Soluble receptors also play a role in Wnt signaling. These are not splice variants or proteolytic endproducts of membrane bound receptors but gene products in their own right. To date, there are Dkks, Norrin, WISE, WIF, Cerberus, and sFRP family members. While soluble "receptors" are often associated with ligand antagonism, soluble proteins associated with the Wnt system are reportedly both facilitatory and inhibitory to Wnt pathway signaling.
The Dkk proteins are four structurally-related, cysteine-rich proteins, the first of which was originally identified as an inducer of head formation. In Xenopus, Dkk-1, together with a mutant BMP-4 receptor, induces the formation of a secondary axis with a complete head. In addition, when Dkk-1 is introduced into Xenopus endomesoderm, it creates an enlarged head. Finally, neutralization of Dkk activity results in no head formation.7,127 It is now known that Dkk-1 is an inhibitor of Wnt-mediated LRP signaling. It does so by forming a "bridge" between LRP and Kremen, which is an endocytosable molecule. Endocytosis of Kremen is accompanied by internalization of Dkk/LRP. This blocks LRP deactivation/destabilization of axin and results in the phosphorylation/degradation of beta-Catenin. 75896 As was noted earlier, Kremen must be membrane-bound as soluble Kremen has no effect. Kremen has no LRP activity in the absence of Dkk. Again, within the context of the Wnt signaling pathway, the Kremen/Dkk removal of LRP interrupts the Wnt-mediated interaction of Frizzled with LRP oligomers, thus blocking the axin arm of the beta-Catenin pathway.7,128,129
Human Dkk-1 is a 35-50 kDa, monomeric, glycosylated polypeptide that contains two CRDs. It is synthesized as a 266 aa precursor that contains a 19 aa signal sequence, a 54 aa, N-terminal CRD, and a 75 aa, C-terminal CRD that resembles colipase. 127,130,131 The two CRDs are distinct and discrete.129 Dkk-1 has multiple single nucleotide polymorphisms, but their significance is unknown.132 Dkk-1 binds both LRP-5 and 6, and Kremen-1 and 2. 7,97,99,128,133 The C-terminal CRD is involved in the binding of Dkk-1 to both LRP-6 and Kremen-2.100134 Notably, the binding of LRP-6 via the C-terminal CRD has the potential to activate LRP and induce downstream beta-Catenin signaling.129 This does not happen, however, and it appears that it is due to the configuration of the N-terminal CRD.7,129 Dkk-1 has been reported to interfere with Wnt-2,131 Wnt-3a,135 Wnt-7a,58 and Wnt-8 signaling.129 Dkk-1 is expressed by mouse E6.5-7.5 (human day 11-20) mesoderm, E8.5 somitic mesoderm (human week 4), and E12.5-E14.5 (human second trimester) interdigital mesenchyme.136 It is also expressed by pancreatic acinar cells137 as well as bone marrow and endometrial stromal cells. 138,139 Dkk-1 is considered a tumor suppressor, possibly because it antagonizes the Wnt signaling pathway140 and/or sensitizes cells to undergo apoptosis.141 As might be expected, Dkk-1 has no effect on the Wnt-related PCP pathway.128 Mouse Dkk-1 has been cloned and found to be 80% aa identical to human Dkk-1 over the entire precursor.127
Human Dkk-2 is synthesized as a 32 kDa, 259 aa precursor that contains a 33 aa signal sequence, a 50 aa N-terminal CRD and a 74 aa, C-terminal CRD. 129,130,135 It has one potential N-linked glycosylation site, but it is doubtful that it is used.135 Dkk-2 is known to undergo proteolytic processing, but the form and extent of the processing are unknown.130 Dkk-2 binds LRP-6 as well as Kremen-1 and 2. 97,129 As with Dkk-1, Dkk-2 binds LRP-6 and Kremen with its C-terminal CRD. 7,100 In the absence of Kremen, Dkk-2 binding to LRP activates the beta-Catenin pathway. However, the levels of LRP must be exceptionally high for Dkk-2 to work. 100,129,134 In the presence of Kremen, Dkk binding to LRP is antagonistic to Wnt signaling. This points out one material difference between Dkk-1 and Dkk-2. In Dkk-1, the N-terminal CRD is, by nature, inhibitory to the C-terminal CRD; in Dkk-2, there is no such antagonism. 100,129 Mouse Dkk-2 has been cloned and found to have 95% aa identity relative to human Dkk-2.136 A rat Dkk-2-like protein has also been documented (GenBank Accession # XP_238256). Its length is 20 aa longer than mouse Dkk-2, so its not clear if it qualifies as a true ortholog to mouse Dkk-2. Dkk-2 has been identified in cardiac mesoderm and interdigital mesenchyme undergoing apoptosis.136 Wnts reportedly blocked by Dkk-2 include Wnt-3a135 and Wnt-8.129
Dkk-3, also known as REIC (reduced expression in immortalized cells), is unique among the Dkks in that its function is undetermined. In human, it is by far the largest of the Dkks at 60 kDa and 336 aa.130,142143 While it contains the two typical CRDs, its C-terminus is rich in acidic amino acids (Asp and Glu) and it demonstrates a polyalanine repeat at its end.130,142 There is no evidence that Dkk-3 binds to either Kremen or LRP-6.97,100 The Dkk-3 gene is unusual in that it contains two distinct promoters of different strength. This perhaps drives tissue-specific expression.143 Whatever its function, it is expressed at a number of sites. These include the zona glomerulosa and medulla of the adrenal gland,144 the myocardium and mesenchyme of the tooth,136 in mesenchyme surrounding new bone,136 in neurons of the cerebral cortex, retina and hippocampus, and lens epithelium.130 Mouse Dkk-3 has been cloned and the molecule shows 83% aa identity to human Dkk-3 over the entire ORF.130,136
Mature human Dkk-4 is a 40 kDa, 206 aa, non-glycosylated polypeptide that contains a 50 aa and a 74 aa CRD.130 There are no potential N-linked glycosylation sites. The molecule is known to be proteolytically processed, creating 31 kDa and 16 kDa fragments. At least one processing site has been identified between two lysines at positions 114 and 115 of the mature molecule.130 Such cleavage does not break up the C-terminal CRD associated with Kremen and LRP binding. Dkk-4 is reported to block Wnt signaling.100,130 It is also known to bind Kremen-2 and presumably LRP.100 Mature mouse Dkk-4 has been identified and found to be 77% identical to human Dkk-4.145
Among the four human Dkk precursors, there is marginal aa identity. Dkk-1 is 50%, 40%, and 45% identical to Dkk-2, 3 and 4, respectively. Dkk-2 is 37% and 46% identical to Dkk-3 and 4, respectively. Dkk-3 is 40% identical to Dkk-4.
Loss of function of the gene encoding Norrin causes Norrie disease, an X-linked disorder characterized by blindness, deafness, and mental retardation. Norrin is a cysteine-knot growth factor that resembles TGF-beta in its tertiary structure.146 The Norrin protein consists of a 133 aa precursor that contains a 39 aa signal sequence (GenBank Accession # Q00604). 146,147,148 The protein is presumed to be secreted, albeit poorly,149,150 and it is described as being matrix-associated,150 suggesting short-range activity. Based on its cysteine-knot motif, it likely exists as a disulfide-linked dimer.146 Mouse Norrin has been identified and found to be 94% identical to the human protein at the aa level. Norrin is found in cerebellar Purkinje cells, retinal ganglion cells, and sensory olfactory epithelium.151
In Norrie disease, incomplete vascularization has been reported to occur in both cochlea and retina.149,152 Thus, it likely plays an important role in development. Recently, Norrin was reported to bind to the CRD of Frizzled-4, and be involved in LRP-5/6 in activating the beta-Catenin pathway.149 It is not clear if it binds LRP, but it does use it to signal. Notably, Norrin acts like a Wnt rather than a soluble receptor.
Very little is known about WISE. Cloned from human, mouse, and Xenopus, it apparently controls eye development and surface ectoderm formation. Like Norrin, it contains a cysteine-knot domain. It is somewhat larger than Norrin at 177 aa for the mature protein. It contains eight cysteines and two potential N-linked glycosylation sites. Mouse to human, there are only four aa differences in the mature segment, resulting in 98% aa identity. Human to Xenopus, there is 80% aa identity.
It is not clear what WISE actually does. It will bind to the first two EGF repeats of LRP-6. It will not bind to Wnt-8 or Frizzled-8. When it is bound to LRP, it blocks the ability of Wnt to recruit LRP into a signaling complex. It has no effect on Wnt interaction with Frizzled, however. Notably, WISE ligation of LRP-6 results in weak beta-Catenin activation. Thus, WISE would seem to be an antagonist of Wnt activity but a protagonist of its own beta-Catenin activity.153
WIF-1 is a Wnt binding protein secreted by a variety of tumor and embryonic tissues. In human, it is synthesized as a 379 aa precursor that contains a 28 aa signal sequence and a 351 aa mature segment. There are two N-linked glycosylation sites, a 139 aa WIF domain that mediates WIF binding to Wnt, and five 30 aa, EGF-like segments.7,154155 The WIF module is related to portions of the extracellular domain of Ryk (related to tyrosine kinase receptor) molecules.156 There are only two cysteines in each WIF module, and it is difficult to correlate any particular aa pattern with function. Nevertheless, the 140 aa stretch must have some "Wnt-a-philic" property, and it suggests that Ryk molecules may be an as yet unappreciated group of Wnt receptors.156 Human to mouse, there is 94% aa identity; human to Xenopus, there is 81% aa identity in the precursor protein.154 Cells known to express WIF-1 include the epithelium of lung, prostate, breast, and bladder,155 endothelial cells,155 primary prostate and breast cancer cells,155 adenoma cell lines,157 and notochord, paraxial presomite mesoderm and neural crest cells.154 WIF-1 interferes with Wnt signaling by blocking Wnt interaction with Frizzled.154 It apparently does so by forming a non-covalent complex with Wnt-8 and Wnt-1.154
Xenopus Cerberus was identified as a molecule whose presence induced multiple heads and suppressed mesoderm formation.158 It is a cysteine-knot protein that binds BMPs, Nodal and Wnt.159 Because of the broad range of binding activities in Xenopus, mouse Cerberus-1 was cloned and tested for similar activity.160 It was found to exist as both a 42 kDa monomer and 85 kDa dimer, that bound BMP and Nodal but not Wnt.161,162 Furthermore, it had very low (approximately 30%) aa identity relative to Xenopus Cerberus and was expressed in non-analogous tissues.160,162 Thus, mouse Cerberus-1 may not be the mammalian ortholog to Xenopus Cerberus and will not be considered part of the Wnt-related group of proteins. 7,161 COCO is a second 25 kDa member of the Cerberus/DAN/Gremlin family of secreted BMP inhibitors.163 Little is known about COCO other than it is not inhibitory to Wnt signaling in mammals.7 As such, its inclusion in any Wnt binding protein group awaits further study.
The sFRPs are a group of five Wnt-binding glycoproteins that resemble the membrane-bound Frizzleds. Their actions are both inhibitory and stimulatory to Wnt activity. In mammals, there are two groups, the first consisting of sFRP-1, 2, and 5, and the second including sFRP-3 and 4. 7,164 All are secreted and derived from unique genes; none are alternate splice forms of the Frizzled family.164 Each sFRP contains an N-terminal CRD. This type of domain has been suggested to be a mediator of ligand-ligand, or ligand-receptor interactions involved in cell fate determinations. Following the CRD, each sFRP terminates in a Netrin-like domain. Netrins are Laminin-related proteins made up of a Laminin B-like domain followed by three EGF-like repeats. At the end of the Netrin molecule lies the Netrin-like motif that exhibits six conserved cysteines and several blocks of hydrophobic residues including a Tyr-Leu-Leu-Leu-Gly (YLLLG) motif.165 The function of the Netrin-like domain is unclear. It exists in TIMPs (tissue inhibitors of metalloproteinases) and may mediate suppression of metalloproteinase activity. In sFRPs, it may participate in matrix-stabilizing activity during tissue-restructuring events.165 Although sFRPs can be both stimulatory and inhibitory to the Wnt pathway, when inhibitory they may simply dictate activity boundaries by limiting the range of Wnt activity.7 Alternatively, sFRPs may simply act as carrier proteins for Wnts, perhaps increasing their range of activity.7
Human sFRP-1, also known as SARP2 (secreted apoptosis-related protein 2) and FrzA (Frizzled in aorta), is a 36 kDa, 286 aa glycoprotein that contains a 110 aa, N-terminal CRD, a 90 aa, Netrin-like, heparin-binding, C-terminal region, and a 26 aa C-terminus.166,167 There are eight intrachain disulfide bonds, five in the CRD and three in the Netrin domain. Notably perhaps, this differs in sFRP-3 and 4, where cysteine shuffling results in a change in Netrin domain disulfide bonding. Here, one disulfide bond almost becomes interdomain in location, subject to a redefinition in boundaries.167 Human sFRP-1 exists in multiple forms. First, in some tumors, mutations in its gene creates a stop codon at precursor position 150. Second, there can be a 1 aa insert at position 13, leading to the addition of an extra alanine residue. Third, there is an alternate splice form that removes the extreme seven C-terminal aa and replaces them with a new stretch of 30 aa. This new addition is suggested to serve as a membrane "anchor", creating a membrane-anchored receptor.168 In addition to gene-directed changes, proteolytic processing creates multiple sFRP-1 isoforms. In particular, the N-terminus is reported to begin at Ser4, Asp14, and Phe23 of the mature segment. The two shortest forms represent up to 17% of all sFRP-1. Further, the C-terminal lysine is known to be selectively cleaved. The significance of this is unknown.167 Mouse sFRP-1 has been cloned and found to be 97% identical to human sFRP-1 over the mature segment.169,170 Mature human sFRP-1 is 98% identical to bovine sFRP-1.171 Cells known to express sFRP-1 include endothelial cells, neurons, myocardium, smooth muscle cells and bronchial epithelial cells,171,172 adrenal cortex and salivary gland epithelium,173 and cervical stratified squamous epithelium.174
Wnts known to bind to sFRP-1 include Wnt-1,175,176 Wnt-2,176,177 Wnt-8,172 Wnt-4, and Wnt-3a.178,179 It apparently does not bind to Wnt-5a.175,177 When sFRP-1 binds to Wnt-1, two molecules of sFRP-1 are suggested to be participate in the binding.176 Notably, the CRD on sFRP-1 that binds Wnt-1 can potentially be used to bind Frizzled-6.176 Whether this qualifies as a competitive phenomenon or even results in receptor activation is unknown. In any event, in binding to Wnts, sFRP-1 would seem to act primarily as an inhibitor of Wnt signaling.175,180 sFRP-1 binding to Wnt-1 is reported to be antagonistic to Wnt activity.176 By sequestering Wnts, sFRP removes the stimulus for beta-Catenin stabilization.175,180 This may or may not translate into an increase in apoptosis. sFRP-1 has been reported to protect cells from apoptosis, but this may be context dependent. 170,178,181,182,183, Other functions associated with sFRP-1 include endothelial cell migration and capillary tube formation,182 myofibroblast recruitment and collagen deposition, and an sFRP-induced decrease in MMP-9 activity.170 Whether this is attributable to the Netrin-like domain of sFRP is not known.
Human sFRP-2, also known as SARP-1 and SDF-5, is synthesized as a 295 aa precursor that contains a 19 or 20 aa signal sequence, a 122 aa CRD, and a 104 aa Netrin-like region.184,185 Mouse and rat sFRP-2 have also been isolated and show 98% and 97% overall aa identity to human sFRP-2, respectively (GenBank Accession # XP227314).186 Mouse and rat sFRP-2 are 99% identical at the aa level. Cells reported to express sFRP-2 include preadipocytes,185 skeletal muscle cells,187 tumorigenic breast epithelium,188 the ependyme of spinal cord and brain, limb bud mesenchyme, and skeletal condensations of face and vertebrae.189 sFRP-2 is believed to bind to Wnt-4189 and Wnt-7a.190
sFRP-2 is often described as being a functional counterpart to sFRP-1. That is, it may act as an anti-apoptotic molecule by an unknown mechanism.179,183188 In chick, it has been proposed that sFRP-2 binds Wnt. Wnt then cannot induce BMP-4, a molecule that otherwise would activate Msx-2, leading to apoptosis.191 sFRP-2 does seem to play a role in skeletal muscle formation. Normally, CD45+ stem cells will form CD45- muscle myoblasts under the influence of Wnt-5a and b. sFRP-2 and 3, potentially derived from new skeletal muscle fibers, block this transition.187
Human sFRP-3, also known as FrzB (Frizzled in bone), FrzB-1, Fritz, and Frezzled, is a 36 kDa, 291 aa glycoprotein that contains a 118 aa CRD, a 121 aa Netrin-like domain, and an 18 aa, Ser/Thr-rich, C-terminal segment. 169,192,193,194 It does not form disulfide-linked dimers. Mouse sFRP-3, however, is reported to form a non-covalently linked homodimer.167,193 As noted above, sFRP-3 has an unusual intrachain disulfide bonding pattern. Five bonds exist in the CRD domain and two in the Netrin-like region. The third typical Netrin disulfide bond links the most distal area of the CRD with the center of the Netrin domain.167 At the C-terminus, there is apparently some proteolytic processing that occurs. Mouse sFRP-3 is 93% identical to human sFRP-3 in the mature segment.192,194 Cells known to express sFRP-3 include fetal chondrocytes of long bones,193 pancreatic islet endocrine cells,195 ectomesenchyme of teeth, postnatal chondrocytes,196 cells of the primitive streak,192 and dorsal root ganglion cells in the peripheral nervous system.194 sFRP-3 potentially binds to both Wnts and Frizzled receptors. Wnts identified as targets for sFRP-3 include Xenopus Wnt-8192 and mammalian Wnt-1.197 sFRP-3 binding to Wnt-1 is antagonistic to Wnt activity. By contrast, sFRP-3 is also reported to bind Wnt-5a, but without any antagonistic effect.197
Human sFRP-4, also known as DDC-4 (differential display coincidence-4), FrzB-2 and FRP-AP, is a 45-50 kDa, 327 aa glycoprotein that exhibits the same domain architecture as other sFRP family members. 198,199,200 After cleavage of a 21 aa signal sequence, there is a 325 aa mature protein with a 113 aa CRD and a 119 aa, Netrin-like domain. As with sFRP-3, the cysteine distribution within the CRD and Netrin-like domain differs from that of sFRP-1.167 In particular, four disulfide bonds are associated with the Netrin-like region. Three mirror the unusual pattern in sFRP-3, with a fourth disulfide added at the end of the Netrin-like domain where an aa extension exists.167 The significance of the Netrin-to-CRD bond is unclear. Rat sFRP-4 has been cloned and found to be 93% identical to human sFRP-4 over the mature segment. 201,202,203 Cells known to express sFRP-4 include prostate epithelium,199 decidual cells of the endometrium,201 granulosa lutein cells,202 myocardial cells,204 and osteoblasts, chondrocytes, fibroblasts, and skeletal muscle.200
The presumed function of sFRP-4 generally parallels that of other sFRPs. It is considered antagonistic to Wnt signaling and induces apoptosis.200,205 However, it also has another function that it shares with FGF-23. It is now reported that sFRP-4 is a phosphaturic factor that promotes phosphate loss through the kidney. It does so in a parathyroid hormone-independent manner by antagonizing renal Wnt signaling.198 Given that circulating levels of sFRP-4 are 30-40 ng/mL, sFRP-4 qualifies as a phosphatonin and may be a key player in osteomalacia.
Little is known about sFRP-5, also known as SARP-3 and FrzB-1 beta. In human, its precursor is 317 aa in length and shows 98% and 95% aa identity to bovine and mouse sFRP-5, respectively. It is produced by anterior visceral endoderm that normally produces Wnt and TGF-beta signal inhibitors.206 Thus, it might be expected to be a Wnt antagonist. Intriguingly, it is suggested to direct the polarity of photoreceptors in the retina.207 If so, it may play a role in the PCP pathway.
There is marginal precursor aa identity between the human sFRPs. SFRP-1 is 38%, 24%, 21%, and 55% identical to sFRP-2, 3, 4, and 5, respectively. sFRP-2 is 25%, 21%, and 39% identical to sFRP-3, 4, and 5, respectively. sFRP-3 is 46% and 24% identical to sFRP-4 and 5, respectively. sFRP-4 is 19% identical to sFRP-5.