Secreted Wnt “inhibitors” are not just inhibitors: Regulation of extracellular Wnt by secreted Frizzled-related proteins


*Author to whom all correspondence should be addressed.


Gradient formation and signaling ranges of secreted proteins are crucial problems to understand how morphogens work for positional information and patterning in animal development. Yet, extracellular behaviors of secreted signaling molecules remain unexplored compared to their downstream pathways inside the cell. Recent advances in bioimaging make it possible to directly visualize morphogen molecules, and this simple strategy has, at least partly, succeeded in uncovering molecular behaviors of morphogens, such as Wnt (wingless-type MMTV integration site family member) and BMP (bone morphogenetic protein) as well as secreted Wnt binding proteins, sFRPs (secreted Frizzled-related proteins), in embryonic tissues. Here, we review the regulation of Wnt signaling by sFRPs, focusing on extracellular regulation of Wnt ligands in comparison with other morphogens. We also discuss evolutionary aspects with comprehensive syntenic and phylogenetic information about vertebrate sfrp genes. We newly annotated several sfrp genes including sfrp2-like 1 (sfrp2l1) in frogs and fishes and crescent in mammals.


Turing proposed the concept “morphogen” as the substance that gives positional information in developmental processes by concentration gradient formation (Turing 1952). Since then, the notion of morphogen has attracted not only experimental but also theoretical biologists in various fields (Wolpert 1969; Tabata & Takei 2004; Zhu & Scott 2004; Ben-Zvi & Barkai 2010). To date, several kinds of molecules have been postulated to act as morphogens. For example, transcription factors like Bicoid in the Drosophila syncytial blastoderm, and peptide growth factors like Sonic hedgehog and small compounds like retinoic acid (RA) in vertebrate embryos are well-known morphogens (Ephrussi & St Johnston 2004; Campo-Paysaa et al. 2008). In this review, we focus on secreted signaling proteins, the Wingless and int-1 family protein (Wnt) (Logan & Nusse 2004; Clevers 2006) and its so-called secreted inhibitors, secreted Frizzled-related proteins (sFRPs) (Kawano & Kypta 2003; Bovolenta et al. 2008). We also refer to the bone morphogenetic protein (BMP), which is postulated to act as a morphogen (De Robertis & Kuroda 2004).

Developmental model systems for studying morphogens

Earlier works of morphogens have largely been done with Drosophila. Although Bicoid is not a secreted protein but a transcription factor, it is one of the best studied morphogens (Kugler & Lasko 2009). Many leading studies of secreted morphogens have been carried out using wing discs of Drosophila, mainly because of its simple structure composed of a sheet of cells and its great advantage of genetics. For instance, the Wingless (Wg) gradient in the wing disc is reportedly formed by diffusion (Strigini & Cohen 2000). Hedgehog (Hh) and Decapentaplegic (Dpp) also act as key morphogens in the wing disc (Tabata & Takei 2004). Their vertebrate homologues, Sonic hedgehog (Shh) (Chiang et al. 1996), bone morphogenetic proteins (BMPs) (De Robertis & Kuroda 2004), and Wnts (Kiecker & Niehrs 2001) have also been suggested to have important roles in various developmental processes as morphogens.

Compared with the Drosophila wing disc, study of morphogens in vertebrate embryos seems to have a lot of space to be explored. Among vertebrates, zebrafish and Xenopus embryos as well as chick limb buds have been used as popular model systems to study morphogens. Studies using chick limb buds have revealed fibroblast growth factors (FGFs) (Martin 1998), Shh (Riddle et al. 1993), and RA (Thaller & Eichele 1987) as molecular substances of morphogens for digit patterning. The animal cap of Xenopus has been used for studying Activin and BMP as morphogens, because of its simple structure and convenience for ectopic expression by mRNA injection (Smith 2009). Gurdon and his colleagues have shown that passive diffusion of Activin and transforming growth factor-β1 (TGF-β1) directly causes concentration-dependent activation of downstream target genes (Gurdon et al. 1994; Dyson & Gurdon 1998; McDowell et al. 2001). In this review, we reintroduce the Xenopus embryo as a good system for examining extracellular regulation of Wnt gradient formation, by co-expression of fluorescent-tagged Wnt ligands and sFRPs.

Regulation of Wnt signaling area by sFRPs

Overview of Wnt and sFRP

Wnts constitute a family of secreted signaling proteins and are thought to have evolved during the establishment of Metazoa after divergence from choanoflagellates (Adamska et al. 2007). Wnt signaling is one of the most intensively studied signaling pathways due to its importance for animal development, such as dorsoventral axis formation (Tao et al. 2005; Cha et al. 2009), and carcinogenesis, such as mammary and colorectal cancers (Bienz & Clevers 2000). Wnt signaling can be divided into the canonical pathway (Clevers 2006) and non-canonical pathways (Veeman et al. 2003). In early vertebrate embryos, canonical Wnt signaling is involved, for example, in the anteroposterior (AP) patterning by generating a posterior-to-anterior gradient as assayed by nuclear β-catenin (Kiecker & Niehrs 2001). Concomitantly, non-canonical Wnt signaling, especially the planar cell polarity (PCP) pathway regulates morphogenetic cell movements including convergent extension movements (CEMs) in the posterior region (Keller 2002).

As a reflection of the importance of Wnt signaling, there are several types of secreted regulatory proteins for Wnt signaling, such as sFRPs (ex. Frzb and Crescent) (Leyns et al. 1997; Wang et al. 1997; Pera & De Robertis 2000; Shibata et al. 2000), Cerberus (Bouwmeester et al. 1996; Piccolo et al. 1999), Dickkopf (Dkk) (Niehrs 2006), and Wnt-inhibitory factor (WIF) (Hsieh et al. 1999a). Of them, sFRPs, Cerberus, and Dkk are expressed in the anterior part of the Spemann-Mangold organizer, called the head organizer (Niehrs 2004), and supposedly prevent posteriorization of the anterior region by Wnt signaling (Kawano & Kypta 2003). Among these Wnt inhibitory proteins, Frzb was identified as the first described secreted Wnt inhibitor (Leyns et al. 1997; Wang et al. 1997). Frzb (Hoang et al. 1996) as well as SDF5 (stromal cell-derived factor 5, a synonym of sFRP2) (Shirozu et al. 1996) was reported as a secreted protein, which has homology with the Wnt receptor Frizzled (Schulte 2010). The shared domain between sFRPs and Frizzled is called the cysteine rich domain (CRD) (Banyai & Patthy 1999). sFRPs also have the Netrin (NTR) domain in common with Netrins and other protein families such as complement proteins, C3, C4, and C5, and type I procollagen C-proteinase enhancer proteins (Banyai & Patthy 1999; Chong et al. 2002).

A well-known function of sFRPs is to inhibit Wnt ligands from signaling by direct binding. Several sFRP members reportedly play such inhibitory roles for Wnt in embryogenesis. For instance, Frzb, which is expressed in the head organizer of Xenopus, inhibits Wnt8, a canonical Wnt ligand as a posteriorizing factor (Leyns et al. 1997; Wang et al. 1997). According to phenotypes of knockout mice, sFRP1, sFRP2, and sFRP5, which have some redundancy in their functions, are suggested to be involved in AP patterning of the vertebrae via the canonical Wnt pathway and morphogenetic cell movements via the Wnt PCP pathway (Satoh et al. 2006, 2008; Warr et al. 2009). Also Crescent and sFRP5 are suggested to regulate CEMs or morphogenetic cell movements as antagonists of Wnt ligands that stimulate the PCP pathway. Crescent is expressed in the head organizer and binds not only to Wnt3a and Wnt8, which mainly activate the canonical pathway, but also to Wnt4, Wnt5a, and Wnt11, which mainly activate non-canonical pathways, but is involved mainly in the regulation of CEMs in the posterior region (Pera & De Robertis 2000; Shibata et al. 2005). Like Crescent, sFRP5 can bind and antagonize Wnt5a and Wnt11 in gut specification and morphogenesis, in which Wnt11 activates both canonical and non-canonical signaling (Li et al. 2008). Thus, sFRPs have been shown to act as counterparts of Wnts in gain-of- and loss-of-function experiments. Furthermore, it should be noted that other functions of some sFRPs have been reported, such as the regulation of axon guidance by binding to Frizzled receptors, the interference of BMP signaling by acting as proteinase inhibitors (see below for Sizzled and Crescent), and the inhibition of apoptosis by interacting with the integrin-fibronectin complex (reviewed in [Bovolenta et al. 2008]). Thus, sFRPs have distinct or multiple functions in a context dependent manner.

Signaling area of Wnt expanded by sFRPs

In spite of the importance of Wnt proteins as morphogens in patterning in the Drosophila wing disc and in the AP patterning of early vertebrate embryogenesis, only a few studies have reported on the regulatory mechanism for the range of Wnt signaling. To elucidate the mechanism of the AP patterning, diffusion properties of Wnt and sFRP, which are expressed posteriorly and anteriorly, respectively, would be a key factor. Therefore, we started with visualization of Wnt ligands. We constructed a simple system to examine the distribution range of secreted proteins (Fig. 1A,B). In this system, we can easily observe Venus-tagged proteins in the extracellular space secreted from their source cells. With this method, we found that Frzb and Crescent spread throughout the embryo, whereas Wnt8 and Wnt11 were not or barely detected in the region distant from the source region (Fig. 1C) (Mii & Taira 2009). This big difference in distribution ranges between sFRPs and Wnts prompted us to examine the distribution range of Wnt ligands with co-expressed sFRPs. Interestingly the distribution range of Wnt8 and Wnt11 was expanded by co-expressed Frzb and Crescent (see Fig. 1C for the combination of Wnt8 and Frzb), except for the combination of Wnt11 and Frzb, which was shown to have a weak interaction (Shibata et al. 2005). Importantly, we have shown that sFRPs can also expand a signaling range of Wnt8, despite their molecular functions as Wnt inhibitors, suggesting another regulatory system for Wnt signaling (Mii & Taira 2009).

Figure 1.

 Measurement for a distribution range of secreted proteins in the Xenopus embryo. (A) Constructs for Venus-tagged secreted Frizzled-related proteins (sFRPs) or Wnts. The signal peptide of transforming growth factor-β1 (TGF-β1) was commonly used. (B) Schematic of the experiment. mRNAs for Venus-tagged proteins and monomeric red fluorescent protein (mRFP) as a cell lineage tracer were co-injected into the animal pole region of a ventral blastomere at the four-cell stage. The injected embryos were cultured until the gastrula stage for observations. In the bottom panel, the left cartoon represents a Venus-tagged protein that widely spreads, whereas the right one represents a Venus-tagged protein that stays near the source region. (C) Expanding distribution of Wnt with sFRP. mRNAs for Venus-tagged Wnt, mRFP, and sFRP were co-injected (the right-most column). A combination of Wnt8 and Frzb is only shown. Upper panels, distribution of mRFP (magenta) as the source region; middle panels, distribution of Venus-tagged proteins (green); lower panels, merged pictures of mRFP and Venus signals. Note that overlay of green and magenta looks white.

Although the molecular mechanism of the range expansion remains uncertain, we propose here a possible mechanism as follows (Fig. 2). Wnt ligands reportedly interact with heparan sulfate proteoglycans (HSPGs) (Lin & Perrimon 1999; Tsuda et al. 1999; Ohkawara et al. 2003) and Frizzled receptors (Bhanot et al. 1996; Hsieh et al. 1999b). Wnts are also lipid-modified (Willert et al. 2003; Takada et al. 2006) and possibly interact with the plasma membrane. Because of these interactions, Wnt ligands tend to stay near the source region. When sFRPs co-exist, however, Wnts become more diffusible, possibly binding of sFRPs inhibits interactions between Wnt ligands and HSPGs and/or the plasma membrane. As a result, the Wnt-sFRP complex might become more diffusible, which leads to a wider distribution. Because protein–protein interactions are generally reversible, the signaling range of Wnt can be expanded by sFRPs.

Figure 2.

 A model for signaling range expansion. See the text for details. Interactions with heparan sulfate (blue distorted lines) and the plasma membrane (gray thick horizontal line) tether Wnt ligands near the source region. However, secreted Frizzled-related protein (sFRP) interferes with the interactions between Wnt and heparan sulfate proteoglycan (HSPG) or the plasma membrane. Therefore, Wnt-sFRP complexes become diffusible and Wnt ligands released from these complexes can activate Wnt signaling by binding to the Wnt receptor Frizzled. The lower panel represents gradients of sFRP, Wnt, and their complex in anteroposterior (AP) patterning as indicated. In Xenopus, sFRPs, such as Frzb and Crescent, are expressed in the anterior part of the Spemann-Mangold organizer, called the head organizer, whereas Wnt8 and Wnt11 are expressed in the posterior region (Mii & Taira 2009).

Gene expression pattern analyses of the early chicken and mouse embryos also show that Wnts are expressed in the posterior region and sFRPs are expressed in the anterior region, (Chapman et al. 2004; Kemp et al. 2005), as shown in Xenopus (see Fig. 2). Studies using amphioxus and ascidian suggest that this arrangement of Wnts and sFRPs is basically conserved among chordates (Sasakura et al. 1998; Holland 2002; Lamy et al. 2006; Yu et al. 2007). Because Wnts and sFRPs are both secreted, it has been assumed that they form “opposing gradients”, posterior-to-anterior and anterior-to-posterior, respectively. So far, the opposing gradients of Wnts and sFRPs have been understood as a postulation that sFRPs are mere inhibitors of Wnt signaling. However, our finding suggests that they could expand ranges of Wnt for AP patterning (Fig. 2).

Interestingly, Wg in Drosophila is reportedly a long-range morphogen (Neumann & Cohen 1997), contrary to our results in Xenopus. As Drosophila seems to have lost sfrp genes (Lee et al. 2006b; Bovolenta et al. 2008), it might gain special mechanisms instead of using sFRPs (we will discuss these mechanisms later).

Besides AP patterning in vertebrates, there are many Wnt ligands and many sFRPs expressed in various parts of the embryo. Although comprehensive binding specificities of Wnts and sFRPs have not been examined yet, there can be various combinations between a Wnt ligand and a sFRP with different binding affinities and with different effects on Wnt signaling. For instance, sFRP3 (a synonym of Frzb) binds to Wnt3a, but it does not inhibit canonical Wnt signaling activated by Wnt3a (Galli et al. 2006). Therefore, it is possible that sFRP3/Frzb function as a conveyer of Wnt3a, as suggested by our study. Besides AP patterning, various Wnt ligands play versatile roles in embryogenesis. It is an attractive idea that, because Wnt itself is a short range morphogen, with the help of sFRPs as range modulators, the Wnt signaling area becomes controllable.

“More or less” sFRPs in cancer

As int-1, the former name of wnt1 gene, was initially identified as an oncogene, excessive Wnt/β-catenin signaling can lead to tumorigenesis (Nusse & Varmus 1982; Rubinfeld et al. 1993; Su et al. 1993). As expected from the fact that sFRPs can modulate Wnt signaling, perturbation of sfrp genes are also involved in tumorigenesis (Table 1). Interestingly, both up- and downregulations of sfrp expression are associated with tumorigenesis. Downregulation of sfrp genes and subsequent activation of β-catenin are predictable, given that they are Wnt inhibitors. However, in some cases, upregulation of sfrp genes is accompanied with nuclear accumulation of β-catenin. One possibility is that sFRPs enhance diffusion and availability of Wnt ligands as we demonstrated in the Xenopus embryos (Mii & Taira 2009), which enhances canonical Wnt signaling, leading to tumorigenesis.

Table 1.   Secreted Frizzled-related proteins (sFRPs) in cancer
GeneKind of cancerSpeciesUp- or downregulation of sfrp expressionReference
sfrp1Colorectal cancerHumanDownCaldwell et al. 2004
Renal cell carcinomaHumanDownDahl et al. 2007
Hepatocellular carcinomaHumanDownHuang et al. 2007
sfrp2Mammary gland tumorsDogUpLee et al. 2004
Renal cell carcinomaHumanUpYamamura et al. 2010
sfrp3/frzbRenal cell carcinomaHumanBoth up and downHirata et al. 2010
sfrp4OsteomalaciaRatUp (inhibits Wnt/β-catenin signaling)Berndt et al. 2003
Colorectal carcinomaHumanUpFeng Han et al. 2006
sfrp5Renal cell carcinomaHumanDownKawakami et al. 2011

Phylogenetic analysis of vertebrate sfrp genes including newly found paralogues

To understand how sfrp genes evolved with wnt genes, we refer here to the evolutional origin of sfrp genes (see reviews [Guder et al. 2006; Lee et al. 2006b] for the evolution of wnt genes). The existence of secreted proteins named “SFRPs” as well as that of Wnts was reported in a demosponge Amphimedon queenslandica, but these proteins lack the NTR domain. “True sFRPs,” which have both the CRD and the NTR domain, were found in another demosponge, Lubomirskia baicalensis (Adell et al. 2007; Adamska et al. 2010), suggesting that sFRPs have regulated Wnt signaling from the early history of Wnt. In the genome of the sea anemone Nematostella vectensis, there are two sfrp genes that consist of the CRD and the NTR domain (Adamska et al. 2010). Considering that sponges appear to be paraphyletic, it is suggested that the common ancestor of eumetazoans evolved from one of the sponges that had “true sFRPs.” Given this evolutionary scenario, it is interesting to note that Drosophila seems to have lost sfrp genes during evolution, because other protostomes including Caenorhabditis elegans and the planarian maintain sfrp genes (Gurley et al. 2010; Harterink et al. 2011).

There are three paralogous groups of sfrp and five to eight sfrp paralogues in vertebrate genomes. These numbers are fewer than those of the Wnt family, which has 12 eumetazoan paralogous groups and total 19 paralogues in the mouse genome ([Garriock et al. 2007], see also the Wnt homepage: The mouse and human reportedly have five sfrp genes, named sfrp1 sfrp2, frzb (sfrp3), sfrp4, and sfrp5. Here we describe the crescent gene in the human and opossum genomes, which has not been reported before (see below). Besides these genes, the frog has two more sfrp genes, sizzled and the one that we recently found in expressed sequence tags (ESTs) of Xenopus laevis and Silurana tropicalis (commonly called Xenopus tropicalis), named sfrp2-like 1 (sfrp2l1). Our annotated genes are shown in Table 2.

Table 2.   Newly annotated sfrp genes
GeneSpeciesAccession numberNotes
  1. *Ensemble peptide ID.

crescentHumanXM_003120401Annotated as sfrp5-like
crescentOpossumNot assigned
sfrp2-like 1Xenopus laevisNM_001105267sfrpx (provisional)
sfrp2-like 1Xenopus tropicalisNM_001097341sfrpx (provisional)
sfrp2-like 1Stickleback*ENSGACP00000010386 
sfrp2-like 1Medaka*ENSORLP00000013980 

The significance of such genetic diversity of this family and their multiple functions mentioned above appears to be a notable feature of the sFRP family. Although the phylogenetic relationship of vertebrate sFRPs has been reported (Jones & Jomary 2002; Kawano & Kypta 2003), syntenic relationships have not been reported yet (see a reference [Garriock et al. 2007] for syntenic analysis of wnt genes). Therefore, we performed comprehensive phylogenetic analysis for all vertebrate sFRPs with their synteny maps (Fig. 3) and a phylogenetic tree generated by the ClustalW alignment tool and the neighbor-joining method (Fig. 4). As previously reported (Bovolenta et al. 2008), vertebrate sfrp genes are classified into three subfamilies, sfrp1/2/5, frzb/sfrp4, and crescent/sizzled (Figs 3, 4). We found that, besides these paralogous members, there are species-specific sfrp genes in fishes and frogs as follows.

Figure 3.

 Conserved synteny of vertebrate sfrp genes. A boat-shape object represents a gene with a direction, in which the tip of boat corresponds to the 3′ of the gene. Genes indicated with a same color mean orthologous genes in the same synteny map group (boxed region), but does not mean this in different map groups. In the blue box, genes with the red outline indicate common paralogues between the sfrp1 and sfrp5 synteny map groups. Otherwise, the same colors between different synteny map groups do not mean the same genes. These maps are drawn based on JGI Metazome data, with some manual editing and corrections. Genes colored in gray are manually annotated as homologous genes; different gray densities mean different homologous genes. Note that synteny maps of the human and opossum around the crescent gene were manually edited, according to synteny maps of surrounding genes with Metazome gene family numbers 1989871, 1973228, and 1997995, because the human and opossum crescent genes had not been identified. Scientific names of species are as follows: human, Homo sapiens; mouse, Mus musculus; opossum, Monodelphis domestica; chicken, Gallus gallus; frog, Xenopus tropicalis; stickleback, Gasterosteus aculeatus; medaka, Oryzias latipes; pufferfish, Takifugu rubripes; zebrafish, Danio rerio.

Figure 4.

 A phylogenetic tree of vertebrate secreted Frizzled-related protein (sFRP) proteins. The tree was constructed by the neighbor-joining method with the ClustalW program. Values beside nodes show the number of times that a node was supported in 1000 bootstrap pseudoreplications. The scale bar indicates an evolutionary distance of 0.1 substitutions per position. Abbreviations for species names are as follows: Hsa, Homo sapiens; Mmu, Mus musculus; Mdo, Monodelphis domestica; Gga, Gallus gallus; Xtr, Xenopus tropicalis; Gac, Gasterosteus aculeatus; Ola, Oryzias latipes; Dre, Danio rerio. *The zebrafish Danio rerio has a unique sfrp gene, tlc (Houart et al. 2002). Our BLAST searches did not find tlc genes in other organisms including medaka, pufferfish, and stickleback. Amino acid sequence alignments with the entire proteins indicated that Tlc has the common ancestor with the sFRP1/5 clade as previously shown (Houart et al. 2002). However, the cysteine rich domain (CRD) sequence of Tlc is closer to that of Crescent than those of sFRP1/5 (Bovolenta et al. 2008). Thus, the evolutionary origin of tlc is obscure. **The sfrp4 gene exists in the X. tropicalis genome (Metazome protein ID: 180191), which was previously reported only in amniotes (Jones & Jomary 2002), but was not found in teleost genomes of stickleback, medaka, pufferfish, and zebrafish, suggesting that the sfrp4 gene evolved probably in the common ancestor of tetrapods.

sfrp2 and sfrp2-like 1 in fishes and frogs

We found that fishes and frogs have another sfrp2 paralogue, named sfrp2-like 1 (sfrp2l1), in their genomes, whereas amniotes seem to have a single sfrp2 gene, because BLAST search did not detect sfrp2l1 in the chick, mouse, and human genomes. Syntenic analysis showed that synteny of sfrp2l1 is conserved between fishes and frogs, suggesting that amniotes lost sfrp2l1.

crescent but not sizzled in mammals

Two closely related genes, crescent and sizzled are reportedly present in the vertebrate except for mammals. However, here we found a crescent gene in human and opossum genomes in the same syntenic region as other vertebrates (Figs 3, 4), using National Center for Biotechnology Information (NCBI) and Metazome tblastn programs with the protein sequence of chicken Crescent as a query. We also found traces of a crescent gene at the corresponding region in the mouse genome (existing in intron 2 of the mammalian P4htm gene); however, the coding sequence has been broken with many stop codons, suggesting that the mouse has lost the crescent gene. We could not find sizzled genes in the mammalian genomes with BLAST searches.

Nomenclature of sfrp genes

Because there is some confusion in the sfrp gene nomenclature, we summarized their synonyms (Table 3). Confusing names are as follows. (i) Independently identified SARPs (secreted apoptosis-related proteins) Sarp1, Sarp2, and Sarp3 turned out to belong to the sFRP family (Melkonyan et al. 1997); However, note that the numbering is not consistent, as shown in Table 3. (ii) frza is another name of sfrp1. (iii) sfrp3 is another name of frzb. X. laevis has two frzb genes, frzb and frzb-1. We compared coding sequences of these X. laevis frzb (Wang et al. 1997) and frzb-1 (Leyns et al. 1997) genes and X. tropicalis frzb using ClustalW (data not shown), and found that frzb and frzb-1 are both orthologues of X. tropicalis. X. laevis is allotetraploidy or amphidiploid, and hence generally has two related genes, called homoeologues. Therefore, frzb and frzb-1 are likely to be “homoeologues” in X. laevis. (iv) frzb-2 is another name of crescent; however, this name is somewhat confusing because crescent is not the closest relative of frzb. More confusingly, frzb-2 was once used for human sfrp4 (James et al. 2000). (v) In X. laevis, two sizzled genes named sizzled and sizzled2, have been reported (Salic et al. 1997; Bradley et al. 2000). We compared coding sequences of these two X. laevis sizzled genes and X. tropicalis sizzled (data not shown), and found that sizzled and sizzled2 are likely to be homoeologues. Overall, we elucidate a comprehensive view of vertebrate sfrp genes.

Table 3.   Nomenclature of sfrp genes in vertebrates
Approved namesProvisional nameSuggested nameAlso known as 
  1. Abbreviations for species names are as follows: Hsa, Homo sapiens; Mmu, Mus musculus; Rno, Rattus norvegicus; Gga, Gallus gallus; Xla, Xenopus laevis; Xtr, Xenopus tropicalis; Dre, Danio rerio. Approved names are according to Mouse Genome Informatics for sfrp1, sfrp2, frzb, sfrp4, and sfrp5, Xenopus laevis and Xenopus tropicalis biology and genomics resource for sizzled and sfrp2l1, and The Zebrafish Model Organism Database for tlc.

sfrp1 (Mmu)  sarp2 (Mmu, Hsa)frza (Bta)
sfrp2 (Mmu)  sarp1(Mmu, Hsa)sdf5 (Mmu)
frzb (Mmu)  sfrp3 (Mmu, Hsa)frzb-1 (Xla; homoeologue)
sfrp4 (Mmu)  frzb-2 (Hsa)ddc-4 (Rno)
sfrp5 (Mmu)  sarp3 (Mmu, Hsa) 
crescent (Gga)frzb2 (Xtr)crescentfrzb-2 (Xla)sfrp5-like (Hsa)
sizzled (Xtr)  sizzled2 (Xla; homoeologue) 
 sfrpx (Xtr)sfrp2-like 1  
tlc (Dre)    

Regulation of BMP signaling range by binding proteins

The regulation of patterning by BMP/Dpp and its antagonist Chordin (Chd)/Short gastrulation (Sog) is one of the most intensively studied systems of morphogens at the molecular level. This situation is similar to that of Wnts and sFRPs, thereby worth discussing here. Vertebrates and Drosophila use the BMP/Sog system to create a morphogen gradient for dorsoventral (DV) patterning in early embryogenesis including neural induction in vertebrates (De Robertis & Kuroda 2004). Compared to the Wnt-sFRP system, the BMP-Chd (Dpp-Sog in Drosophila) system seems to be more complicated, because this system involves additional factors, such as Twisted gastrulation (Tsg) and Tolloid (Tld). Tld and its related proteases are necessary for the cleavage of Chordin to release BMP ligands (Blader et al. 1997; Marques et al. 1997; Piccolo et al. 1997). Sog and Tsg facilitate the diffusibility of Dpp (Eldar et al. 2002). Shimmi et al. (2005) have shown that the facilitated diffusion of a Dpp and Screw (Scw) heterodimer to the dorsal midline by forming a complex with Sog and Tsg together, leading to the robustness of the DV patterning in Drosophila. Another BMP type ligand, ADMP (anti-dorsalizing morphogenetic protein) is expressed in the dorsal side of the Xenopus embryo, which is opposite to ventral expression of other BMP ligands (BMP2/4/7), and this regulates the total BMP signaling level by feedback loops (Reversade & De Robertis 2005). Sizzled, a member of sFRPs, inhibits the protease activity of Tld (Lee et al. 2006a; Muraoka et al. 2006). Recently, Crescent has been shown to inhibit Tld (Ploper et al. 2011), being reasonable because crescent is the closest paralogue of sizzled as mentioned above. Taking the facilitated transport of the BMP ligand and its regulatory factors into consideration, Ben-Zvi et al. (2008) proposed “shuttling” of the BMP ligand and “scaling” of DV patterning along with the embryo body size. In addition, a recent study elucidated another secreted protein, ONT1 (Olfactomedin–Noelin–Tiarin protein 1) also contributes to the robustness of DV patterning as a scaffold for Chordin degradation (Inomata et al. 2008). These studies demonstrate impressively multiple implements of robustness and autonomy in the DV patterning of Xenopus and Drosophila. Compared to the BMP-Chd system, the Wnt-sFRP system is much simpler. However, without other regulatory proteins, binding equilibrium of Wnt to Frizzled and sFRP could occur in a similar order, because sFRPs contain the CRD in common with the Frizzled receptor, and reported affinity of Wnt-Frizzled and Wnt-sFRP are the same order (Hsieh et al. 1999b; Wawrzak et al. 2007). Alternatively, other factors that are involved in the Wnt-sFRP system might be awaiting discovery. Nevertheless, it is common between Wnt-sFRP and BMP-Chd systems that so-called binding “inhibitors” enhance the diffusion of the ligands, suggesting that such systems have some advantages including robustness or versatility, or both, in general.

Mechanisms proposed for morphogen gradient formation

Although the idea of morphogen has been widely accepted, several mechanisms are proposed for gradient formation or transportation of morphogens, especially in Drosophila (see a review [Zhu & Scott 2004]). Here we focus on the extracellular regulation of gradient formation. In the Drosophila wing disc, there are a variety of mechanisms including passive diffusion (Strigini & Cohen 2000), “argosomes/lipoprotein particles” (Greco et al. 2001; Panakova et al. 2005), “cytonames” (Roy et al. 2011), and “transcytosis” (Entchev et al. 2000). Even in the situation where passive diffusion is dominant, things do not seem to be so simple, because there are various molecules that have interactions with Wg/Wnt, Dpp/BMP, and Hh ligands. Interestingly, several groups reported that mutations of sugar chain synthesis or modifying enzymes or mutations of heparan sulfate proteoglycans (HSPGs) such as Dally and Dally-like severely reduce extracellular levels of ligands (Dpp, Wg, and Hh) in the Drosophila wing disc (Kirkpatrick et al. 2004; Takei et al. 2004; Han et al. 2005). These phenomena suggest that those morphogens are basically bound to HSPGs in the extracellular space and might spread into tissues through “restricted diffusion”, in which the ligand can diffuse only when it is released from HSPGs or other extracellular matrix (ECM) components. Therefore, ECMs (especially HSPGs) seem important to define a diffusible field of morphogens, or it may be a basic requirement for forming a gradient (see reviews [Yan & Lin 2009; Nishihara 2010]). From this point of view, HSPGs are suggested to keep suitable concentrations for ligand-receptor interactions in addition to their functions as co-receptors (reviewed by [Bernfield et al. 1999]).

Receptors not only transduce signaling but also could affect ligand distribution as suggested by Patched, which is upregulated by Hh as a negative feedback regulation (Chen & Struhl 1996), and overexpression of Fz2 in the Drosophila embryo (Cadigan et al. 1998). However, in other situations, receptors indirectly affect gradient formation. For example, removal of both Fz and Fz2, or their co-receptor Arrow (LRP5 and LRP6 in vertebrates) upregulates Dally-like and hence increases a extracellular Wg level (Han et al. 2005). Furthermore, some ligands such as Wnt and Hh are shown to be modified with lipid (Porter et al. 1996; Pepinsky et al. 1998; Willert et al. 2003; Takada et al. 2006). By hydrophobic interactions of ligands with the plasma membrane, lipid modifications may regulate ligand concentrations and signaling areas as HSPGs seem to do (reviewed by [Miura & Treisman 2006]). Besides these extracellular regulations mentioned above, intracellular processes, such as internalization of ligands by endocytosis, can also affect morphogen gradients (Dubois et al. 2001; Le Roy & Wrana 2005). Thus, many factors are directly or indirectly involved in the extracellular distribution of ligands.

Although there are various mechanisms involved in morphogen gradient formation in Drosophila as mentioned above, passive diffusion seems dominant in vertebrate embryo as exemplified by Activin and Nodal proteins (McDowell et al. 2001; Williams et al. 2004). Compared to Drosophila, there are not many studies in morphogen and heparan sulfate in vertebrate embryogenesis; however, here we refer to two excellent works. Using the Xenopus animal cap, Ohkawara et al. (2002) have shown that the signaling range of BMP4, which acts in a short range, is defined by its N-terminal basic amino acid core, which is suggested to bind to HSPGs. Harada et al. (2009) have proposed the model that the affinity of FGF9 to HSPGs regulates the distribution and signaling range in mice. Thus, binding affinities of ligands to the ECM or the cell membrane are probably important factors for defining distribution and signaling ranges.

In conclusion, morphogen gradients are formed by various mechanisms involving various molecules and their interactions. However, analyses of morphogen gradient formation using vertebrate model animals seem to be still limited.

Quantitative and theoretical ways to analyze gradient formation

Among many biological processes in embryogenesis, morphogen diffusion and the subsequent patterning have a close relationship with quantitative biology and theoretical biology ever since Turing. In many published reports, the range of morphogen signaling is often described as “long” or “short” (Williams et al. 2004; Clevers 2006). Apparently a signaling range is an important factor for understanding the patterning governed by a morphogen, and seems to correspond to the range of ligand distribution (Takei et al. 2004; Mii & Taira 2009). Thus, visualizing a ligand with a fluorescent protein or an epitope tag is an informative way to examine the signaling range. However, the range of ligand distribution itself is determined by many parameters. For example, “slow diffusion and slow degradation” and “fast diffusion and fast degradation” can result in a similar distribution range. Therefore, it will be more important to describe the nature of morphogen in quantitative ways. At the molecular level, the ligand distribution can be thought to be determined by a combination of intrinsic diffusion property, production and degradation rates, a shape of extracellular space, and binding affinity to the ECM and the plasma membrane. These phenomena can be described in some physical or chemical constants or coefficients. With the technical development of live imaging, it has become possible to handle the dynamic behavior of morphogen gradient formation (Mavrakis et al. 2010). We discuss recent achievements on quantitative and theoretical approaches in this field.

To date, fluorescence recovery after photobleaching (FRAP) is a standard technique to analyze the diffusion coefficient and related properties of a protein of interest (Sprague & McNally 2005), but, in most studies, have been used to analyze proteins inside the cell. Among others, Kicheva et al. (2007) applied for the first time FRAP assays for extracellular morphogen gradient formation in the fly wing disc, showing that FRAP is a useful method to analyze dynamics of the Dpp gradient. Alternatively, time-lapse imaging of exogenous labeled proteins can be used to measure diffusion coefficient. Thorne et al. (2008) measured a diffusion coefficient of labeled lactoferrin in a brain tissue, which had been injected using a micropipette. Their quantitative analyses have also shown that heparan sulfate has a high capacity and a low affinity for ligands binding, whereas receptors have a low capacity and a high affinity in the extracellular space.

Fluorescence correlation spectroscopy (FCS) is another method to analyze protein diffusion. Using FCS, a diffusion coefficient of FGF8 has been recently measured in the zebrafish embryo (Yu et al. 2009). They reported that most of the ligand freely diffused and the diffusion coefficient of FGF8 measured by FCS was much higher than those of Dpp and Wg reported in the Drosophila embryo measured by FRAP (Kicheva et al. 2007). With the effect of endocytosis on gradient formation, they proposed the “source-sink” mechanism that freely diffusing morphogen and degradation via endocytosis define the shape of gradient. This study opens the interesting question as to what makes the measured diffusion coefficient of FGF8 so different from that of Dpp and Wg (Schier & Needleman 2009). These kinds of quantitative studies are important because they provide parameters, which can be used for model constructions and simulations of theoretical studies. Also, as Ben-Zvi et al. (2008) have proposed the BMP-Chd shuttling model, combinations of experimental and theoretical approaches are a powerful way to study pattern formation. These approaches have just recently emerged, but will soon become more popular and more important in the field of developmental biology.

Conclusions and perspectives

Intracellular signal transduction after the reception of a signaling ligand/morphogen has been intensively studied, thanks to molecular biology and biochemistry. On the other hand, extracellular regulation for morphogens has remained relatively unexplored. To date, various extracellular behaviors of secreted proteins have been reported as we reviewed above. However, we have not fully understood the relationship among “molecular movement”, “affinity to ECM or cell surface”, and “signaling range.” For investigating this attractive field, new methods, technologies, and ways of thinking will be required. Especially, live-imaging using fluorescence, quantitative analysis and physical or mathematical analyses will become more and more important. To understand extracellular events, computational simulation and mathematical modeling will be also necessary. With combination of “wet” experiments and “dry” analyses will solve problems outside the cell.


This work was supported in part by a Grant-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (JSPS). Y. Mii is a JSPS Research Fellow.