Kremen is required for neural crest induction in Xenopus and promotes LRP6-mediated Wnt signaling

The canonical Wnt/β-catenin signaling cascade is an evolutionarily conserved and ancient pathway playing major roles during animal development,e.g. in axis formation, anteroposterior (AP) patterning and neural crest formation (Moon et al., 1997; Wodarz and Nusse, 1998). Misregulation of Wnt signaling is implicated in cancer formation(Bienz and Clevers, 2000). During Wnt/β-catenin signaling, Wnt glycoproteins bind to Frizzled receptors and LRP5/6 co-receptors, and activate a complex signaling cascade to stabilize β-catenin, a transcriptional co-activator(Wodarz and Nusse, 1998; Nusse, 2005).

LRP5/6 proteins are essential components for Wnt/β-catenin signal transduction, and their inactivation phenocopies loss of Wnt signaling in vertebrates and invertebrates (Pinson et al., 2000; Tamai et al.,2000; Wehrli et al.,2000). LRP5/6 proteins are members of the low-density lipoprotein receptor family (LDLR), transmembrane receptors characterized by an extracellular domain containing LDLR repeats and YWTD β-propeller/EGF modules (He et al., 2004). The YWTD β-propeller/EGF modules are cysteine-rich domains of around 35 kDa,which require specific folding and maturation assistance in the endoplasmic reticulum (ER), mediated by the chaperone Mesd(Culi and Mann, 2003; Hsieh et al., 2003; Culi et al., 2004). Mesd is a resident protein of the ER that binds to and promotes cell surface localization of LRP5/6 by reducing receptor aggregation(Hsieh et al., 2003).

Wnt/LRP5/6 signaling is antagonized by Dickkopf (Dkk) proteins, which bind and inhibit LRP5/6 (Bafico et al.,2001; Mao et al.,2001; Semenov et al.,2001). During early vertebrate development, Dkk1 is expressed in anterior endomesoderm and plays an important role in AP patterning by inhibiting Wnt/LRP6 signaling in the head organizer and prechordal plate(Glinka et al., 1998; Hashimoto et al., 2000; Shinya et al., 2000; Mukhopadhyay et al.,2001).

Other Dkk receptors are the transmembrane proteins Kremen1 and 2 (Krm1 and 2, collectively termed Krms). Dkk1 forms a ternary complex with Krm1 or 2 and LRP6, which is cleared from the cell surface, thereby shutting down Wnt signal transduction (Mao et al.,2002). Krms are evolutionary conserved in vertebrates and are differentially expressed during embryonic development in mouse and frog(Nakamura et al., 2001; Davidson et al., 2002). In Xenopus, Krms and Dkk1 are co-expressed in the prechordal plate and functionally cooperate in Wnt inhibition to regulate AP patterning of the central nervous system (CNS) (Davidson et al., 2002). Recently, it was shown that Krm1 is required for formation of thymic architecture in mice by acting as Wnt inhibitor(Osada et al., 2006).

While it is established that Krms and Dkk1 cooperate in Wnt inhibition during Xenopus AP patterning, it is unknown whether Krms play other roles during early development. In Xenopus embryos, krm2 is expressed in various regions that do not overlap with dkk1 expression domains (Davidson et al.,2002). We therefore investigated the possibility of Dkk1-independent functions of Krm2.

Here we report that Krm2 plays a Dkk1-independent role in neural crest (NC)formation. Krm2 is positively regulated by zygotic Wnt signaling and is required for NC formation. We show that in the absence of Dkk1, Krms stimulate Wnt signaling and promote, via direct binding, cell-surface localization of LRP6. Furthermore, Krm2 knockdown specifically reduces LRP6 protein levels in NC cells. The results suggest that Krms act as inhibitor or activator of Wnt signaling, dependent on the presence or absence of Dkk,respectively.

In vitro fertilization, embryo culture, staging and microinjection were carried out as described (Gawantka et al.,1995). Preparation of mRNA for Xenopus injections was carried out in the presence of cap analog using the MegaScript in vitro transcription kit (Ambion), according to the manufacturer's instructions. Animal cap explants and conjugates were prepared as described(Villanueva et al., 2002).

Whole-mount in situ hybridizations were carried out essentially as described (Gawantka et al.,1995). For lineage tracing, lacZ mRNA (250 pg per blastomere) was co-injected and β-galactosidase staining was performed as described (Sive et al., 2000),using blue (X-Gal) or red (Magenta-Gal) substrate. RT-PCR assays were carried out in the exponential phase of amplification and with primers as described[H4 (Niehrs et al.,1994); chordin (Sasai et al., 1994); XWnt8(Christian et al., 1991); Xhox3 (Ruiz i Altaba and Melton,1989)]. Other primers were: Xenopus krm2 (forward,GGAACCAGACCACACAGCACTTG; reverse, CCGCCTCCACACCTGCATACT) and Xenopus brachyury (forward, CACAGTTCATAGCAGTGACCG; reverse,TTCTGTGAGTGTACGGACTGG).

Krm2 MO-1 targeting the ATG start codon of Xkrm2 was as described(Davidson et al., 2002); Krm2 MO-2 [targets 5′ untranslated region (UTR) of Xkrm2]:ATCCTCACATGAAGACGTGCTGGAA; LRP6 MO [targets 5′ UTR of X. laevis(AF508961) and X. tropicalis (CX889920) LRP6]:CCCCGGCTTCTCCGCTCCGACCCCT. Control morpholino: standard control morpholino oligo designed by Gene Tools, LLC.

For neural crest immunostaining experiments, two-cell stage embryos were injected with 7.5 and 12.5 ng CoMO or Krm2 MO-2, 0.2, 2 and 5 ng LRP6 MO, or 1.5, 3 and 5 ng Krm2 MO-1 per blastomere, respectively(Fig. 7B). Anterior NC explants were prepared from neurulae by peeling off the epidermal layer before dissection with eyebrow knives. Immunostaining procedures were essentially carried out as described (Unterseher et al., 2004), and embryos were mounted in Mowiol. Primary antibodies used were rabbit anti-LRP6 T1479 (dilution 1: 50)(Davidson et al., 2005) and mouse anti-β1-integrin (dilution 1:10) (8C8, DSHB). Secondary antibodies were goat anti-rabbit Alexa 488 and goat anti-mouse Alexa 546 (Molecular Probes). Nuclei were counterstained with Hoechst. Explants were examined on a confocal laser-scanning microscope (Nikon C1Si). For statistics, all images were transformed into monochrome RGB images (red for Alexa 546, green for Alexa 488, blue for Hoechst) and processed using ImageJ. For each neural crest explant, fluorescence intensity was measured in red (R) and green (G) channels at three random plasma membrane positions and averaged. The average R/G signal intensity of all explants per sample was determined. LRP6 protein levels were classified as unchanged when R/G <1.5, as moderately reduced (+) when R/G=1.5-3, or as strongly reduced (++) when R/G >3.

HEK293T cells were maintained in DMEM supplemented with 1% L-Glutamine, 1%PEN-STREP, and 10% FCS and grown at 10% CO₂. Luciferase reporter gene assays were carried out in triplicates in 96-well plates using Promega's Dual-Luciferase Reporter Assay System as described(Wu et al., 2000). In all experiments a total of 50 ng DNA per well was transfected; 1 ng pCMV-SPORT6-mesd (mouse) or 5 ng pCS2-Xkrm1 DNA were used as indicated. For Wnt reporter assays transfected DNAs per well were: 10 ng TOPFLASH and 1 ng pTK-Renilla reporter plasmids; 12 ng(Fig. 4B) or 24 ng(Fig. 4A,D) human LRP6; 2 ng mouse frizzled8 (fz8); 0.25 ng human dishevelled1 (dvl1); 0.5 ng human LRP6ΔE1-4 (Mao et al., 2001); 5 ng/3 ng mouse wnt1/human LRP6(Fig. 4D); 0.5 ng Xdkk1.

For BMP responsive reporter assays transfected DNAs per well were: 20 ng BREx4-E1b-dLuc plasmid [modified from Hata et al.(Hata et al., 2000)] and 1 ng pTK-Renilla; 10 ng pcDNA3.1-BMP4 as indicated.

After transfection, cells were grown for 48 hours, then lysed in 50 μl passive lysis buffer (Promega) per well. Luciferase activity was normalized against Renilla activity.

For co-immunoprecipitation (Co-IP) assays, HEK293T cells were transfected in 10 cm plates with 0.1 μg pCS2-krm1-V5 or pCS2-krm2-V5(both mouse) together with 1.5 μg pCS2-flag-LRP6, 0.1 μg pCS2-flag-LDLRΔC (both human), 0.5 μg pCS2-flag-XFLRT3 (Xenopus fibronectin leucine-rich transmembrane protein 3) or 1.5 μg empty vector pCS2 using Fugene6 (Roche). After 48 hours, cells were washed in PBS and lysed in NP-40 buffer containing 150 mM NaCl, 50 mM Tris pH 7.4, 7.5% glycerol, 1 mM EDTA, 1 mMβ-mercaptoethanol, 25 mM NaF, one protease inhibitor cocktail tablet/25 ml (Roche) and 0.8% NP-40. Lysates were subjected to Co-IP with anti-flag antibody beads (Sigma) overnight at 4°C. Co-IPs were washed with NP-40 buffer and analyzed by SDS-PAGE and western blotting.

In vitro binding assays were carried out essentially as described(Cruciat et al., 2006). Recombinant proteins were produced as conditioned media by transient transfection of HEK293T cells with pCS2-krm1ΔTMC-V5,pCS2-krm2ΔTMC-V5, pCS2-V5-dkk3 (all mouse) or pCS2-myc-LRP6ΔTMC (human) in serum-free media (Optimem I, Gibco). Media were concentrated about 50-fold using Centricon Plus-20 filters (Millipore). Equal amounts of V5-tagged proteins were resuspended in NP-40 buffer containing 0.2% (w/v) NP-40 and incubated with anti-V5 antibody beads (Sigma) under gentle shaking overnight at 4°C. IPs were washed and incubated for 5 hours with media containing pCS2-myc-LRP6ΔTMC. Co-IPs were washed again and analyzed by SDS-PAGE and western blotting.

For deglycosylation of LRP6, HEK293T cells were grown in 10 cm plates and transfected with 1 μg flag-LRP6 together with 0.4 μg pCS2-myc-dkk3, pCMV-SPORT6-mesd, pCS2-krm2-V5 (all mouse) or empty pCS2. After 2 days, cells were washed in Hank's buffer and resuspended in 1 ml hypotonic buffer containing 1 mM EDTA, 5 mM HEPES, pH 7.5,0.1 mM PMSF and one protease inhibitor cocktail tablet per 25 ml. Samples were dounced 25 times, and after removal of cell debris by centrifugation at 2500 rpm (1050 g), membranes were pelleted by centrifugation at 30,000 rpm (40,300 g). Pellets were lysed in NP-40 buffer(without NaF) containing 2% NP-40 and subjected to EndoH (Roche) treatment(0.25 U/ml) in 100 mM NaOAc, pH 5.5, for 30 minutes at 37°C. Samples were analyzed by SDS-PAGE and western blotting.

For cell surface biotinylation, HEK293T cells were transfected in 6 cm dishes with 0.25 μg pCS2-flag-LRP6 together with 0.1 μg pCS2-myc-dkk3, pCMV-SPORT6-mesd or pCS2-krm2-V5, or pCS2-flag-NME (human). After 2 days, cells were surface biotinylated by using 0.5 mg/ml sulpho-NHS-LC-biotin (Pierce) according to the manufacturer. After preparation of membranes as described above, samples were immunoprecipitated with anti-flag antibody beads and analyzed by SDS-PAGE and western blotting.

We previously reported that Xkrm1 and 2 are co-expressed with dkk1 in the Xenopus prechordal plate during mid-neurula stages, where they cooperate to regulate AP CNS patterning(Davidson et al., 2002). However, krm2 showed additional expression in regions devoid of dkk1 transcripts. At gastrula stages Xkrm2 expression occurred in the ventrolateral marginal zone of the embryo(Fig. 1A). During neurula stages krm2 was expressed in the lateral neural plate, overlapping with the neural crest marker slug(Fig. 1B). Furthermore, krm2 was co-expressed with Wnt genes such as Wnt8 and Wnt3a (Fig. 1A),raising the possibility that krm2 expression is regulated by Wnt signaling. Indeed, local inhibition of Wnt signaling by injection of dominant-negative XWnt8 mRNA abolished the expression of krm2 in the marginal zone (Fig. 1C,E). Conversely, injection of Wnt8 and Wnt3aDNA led to ectopic expression of krm2(Fig. 1D,E).

Furthermore, we analyzed the effect of Wnt and BMP pathway activation on krm2 expression in animal caps. In control caps krm2 was expressed at moderate levels. Wnt8 and Wnt3a mRNA injections increased krm2 expression, whereas BMP4 mRNA had no effect(Fig. 1F). LiCl treatment of early embryos, which leads to dorsoanteriorization, downregulated krm2 expression along with Wnt8 and other ventrolaterally expressed genes (Fig. 1G, and C.H. and C.N., unpublished).

We conclude that: (1) at gastrula stages krm2 is expressed and regulated like a classical ventrolateral gene; (2) krm2 is co-expressed with Wnts and regulated by zygotic Wnt signaling; (3) at neurula stages krm2 shows differential expression in the neural crest.

To study potential Dkk1-independent roles of Krm2, we first analyzed gain-of-function effects. Localized injection of krm2 mRNA into prospective anterior embryonic regions induced ectopic cement glands(Fig. 2A, arrowhead) and retinal pigment epithelium (Fig. 2B, arrowhead). Ventral injection of krm2 mRNA led to induction of protrusions containing melanocytes and fin-like structures(Fig. 2C, arrowhead). Widespread krm2 overexpression led to strongly hyper-pigmented embryos due to overproduction of melanocytes(Fig. 2E). Embryos frequently also showed eye and tail defects (not shown).

Melanocytes are NC derivatives, which in Xenopus are characteristically overproduced when NC regulators such as slug or sox10 are overexpressed (LaBonne and Bronner-Fraser, 1998; Aoki et al., 2003; Honore et al.,2003). Indeed, Krm2 overexpression affected NC markers in a region-specific fashion: posterior krm2 mRNA injection led to expansion of slug and sox10(Fig. 2J,L, black arrowheads). By contrast, anterior krm2 injection reduced slug and sox10 expression (Fig. 2K,M, black arrowheads). Thus, the region of krm2expression determines its effect on NC markers (summarized in Fig. 2N). This may be explained by the interaction of anteriorly overexpressed Krm2 with Dkk1, to inhibit Wnt signaling and NC induction. In posterior regions, where Dkk1 is absent, Krm2 had a stimulatory effect on NC development. Furthermore, krm2 mRNA injection can also induce ectopic expression of the NC marker sox10(Fig. 2O,P, arrowheads).

As krm2 is expressed in the prospective NC region and is sufficient to induce NC tissue, we next asked if it is also required for NC development. We made use of the previously characterized morpholino antisense oligonucleotide targeting the ATG codon of Xkrm2 (MO-1)(Davidson et al., 2002). Besides the previously described microcephaly, we found that Krm2 MO-1 injection strongly reduced the pigmentation of embryos(Fig. 3A). Co-injection of V5-Xkrm2 mRNA (Davidson et al.,2002), a construct lacking the MO-binding site, partially restored the pigmentation (Fig. 3A),showing the specificity of the morpholino. Furthermore, Krm2 MO-1 inhibited slug expression (Fig. 3B,C). This is also the case for a second Krm2 MO (Krm2 MO-2),which targets the 5′ UTR of krm2(Fig. 3B,C), thus corroborating the specificity of the NC inhibition.

To analyze if Krm2 is required specifically in the ectoderm for NC formation, or in the underlying, inducing mesoderm, we combined Krm2 MO-1-injected animal caps with uninjected dorsolateral marginal zones (DLMZs)and analyzed NC induction. DLMZs induced slug in control caps in 62%of cases (n=42). Slug induction was reduced upon injection of Krm2 MO-1 into the responding ectoderm (38%, n=24). By contrast,injection of Krm2 MO-1 into DLMZs did not affect slug induction in animal caps (86%, n=13) (Fig. 3D). These results indicate that Krm2 is required directly in the ectoderm for NC formation, where it is also expressed.

It is well established that canonical Wnt signaling is required for NC induction (LaBonne and Bronner-Fraser,1999; Wu et al.,2003; Barembaum and Bronner-Fraser, 2005). In Xenopus, overexpression of Wnts or Wnt receptors, as well as downstream signaling components, can all induce NC. Conversely, inhibition of Wnts, Wnt receptors and β-catenin blocks NC induction (Yanfeng et al.,2003; Barembaum and Bronner-Fraser, 2005; Wu et al., 2005; Abu-Elmagd et al.,2006).

As Krm2 is a negative Wnt modulator and is itself Wnt-regulated, we therefore hypothesized that Krm2 may also play a positive role in Wnt/LRP6-mediated NC induction. Hence, we tested whether LRP6 knockdown mimics Krm2 loss-of-function during NC development.

Injection of an LRP6 MO targeting both X. laevis and X. tropicalis genes resulted in strongly anteriorized embryos with enlarged cement gland, shorter and ventrally bent tail and triangular body shape(Fig. 3E). This phenotype closely mimics the anteriorization induced by overexpression of the LRP6 antagonist dkk1 (Fig. 3E). Of note, a morphologically highly sensitive structure to injection of low doses of LRP6 MO is the dorsal fin (see Fig. S1 in the supplementary material). The LRP6 MO phenotype was fully rescued by co-injecting human LRP6 mRNA (Fig. 3Ec,e), confirming specificity of the MO. Furthermore, injection of LRP6 MO in X. tropicalis embryos resulted in the same phenotype as in X. laevis (Fig. 3F).

Similarly to Krm2 MO, LRP6 MO inhibited slug expression(Fig. 3G). Thus, MO-mediated knockdown of both LRP6 and its modulator Krm2 result in inhibition of NC formation.

The results obtained so far suggest that Krm2, besides its well-established role in Wnt inhibition, may also, in the absence of Dkk1, positively regulate Wnt signaling. Previously we found no effect of krm1 or 2 in Wnt responsive reporter assays triggered by transfection of wnt/fz or wnt/fz/LRP6 in HEK293T cells. We now find that Xkrm1, but also Xkrm2, significantly stimulated Wnt signaling when co-transfected with LRP6 alone, whereas no effect was observed with frizzled8 or dishevelled(Fig. 4A and see Fig. S2 in the supplementary material). We conclude that Krms can promote LRP6-mediated signaling.

One possibility for how Krm may stimulate LRP6 signaling is by promoting its trafficking or stability. We therefore compared its effects to Mesd, an LRP6 chaperone, which also stimulates LRP6-mediated Wnt signaling(Culi and Mann, 2003; Hsieh et al., 2003)(Fig. 4A). As previously shown,Mesd is required for maturation of β-propeller/EGF modules of LDLR family members (Culi et al., 2004). Consequently, mesd transfection did not promote signaling stimulated by LRP6ΔE1-4, a constitutive active LRP6 construct with truncated extracellular domain (Fig. 4A). In these experiments, LRP6ΔE1-4 was intentionally transfected at low amounts to sensitize the reporter assay for synergistic effect. By contrast, krm1 does moderately stimulate LRP6ΔE1-4-mediated signaling (Fig. 4A). Co-transfection of krm1 and mesd shows a merely additive effect (Fig. 4B), suggesting that these genes do not functionally interact.

To test the specificity of the Wnt promoting effect of krm1, we analyzed a BMP responsive reporter and found it unaffected(Fig. 4C).

We next analyzed the effect of Xkrm1 and mesd on Wnt signaling in the presence of a very low amount of dkk1, which alone was insufficient to block LRP6 signaling. As reported(Mao et al., 2002; Li et al., 2005),co-transfection of dkk1 with Xkrm1, but not mesd,reduces the LRP6 signal to background levels(Fig. 4D). When Wnt1 was co-transfected additionally, the same result was observed.

Taken together, these data indicate a context-dependent role of Krms in Wnt signaling. Together with Dkk1, Krms inhibit Wnt/LRP6 signaling; however, in the absence of Dkk1, Krms promote LRP6 signaling.

We previously reported that Dkk1 binds to both Kremen and LRP6, thereby bridging the two receptors in a ternary complex, which is then removed from the cell surface (Mao et al.,2001; Mao et al.,2002). As our results indicate that Krms can also function without Dkk1, we tested whether they might bind directly to LRP6, in the absence of Dkk1. We used HEK293T cells, which express very low levels of dkk1(our unpublished observations). In co-immunoprecipitation experiments Krm1 and 2 were specifically precipitated with LRP6(Fig. 5A,A′, lane 2), but not with the control transmembrane proteins FLRT3 and LDLRΔC(Fig. 5A,A′, lanes 3,4). This was also the case in Co-IPs with added anti-Dkk1 antibody, to block any endogenous Dkk1 protein (C.M.C. and C.N., unpublished). To corroborate the directness of binding and to demonstrate extracellular interaction we also performed in vitro binding assays using secreted, recombinant proteins, which show that LRP6 co-precipitates with both Krm1 and 2, but not Dkk3(Fig. 5B). We conclude that Krm1 and 2 specifically and directly bind to LRP6.

To explore the possibility that Krm may influence protein expression or intracellular transport of LRP6, we performed co-transfection experiments. In western blot analysis of transfected HEK293T cell lysates, LRP6 was detected as an upper and a lower band, which are thought to correspond to the mature cell surface (ma), and immature cytoplasmic (im) forms of LRP6(Fig. 6A, lane 1)(Hsieh et al., 2003). We confirmed this by treatment with Endoglycosidase H (EndoH), which cleaves immature glycans that have not yet traversed the Golgi apparatus. Only the lower band was EndoH-sensitive and downshifted due to deglycosylation (dg)(Fig. 6A, lane 2).

Co-transfection of mesd increased mature LRP6, consistent with it being a reported chaperone (Culi and Mann,2003; Hsieh et al.,2003) (Fig. 6A,lanes 5,6). Co-transfection of krm2 had a very similar effect on LRP6. Mature LRP6 increased at the expense of the immature form(Fig. 6A, lanes 7,8), the total amount of LRP6 protein being mostly unaffected. Co-transfection of empty vector (Fig. 6A, lanes 1,2) or dkk3 (Fig. 6A, lanes 3,4), a gene not affecting the Wnt pathway(Niehrs, 2006), had no effect on LRP6 protein expression.

To corroborate this finding, we monitored plasma membrane levels of LRP6 by cell surface biotinylation. Following co-transfection with krm2, cell surface levels of LRP6 were increased, while the total LRP6 was mostly unaffected (Fig. 6B). This effect mimicked mesd co-transfection(Fig. 6B). The cytoplasmic protein Nucleoside diphosphate kinase A (NME1) serves as control and was not biotinylated (Fig. 6B′). These data indicate that Krm2 promotes cell-surface localization of the Wnt co-receptor LRP6.

Our embryological data indicate a requirement of Krm2 for NC induction, and the cell culture data suggest that this may be due to a role of Krm2 promoting LRP6 cell-surface localization and thus Wnt signaling. To corroborate that it acts on LRP6 in vivo, we analyzed by immunofluorescence microscopy whether Krm2 is required for the cell-surface localization of endogenous LRP6 in NC explants. We used an antibody against the intracellular LRP6 domain, which is likely to recognize both mature and immature forms of the protein(Davidson et al., 2005) (data not shown).

In NC cells, LRP6 was predominantly localized at the plasma membrane(Fig. 7A). This signal was abolished following LRP6 MO injection, attesting to the antibody specificity. Injections of Krm2 MO-1 or Krm2 MO-2 both reduced LRP6 levels in a concentration-dependent manner, without affecting β1-Integrin levels(Fig. 7A,B). Unlike in NC explants, in animal caps Krm2 MO injection did not affect LRP6 protein levels(Fig. 7A), indicating that Krm2 is specifically required to maintain LRP6 protein levels in NC.

The Wnt signaling pathway is reiteratively used during NC induction,delamination, migration and differentiation(Kalcheim and Burstyn-Cohen,2005; Raible and Ragland,2005; Taneyhill and Bronner-Fraser, 2005). During NC induction BMP and Wnt signaling are thought to act together to specify NC fate. In Xenopus,overexpression of various Wnts leads to NC induction(Saint-Jeannet et al., 1997; Chang and Hemmati-Brivanlou,1998; LaBonne and Bronner-Fraser, 1998). Conversely, inhibition of Wnt signaling,e.g. by dominant-negative versions of Wnt1/8, Tcf-3 and LRP6(Dorsky et al., 1998; LaBonne and Bronner-Fraser,1998; Tamai et al.,2000; Garcia-Castro et al.,2002), or MO-mediated depletion of Frizzled3/7 or β-catenin(Deardorff et al., 2001; Wu et al., 2005; Abu-Elmagd et al., 2006),blocks NC induction.

Krm2 is differentially expressed in the NC precursors and is required for NC formation. As Krm2 is both regulated by Wnts and promotes LRP6 activity,this suggests that in the context of NC induction Krm2 functions by promoting Wnt signaling. Consistent with this, depletion of LRP6 inhibited NC induction similarly to Krm2 knockdown. Incidentally, this is the first evidence indicating that LRP6 acts non-redundantly from LRP5 in NC formation in Xenopus.

The finding that anteriorly overexpressed Krm2 inhibited, while posterior overexpression enhanced, NC formation suggests a dual Krm2 action. Anteriorly overexpressed Krm2 probably cooperates with Dkk1 in Wnt inhibition, while in posterior regions Krm2 on its own promotes Wnt signaling. This dual activity is also supported by our cell culture experiments.

We cannot exclude the possibility that Krm2 may act also on other pathways that are involved in NC formation, e.g. BMP signaling. However, in cell culture reporter assays BMP signaling was unaffected by Krms.

Krm1 and 2 show differential expression at all stages of development both in Xenopus and mouse(Nakamura et al., 2001; Davidson et al., 2002), and analyses of their developmental role will need to take into account a possible positive action on Wnt signaling. Interestingly, based on EST expression data, KRM2 mRNA is significantly upregulated in several human cancers, e.g. brain, testis, kidney and gastrointestinal tract tumors(http://cgap.nci.nih.gov/),and it is a Wnt target (this study), raising the possibility that Krm2 may be a tumor associated gene or may even have an oncogenic (or tumor suppressive)role.

One surprising result of this study is that Krms not only act as negative,but also as positive, Wnt modulators, providing an explanation for their role in Wnt-mediated NC induction. Krm2 overexpression promoted cell surface localization of LRP6 in cultured cells; in embryos, Krm2 was required to maintain LRP6 protein levels in NC. Taken together, these data raise the possibility that Krm2 regulates LRP6 intracellular transport and turnover.

This is reminiscent of the LRP6 chaperone Mesd, which promotes LRP6 folding and cell surface localization (Culi and Mann, 2003; Hsieh et al.,2003) (this study). Do Krms then act as ER chaperones of LRP6?Three lines of evidence argue against this: (1) Unlike Mesd, Krms are not ER-resident (Culi and Mann,2003; Hsieh et al.,2003; Cruciat et al.,2006); (2) artificially ER-trapped Krm2 has a strongly reduced effect on LRP6 surface localization compared to wild type Krm2 (C.H. and C.N.,unpublished), indicating that a subcellular localization other than in the ER is required for Krm to exert its full effect on LRP6; (3) Krms can, albeit weakly, promote Wnt signaling induced by LRP6ΔE1-4 (see Fig. 4A), a construct lacking all four β-propeller/EGF regions, which are the target of Mesd(Culi et al., 2004). LRP6ΔE1-4, in contrast to full-length LRP6, is predominantly cell surface localized (C.H. and C.N., unpublished)(Cong et al., 2004), indicating that this construct does not require support in trafficking.

What, then, could be the mechanism of action of Krms? Krms may attenuate LRP6 endocytosis and degradation, thus promoting cell surface localization of LRP6. This would be the reverse of Krm action in the presence of Dkk1, which induces rapid LRP6 internalization (Mao et al., 2002). Krms may therefore be context-dependent endocytosis regulators of LRP6. Indeed, a hallmark of LDLR family members is their regulated endocytosis (Howell and Herz,2001; May et al.,2003; Schneider and Nimpf,2003).

We thank E. Karaulanov, M. Eilers, X. He, W. Wu and H. Steinbeisser for reagents, E. Karaulanov for critically reading the manuscript, and the Nikon Imaging Center at the University of Heidelberg for sharing their facility. C.H. was supported by a DFG fellowship (GRK 791).