicalis suggest they may well be involved in regulating the levels of RNAs encoding a number of secreted factors (Table 4) including a number of ADAM and matrix metalloproteinases. While we are currently in the process of confirming the regional effects of sizzled loss of function in X. laevis, as well as the role of particular Sizzled-regulated target genes in axial patterning and mesodermal and neural crest differentiation, the broader significance of these observations involves the linkage between snail2/slug, a gene well known to be involved in tumor metastasis (Alves et al., 2009; Hanahan and Weinberg, 2011; Hugo et al., 2011; Micalizzi et al., 2010; Shirley et al., 2010), and metastatic behavior (Klymkowsky and Savagner, 2009). There is increasing evidence for the importance of secreted factor mediated tumor-stroma interactions (Chaffer and Weinberg, 2011; Hanahan and Weinberg, 2011), and it seems plausible that Sizzled is part of a network of interactions that regulates how cells interact with their extracellular environment.

X. laevis embryos were staged, and explants and co-explants were generated following standard procedures (Klymkowsky and Hanken, 1991; Nieuwkoop and Faber, 1967; Sive et al., 2000; Zhang et al., 2003). Similar studies were carried out using X. tropicalis embryos following methods posted on the Harland (http://tropicalis.berkeley.edu/home/) and Khokha (http://tropicalis.yale.edu/) lab web sites, using animals purchased from Xenopus I. Capped mRNAs were transcribed from linearized plasmid templates using mMessage mMachine kits (Ambion) following manufacturer's instructions. For two-cell stage studies, embryos were injected equatorially; for 32-cell stage studies, C2 and C3 blastomeres were routinely co-injected with RNA encoding GFP and examined at stage 10 to confirm the accuracy of injection. Animal caps were isolated from stage 8–9 blastula embryos in 0.5 MMR (Sive et al., 2000) transferred into wells of an 2% agarose plate (one animal cap per well) and combined with DLMZ isolated from stage 10.5 gastrula embryos according to Bonstein et al. (Bonstein et al., 1998). Explant recombinants were harvested when siblings reached stage 18. Images were captured using either a Nikon CoolPix 995 Camera on a Leica M400 Photomicroskop or a Nikon D5000 camera on a Wild stereomicroscope. Images were manipulated with Fireworks CS4 software (Adobe) using “auto levels”, “curves” and “levels” functions only.

For this study, morpholinos against snail2, snail1, and twist1 RNA are as described previously (Shi et al., 2011; Zhang et al., 2006; Zhang and Klymkowsky, 2009). The snail2 morpholino matches the X. tropicalis snail2 sequence at 24 of 25 positions. We generated a new morpholino, designed by Gene Tools, Inc (5′ AGAGGAGCAGGAAGACTCCGGACAT 3′) to block the translation of X. laevis sizzled RNA. This sizzled morpholino matches the X. tropicalis sizzled RNA at 19 consecutive bases, with two mismatches (out of 25). Plasmids encoding HA-tagged forms of Sizzled were supplied by Eddy DeRobertis (UCLA); we constructed a plasmid encoding sizzled-GFP using the pCS2mt-GFP plasmid (Rubenstein et al., 1997). RNA injected embryos were selected based on GFP-based fluorescence. Eddy DeRobertis and Richard Harland supplied plasmids encoding Sizzled (Salic et al., 1997), Noggin (Smith and Harland, 1992), and Dickkopf (Glinka et al., 1998).

RNA SEQ was carried out on a next-generation Illumina HiSeq instrument, part of the Colorado Institute for Molecular Biotechnology facility run by Dr Jim Huntley. Standard Trizol and Qiagen RNeasy methods (Akkers et al., 2009) were used to isolate total RNA from stage 11 control and sizzled morphant embryos (injected in both blastomeres of the two cell stage.) Poly-adenylated mRNA was selected using a polyA Spin mRNA Isolation Kit (NEB). 4 µg RNA was taken into each mRNA-SEQ library preparation. Indexed mRNA-SEQ libraries were prepared using Illumina's TruSeq RNA Sample Preparation Kit according to manufacturer's instructions. cDNA was prepared using random hexamer priming and Superscript III (Invitrogen), with second strand synthesis using DNA polymerase I and T4 DNA polymerase. cDNA quality was checked at this stage by both Nanodrop spectroscopy and PCR. The resultant cDNA was fragmented using a nebulizer, ends blunted using Klenow and T4 DNA polymerases, adenylated, ligated to Illumina primers, and subjected to limited PCR amplification; 120–170 bp fragments were selected using agarose gel electrophoresis and then sequenced (Wilhelm et al., 2010). These steps were carried out using the Illumina mRNA-Seq 8-Sample Prep Kit. The quality of the libraries was assessed via Bioanalyzer (Agilent), diluted to approximately 10 nM and were combined two-to-three libraries per pool. The pooled libraries were applied to a HiSeq flow cell (version 1.5 flow cell and version 2 cluster generation reagents) at 4 pM resulting in 560–590 clusters/mm². In our control and sizzled morphant study we obtained ∼30 million, ∼100 base reads per sample. Data was filtered (Wilhelm et al., 2010) to remove uninformative sequences. Quality reads were mapped to the reference genome (Xenopus tropicalis, Joint Genome Institute (JGI) assembly version 4.2 from Ensembl) using Bowtie (Langmead et al., 2009) and TopHat (Trapnell et al., 2009). Transcripts were assembled and quantified using Cufflinks (Trapnell et al., 2010).

For in situ hybridization studies, digoxigenin-UTP labeled antisense probes were made following standard methods: specific probes for brachyury, endodermin, myoD, snail2, chd7, c-myc, sox9, and twist1 RNAs were used. Monoclonal anti-tropomyosin was purchased from Sigma and 4A6 (Vize et al., 1995) was a gift from Elizabeth Jones. Whole-mount immunocytochemistry was carried out as described by Dent et al. (Dent et al., 1989), and embryos were cleared with either benzyl alcohol:benzyl benzoate or methyl salicyate. In many cases embryos were co-injected with RNA (50 pg/embryo) encoding β-galactosidase; β-galactosidase activity was visualized in fixed embryos using a brief Red-Gal (Research Organics) reaction, in order to identify successfully injected embryos.

RNA isolation, RT-PCR and quantitative RT-PCR analyses were carried out as described previously (Zhang et al., 2003; Zhang et al., 2006). Primers for RT-PCR analyses were ornithine decarboxylase (odc) [U 5′-CAG CTA GCT GTG GTG TGG-3′ D 5′-CAA CAT GGA AAC TCA CAC-3′]; sox9 [U 5′-TGC AAT TTT CAA GGC CCT AC –3′ D 5′- GCT GCC TAC CAT TCT CTT GC –3′]; sizzled [U 5′-CAT GTC CGG AGT CTT CCT GC –3′ D 5′-GGA TGA ACG TGT CCA GGC AG –3′]; cerberus [U 5′-CCT TGC CCT TCA CTC AG –3′ D 5′-TGG CAG ACA GTC CTT T –3′]; wnt8 [U 5′-TGA TGC CTT CAC TTC TGT GG-3′ D 5′-TCC TGC AGC TTC TTC TCT CC –3′]; bmp4 [U 5′-TGG TGG ATT AGT CTC GTG TCC –3′ D 5′-TCA ACC TCA GCA GCA TTC C –3′]; dkk [U 5′-ACG GAA GAT GAT GAC TGT GC –3′ D 5′-CTC TTG ATC TTG CTC CAC AGG –3′]; noggin [U 5′-AGT TGC AGA TGT GGC TCT –3′ D 5′-AGT CCA AGA GTC TCA GCA –3′]; frzb1 [U 5′- TGG ACT CAT TCC TGC TAC TGG-3′ D 5′-AAT TGC CAG GAT AGC ATT GG –3′]; sfrp2 [U 5′-TCT GTG TGA GCA GGT GAA GG –3′ D 5′-GTC ATT GTC ATC CTC GTT GC –3′]; chordin [U 5′-CCT CCA ATC CAA GAC TCC AGC AG –3′ D 5′-GGA GGA GGA GGA GCT TTG GGA CAA G –3′].

We thank Courtney Severson, Jianxia Yang, and Chi Zhang for their work on the project, Doris Wedlich and Hazel Sive for discussion, and Jim Huntley for RNA SEQ runs. This work was supported by NIH ARRA grant GM84133. We thank the Xenopus community, particularly Eddy DeRobertis and Richard Harland for reagents.