Carboxypeptidase Z (CPZ) modulates Wnt signaling and regulates the development of skeletal elements in the chicken

Carboxypeptidase Z (CPZ) is a member of the carboxypeptidase E subfamily of metallocarboxypeptidases (Song and Fricker, 1997). Although these Zn-dependent enzymes have generally been implicated in intra- and extracellular processing of proteins(Skidgel, 1988) (reviewed by Fricker, 1998: Reznik and Fricker, 2001) not much is known about the specific substrates of CPZ. Novikova and Fricker(Novikova and Fricker, 1999)found that CPZ cleaves a C- terminal arginine present in synthetic peptide substrates with maximal catalytic activity at neutral pH(Novikova and Fricker, 1999). This is consistent with an enzymatic function in the extracellular matrix. When expressed in cultured cells CPZ is secreted and associates with the matrix (Novikova et al.,2000).

CPZ harbors a cysteine-rich-domain (CRD) N-terminal to the catalytic domain(Song and Fricker, 1997; Xin et al., 1998). A CRD is characterized by a series of 10 cysteine residues and is found in several proteins including Frizzled, Frizzled related proteins, Smoothened, the receptor tyrosine kinase MuSK and CPZ. In the case of Frizzled and Frizzled related proteins the CRD has been shown to act as a ligand-binding domain for Wnts (Bhanot et al., 1996; Rattner et al., 1997). Wnt proteins are secreted molecules involved in many developmental processes(reviewed by Cadigan and Nusse,1997) including patterning of somites and limb development. The presence of a CRD has implicated CPZ in Wnt signaling during development(Reznik and Fricker, 2001). However, experimental evidence has not yet been provided to support this proposal.

Somites are segmental units of the paraxial mesoderm. They form by epithelialization of mesenchymal cell clusters in the anterior region of the unsegmented paraxial mesoderm. Thereafter, epithelial somites are regionalized into a ventral compartment, the sclerotome, from which the axial skeleton forms, and a dorsal compartment, the dermomyotome giving rise to dermis and skeletal muscle (Keynes and Stern,1988). Somite patterning is controlled by signals from adjacent tissues including the notochord, neural tube, surface ectoderm and lateral plate mesoderm (Brand-Saberi et al.,1993; Pourquie et al.,1993; Fan and Tessier-Lavigne,1994; Kuratani et al.,1994; Spence et al.,1996). Several members of the Wnt family are expressed in these tissues and have been shown to induce the expression of dermamyotomal genes such as the paired-box transcription factor Pax3(Fan et al., 1997; Cossu and Borello, 1999). Sonic hedgehog is another major axial signal that is responsible for induction and differentiation of the sclerotome(Marcelle et al., 1999). Signaling activity of these secreted proteins may be regulated by proteolytic processing.

The present study uses a combination of strategies to unravel the developmental function of CPZ. In situ hybridization in chick embryos revealed regionalized expression of CPZ in somites, sclerotome, paraxial head mesoderm and the apical ectodermal ridge. Retrovirus-mediated ectopic CPZ expression in the chick was used to investigate the role of CPZ during embryogenesis. Overexpression in the somites resulted in upregulation of Pax3 in the hypaxial dermomyotome, in a downregulation of Pax1 in cells fated to form the scapula and in a partial loss of the scapula and ribs. CPZ increased Wnt4-mediated induction of the homeobox gene Cdx1 in vitro, and immunoprecipitation experiments showed that the CRD of CPZ can bind to Wnt4. Collectively, these experiments suggest that CPZ has a role in Wnt signaling.

A 750 bp cDNA fragment corresponding to chicken CPZ was used to screen a chick cDNA library. This 750 bp fragment had originally been isolated in a screen for retinoic acid-induced genes (Swindell and Eichele, unpublished data). Ten cDNAs were isolated, all of which lacked the predicted 5′ end of the coding region. To isolate the 5′ end, SMART 5′ RACE was carried out (Clontech). The murine CPZ ORF was isolated by RT-PCR starting with the amplification of a 300 bp fragment using the primer pair 5′CCCAGTACTGTGCTC(C/T)GAGT3′;5′CCGAATTTCTCTGTCACC(A/T)CAC3′.

The rest of the ORF was obtained by 5′ and 3′ RACE-PCR using the SMART RACE cDNA Amplification Kit (Clontech, USA). GenBank accession numbers: chicken CPZ AF351205; murine CPZ AF356844.

Whole-mount in situ hybridization (WMISH) and subsequent sectioning of embryos were carried out as described previously(Albrecht et al., 1997). The entire cCPZ coding region was used as template for riboprobe synthesis. In situ hybridization analysis on sections was performed as described previously (Swindell et al.,2001). For Pax1, Pax3, myf5, myogenin and MyoDfull-length cDNAs were used as templates for riboprobe production.

A single nucleotide change (G1405 to C1405) was inserted into cCPZ using the QuikChange site directed mutagenesis protocol (Stratagene). Primers:5′GCTTTGAAGTTACTGTGCAGGTAGGATGTG3′,5′CACATCCTACCTGCACAGTAACTTCAAAGC3′. This mutagenesis resulted in a single amino acid change (Glu469 to Gln469). The corresponding mutation was also inserted into the murine CPZ with the following primers:5′GCTTTGAGATCACCGTGCAACTGGGCTGTGTGAAGTTC3′,5′GAACTTCACACAGCCCAGTTGCACGGTGATCTCAAAGC3′.

This nucleotide mutation resulted in the single amino acid change Glu477 to Gln477.

Full-length chicken CPZ and mutant chicken CPZ(cCPZ^E469Q) were cloned into the RCAS-BPA vector. The virus was produced and concentrated as described by Logan and Tabin(Logan and Tabin, 1998). The virus was injected into the segmental plate of HH stage 10 embryos at the level of the presumptive wings (Chaube,1959). Injection of the virus to other sites had no effects. Embryos were then collected for WMISH or Alcian Blue staining at the times noted.

Day-10 chick embryos were collected and fixed in 5% TCA. Embryos were then stained with 0.1% Alcian Blue, unspecifically bound dye was washed off with 1%HCl/70% ethanol, followed by dehydration in 100% ethanol and clearing in methyl salicylate in order to visualize the skeleton.

HEK-293 cells were transfected by lipofection (Effectene, Qiagen) using linearized pcDNA3.1/myc-HisA (Invitrogen) containing the full-length coding region of the murine CPZ cDNA either in its native form or carrying a glutamate to glutamine mutation. Cells were split 24 hours after transfection and grown in 6-well plates under selective conditions (DMEM, 10% FCS, and 1 mg/ml G418). Clones were tested for native and mutant CPZ expression by western blot analysis using mouse anti-myc antibody (Invitrogen). CPZ-containing extracellular matrix (ECM) was prepared as described previously(Novikova et al., 2000).

Analyses are based on multiple experiments using different CPZ and CPZ^E477Q cell lines. CPZ-expressing HEK-293 cells or wild type HEK-293 cells were seeded into 6 cm tissue culture dishes and grown for 1 day. These cultures were always done in duplicate. To generate plates coated with normal ECM or ECM spiked with CPZ, normal HEK-293 cells or CPZ-producing HEK-293 cells were detached with 1 mM EDTA in PBS. Into these conditioned plates were placed either untransfected HEK-293 cells or CPZ-expressing HEK-293 cells and NIH-3T3 fibroblasts stably transfected with different Wnt cDNAs (Kispert et al.,1998). Equal numbers of HEK-293 and NIH-3T3 cells were used to give a total cell number of 3×10⁶ cells per plate. HEK-293 and NIH-3T3 cells were cultured for ∼4 hours after which time ES cells were seeded on top of these cells as previously described(Lickert et al., 2000). Depending on the Wnt-expressing cell line used and the passage number of the ES cells, the co-cultures were grown for between 6 and 12 hours. Thereafter,RNA was isolated with RNAzol (WAK Chemie) and cDNA was generated using Superscript II Reverse Transcriptase (Invitrogen). Cdx1 expression levels were detected with quantitative RTPCR as described(Fruman et al., 2002) with the housekeeping gene elongation factor 1 alpha (EF1α) to standardize Cdx1 expression levels. The following primer pairs were used: EF1α-forward 5′GTCCCCAGGACACAGAGACTTCA3′, EF1α-reverse 5′AATTCACCAACACCAGCAGCAA3′, Cdx1-forward 5′TACAGCCGGTACATCACTAT CCG3′, Cdx1-reverse 5′CTGTTTCTTCTTGTTTACTTTGCGC3′. Co-cultures of lacZ-NIH-3T3, untransfected HEK-293 cells and ES cells display basal levels of Cdx1 expression(Lickert et al., 2000). Hence the increase of Cdx1 expression in the presence of inducers (Wnts,CPZ) was calculated as the ratio of expression in the presence of inducers and basal level of expression resulting in a `fold-induction'. The co-culture experiment were repeated multiple times and data shown in Fig. 6B are typical.

2.4×10⁶ HEK-293 cells were plated on 10 cm dishes. 24 hours later they were transfected with 4 μg of a plasmid mixture using Effectene in a standard reaction set-up. This mixture contained 2.4 μg of pcDNA3.1-Wnt4-HA (Lescher et al.,1998) and 1.6 μg of either the positive control pcDNA3.1-sFRP2-myc (Lescher et al.,1998) or one of the CPZ constructs. 48 hours after transfection,cells were washed twice with PBS at 37°C and harvested in 350 μl of lysis buffer. After cell lysis the cotransfected proteins were precipitated with a monoclonal mouse anti-HA antibody (Babco) and protein-G agarose (Roche)using standard protocols. The coprecipitated proteins were detected with a monoclonal mouse anti-myc antibody (Invitrogen).

A 750 bp chicken cDNA fragment with homology to human and rat CPZwas used to screen a Hamburger Hamilton (HH) stage 14-17 chick cDNA library(Hamburger and Hamilton,1951). Ten putative chicken CPZ cDNAs (cCPZ)were isolated, but all lacked the 5′ end of the protein coding region. The missing 5′ end of the cCPZ cDNA was obtained using a 5′ RACE-PCR procedure. The assembled cDNA encodes a protein with 77%sequence similarity (identity 64%) to human and rat CPZ(Fig. 1). Similar to its putative mammalian orthologs, cCPZ harbors an N-terminal signal peptide, a cysteine-rich-domain (CRD) and a carboxypeptidase domain(Fig. 1). The CRD and carboxypeptidase domain of cCPZ show similarities of 84% and 86% (identity:69% and 73%), respectively, with the human ortholog. Like the rat and human CPZs, the CRD of the chicken enzyme contains 10 conserved cysteine residues. Amino acid residues important for substrate binding and catalytic activity(His245, Glu248, Arg320, His377, Glu469) are also fully conserved. Amino acid sequence conservation to CRD domains of other proteins amounts to 25% and 32%,supporting the notion that the isolated cDNA represents the chicken ortholog of mammalian CPZ.

In situ hybridization analysis was used to determine the expression pattern of cCPZ during early developmental stages of the chick(Fig. 2). Weak cCPZexpression was first observed at HH stage 7 in the developing somites (not shown). In subsequent stages, when progressively more somites form and differentiate, cCPZ expression is maintained in these structures(Fig. 2. A-D,F,H). Transverse sections demonstrate the presence of cCPZ transcripts throughout the entire epithelial somite (Fig. 2D,H). As somites differentiate into sclerotome and dermomyotome cCPZ expression becomes restricted to the sclerotome(Fig. 2F). cCPZtranscripts are present throughout the sclerotome unlike, for example, Pax1 that is expressed in the ventromedial portion of the sclerotome(Müller et al., 1996). By HH stage 22, when somite formation ceases, all sclerotomes express cCPZ. With condensation of the sclerotomes CPZ expression is lost in a rostral to caudal progression (not shown). Additionally, cCPZ is expressed in the unsegmented paraxial head mesoderm surrounding the notochord(Fig. 2E). cCPZ was also detected in the apical ectodermal ridge (AER, Fig. 2C,G), a transient signaling tissue mediating limb outgrowth.

It has been shown for several genes including Pax1 that sclerotomal expression depends on signals released by the notochord and/or neural tube (Fan and Tessier-Lavigne,1994; Marcelle et al.,1999); this is, however, not the case for cCPZ. When presomitic mesoderm not expressing cCPZ was separated from axial structures and the operated embryos were harvested 14 h later, all somites,even those deprived of axial signaling still expressed cCPZ(Fig. 2I). We conclude that cCPZ expression is not regulated by signals emanating from the notochord or neural tube but by factors intrinsic to the somitic mesoderm and/or derived from the surface ectoderm.

The expression of cCPZ in somites suggests a role for this enzyme in the development of the axial skeleton. In order to test this we reasoned that ectopic expression of CPZ in the chick embryo might specifically affect the development of these structures. RCAS virus containing the cCPZopen reading frame was injected into the segmental plate of HH stage 10 chick embryos in the presumptive wing region. Embryos were harvested 48-60 hours after injection. In most cases they showed a high level of cCPZexpression across 2-4 somites and in the lateral plate at the level of the wing bud (Fig. 3A). Expression of cCPZ was not seen in the somites on the non-injected side of the embryo (Fig. 3B,C). Transverse sections through whole mounts showed that virally mediated cCPZexpression occurred in epaxial, central and hypaxial dermomyotome but not in the sclerotome (Fig. 3C). Such targeted expression to dermomyotome by RCAS virus injected into the segmental plate has also been reported for sonic hedgehog(Johnson et al., 1994). The ectopic expression of cCPZ in the dermomyotome prompted us to search for changes of expression of dermamyotomal marker genes. Expression of myoD, myf5 or myogenin was not changed (n=10 for each gene, data not shown), but the expression of Pax3 was markedly affected. At the wing level, Pax3 is normally expressed in the epaxial portion of the dermomyotome (Fig. 3F). Overexpression of cCPZ in the dermomyotome resulted in ectopic expression of Pax3 in the hypaxial dermomyotome(Fig. 3D,F; 16 out of 31 injected embryos). Of note, overexpression of cCPZ in dermomyotome did not alter either the normal expression of Pax1 in the sclerotome or, at this stage, induce Pax1 in the dermomyotome (n=10,data not shown).

Metallocarboxypeptidases are characterized by a conserved glutamic acid residue that is required for enzymatic activity. Substitution of this Glu residue with a Gln abolishes the activity of CPE, but does not affect the binding of synthetic peptide substrates(Qian et al., 1999). If cCPZ functions as an enzyme, one would expect that loss of its catalytic activity should manifest itself by an inability to induce Pax3 expression. We therefore mutated the corresponding residue Glu⁴⁶⁹ of cCPZ to a Gln, and produced an RCAS virus capable of expressing this mutant form of CPZ. Virus overexpressing CPZ^E469Q was injected in the same site as the normal CPZ virus. 48-60 hours after injection, we observed levels of mutant CPZ expression similar to those described for the wild-type virus (not shown). However, we did not detect ectopic expression of Pax3 in hypaxial dermomyotome(n=15, data not shown).

The appearance of Pax3 mRNA in the hypaxial dermomyotome may reflect the possibility that ectopic CPZ evokes a change in the fate of hypaxial mesodermal cells. In turn, this may affect the development of the scapula blade known to derive from this tissue(Huang et al., 2000). In the chicken, the scapula consists of a head (acromium) and a blade that are connected by the `neck' of the scapula(Baumel and Witmer, 1993; Ede, 1964). When cCPZ-injected embryos were examined at day 10, 55% (12 out of 22 injected embryos) showed a truncation of the scapular blade(Fig. 4A,B). We also noticed that ectopic expression of cCPZ causes truncation or loss of rostralmost ribs(10 out of 22 injected embryos; Fig. 4A,D). Injection of a retrovirus encoding alkaline phosphatase as a control had no effect on morphogenesis of the scapula or ribs(n=22, not shown). When embryos injected with CPZ^E469Q were allowed to develop to day 10, we observed a much lower frequency and severity of skeletal malformations (3 out of 22 injected embryos). One embryo had a partial loss of the distalmost part of the blade of the scapula and the other two embryos exhibited a slight outward bending of the scapula (not shown).

At HH stage 26, Pax1 is expressed in a stripe of mesenchymal cells located lateral to somites 17-20. These Pax1-positive cells derive from hypaxial dermomyotome and are thought to give rise to the blade of the scapula (Huang et al., 2000). Because in cCPZ-treated embryos the hypaxial dermomyotome expresses Pax3 and the blade of the scapula is missing, one would predict that the Pax1-positive stripe may also be affected. Indeed, we noted absence of the stripe of Pax1 expression in 50% of injected embryos(n=12) (Fig. 4C,D), a frequency similar to that seen for skeletal defects.

The above experiments demonstrate striking effects of ectopic CPZ expression on Pax3 expression in hypaxial dermomyotome. Pax3had previously been shown to be regulated by Wnt signals(Fan et al., 1997). This finding and the presence of a CRD in CPZ, which in other proteins was shown to bind Wnt ligands (see Introduction), prompted us to hypothesize that CPZ plays a role in Wnt signaling. To assess whether CPZ can influence Wnt signaling, we adapted a paracrine in vitro Wnt assay(Lickert et al., 2000). In this assay transfected Wnt secreting NIH-3T3 feeder cells were cocultured with murine ES cells. Wnts secreted by the feeder cells induce the homeobox gene Cdx1 in ES cells. In order to test whether CPZ modulates Wnt signaling we added HEK-293 cells stably expressing murine CPZ to the culture and measured Cdx1 induction. All CPZ HEK-293 cell lines generated exhibited similar levels of CPZ protein expression (see Fig. 5A for five representative cell lines). Immunolocalization studies of CPZ-producing HEK-293 cells showed that CPZ localizes to the endoplasmic reticulum (not shown). Cell extraction further demonstrated that CPZ is present in the extracellular matrix(Fig. 5A).

Next, CPZ HEK-293 or control HEK-293 cells were grown in culture dishes for a day. Cells were washed off with 1 mM EDTA in PBS leaving ECM on the plates. This created two types of conditioned plates, one containing ECM with attached CPZ (Fig. 5A), and a second type with unmodified ECM. The former plates were seeded with Wnt4-expressing NIH-3T3 cells, CPZ HEK-293 cells and ES cells, while the latter plates were seeded with Wnt4-expressing NIH-3T3 cells, HEK-293 cells and ES cells. After 8 hours of co-culture Cdx1 induction was quantified by quantitative PCR. We found that Cdx1 induction was increased by as much as 50 percent in the presence of CPZ-containing ECM and CPZ producing HEK-293 cells(Fig. 5B, compare bars 4 and 5). We also seeded CPZ HEK-293 conditioned plates with Wnt-4 producing NIH-3T3 cells and ES cells. This still resulted in Cdx1 induction(Fig. 5B, bars 4 and 7). These results indicate that CPZ protein present in the extracellular matrix is sufficient to enhance Wnt4 signaling. Cells expressing CPZ bearing an active site glutamate to glutamine substitution did not potentiate Cdx1induction (Fig. 5B, bars 4 and 6) suggested that the catalytic activity of CPZ is required for this effect. Cdx1 induction in ES cells strictly depends on Wnt4. No Cdx1induction was observed in ES cells cocultured with CPZ HEK-293 cells and NIH-3T3 cells containing lacZ instead of Wnt4(Fig. 5B, compare bars 1 and 2). Similarly, ECM generated by CPZ HEK-293 cells did not induce Cdx1in this system (Fig. 5B,compare bars 1 and 3). Wnt1 and Wnt3a were also tested in this assay but Cdx1 induction was not potentiated by CPZ (data not shown).

The potentiation of Wnt4 signaling by CPZ raised the question of whether CPZ can directly bind to Wnt4 and if so, which part of the protein may mediate this interaction. To answer this questions we performed co-immunoprecipitation experiments from cells co-expressing HA-tagged Wnt4^HA(Lescher et al., 1998) and myc-tagged CPZ (Fig. 6A,CPZ^myc). pcDNA3.1 expression vectors containing the appropriate ORFs were cotransfected into HEK-293 cells and complexes were precipitated using anti-HA antibody. Precipitated proteins were separated by PAGE, western blotted and an antibody directed against the myc-epitope was used to detect CPZ. Analyses of cell lysates demonstrated that CPZ^myc of the correct size was expressed (Fig. 6B, lane 2). Importantly, analysis of the co-precipitate demonstrated the presence of a CPZ^myc/Wnt4^HA complex(Fig. 6C, lane 2). A myc-tagged sFRP-2 (Fig. 6A) served as a positive control [specific binding of sFRP-2 to Wnt4 had been described previously (Lescher et al.,1998)]. As shown in Fig. 6B,C sFRP-2 was produced (lane 1) and co-precipitated with Wnt4(lane1). CPZ^myc/E477Q, bearing the above described inactivating point mutation, co-precipitated with Wnt4^HA suggesting that the catalytic activity of CPZ is not required for Wnt4 binding(Fig. 6B,C, lanes 3). However,the CRD of CPZ is required for the interaction with Wnt4. In the absence of this domain, as in the constructs CPZ^ΔCRD/myc(Fig. 6A) and CPZ^{ΔCRD/myc/E477Q}, CPZ could not be co-precipitated with Wnt4^HA (Fig. 6C,lanes 4 and 5) although both mutant proteins were present in the lysate(Fig. 6B, lanes 4 and 5). In contrast, Wnt4^HA co-precipitated with a CPZ lacking the carboxypeptidase domain (CPZ^ΔCPD/myc) (see Fig. 6B,C, lanes 6). None of the co-immunoprecipitated proteins was unspecifically bound to Protein-G agarose or was unspecifically precipitated by the anti-HA antibody alone (data not shown).

Taken together these data suggest that Wnt4 and CPZ can interact, that this interaction is mediated by the CRD of CPZ and that catalytic activity is not required for binding per se (although such activity is obviously required for CPZ function, see above). Since co-precipitation experiments were performed from cell lysates we cannot rule out the possibility that CPZ-Wnt4 complexes contain additional factors mediating CPZWnt4 binding.

We report a series of experimental studies that investigated the developmental and biochemical function of carboxypeptidase Z (CPZ). We show that cCPZ expression is restricted to epithelial somites, sclerotome,paraxial head mesoderm and the AER of the limb bud. Ectopic expression of cCPZ in the dermomyotome at the wing level has dramatic effects on the expression of Pax1 and Pax3, and on the morphogenesis of scapula and rostral ribs.

It has previously been shown that hypaxial dermomyotome-derived Pax1-expressing cells give rise to the scapular blade(Huang et al., 2000). In embryos expressing cCPZ in the dermomyotome, ectopic Pax3expression was induced in presumptive scapula cells of the hypaxial dermomyotome, concomitant with a loss of Pax1 expression in the descendants of these cells. These changes in the gene expression program are likely to underlie the severe dysmorphogenesis of the blade of the scapula and of the distal portion of the ribs. Pax3 expression in paraxial mesoderm is known to be regulated by Wnt signals(Fan et al., 1997). Such a Wnt signal may be affected by ectopically expressed cCPZ. Support for this possibility comes from our observations that CPZ enhances Wnt4 signaling and that it binds to Wnt4 via its cysteine-rich domain (CRD). Except for binding to Wnt4, all effects described in this study require CPZ to be catalytically active.

It has recently been suggested that CPZ processes Wnt signals (see Reznik and Fricker, 2001). This view is based on the fact that CPZ harbors a CRD domain. Such a domain is found in several other proteins (including Frizzled and sFRPs) known to directly interact with Wnts. The present study provides three lines of experimental evidence for a distinct role of CPZ in Wnt signaling. First, it is shown that CPZ potentiates the activation of a Wnt reporter gene, Cdx1, in an in vitro assay. In addition, ectopic expression of CPZ in dermomyotome induces Pax3, a Wnt response gene(Fan et al., 1997; Lee et al., 2000). Finally,evidence is provided that CPZ binds to Wnt4 and that this interaction occurs through the CRD of CPZ.

In the following section, we discuss mechanisms by which CPZ could participate in Wnt signaling. Wnt molecules are locally released from cells,diffuse into the extracellular space, bind to Frizzled transmembrane receptors and through a β-catenin or a `non-canonical' pathway regulate gene expression (for reviews, see Wodarz and Nusse, 1998; Borycki and Emerson, 2000). In the extracellular space, Wnts also bind to soluble Frizzled related proteins (sFRPs) that sequester Wnt from binding to their cognate receptors (for reviews, see Wodarz and Nusse, 1998; Bejsovec, 2000; Pandur et al., 2002; Lee et al., 2000). Where in this complex inter- and intracellular signaling process could CPZ play a role?Three obvious, not necessarily mutually exclusive mechanisms of action can be envisaged. (1) CPZ could degrade components of the extracellular matrix,thereby enhancing the availability of Wnt molecules [for an example, see Dhoot et al., 2001)]. (2) CPZ could proteolytically process sFRPs and thereby affect their affinity for Wnts. A precedent for such a mechanism is provided by the BMPs that bind to chordin which is proteolytically cleaved by the protease BMP-1/tolloid,allowing BMPs to bind to their cognate receptor (for review, see Nakayama et al., 2000). (3)Wnts and CPZ could directly interact. Our experiments provide evidence for the latter mechanism, as we show that Wnt4 and CPZ can be co-precipitated from mammalian cell extracts. Of course, we cannot rule out that other components are required for a Wnt4-CPZ-interaction to occur. A direct in vitro binding study could address this issue, but in vitro production or purification of Wnt proteins has remained elusive. Binding of Wnt4 to CPZ may result in quenching of Wnt signaling. We deem this less likely, because CPZ does not abolish, but enhances Wnt4 signaling in our in vitro assay. An additional argument against a quenching mechanism arises from our observation, that enzymatically inactive CPZ fails in our functional studies. Quenching would not depend on such a catalytic activity. Hence we favor a mechanism in which binding of Wnt4 to CPZ represents a first step followed by proteolytic processing of the Wnt4 ligand. Song and Fricker (Song and Fricker,1997) have shown that CPZ effectively hydrolyses peptides carrying a C-terminal arginine. Intriguingly, among the three Wnt molecules tested in our in vitro assay (Wnt1, Wnt3a and Wnt4) only Wnt4 has a C-terminal arginine and only the combination of Wnt4 and CPZ potentiates Cdx1 induction. Wnt8C from chicken is the only other Wnt carrying a C-terminal arginine residue Hume and Dood,1993).

Most of the dermomyotome should give rise to muscles with exception of the hypaxial dermomyotome at somite levels 17 to 24 from which the blade of the scapula arises (Huang et al.,2000). It has been proposed that downregulation of Pax3expression in the hypaxial dermomyotome prevents this tissue from committing to a myogenic fate (Huang et al.,2000). Instead, signals from ectoderm would trigger a chondrogenic fate in the hypaxial dermomyotome, as shown by the finding that cells descending from this region switch on Pax1(Huang et al., 2000). In embryos expressing CPZ throughout the dermomyotome, Pax3 is induced in the hypaxial portion of the dermomyotome suggesting that these cells do not acquire a chondrogenic fate. It has been shown that Wnt molecules, including Wnt4, induce Pax3 in the presomitic mesoderm(Fan et al., 1997; Lee et al., 2000). Ectopic expression of CPZ in the dermomyotome may thus lead to greater activation of a Wnt signal emanating from surrounding tissues, most probably the ectoderm. This subsequently causes a change in the developmental fate of this tissue and thus prevents morphogenesis of the blade of the scapula. At the molecular level this is reflected by ectopic activation of Pax3 in earlier stages (HH 21-22) followed by the downregulation of Pax1 in cells originating from the hypaxial dermomyotome and fated to form the scapula blade. It remains to be seen whether these changes in Pax gene expression are indeed causing the observed morphological defects. Of note, Pax1^-/- mice lack part of the spine of the scapula, a structure homologous to the avian scapular blade(Wilm et al., 1998).

Ectopic CPZ expression also results in the partial (i.e. distal) or complete loss of the rostralmost ribs. Ribs are thought to either exclusively derive from the sclerotome (Huang et al.,2000b; Evans,2003) or from sclerotome and dermomyotome(Kato and Aoyama, 1998). Sclerotome normally expresses CPZ and it is thus unlikely that ectopic CPZ per se causes the observed rib defects. Since ectopic CPZ is expressed in dermomyotome, one possibility is that this leads to an excessive or `ectopic' activation of Wnt signaling. In turn, this may influence the differentiation of the lateral part of the sclerotome, resulting in partial or complete absence of ribs. If one assumes that part of the ribs derive from the dermomyotome (Kato and Aoyama, 1998), the rib defects can be readily explained along the lines discussed for the blade of the scapula. Ectopic CPZ may upregulate Pax3 and thereby abolish rib chondrogenesis.

So far the developmental function of CPZ has been discussed in the context of overexpression in the dermomyotome. However, CPZ transcripts are normally expressed in the epithelial somites and in the sclerotome. Because CPZ protein is secreted, it may also act in tissues adjacent to the sclerotome. In fact, our in vitro studies find CPZ in the extracellular matrix of CPZ-producing HEK-293 cells. Since Wnt molecules are also secreted factors acting over a certain distance (Fan et al., 1997) CPZ could encounter Wnts that are released from surrounding tissues, e.g. the surface ectoderm (see below) or it could interact with Wnt molecules that are directly expressed in the somites, such as Wnt5a and Wnt11(Cauthen et al., 2001). If one assumes that CPZ functions in tissue adjacent to the sclerotome, such as in the dermomyotome or mesenchymal cells in this region, CPZ could interact with a number of Wnt molecules. Wnt4 is expressed in the dorsal neural tube of the mouse and chicken embryo as well as the surface ectoderm of the mouse embryo (Parr et al.,1993; Cauthen et al.,2001). In addition to Wnt4, several other Wnts are expressed in, and hence released by, the surface ectoderm, the dorsal neural tube [e.g. Wnt6 and Wnt7a, (see Parr et al., 1993; Cauthen et al., 2001; Tajbakhsh et al., 1998)], and in the dermomyotome [e.g. Wnt11 (see Tanda, 1995; Marcelle et al., 1997)].

A future challenge remains: the identification of endogenous substrates of CPZ. Although our data suggest that Wnt4 may represent such a substrate,definitive biochemical proof is still lacking. In addition, the relevance of proteolytic processing for binding of Wnts to their cognate receptors remains to be explored.

This work was supported by the Max Planck Society, the SFB 271 and the NIH grant HD20209 to G. Eichele. We thank M. Yaylaoglu for help in performing section in situ hybridization analyses, R. Kemler and H. Lickert for advice in the Cdx1 in vitro induction assay, and to H. Oster for critically reading this manuscript. A. Vortkamp and colleagues provided us with injection needles for the viral overexpression experiments and M. Leitges and U. Braun assisted in ES cell culture.