Dickkopf (Dkk) genes comprise an evolutionary conserved small gene family of four members (Dkk1-4) and a unique Dkk3-related gene, Dkkl1 (soggy). They encode secreted proteins that typically antagonize Wnt/β-catenin signaling, by inhibiting the Wnt coreceptors Lrp5 and 6. Additionally, Dkks are high affinity ligands for the transmembrane proteins Kremen1 and 2, which also modulate Wnt signaling. Dkks play an important role in vertebrate development, where they locally inhibit Wnt regulated processes such as antero–posterior axial patterning, limb development, somitogenesis and eye formation. In the adult, Dkks are implicated in bone formation and bone disease, cancer and Alzheimer's disease.
Wnts are an evolutionary conserved family of growth factors, whose signaling is involved in numerous processes in development, in the adult and in disease (Cadigan and Nusse, 1997; Moon et al., 2004; Cadigan and Liu, 2006). Wnts are subject to negative and positive regulation by a wide range on effectors that act either intracellularly to modulate components of the signal transduction machinery, or extracellularly to modulate ligand receptor interactions. Five families of extracellular Wnt antagonists are currently known (Kawano and Kypta, 2003), the secreted frizzled-related protein (sFRP), Wnt inhibitory factor 1 (Wif1), Xenopus Cerberus, Wise and the Dickkopf (Dkk) family of secreted proteins, and the latter will be reviewed here.
Physical properties and structure of Dkk proteins
The dickkopf (Dkk) family encodes secreted proteins and consists of four main members in vertebrates (Dkk1,2,3,4). Dkks are glycoproteins of 255–350 amino acids (aa) and a calculated MW between 24 and 29 kDa for Dkk1, 2 and -4 and 38 kDa for Dkk3. On sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) analysis Xenopus Dkk1 secreted from oocytes and human Dkk1 produced in SK-LMS-1 and HEK 293T cells, have an apparent MW of 40, 35, and around 45 kDa, respectively (Glinka et al., 1998; Fedi et al., 1999; Krupnik et al., 1999). The protein is probably N-glycosylated at a single site at the N-terminus (Fedi et al., 1999; Krupnik et al., 1999). This is supported by mass spectroscopy analysis where Xenopus Dkk1 immunoisolated from frog liver has a MW of 29237+/−34 Da, and is thus 373 Da larger than predicted for the full-length protein (C Kiecker, W-D Lehmann and CN, unpublished research). As the 30 aa signal sequence is likely to be cleaved off, Dkk1 from frog liver therefore contains an extra mass of 3 around kDa due to posttranslational modifications.
For Dkk2, 3 and -4 significant amounts of smaller MW proteins are observed in SDS–PAGE, indicative of proteolytic processing, which probably occurs at the various dibasic cleavage sites that Dkk proteins harbor and which may be substrates for Furin type proteases. Interestingly, N-terminal processing is likely to liberate a fragment containing the C-terminal colipase domain, which in isolation has properties differing from the full-length protein (see below). The possibility that Dkk function is modulated by proteolytic processing in vivo is still unexplored.
Dkks contain a signal sequence and share two conserved cysteine-rich domains, each of which displays a characteristic spacing of cysteines and other conserved aa, which give the family its uniqueness (Figure 1). Outside of these, Dkks show little sequence similarity (Glinka et al., 1998). The N-terminal cysteine-rich domain, Dkk_N (formerly called Cys1) is unique to the Dkks while the C-terminal cysteine-rich domain (formerly called Cys2) has a pattern of 10 cysteines related to that of the colipase fold (Aravind and Koonin, 1998). Protein sequence and structural analysis suggests that Dkks and colipases have the same disulfide-bonding pattern and a similar fold. Colipases are necessary for lipid hydrolysis by pancreatic lipases and interact with lipid micelles (van Tilbeurgh et al., 1999). The colipase fold is found in a wide range of proteins other than colipases, including, for example, toxins from funnel web spider (Szeto et al., 2000), black mamba (Boisbouvier et al., 1998) and frog skin (Bv8) as well as in protease inhibitors (Kaser et al., 2003). The structure of the colipase fold is solved and it consists of short β-strands connected by loops and stabilized by disulfide bonds, resulting in finger-like structures, that may serve as interactive surfaces (van Tilbeurgh et al., 1999).
All vertebrate Dkks contain both the Dkk_N and colipase fold domains, separated by a non-conserved linker region that spans 50–55 aa in Dkk1, Dkk2 and Dkk4, but only 12 aa in Dkk3. In some invertebrate Dkks such as the Hydra ortholog Hydkk1/2/4, only the colipase fold occurs (Guder et al., 2006) and structure-function analysis of vertebrate Dkks has shown that this domain alone is sufficient for biological activity, that is, Wnt inhibition (Brott and Sokol, 2002; Li et al., 2002; Mao and Niehrs, 2003). The colipase fold also shows the greatest sequence conservation among the Dkks.
Of note, Bv8, a small colipase fold protein from frog skin, which is able to potently stimulate intestinal contraction in vitro, has no effect in standard Wnt signaling assays (G Kreil, A Glinka and C Niehrs, unpublished). Another gene family with C-terminal colipase fold is the prokineticins, encoding cysteine-rich secreted proteins that regulate diverse biological processes, including gastrointestinal motility, angiogenesis, and circadian rhythms. Prokineticins are inactive in the Bv8 intestinal contraction assay and substitution of their colipase fold by the homologous region from colipase or Dkk1 inactivates their ability to mobilize Ca2+ (Bullock et al., 2004). Taken together these results suggest that Dkks share weak similarity to the colipase fold, a protein scaffold that can mediate unrelated molecular interactions in different protein families.
Evolution of Dkk genes
Dkk1, the founding member of the family, was originally identified as embryonic head inducer and Wnt antagonist in Xenopus (Glinka et al., 1998). Since then Dkks were identified in other vertebrates including humans as well as in invertebrates such as Dictyostelium, cnidarians, urochordates and ascidians (Hino et al., 2003; Fedders et al., 2004; Guder et al., 2006), but not in Drosophila and Caenorhabditis elegans. Thus, Dkks are an evolutionarily ancient gene family that was already present in the last common ancestor of cnidarians and bilaterians and which was probably secondarily lost during evolution in protostomes. A distant Dkk family member, soggy (sgy; also called Dickkopf-like protein 1 DKKL1), was described in vertebrates (Krupnik et al., 1999), which shows unique homology to Dkk3.
Unlike in vertebrates, in Hydra Hydkk1/2/4 only the colipase domain occurs, and this protein is biologically active in Wnt inhibition (Guder et al., 2006). Hydra Dkk3 harbors both conserved domains (Fedders et al., 2004).
By a number of criteria Dkk3 appears to be a divergent member of the Dkk family. (1) By DNA sequence similarity, vertebrate Dkk1, 2 and -4 are more related to each other than they are to Dkk3 (Glinka et al., 1998). (2) Soggy, the distant member of the Dkk family, shares sequence similarity with Dkk3 but not with other Dkks (Krupnik et al., 1999). The similarity is most pronounced outside the two conserved Dkk cysteine-rich domains, raising the possibility that the gene arose from an ancestral Dkk3 precursor. (3) Dkk1, 2, and -4 all regulate Wnt signaling and bind the same effectors, unlike Dkk3 (see below). (4) Two Cnidaria, Hydra and Nematostella, each have only two Dkk genes, one related to vertebrate Dkk1, 2 and -4 (Guder et al., 2006) and one related to vertebrate Dkk3 (Fedders et al., 2004). (5) Human DKK1, 2 and -4 are located within the well-characterized chromosome 4/5/8/10 paralogy group (DKK1 maps to 10q11, DKK2 to 4q25 and DKK4 to 8p11). Genes within this paralogy region were duplicated early in vertebrate evolution (Pollard and Holland, 2000; Luke et al., 2003). In contrast, human DKK3 maps to 11p15.3, which is not part of the same paralogous chromosome group.
Taken together, these observations suggest a deep split into Dkk3 and Dkk1/2/4 gene families during early metazoan evolution, and more recent gene duplications accompanied by functional divergence of the two Dkk subfamilies (Guder et al., 2006).
Dkks modulate Wnt signaling
The hallmark of the Dkk1/2/4 family is their ability to modulate Wnt signaling. Mostly their effect is inhibitory but there is evidence that Dkk2 can also activate Wnt signaling (see below). The founding member of the family, Dkk1, was discovered by its ability to block Wnt signaling during early Xenopus embryogenesis, which is required for head induction (Glinka et al., 1998). Since then Dkk1 has been shown to block the Wnt pathway using all kinds of assays, including inhibition of Wnt-induced stabilization of β-catenin (Fedi et al., 1999) and in many cell types and vertebrate species.
Wnts trigger at least two, likely three pathways that employ Wnt receptors of the frizzled seven transmembrane class. These are (1) the canonical, or Wnt/β-catenin pathway (Cadigan and Liu, 2006), (2) the planar cell polarity pathway (PCP), which does not involve β-catenin but recruits small GTPases of the rho/cdc42 family to activate Jun kinase (JNK) and which at least in vertebrates appears to be triggered by Wnts (Tada et al., 2002) and (3) the Wnt/Ca2+ cascade, which is still controversial and may be partly overlapping with the PCP pathway (Kohn and Moon, 2005). Of the various pathways employed by Wnts, Dkks specifically affects the canonical or Wnt/β-catenin cascade. Dkk1 for example does not interfere with PCP pathway in Xenopus nor with a Xwnt5a-induced Ca2+ activation as measured by CaMKII autophosphorylation (Semenov et al., 2001). This is because Dkks bind and modulate Wnt coreceptors of the lipoprotein receptor-related protein 5/6 class, which are indispensable for routing the Wnt signal to the β-catenin pathway (He et al., 2004). Lrp5/6 are coreceptors that form a ternary complex with Wnt and Fz proteins (Figure 2a).
In addition to Fz and Lrp6, there are two other Wnt coreceptors, Ror2, an orphan tyrosine kinase possessing a CRD domain, which is bound by Wnt5a and activates JNK (Oishi et al., 2003; Mikels and Nusse, 2006), and the tyrosine kinase Derailed/Ryk (Yoshikawa et al., 2003). Their mechanism of signal transduction is still poorly understood and possible effects of Dkks on them are still unknown.
Wnt-independent functions of Dkks
An overwhelming body of evidence supports the function of Dkk1 in Wnt/β-catenin signaling, but it has been suggested that Dkk1 and -2 may have also β-catenin-independent functions. For example, the mesothelioma cell line H28 contains a homozygous deletion of the β-catenin gene and yet Dkk1 overexpression induces apoptosis and growth suppression (Lee et al., 2004). Similarly in HeLa cells, there are no significant changes in either cellular β-catenin localization, expression of Wnt target genes or tcf-reporter activity after ectopic expression of Dkk1 (Mikheev et al., 2004). However, activation of Lrp6 by Wnts may have effects which are β-catenin independent (Orme et al., 2003) and still be Dkk1 modulated. Comparing Dkk1 to dominant-negative Lrp6 or to other receptor antagonists such as Wise may reveal Lrp6-independent functions of Dkks.
JNK activation may be one such β-catenin-independent effect of blocking Wnt-LRP6 signaling. Wnt11 (via the PCP pathway) as well as Dkk1 and another secreted Wnt antagonist crescent, all can activate JNK (Pandur et al., 2002). Furthermore, inhibitors of JNK reduce apoptosis induced by Dkk1 overexpression in H28 cells (Lee et al., 2004). This suggests a crosstalk upstream of β-catenin between the canonical and the Wnt/PCP pathway.
Taken together, while it cannot be ruled out that Dkk1 and -2 may affect signaling cascades independent of their ability to regulate LRP6, there is no convincing evidence for this at present.
Dkk3 and sgy are divergent members of the Dkk family which do not seem to function in Wnt signaling (Glinka et al., 1998; Krupnik et al., 1999; Mao et al., 2001a). It was suggested that an N-terminally truncated Dkk3 encodes a substrate binding subunit, ‘p29’, of the type II iodothyronine 5′-deiodinase (D2) in rat (Leonard et al., 2000). This suggestion has not been supported (a) because of the seleno-nature of all other cloned deiodinases that act without substrate binding subunits, (b) because there is poor correlation between Dkk3/p29 and the D2 expression patterns in rat brain (Montero-Pedrazuela et al., 2003) and (c) since Dkk3 knockout mice have normal deiodinase and thyroid hormone status (del Barco Barrantes et al., 2006). While Dkk3 mouse mutants are viable, they show changes in the frequency of NK cells, immunoglobulin M, hemoglobin and hematocrit levels, as well as lung ventilation. Furthermore, Dkk3-deficient mice display hyperactivity.
Vertebrate Lrp5 and Lrp6 are closely related type I transmembrane proteins, of about 180 kDa molecular weight. They are members of the low-density lipoprotein receptor (LDLR) superfamily and function as Wnt coreceptors, whose activity is modulated by Dkks (He et al., 2004). The Lrp5/6 extracellular regions have three basic domains which occur as repeats: a YWTD type β-propeller domain, an epidermal growth factor (EGF)-like domain, and the LDLR type A domain. The cytoplasmic domain of Lrp5/6 contains five phosphorylated PPPSP motifs with flanking S/T clusters, which are targets for Casein kinase 1γ phosphorylation. Upon Wnt-Fz interaction, Lrp6 becomes phosphorylated by Casein kinase 1γ, which allows Axin to bind to Lrp6 (Davidson et al., 2005; Zeng et al., 2005).
While the affinity of Lrp5/6 for Wnts is poor, Dkk1 and -2 bind to Lrp6 with high affinity and an apparent Kd in the range of 10−10 M (Bafico et al., 2001; Semenov et al., 2001; Mao et al., 2001a). Crosslinking experiments indicate a binary 1:1 complex between Dkk1 and Lrp6 (Bafico et al., 2001). Dkk1 binding to Lrp6 interferes with the receptors ability to interact with Wnt-Fz and thereby blocks signal transduction (Semenov et al., 2001). Conversely, fusion proteins between Fz5 and Dkk1 (Holmen et al., 2005) or between Wnt5 and Dkk2 are potent activators of Wnt signaling when co-transfected with LRP6 (Liu et al., 2005), presumably by promoting Lrp6-Fz interaction.
Dkk1 binds both Lrp5 and Lrp6, but does not interact with other members of the LDLR family such as LDLR, VLDLR, ApoER2 or LRP (Semenov et al., 2001; Mao et al., 2001a). As for Dkk1, a biochemical or at least functional interaction with Lrp6 has been demonstrated for Dkk2 and -4, in contrast to Dkk3, which neither binds Lrp6 nor significantly affects Wnt signaling (Brott and Sokol, 2002; Li et al., 2002; Mao and Niehrs, 2003).
Within Lrp6, the C-terminal EGF domains 3 and 4 mediate binding to Dkk1, while the N-terminal domains 1 and -2 mediate Wnt-Fz interaction (Mao et al., 2001a; Itasaki et al., 2003). Within Dkk1, the colipase fold but not the DKK_N domain is sufficient for Lrp6 binding and Wnt inhibition (Brott and Sokol, 2002; Mao and Niehrs, 2003) and mutation of the conserved Cys220 of the colipase fold abolishes the interaction (Semenov et al., 2001). A 21-mer peptide derived from the colipase fold (L184-Ser204) has a high affinity for Lrp6, and another colipase fold derived peptide (Cys233–Cys253) mimics effects of full-length Dkk1 in reducing β-catenin and inducing proliferation in human mesenchymal stem cells (Gregory et al., 2005). Such peptides may thus serve as useful LRP5/6 inhibitors.
Genetic evidence for a Dkk1-Lrp6 interaction comes from the striking correction of severe developmental defects of Lrp6 null and Dkk1 null mice observed in Dkk1/Lrp6 double mutants, which supports a direct role of Dkk1 in reducing Wnt signaling through the Lrp6 receptor (MacDonald et al., 2004). A partial rescue of hypomorphic Dkk1 mutant phenotype is also observed in Lrp5 double mutants.
A second class of high affinity Dkk receptors are the Kremen1 and 2 (Krm1/2) single-pass transmembrane proteins that contain single Kringle, WSC and CUB domains, which are all required for Dkk1 interaction, and an intracellular domain without obvious sequence motifs (Mao et al., 2002). Kremens bind both Dkk1 and 2 but not Dkk3 with an apparent Kd in the nM range. As is the case for the LRP6 interaction, it is the colipase fold of Dkk1, which is necessary and sufficient for Kremen binding (Mao and Niehrs, 2003). Krm1/2 greatly potentiate the ability of Dkk1 to block Wnt signaling. Membrane attachment of the Krm2 extracellular domain is critical for its function, as a secreted form is inactive, while a GPI-anchored extracellular Kremen domain confers Dkk1 binding and Wnt inhibition. This suggests that Krm cytoplasmic domain is of minor importance to inhibit Wnt signaling. Krm2 forms a ternary complex with Dkk1 and Lrp6, and induces rapid endocytosis and removal of Lrp6 from the plasma membrane, thereby presumably blocking β-catenin signaling (Figure 2c). There are no Krm homologs in Drosophila, but heterologous expression of vertebrate Dkk1 and Krm2 together, but not either gene alone, results in inhibition of Wg signaling in the fly wing (Mao et al., 2002).
Antisense inhibition of Krm1/2 in Xenopus embryos induces embryonic head defects, which can be rescued by Dkk1 mRNA overexpression. Furthermore, there is strong enhancement of head defects when both Dkk1 and Krm1,2 are inhibited (Davidson et al., 2002). This argues for a functional interaction between these proteins in vivo and is consistent with overlapping expression patterns during development (Monaghan et al., 1999; Nakamura et al., 2001; Davidson et al., 2002). However, it is likely that Dkks are also able to modulate Wnt/β-catenin signaling in the absence of Krm, since Dkk1 can directly block the Lrp6-Wnt interaction (Semenov et al., 2001) (Figure 2b). Mouse mutants will be informative to reveal the general importance of Krm1/2 for Dkk action.
Wnt activation by Dkk2
While Dkk1 acts as a pure inhibitor of Wnt/β-catenin signaling, Dkk2 can either activate or inhibit the pathway, depending on cellular context. Thus, when overexpressed in Xenopus, Dkk2 synergizes with coexpressed XFz8 (Wu et al., 2000) or Lrp6 (Brott and Sokol, 2002) to induce Wnt/β-catenin signaling. This activating effect of Dkk2 is blocked by Dkk1 (Wu et al., 2000). In contrast, in HEK 293 or NIH 3T3 cells Dkk2 will inhibit a Wnt-Fz signal but activate the pathway, when cotransfected with Lrp5 or -6 (Wu et al., 2000; Li et al., 2002; Mao and Niehrs, 2003). Synergy between Dkk2 and Lrp6 is also seen in PC12 cells (Caricasole et al., 2003). Not surprisingly, the colipase fold of Dkk2 is sufficient for this synergy (Li et al., 2002). Thus, at low and high Lrp5/6 levels Dkk2 functions as Wnt inhibitor or activator, respectively. This may be explained by assuming that Wnt by virtue of its ability to activate both Lrp6 as well as Fz is a more potent pathway activator than Dkk2, which only engages Lrp5/6. At low Lrp6 doses Dkk2 would compete with Wnt and thus lower the signaling output, while at high dose Dkk2-Lrp6 signaling can overcompensate for lack of Wnt-Fz-Lrp5/6 interaction (Figure 2d).
The ability of Dkk2 to act either as agonist or antagonist of Wnt/Lrp6 signaling can also be regulated by the Dkk receptor Krm2, which invariably turns Dkk2 into a Wnt inhibitor (Mao and Niehrs, 2003). It was suggested that the Lrp6 extracellular domain is autoinhibitory because its deletion yields a constitutively active receptor (Mao et al., 2001a, 2001b). Dkk2 binding to Lrp6 may induce a conformation where this autoinhibition is blocked, leading to receptor activation. However, in the presence of Krm2 and Dkk2, the receptor complex may be internalized as is the case for Dkk1/Krm/Lrp6 (Mao et al., 2002) and thus Lrp6-mediated signaling is inhibited.
Interestingly, neither overexpression of axin or of Gsk3, nor inhibition of disheveled significantly block the ability of the Dkk2-colipase fold to activate Wnt signaling in cooperation with Lrp6 (Li et al., 2002; Liu et al., 2005) while all three will block a Wnt-Fz-mediated signal. This result supports the idea of two distinct receptor events, mediated by Lrp5/6 and Fz, respectively, which converge on β-catenin (Cadigan and Liu, 2006; Li et al., 2002).
What is it about Dkk2 that makes it different from Dkk1 in its modulating activity towards Wnt signaling? A structure–function analysis revealed that the colipase fold of Dkk1 in isolation can also synergize with Lrp6 to induce Wnt signaling and that its DKK_N domain is important to impart a purely inhibitory effect. The DKK_N domain of Dkk2 is neutral in this regard (Brott and Sokol, 2002). Again, the DKK_N and colipase fold of Dkk3 have no effect in Wnt signaling assays. Thus, while the colipase fold of Dkk1/2 mediates binding of Dkks to Lrps, the DKK_N domain modulates the outcome of this interaction (Brott and Sokol, 2002).
It is important to note that these studies all involve Dkk2 and Lrp6 overexpression and it is not clear whether physiologically agonistic or antagonistic effects dominate. During eye formation Dkk2 acts as Wnt antagonist but during bone formation Dkk1 and Dkk2 have very different properties, which raises the possibility of an agonistic action (see below).
Role of Dkks in embryonic development
Given the paramount role that Wnts play during embryonic development it is not surprising that one main function of Dkks is to control cell fate in vertebrates as they show highly regionalized expression (Grotewold et al., 1999; Monaghan et al., 1999; Hashimoto et al., 2000; Shinya et al., 2000; Chapman et al., 2004; Diep et al., 2004; Idkowiak et al., 2004; Fjeld et al., 2005; Nie, 2005; Nie et al., 2005). Dkk1, 2 and -3 mouse mutants are available and a summary of their phenotypes is presented in Table 1.
Dkk1 and antero-posterior (a–p) axial patterning
Wnt/β-catenin signaling plays an important role in a–p patterning of the primary embryonic axis. One of the functions of Spemann's organizer in Xenopus, the zebrafish shield and the mouse anterior mesendoderm is to antagonize Wnt signals, which together with bone morphogenetic proteins (BMPs) and Nodals, are inhibitory to anterior development in general, and in particular to the anterior central nervous system (De Robertis and Kuroda, 2004; Niehrs, 2004b). The different expression domains of these growth factors and their antagonists, such as Dkk1, create signaling gradients, which pattern the early embryo in a combinatorial fashion and orchestrate regionally specific induction of axial structures, including the head.
Dkk1 was initially identified as a gene conferring Spemann's head organizer activity in Xenopus embryos. When coexpressed with BMP inhibitors, Dkk1 will induce entire ectopic heads in Xenopus embryos, while injection of anti-Dkk1 antibodies yields microcephalic or even headless Xenopus embryos (Glinka et al., 1998). Consistent with a physiological role as head inducer, Dkk1 is specifically expressed in the anterior mesendoderm (prospective prechordal plate), which harbors potent head inducing ability when transplanted. Dkk1 is a direct transcriptional target of β-catenin in Xenopus and thus acts as a negative feedback regulator (Chamorro et al., 2005).
In zebrafish, Dkk1 is also expressed in organizer derivatives of gastrulating embryos (Hashimoto et al., 2000; Shinya et al., 2000). Overexpression of Dkk1 suppresses defects in the development of forebrain, eyes, and notochord in boz mutants, which have organizer deficiencies. Likewise, Dkk1 overexpression promotes anterior neural development in embryos injected with antivin RNA, which lack most mesendoderm (Hashimoto et al., 2000). These data support the model that Wnt/β-catenin signaling inhibits anterior embryonic development, which is prevented by Dkk1.
In mouse embryos, Dkk1 is expressed from gastrula to neurula in the anterior visceral endoderm, the anterior mesendoderm, and foregut endoderm, respectively, tissues which are all associated with anterior specification. Mice mutant for Dkk1 lack head structures anterior to the mid-hindbrain boundary and already at E7.5 the anterior marker Hesx1 fails to be expressed (Mukhopadhyay et al., 2001). Analysis of chimeric embryos composed of Dkk1-expressing epiblast cells developing within the confines of a Dkk1−/− visceral endoderm showed that Dkk1 function in the anterior visceral endoderm (AVE) is not essential for proper anterior morphogenesis. Thus, Dkk1 expression appears to be required in the definitive anterior mesendoderm for early anterior development in mouse (Mukhopadhyay et al., 2001).
In the transgene-induced mouse mutant doubleridge, which harbors a hypomorphic Dkk1 allele (Dkk1d) with 10% normal expression levels, head development is normal. However, Dkk1d/– compound heterozygotes on a C57BL/6J genetic background, show a variety of head defects, from hydrocephaly with micrognathia, to loss of anterior head structures and eyes (MacDonald et al., 2004), which are still less severe than the truncation of anterior head structures observed in the Dkk1 null mice.
While Dkk1 heterozygous mice are normal and fertile, mice double heterozygous for Dkk1 and the BMP inhibitor Noggin show head defects similar to those of Dkk1 homozygous mutants (del Barco Barrantes et al., 2003). This strongly supports the idea that double inhibition of BMP and Wnt signals is required for head development (Glinka et al., 1997). Interestingly, homozygous Noggin/Dkk1 mutant embryos were never obtained, at least not by E6. This raises the possibility that Dkk1 and Noggin cooperate even before gastrulation in some unknown process.
A very elegant study recently revealed a pregastrulation role for Dkk1 at E5.5–E6, when the initial proximal–distal axis of the mouse embryo is converted to a–p polarity, by a process involving coordinated cell migrations (Kimura-Yoshida et al., 2005). The homeobox gene Otx2 is essential for this process, and since one of its target genes is Dkk1 (Zakin et al., 2000), it suggests that some of the head defects observed in Otx2 mutants are mediated by Dkk1. To test this, the authors inserted Dkk1 in the Otx2 locus and indeed, defective axis conversion due to Otx2 deficiency was rescued (Kimura-Yoshida et al., 2005). Painstaking cell transplantation experiments in these early mouse embryos revealed that canonical Wnt signals and Dkk1 function as repulsive and attractive guidance cues, respectively, in the migration of visceral endoderm cells. This supports a model where a signaling gradient of Wnt/β-catenin is set up by Wnts and localized Dkk1, which mediates a–p axis polarization by guiding cell migration toward the prospective anterior in the pregastrula mouse embryo.
Another possible role of Dkk1 in a–p patterning concerns formation of the embryonic heart, which arises from progenitors localized in bilateral anterior domains of the embryo. Dkk1 and crescent (a Wnt inhibitor of the sFRP family) are able to induce myocardial tissue formation from putative noncardiogenic mesodermal tissue (Marvin et al., 2001; Schneider and Mercola, 2001). While this supports the idea that Wnt inhibition is required for formation of yet another organ of anterior origin, Dkk1 may not be essential for this process, or act redundantly with other Wnt inhibitors, since Dkk1 mutant mice have apparently normal hearts.
While loss of Dkk1 function is associated with reduction of anterior structures, mice double mutant for the Dkk receptors LRP5 and -6 show a somewhat opposite phenotype, with expansion both of anterior primitive streak derivatives and anterior neurectoderm (Kelly et al., 2004). This supports the general conclusion that maintenance of anterior identity by antagonizing Wnt-Lrp5/6 signaling is a key role of Dkk1 during early vertebrate development.
Dkk1 and limb formation
Wnt signaling plays an important role in regulating patterning and growth of the vertebrate limb (Church and Francis-West, 2002). During early limb development, Wnt/β-catenin signaling by Wnt3 is required for formation of the apical ectodermal ridge (AER), since Wnt3 mutants have a reduced AER (Barrow et al., 2003; Soshnikova et al., 2003), a signaling center that dictates limb growth. Similarly, overexpression of Dkk1 in chicks induces limb truncation and this is accompanied by apoptosis (Mukhopadhyay et al., 2001; Grotewold and Rüther, 2002). Conversely, overactivation of Wnt signaling by expression of a gain-of-function mutation of β-catenin results in expansion of the AER (Soshnikova et al., 2003). An expansion of the AER is observed in both Dkk1 null and hypomorphic Dkk1d/d mutants, suggesting a genetic interaction with Wnt3 (Mukhopadhyay et al., 2001; Adamska et al., 2004; MacDonald et al., 2004). In the control of the AER, Wnt3 lies upstream of BMP ligands (Barrow et al., 2003) and Dkk1 interacts with BMPs in this context, as they can mutually induce each others expression (Mukhopadhyay et al., 2001; Grotewold and Rüther, 2002).
Dkk1 null and Dkk1d/d mutants also display postaxial polysyndactyly in the forelimbs (Mukhopadhyay et al., 2001; MacDonald et al., 2004). Consistent with the polydactyly in Dkk1d/d mice being the result of excess Wnt signaling through Lrp6, normal digit number is restored in Dkk1d/dLrp6+/– mice (MacDonald et al., 2004). Polydactyly in Dkk1 mutants is due to an interaction with Wnt7a, which regulates digit number (Adamska et al., 2004). One of the consequences of reduced Wnt signaling in Wnt7a null mice are missing digits, while the Dkk1 mutants show extra digits, similar to null mice mutant for Engrailed (En1), a transcriptional repressor of Wnt7a. That Dkk1 and Wnt7a genetically interact is shown in Dkk1d/+Wnt7a/ double mutants, in which digit loss is prevented. In contrast, Dkk1d/dEn−/− double mutants show enhanced digit phenotypes not seen in either single mutant, which indicates that Dkk1 and En1 interact in negative Wnt7a regulation (Adamska et al., 2004).
In conclusion, during mouse forelimb development Dkk1 controls both digit patterning through Wnt7a and limb growth via Wnt3.
Dkk1 and vertebral development
During vertebrate embryogenesis, the somites give rise to the vertebrae or ribs. A new pair of somites forms during development in a periodic process, which is under control of a segmentation clock that drives oscillating mRNA expression of a number of cyclic genes in the unsegmented presomitic mesoderm (PSM). This oscillation is regulated by an interaction between the Notch and the Wnt/β-catenin pathways (Aulehla and Herrmann, 2004).
In the vestigial tail (vt) mouse mutant, a hypomorph of Wnt3a, caudal somites are missing (Greco et al., 1996), and oscillating expression of the Wnt target Axin2 is lost (Aulehla et al., 2003). The phenotype of Lrp6 null mutant mice with loss of caudal somites is very similar (Pinson et al., 2000). In the hypomorphic Lrp6 mutant ringelschwanz (rs), a–p polarity of the somites is affected, and no discrete somite a–p compartments are established from the lumbar region onwards (Kokubu et al., 2004). As the cycling Lfng expression in the PSM is abolished in Lrp6rs mutants, this indicates that during somitogenesis Lrp6 is required in the segmentation clock.
The expression of Dkk1 in the PSM overlaps the Wnt3a expression domain, where it shows a segmented expression pattern (Glinka et al., 1998). Intriguingly, Dkk1 came out of a microarray analysis of Wnt target genes in PSM as the gene with most pronounced periodic expression (O Pourquie, personal communication). In Dkk1 null mice kinked tails and fused vertebrae are observed (MacDonald et al., 2004). Somite defects are also observed in Crooked tail (Cd) a hypomorphic Lrp6 mutant. Lrp6Cd is a gain of function mutant, which still can interact with Wnts, but which is unable to respond to Dkk1 and is hyperactive (Carter et al., 2005). Taken together, these results suggest a role of Dkk1 as a negative feedback regulator in the Wnt-Notch oscillator.
Dkk1 and 2 in bone formation
The Wnt/β-catenin pathway plays a crucial role in bone formation and generally promotes an increase in bone mass by various mechanisms including renewal of stem cells, osteoblast proliferation, induction of osteoblastogenesis and inhibition of osteoblast and osteocyte apoptosis (Krishnan et al., 2006).
Lrp5 is a key regulator for bone mass. In humans, LRP5 loss-of-function mutations lead to osteoporosis-pseudoglioma syndrome, which is characterized by low bone density (Gong et al., 2001), while gain-of-function mutations are associated with high bone mass (Boyden et al., 2002). Interestingly, the N-terminal LRP5 gain-of-function mutations (e.g. G171V) show reduced affinity for Dkk1, and thus presumably are due to Lrp5 derepression (Boyden et al., 2002; Ai et al., 2005).
Similar to humans, Lrp5–/– mice have a low bone mass phenotype due to reduced proliferation of precursor cells (Kato et al., 2002), which is enhanced by additional loss of one Lrp6 allele (Holmen et al., 2004). Conversely, mice that overexpress the G171V LRP5 mutant, have a high bone mass phenotype (Babij et al., 2003). Unlike homozygous mutants, heterozygous Dkk1+/− mice are viable and they show an increase in bone mineral density (Morvan et al., 2006). Conversely, transgenic mice overexpressing Dkk1 in bone develop osteopenia (Li et al., 2006). This supports the notion that Wnt-Lrp5 signaling is under negative control of Dkk1 to regulate physiological levels of bone mass. The Dkk1-Lrp5 interaction has therefore become an interesting pharmaceutical target for drug development to treat osteoporosis.
In light of the results with Dkk1+/− mice with elevated bone density it was surprising that Dkk2−/− mice show a somewhat opposite phenotype, namely low bone mineral density and osteopenia due to major defects in mineralization (Li et al., 2005). Without a significant change in the number of osteoblasts, the osteoid surface is increased in these animals and it was proposed that Dkk2 is involved in terminal osteoblast differentiation. Although the precise mechanisms by which Dkk2 is involved in osteoblast differentiation remains unclear, the results suggest that Dkk2 may act by antagonizing Wnt/β-catenin signaling at a late stage of bone formation since some of the effects of Dkk2 deficiency were partially reversed by expression FRP3, a different Wnt antagonist. RNAi experiments support the idea that Dkk1 and Dkk2 may have distinct functions in osteoblast differentiation (van der Horst et al., 2005).
Dkk1 and 2 in eye and skin
When expression of Dkks is surveyed in adult mice tissues, the eye is the one organ which shows the most prominent expression of Dkk1, 2 and -3. These Dkks display suggestive expression patterns in eye primordia, including pigmented epithelium, choroid and retina (Monaghan et al., 1999). Wnt/β-catenin signaling plays an important role in various stages of eye formation, including early patterning (Esteve and Bovolenta, 2006), retinal development (Van Raay and Vetter, 2004), lens formation (Lovicu and McAvoy, 2005) and eye angiogenesis (Niehrs, 2004a). In humans, loss of LRP5 function causes congenital blindness (pseudoglioma) (Gong et al., 2001).
Coinjection of Dkk1 with BMP antagonists mRNAs typically induces complete heads containing two well-formed eyes. In contrast, both frzb (a sFRP member) and dominant-negative Xwnt8 together with BMP inhibitors induce only one ectopic eye. The induction of bilateral eyes can be explained by the ability of Dkk1 but not Frzb (which may not inhibit all Wnts) to promote prechordal plate formation, which is essential to split the primary eye field via hedgehog signaling (Kazanskaya et al., 2000). Consistent with a role in prechordal plate specification, Dkk1 null mutant mice express no or reduced levels of Hesx1 (Mukhopadhyay et al., 2001), as is the case for Xenopus embryos injected with anti-Dkk1 antibody (Dkk1 Ab) (Kazanskaya et al., 2000). The cyclopia observed in Dkk1 null mutant mice and in Dkk1-Ab injected Xenopus embryos is therefore most likely due to defective eye field splitting.
While mouse Dkk1 is also expressed in specific regions of the developing eye, the severity of head defects in null mutants precludes analysis of a more direct role that Dkk1 may play. In contrast, Dkk2 null mutant mice are viable but they are blind, because they grow hair on their cornea (Mukhopadhyay et al., 2006). In fact, in the absence of Dkk2 there occurs a complete transformation of the corneal epithelium into a stratified epithelium that expresses epidermal-specific differentiation markers and develops appendages such as hair follicles. Analysis of Wnt reporter in these mutants indicates that the loss of Dkk2 activates Wnt/β-catenin signaling (Mukhopadhyay et al., 2006). Furthermore, conditional ablation of Notch1 leads to a similar phenotype in mice, which is also due to activation of the Wnt pathway (Nicolas et al., 2003). Thus, in the context of eye development, Dkk2 seems to function as Wnt antagonist to inhibit stratified epithelium and hair follicle formation. Indeed, transgenic mice overexpressing Dkk1 in the skin are almost hairless, because of inhibition of follicle formation (Andl et al., 2002).
Dkks and cancer
Patients with multiple myeloma frequently show painful bone lesions and in a recent analysis an increase in Dkk1 in the serum of those patients was noted, whose levels correlated with lesion occurrence (Tian et al., 2003). Bone marrow serum containing an elevated level of Dkk1 inhibited the differentiation of osteoblast precursor cells in vitro. The authors propose that Dkk1 produced by myeloma cells blocks osteoblast differentiation, thereby causing the lytic bone lesions. Supporting this idea, treatment of myeloma patients by autologous stem cell transplantation induces a decrease in Dkk1 levels, which is accompanied with elevation of bone formation markers (Politou et al., 2006). This raises again interesting therapeutic approaches for interference with the Lrp5/6-Dkk1 interaction.
A number of studies have shown changes of Dkk expression in tumor cell lines or tissues (Table 2). In colon cancer, DKK1 is a downstream target gene of β-catenin (Gonzalez-Sancho et al., 2005), as it is in human ovarian endometrioid adenocarcinomas (Chamorro et al., 2005). DKK1 is typically silenced in colon cancer by DNA hypermethylation and this is correlated with advanced stages of colorectal tumorigenesis (Gonzalez-Sancho et al., 2005; Aguilera et al., 2006). Overexpressing DKK1 in colon cancer cells (Aguilera et al., 2006) or Hela cells (Mikheev et al., 2004) reduces colony formation and tumor growth in xenografts, suggesting a tumor-suppressor function for DKK1.
Dkk3 has also been proposed to act as tumor suppressor, as it is downregulated in a number of tumor cells (Table 2). Hence, DKK3 is also known as ‘REIC’ (Reduced Expression in Immortalized Cells) (Tsuji et al., 2000). Furthermore, DKK3 overexpression suppresses tumor cell growth (Hoang et al., 2004; Abarzua et al., 2005; Lodygin et al., 2005; Kawano et al., 2006). While hypermethylation of human DKK3 correlates with certain cancers (Kobayashi et al., 2002; Roman-Gomez et al., 2004; Lodygin et al., 2005), the physiological relevance of altered DKK3 expression in tumors and its potential growth inhibitory effect are unknown. Of note, Dkk3 mutant mice show no enhanced tumorigenesis (del Barco Barrantes et al., 2006).
Other roles of Dkk1
A physiological role for Dkk1 in skin concerns the pigmentation pattern of the human hand. Palmoplantar fibroblasts secrete Dkk1, which inhibits melanocyte growth and differentiation, and this may explain the characteristic lighter pigmentation of the inner- compared to the outer hand (Yamaguchi et al., 2004).
Dkk1 has recently been implicated in neurodegenerative disease. Dkk1 is induced in degenerating neurons from Alzheimer patients as well as in cultured neurons challenged with β-amyloid peptide. Dkk1 may promote apoptosis in Alzheimer neurons by enhancing Gsk3-mediated phosphorylation of the Tau protein in β-amyloid-treated neurons (Caricasole et al., 2004). Dkk1 is also induced in neurons subjected to ischemic insults, and antisense knockdown of Dkk1 shows some protection against neuronal apoptosis (Cappuccio et al., 2005). Hence, induction of Dkk1 may be involved in an apoptosis pathway. This is supported by the observation that Dkk1 is a transcriptional target of p53 (Wang et al., 2000) and is proapoptotic in limb development (Grotewold and Rüther, 2002), in mesothelioma (Lee et al., 2004), as well as in human glioma cells, where it is induced by DNA damage (Shou et al., 2002).
In the 8 years since the discovery of the Dkk family of Wnt modulators we have learned a good deal about their biochemical function, as well as the role that these proteins play both in development and disease. The next years will likely see more examples of biological processes in which these proteins are involved, given the overwhelming importance of Wnts. Dkks are an interesting target for the pharmaceutical industry, with applications in osteoporosis and cancer. Development of small molecules that interfere with Dkk-Lrp5/6 interactions should be aided by X-ray crystallography. A major open question regards the receptor and signaling cascade as well as the roles of the Dkk3/sgy subfamily, which is still poorly understood.
I thank Kristina Ellwanger, Andrei Glinka and Sonia Pinho for critically reading the manuscript and B Engelhardt for artwork. This work was supported by the Deutsche Forschungsgemeinschaft (Ni 286/12-1).
About this article
Nature Communications (2018)