Sfrp1 is expressed in the anterior neural plate and is required to establish the prospective eye territory 1213. In line with this idea, Sfrp1 (Figure 1) over-expression in the medaka fish leads to a morphologically evident enlargement of the forebrain, posterior truncations and axial duplications (Figure 2b; Table 1). These defects correlate with the expansion of the expression domains of telencephalic, optic vesicle and diencephalic markers such as fgf8, rx3 and pax6 (Figure 2f,J,n), the alteration of the axial mesoderm marker foxa2 (Figure 2j) and the loss of the posterior domain of pax6 (arrow in Figure 2n). To determine whether the NTR domain of Sfrp1 contributed to this effect, we generated expression constructs encoding truncated (Sfrp1_CRD) or chimerical peptides (Sfrp1_NTR; harbouring its own signal peptide to ensure proper secretion; Figure 1) that comprised the two independent domains in which the protein has been shown to fold 5. Notably, injections of equimolar concentrations of Sfrp1_NTR mRNA led to the enlargement of the forebrain and the expansion of anterior markers (Table 1; Figure 2d,h,l,p), as observed after the over-expression of full-length Sfrp1. Although all peptides seemed to be produced at comparable levels (Figure 3; see below), higher concentrations of Sfrp1_NTR mRNA were necessary to induce posterior truncations or axial duplications (data not shown), suggesting a differential requirement of Sfrp1_NTR along the antero-posterior axis. Alternatively, the peptide was less effective than the entire Sfrp1 protein, perhaps due to a difference in maturation and half-life or diffusion range. Another possible explanation is that monomeric Sfrp1_NTR is less effective than the full-length protein, since protein dimerization through the CRD motif has been previously described 810.

Quite surprisingly, over-expression of Sfrp1_CRD, the domain postulated to mediate SFRP-Wnt interactions, did not result in comparable phenotypes (Table 1). Instead, Sfrp1_CRD mRNA injected embryos presented a small but appreciable reduction of the forebrain (Figure 2c), which was associated with a diminished expression of prosencephalic markers (Figure 2g,k,o). Forebrain reduction was more evident at earlier stages of differentiation even with lower doses of mRNA (data not shown), supporting that the Sfrp1_CRD gain-of-function phenotype did not reflect lower levels of peptide expression. Accordingly, Western blot analysis of embryos injected with haemagglutinin (HA)-tagged versions of the peptides indicated that Sfrp1 and Sfrp1_NTR mRNA were efficiently translated at comparable levels while the Sfrp1_CRD mRNA was produced in a larger amount, which existed in a monomeric and possibly a dimeric form (Figure 3).

Morpholino (Mo)-based knock-down of Sfrp1 expression results in embryos with a reduced eye field associated, in the most affected embryos, with a shortening and widening of the antero-posterior axis 13 (compare Figure 4b,b' with Figure 4a,a'). Low concentrations of Sfrp1 mRNA are sufficient to completely rescue this phenotype in a large part of the embryos 13 (Figure 4c,c',f). If Sfrp1_NTR can mimic the effect of the entire molecule, it should also be able to rescue the effects of Mo interference. Supporting this hypothesis, co-injection of Mo-Sfrp1 and Sfrp1_NTR mRNA rescued the size of the eye field of the treated embryos with efficiency similar to that of Sfrp1 (Figure 4e,e',f). In contrast, Sfrp1_CRD mRNA did not counteract the Mo-Sfrp1 induced phenotype (Figure 4d,d',f) and even appeared to exacerbate it, in line with the over-expression studies.

Together, these data suggested that the molecular events induced by the two domains of Sfrp1 were probably different in nature. The Sfrp1_CRD-induced phenotype was difficult to explain according to the generally accepted view that this domain binds Wnt ligands and antagonizes their activity. In contrast, the strong anteriorisation observed after Sfrp1 and Sfrp1_NTR over-expression could be easily explained as the result of an early and generalized antagonism of the canonical Wnt pathway, since inhibition of this pathway induces similar anteriorised and dorsalised phenotypes in both fish and Xenopus embryos 1213.

To investigate this possibility, we next assayed whether injection of Sfrp1 and Sfrp1_NTR could alleviate the phenotypes caused by Wnt8-mediated activation of canonical Wnt signalling. As previously shown in other species 1415, Wnt8 over-expression in medaka fish embryos led to a strong reduction of the forebrain associated with loss of the rx3-positive optic vesicles (Table 2; compare Figure 5a,e with Figure 5b,f). These anterior defects were similar to those observed after Sfrp1_CRD injections (Figures 2c,k and 5c,g; Table 2) but opposite to those induced by Sfrp1 or Sfrp1_NTR over-expression (Figures 2b,d,j,l and 5d,h; Table 2). Upon co-injection, Wnt8 and Sfrp1 mRNAs appeared to counteract each other's activity, resulting in mildly anteriorised embryos (Figure 5i,l; Table 2) that, however, still presented partial posterior truncations or axis duplications (Figure 5i,l). This suggests that, in the concentration range tested, Wnt8 cannot completely counteract Sfrp1-induced axial defects. In agreement with our previous observations, Sfrp1_NTR mRNA abrogated the Wn8-induced phenotype, restoring almost completely the size of the rx3 expression domain (Figure 5k,n; Table 2; compare to control embryos in Figure 5a,b). In contrast, Sfrp1_CRD, rather than counteracting, accentuated the reduction of the forebrain induced by Wnt8 (Figure 5j,m).

Altogether, these results challenged the view that the CRD domain of the Sfrp1 protein plays an important role in Wnt antagonism. To exclude the possibility that inadequate folding or destabilization of the Sfrp1_CRD construct could mislead this interpretation, we designed an additional construct encoding the CRD and the entire linker region (Sfrp1_CRD2; Figure 1) to ensure proper folding of the Sfrp1 CRD domain 5. Over-expression of this new construct, Sfrp1_CRD2, caused phenotypes similar to those observed upon Sfrp1_CRD injection (Additional file 1). As an alternative explanation, the behaviour of the Sfrp1_CRD could reflect a peculiarity of this specific member of the SFRP family. Therefore, the CRD domain of Sfrp3 (Sfrp3_CRD; Figure 1), the family member that diverges the most from Sfrp1 13, was also analyzed. Interestingly, over-expression of Sfrp3_CRD had no morphologically evident effects on embryonic development, even at high concentrations (Additional file 1; Table 1) and, in contrast to Sfrp1_CRD, failed to enhance Wnt8-induced phenotype (Additional file 1; Table 2).

As a third possibility, we considered that our results could reflect differential affinities between SFRPs and this particular Wnt ligand 16. Therefore, co-injection studies were repeated using two different Wnts: Wnt1, another canonical Wnt that, like Wnt8, can induce posteriorisation of the embryos 17, and Wnt5, which is thought to activate preferentially the non-canonical Wnt signalling pathway 18. As shown in Figure 6, injections of Sfrp1 and Sfrp1_NTR counteracted the phenotype caused by Wnt1-induced phenotype with efficiencies that were very comparable to those observed with Wnt8, while Sfrp1_CRD did not. Wnt5 over-expression in fish and Xenopus embryos leads to variable phenotypes 1819, including defects in axial extension and reduction of the optic vesicle size, albeit less dramatic than those observed with Wnt8 (Additional file 1). Co-injection of Wnt5 with Sfrp1_CRD or Sfrp3_CRD did not rescue the Wnt5-induced phenotype (Additional file 1; Table 2), thus diminishing the relevance of the Sfrp_CRD as a Wnt ligand antagonist. In contrast, our results suggest a relevant role of Sfrp1_NTR in antagonizing Wnt activity.

To explore this possibility further, we next investigated whether the relevance of the NTR domain in antagonizing Wnt ligands could be extended to other SFRP family members or even to distantly related NTR domains 11. According to phylogenetic analysis, the SFRP family is composed of three subfamilies: Sfrp1/2/5, Tlc/Sizzled and the very divergent Sfrp3/4 13. We thus compared the activity of Sfrp1 and Sfrp1_NTR with equivalent constructs from Sfrp2 and Sfrp3 (Figure 1), close and a divergent members of the SFRP family, respectively. Furthermore, we also chose to analyze the NTR domain of Netrin-1 (Figure 1), a secreted protein involved in axon guidance where the NTR domain was first identified 1120 as a distantly related module. When assayed for their ability to reproduce the Sfrp1 over-expression phenotype (Figures 2b and 7b), Sfrp2 and Sfrp2_NTR displayed a significant anteriorising activity almost identical to that of Sfrp1 and Sfrp1_NTR, respectively (Figure 7c,f; Table 1), while Sfrp3 and Sfrp3_NTR had a much weaker activity and expansion of anterior markers was only observed upon injection of high mRNA concentrations (Figure 7d(inset),g; Table 1). Intriguingly, Netrin-1_NTR mRNA injections led to a mild expansion of the forebrain at lower frequency than those of Sfrp1_NTR (Figure 7i; Table 1). These results indicate that, despite the evolutionary distance, this module can mimic SFRP function, presumably by binding to endogenous Wnt.

We next asked whether the tertiary structure of Sfrp1_NTR was important for its function. The NTR motif is, in general, poorly conserved and mainly defined by the presence of six conserved cysteine residues that form three disulfide bonds 511. Mutations of the first two of these residues (Cys177 and Cys180) are predicted to disrupt two disulfide bonds, thus destabilizing the tertiary structure of the NTR domain. Indeed, over-expression of such a mutated construct (Sfrp1_{NTR-C177S;C180S}; Figure 1) did not alter medaka embryonic development (Figure 7h; Table 1), indicating that intact tertiary structure of the NTR motif is required for Sfrp1 activity. Notably, analogous mutations of the first two conserved cysteines of the Netrin-1_NTR (Netrin-1_{NTR-C471S;C475S}; Figure 1) also interfered with, but surprisingly not totally abolished, the anteriorising activity of this domain (Figure 7j; Table 1).

Altogether, these data strongly support that the NTR domain has a relevant role in mediating SFRP function and that this role is conserved also in distantly related domains.

In agreement with our finding that NTR domains of SFRPs are functionally relevant to Wnt signalling modulation, in vitro studies of the interaction between Sfrp1 and Wingless have mapped the relevant SFRP binding site to the carboxyl terminus of the protein 4. To assess whether a similar biochemical interaction between Wnt8 and Sfrp1_NTR could explain our over-expression experiments in medaka fish embryos, we challenged Wnt8 interaction with the two Sfrp1 domains.

To mimic the physiological conditions of the extracellular interaction between Wnt and SFRPs, we collected conditioned media derived from HEK 293T cells separately transfected with Wnt8-HA, Sfrp1-myc, Sfrp1_NTR-myc or Sfrp1_CRD-myc. The levels of proteins present in the conditioned media were carefully evaluated and equivalent amounts of Wnt8 (Figure 8a(iii)) were incubated with comparable quantities of either Sfrp1 or its derivatives (Figure 8a(ii)) and used for co-immunoprecipitation assays. Pull-downs with anti-HA IgG revealed that both Sfrp1-myc and Sfrp1_NTR-myc specifically interacted with Wnt8-HA, while Sfrp1_CRD-myc did not (Figure 8a(i)). Comparable levels of Sfrp1 and its derivatives were pulled down with anti-myc monoclonal antibodies (Figure 8a(iv)), minimising the possibility that the lack of Sfrp1_CRD-Wnt interaction might be due to a less efficient immunoprecipitation of the Sfrp1_CRD. Reverse pull-downs with a polyclonal anti-myc antiserum confirmed these results (Additional file 2).

To further test the functionality of this interaction in β-catenin-mediated Wnt signalling and to compare it with that of other NTR domains, we performed TCF-luciferase reporter-based assays in embryonic retinal cells, where β-catenin-mediated transcriptional activity is physiologically low 12. We thus transfected retina cells with Fz5, a Wnt β-catenin associated receptor expressed in the anterior neural plate 21 to ensure Wnt8-mediated signalling activation 22. Fz5 alone or in combination with Sfrp1, Sfrp1_CRD or Sfrp1_NTR did not modify basal β-catenin activity (Additional file 3). Instead, co-transfection or addition of Sfrp1, Sfrp1_NTR or Netrin-1_NTR conditioned media strongly inhibited reporter activity induced by Wnt8 and Fz5 over-expression (Figure 8b; Additional file 3). Equivalent amounts of Sfrp3 or Sfrp3_NTR were less effective (Figure 8b), in good agreement with what is observed in medaka fish embryos (Figure 7). In apparent contrast with immunoprecipitation experiments, co-transfection of Sfrp1_CRD also resulted in a significant decrease in reporter activity (Figure 8b). Notably, co-transfection with Sizzled or Sizzled_CRD, a SFRP family member that does not appear to interfere with Wnt signalling 23, had a weaker activity (Additional file 3).

Sfrp1 has been shown to form complexes with Fz6 8 and Fz2 24, while crystallographic studies have shown that Fz8_CRD and Sfrp3_CRD can form dimers 10. It was possible, therefore, that Sfrp1_CRD-mediated inhibition of β-catenin transcriptional activity could result from Sfrp1_CRD binding to the Fz5 receptor, thus preventing signal activation as previously proposed 8. To test this possibility, we performed co-immunoprecipitation studies using cell lysates from HEK 293T cells transfected with Fz5-HA, Sfrp1-myc or its derivatives or co-transfected with Fz2-HA, as a positive control 24, and Sfrp1-myc or its derivatives. As shown in Figure 8c, both Sfrp1 and Sfrp1_CRD, but not Sfrp1_NTR, interacted with Fz5-HA, supporting the possibility that Sfrp1_CRD could impede Fz5 activation in TCF-luciferase reporter-based assays by competing with Wnt8 for binding to the Fz receptor. A similar interaction was also observed between Fz2-HA and Sfrp1_CRD-myc as well as with the entire protein (Additional file 2), confirming and extending previous studies 24.

Whole-mount in situ hybridizations were performed in medaka embryos using digoxigenin- and fluorescein-labelled riboprobes. A minimum of 40 embryos were hybridized for each marker and condition. All embryos shown correspond to Iwamatsu stage 19–20 40.

olSfrp1, mWnt8a, zWnt5 and zWnt1 expression constructs have been described 13364142. zSizzled was a kind gift of Dr Hibi and xSizzled of Dr E De Robertis. Medaka Sfrp2 full length clone corresponds to the expressed sequence tag MF01SSA080C03, kindly provided by Dr. Takeda. zSfrp3 and olNetrin-1 where cloned by RT-PCR using specific primers. Full length, truncated and chimerical coding sequences of Sfrp1, Sfrp2, Sfrp3 and Netrin-1 where cloned by PCR into pCS2+. All chimerical constructs where designed so that the signal peptide of the corresponding protein was fused in frame with the linker region that precedes the NTR domain, ensuring proper secretion of the corresponding peptide (Figure 1). Cysteine to serine mutations were introduced into the NTR of both Sfrp1 and Netrin-1 by PCR. Given the structural similarity between serine and cysteine, this substitution is expected to disrupt di-sulphide bridge formation without altering the secondary structure of the peptide. Carboxy-terminal 3xHA tagged constructs of Sfrp1, Sfrp1_CRD and Sfrp1_NTR were generated with linker oligos. All constructs were fully sequenced to ensure in-frame fusions.

pCS2 plasmids were linearised and transcribed in vitro using the SP6 Message mMachine kit (Ambion, Austin, TX, USA). The synthesized mRNA was purified and injected into two-cell stage embryos at different concentrations (titration curve: 50–300 ng/μl) and the severity of the induced phenotypes was dose dependent in all the cases. Injection solutions included 30 ng/ml of hGFP mRNA as a lineage tracer. Selected working concentrations correspond to equimolecular amounts of the different Sfrp mRNAs (full length, truncated and chimerical) to obtain equivalent protein levels (Tables 1 and 2). Mo studies were performed as previously described 13 using the following tested Mo (Gene Tools, LLC, Philomath, OR, USA) designed against olSfrp1: 5'-CTGTGTTT GTAGGAACCTCGACTGG-3'. Mo were injected at the final concentration of 0.3 mM into one blastomere of embryos at the two-cell stage. For co-injection experiments, 60 ng of Sfrp1 or 30 ng of Sfrp1_CRD or 35 ng of Sfrp1_NTR mRNAs were used. At least three independent experiments were conducted for each marker and condition.

To determine the efficiency of translation of the Sfrp1 and its derivatives, triply-HA tagged constructs were generated (see above) and their respective mRNAs were injected into medaka embryos in equimolecular amounts (Sfrp1-3HA, 200 ng/μl; Sfrp1_CRD-3HA, 100 ng/μl; and Sfrp1_NTR-3HA, 120 ng/μl) together with GFP mRNA as a tracer. For each construct, 30 embryos were treated with lysis buffer (150 mM NaCl; 1% NP40; 50 mM Tris pH 8; 10 μg/ml aprotinin; 10 μg/ml leupeptin and 1 mM phenylmethanesulphonylfluoride (PMSF). Lysates were precipitated with a polyclonal anti-HA (Sigma-Aldrich, St Louis, MI, USA) and Protein G-Sepharose for enrichment. The protein complex present in each of the pellets was re-suspended in 2 × SDS sample buffer containing 1 M urea. The proteins were resolved by SDS-PAGE blotted and the membranes probed with a monoclonal anti-HA (Sigma-Aldrich). Proteins from total cell extracts were subjected to SDS-PAGE, blotted and the membranes probed with an anti-GFP antibody (Molecular Probes, Invitrogen, Carlsbad CA, USA) and a secondary anti-rabbit-POD antibody.

Sub-confluent HEK 293T cells were transiently and separately transfected with constructs encoding chick Wnt8c-HA, chick Sfrp1-myc or Sfrp1_CRD-myc or Sfrp1_NTR-myc in 2% fetal calf serum. After 2 days, the conditioned media were collected and clarified by centrifugation. The amount of protein present in the conditioned media was evaluated by western blot and similar amounts of peptides derived from each Sfrp1-myc present in the conditioned media were mixed with conditioned medium from Wnt8-HA or mock transfected for 2 hours. Sample volumes were adjusted to 600 μl with lysis buffer (as above). Proteins from conditioned media were precipitated with 3 μg of an anti-HA polyclonal antibody (Sigma-Aldrich) and Protein G-Sepharose. After four washes with lysis buffer, the protein complex was subjected to SDS-PAGE, blotted and the membranes probed with a monoclonal anti-myc antibody (9E10) and a secondary anti-mouse-POD antibody. Signal was detected with the Advanced ECL Western blotting detection Kit analysis (GE Healthcare Life Sciences, Pollards Wood, Buckinghamshire, UK). Reverse inmunoprecipitation experiments were performed using similar incubations of conditioned media. Proteins were precipitated with a polyclonal anti-myc antibody (SIGMA). The immunocomplexes were subjected to SDS-PAGE, blotted and the membranes probed with a monoclonal anti-myc antibody (9E10) and a secondary anti-mouse-POD antibody.

For Fz2 and Fz5 immunoprecipitations, HEK 293T cells were transiently transfected with mouse Fz2-HA, chick-Sfrp1-myc or Sfrp1_CRD-myc or Sfrp1_NTR-myc or cotransfected with mouse Fz5-HA and chick-Sfrp1-myc or Sfrp1_CRD-myc or Sfrp1_NTR-myc expression constructs. After 2 days, cells were scraped in lysis buffer (as above). Immunoprecipitations were performed as previously described 24.

Dissociated cells from embryonic day (E)5 central retinas were prepared as described 29, seeded in 24-well plates and transfected 3 hours later using the FuGENE HD Transfection Reagent (Roche, Nutley, NJ, USA). In each case the 700 ng/well of total DNA contained 200 ng of a plasmid containing a 4xLef-1 responsive luciferase reporter and 50 ng of pRL-TK (Promega, Madison, WI, USA) together with variable amounts of the effector plasmids or the empty vector. After 24 hours, luciferase activities were determined using a dual-luciferase assay system (Promega). The LEF-1 reporter luciferase activity was normalized with that of the Renilla luciferase to account for transfection efficiency. Data were statistically evaluated using the SPSS v15.0 software (SPSS Inc., Chicago, Illinois, USA) applying a one-way ANOVA test plus post hoc test (Dunnet test).

Live embryos were visualized at room temperature under a Leica stereomicroscope equipped with a PLANAPO objective. Embryos processed for in situ hybridization were fixed with 4% paraformaldehyde (PFA) and equilibrated in 80% glycerol. After removal of the yolk, embryos were mounted and visualized under a Leica microscope. In all cases, images were captured with a Leica digital camera controlled by the Leica software.