Kremen proteins interact with Dickkopf1 to regulate anteroposterior CNS patterning

During early patterning of the vertebrate central nervous system (CNS),neural inducers and modifiers establish a crude anteroposterior (AP) pattern before and during gastrulation that becomes refined at later stages(Shimamura et al., 1995;Lumsden and Krumlauf, 1996;Sasai and de Robertis, 1997;Chang and Hemmati-Brivanlou,1998; Stern, 2001;Chapman et al., 2002). Instructive signalling through the FGF(Hongo et al., 1999;Shanmugalingam et al., 2000;Streit et al., 2000;Wilson et al., 2000;Shinya et al., 2001) and IGF(Pera et al., 2001;Richard-Parpaillon et al.,2002) pathways are involved in anterior neural specification. Furthermore, inhibition of BMP and nodal signalling is a prerequisite for anterior neural induction, and this antagonism is mediated by secreted anti-BMPs (noggin, chordin and follistatin)(Smith and Harland, 1992;Hemmati-Brivanlou et al.,1994; Sasai et al.,1994; Bachiller et al.,2000) (reviewed by Harland and Gerhart, 1997; Streit and Stern, 1999) and anti-nodals (antivin, Cerberus and lefty)(Fainsod et al., 1997;Meno et al., 1999;Piccolo et al., 1999;Thisse and Thisse, 1999;Thisse et al., 2000) (reviewed by Schier and Shen, 2000). These factors are emitted from the Spemann organiser of amphibia and its equivalents in other vertebrates. As for anterior CNS formation, posterior CNS formation also depends on BMP inhibition, but in addition requires posteriorising signals such as FGF, retinoic acid and Wnt(Gilbert and Saxen, 1993;Doniach and Musci, 1995;McGrew et al., 1995;Nieuwkoop, 1997;Sasai and de Robertis, 1997;Dupe and Lumsden, 2001).

We and others recently showed that a gradient of Wnt/β-catenin signalling regulates anteroposterior (AP) patterning of the entire neural plate during Xenopus gastrulation(Kiecker and Niehrs, 2001;Nordstrom et al., 2002). This gradient is high in posterior and low in anterior regions of the embryo, a likely consequence of Wnt and Wnt inhibitor expression domains being predominantly posterior and anterior, respectively. The anterior source of secreted Wnt antagonists is formed in anterior endomesoderm with the expression of cerberus(Bouwmeester et al., 1996),sFRPs (Leyns et al.,1997; Rattner et al.,1997; Wang et al.,1997) and dkk1(Glinka et al., 1998). Indeed,a distinguishing feature of organising centres involved in anterior neural induction in vertebrates is their expression of Wnt antagonists. Three lines of evidence support the theory that Wnt antagonism plays a central role in anterior specification: (1) co-expression of Wnt and BMP antagonists induces ectopic heads including anterior CNS while BMP antagonists alone induce only trunk structures (Glinka et al.,1997; Glinka et al.,1998); (2) overexpression of Wnt inhibitors in Xenopusand zebrafish embryos induces enlarged heads and forebrain(Itoh et al., 1995;Hoppler et al., 1996;Pierce and Kimelman, 1996;Glinka et al., 1997;Leyns et al., 1997;Wang et al., 1997;Deardorff et al., 1998;Glinka et al., 1998;Hsieh et al., 1999;Fekany-Lee et al., 2000;Hashimoto et al., 2000;Heasman et al., 2000;Shinya et al., 2000); (3) in loss-of-function studies Wnt inhibitors were shown to be necessary for formation of anterior neural structures. Inactivation of the secreted Wnt antagonist Dickkopf1 (Dkk1) in Xenopus embryos using neutralising antibodies (Glinka et al.,1998; Kazanskaya et al.,2000) or targeted deletion of the dkk1 gene in mouse(Mukhopadhyay et al., 2001),as well as inactivation of the intracellular Wnt pathway inhibitorstcf3 and axin1 in the zebrafish headless andmasterblind mutants, respectively(Kim et al., 2000;Heisenberg et al., 2001;van de Water et al., 2001),all result in microcephalic embryos.

As regulation of Wnt/β-catenin signalling is crucial for AP neural patterning, it is important to understand the regulatory network interacting with Wnt antagonists, because it will have a bearing on the AP patterning process. We focus on the regulation of the Wnt antagonist Dickkopf1 (Dkk1),member of a multigene family of secreted glycoproteins with at least four different members in human (Glinka et al.,1998; Krupnik et al.,1999). Dkk1 is expressed in the Spemann organiser and the presumptive prechordal plate and acts as a head inducer during vertebrate gastrulation (Glinka et al.,1998; Hashimoto et al.,2000; Kazanskaya et al.,2000; Shinya et al.,2000; Mukhopadhyay et al.,2001). The mechanism of Dkk1 action is unlike that of other extracellular Wnt inhibitors belonging to the sFRP(Leyns et al., 1997;Rattner et al., 1997;Wang et al., 1997), WIF(Hsieh et al., 1999) and Cerberus (Glinka et al., 1997;Piccolo et al., 1999) class,which directly bind and inactivate Wnt proteins. Dkk1 neither interacts with Wnts nor affects Wnt/Fz interactions. Instead, it binds as a high affinity antagonist to Wnt receptors of the lipoprotein receptor-related protein (LRP)5 and 6 class (Bafico et al.,2001; Mao et al.,2001; Semenov et al.,2001). Owing to its mechanism of action(Wehrli et al., 2000), Dkk1 is a selective inhibitor of the Wnt/β-catenin pathway, and it does not affect the Wnt/planar cell polarity (PCP) pathway that drives convergent extension movements in Xenopus(Semenov et al., 2001).

In addition to LRPs, Dkk1 interacts with another recently identified receptor class, the transmembrane proteins Krm1 and Krm2, to synergistically inhibit LRP6 (Mao et al.,2002). Mouse Krm1 (Kremen) had previously been identified as a differentially expressed gene without known function(Nakamura et al., 2001). Murine Krm proteins strongly cooperate with Dkk1 to inhibit Wnt signalling both in the mammalian cell line HEK 293 as well as in Drosophilawings, when expressed as heterologous transgenes. Upon binding to Dkk1, Krm proteins are recruited into a complex with LRP6, which leads to rapid endocytosis and removal of this Wnt receptor from the plasma membrane(Mao et al., 2002).

While this suggests that Krm proteins function in Dkk1-mediated Wnt inhibition, it is unknown what role these transmembrane receptors play physiologically, e.g. during embryogenesis. To investigate the physiological relevance of their interaction with Dkk1 and to study their role during embryogenesis, we have cloned and functionally characterised theXenopus homologues of krm1 and krm2. We show that Krm proteins functionally interact with Dkk1 during Wnt inhibition inXenopus embryos and that they are required for formation of the anterior CNS.

In vitro fertilisation, embryo culture, staging, microinjection, culture of explants and whole-mount in situ hybridisation of Xenopus embryos were carried out as described (Gawantka et al., 1995). For vibratome sectioning, embryos were placed in embedding medium (0.4% gelatine, 30% albumin, 20% sucrose in PBS), and, after hardening in presence of 2% gluteraldehyde, sectioned using a VT100E vibratome(Leica). Whole-mount in situ hybridisation of mouse embryos was performed according to previously described procedures(Koop et al., 1996).

Xenopus krm1 and krm2 cDNA fragments were amplified by RT-PCR using degenerate oligos (forward, AATGGNGCNGAYTAYMGAGG; reverse,CCRCARAARCANGCRTAWCC) and mRNA from stage 18 Xenopus embryos. Two 400 bp krm1 and krm2 fragments were obtained and used as probes to obtain full-length krm cDNAs (Accession Number, AY150813). These were subcloned into pCS2+ to obtain pCS-Xkrm1 and pCS-Xkrm2. C-terminal hemaglutinin-(HA) and V5-tagged pCS-Xkrm1HA and pCS-Xkrm2V5, as well as N-terminal V5-tagged pCS-V5Xkrm2were created by PCR.

The 5′ nucleotide sequences of additional (pseudo-) alleles for bothXenopus krm1 and krm2 genes were obtained using 5′RACE (GeneRacer kit, Invitrogen). Based on these sequences, antisense oligonucleotides with optimal complementary to both alleles around the ATG start codon were designed: krm2, ACCACAGCATCTCCACCAACATTGT;krm1, TGAAATTGTCCAAATATCCATCACC.

Preparation of mRNA for Xenopus injections was carried out using the MegaScript in vitro transcription kit (Ambion), according to manufacturer's instructions. For western blot immunological detection of tagged Krm proteins, either anti-hemagglutinin (HA) (1:10,000, Roche) or anti-V5 (1:10,000, Invitrogen) monoclonal antibodies were used. Chemiluminescence detection (SuperSignal® solution, Pierce) was carried out according to manufacturer's instructions after incubation of blots with anti-mouse IgG-HRP (1:10,000, Pierce).

RT-PCR assays were carried out in the linear phase of amplification and with primers as described (Glinka et al.,1997). Other primers were: mouse Krm1 (forward,GTGCTTCACAGCCAACGGTGCA; reverse, ACGTAGCACCAAGGGCTCACGT); mouse Krm2(forwards, AGGGAAACTGGTCGGCTC; reverse, AAGGCACGGAGTAGGTTGC); Xenopus krm1 (forward, CACTAGATGGTGGGAAGCCTTGC; reverse, CCTCCAGCCCAGCTAGCTTGT);and Xenopus krm2 (forward, CCCGACAATGTTGGTGGAGATGC; reverse,GGTGCCTACGTCTGATGGATCGC).

Full-length cDNA clones of Xenopus krm1 and krm2 were isolated from stage 31 head cDNA libraries. The genes are predicted to encode proteins of 452 amino acids/50 kDa (Xkrm1) and 421 amino acids/46 kDa(Xkrm2). Like their mouse homologues, they contain an extracellular kringle-, WSC- and CUB domain, followed by a transmembrane domain and a cytoplasmic domain (Fig. 1A). Kringles are autonomous structural domains, found predominantly in blood clotting and fibrinolytic proteins (Patthy et al., 1984; Castellino and McCance, 1997), as well as some other serine proteases(Kurosky et al., 1980;Thery and Stern, 1996;Gschwend et al., 1997). Kringle domains are believed to play a role in binding mediators such as membrane phospholipids (Church et al.,1991) and proteoglycans(Goretzki et al., 2000). WSC domains are present in yeast cell wall integrity and stress response component proteins (Lodder et al.,1999). The CUB domain is involved in protein-protein and glycosaminoglycan-protein interactions, and is found in a number of proteins involved in development and differentiation(Bork and Beckmann, 1993). Although amino acid sequence homologies between vertebrate Krm1 and Krm2 are only between 35-40%, both the occurrence and the order of their domains is conserved in all orthologues. Krm1 homologues are more closely related to each other than are Krm2 homologues (Fig. 1B). The cytoplasmic domains show no homologies to known structural motifs, but there is a high degree of sequence conservation within these domains between Krm1 homologues.

In the mouse, both Krm1 and Krm2 are expressed in a variety of adult tissues, particularly in heart, eye and reproductive organs as seen by RT-PCR (Fig. 2A).Krm1 and Krm2 transcripts were first detected at embryonic day (E) 8 by RT-PCR (data not shown). At this time, which corresponds to early headfold stages, Krm1 (Fig. 2B) and Krm2 (similar to Krm1 and therefore not shown) are expressed predominantly in the early anterior neural ridge(arrowheads). Krm1 and Krm2 show differential expression in various neural and mesodermal derivatives in midgestation mouse embryos(Fig. 2B)(Nakamura et al., 2001). At E10.5, prominent co-expression of Krm genes is evident in the branchial arches, the apical epidermal ridge (AER) of limb buds and nasal placode, with lower level co-expression seen in somites(Fig. 2C). Additional expression is seen in the forebrain and otic and optic vesicles(Krm1) and mesonephros (Krm2)(Fig. 2C).

In Xenopus, RT-PCR analysis shows that Krm mRNAs are present throughout embryogenesis, because of both maternal contribution(krm1) and zygotic expression(Fig. 3A). Zygotic expression starts at early (krm2) and late gastrula (krm1) and remains relatively constant throughout neurulation and organogenesis(Fig. 3A).

By whole-mount in situ hybridisation, krm2 expression is observed in the gastrula marginal zone (both deep and superficial, not shown) with exception of the Spemann organiser (Fig. 3G). At early neurula stage, krm2 expression is seen in two longitudinal stripes along the lateral neural plate(Fig. 3H). Longitudinal(Fig. 3I) and sagittal(Fig. 3J) sections at these stages show staining in anterior mesoderm, but not anterior neuroectoderm. In mid neurulae neural tubes, expression is seen in the dorsal midline as well as two longitudinal stripes, and in sagittal sections prominent expression is observed in the prechordal plate (Fig. 3K,L). In tailbud embryos expression is seen in hatching gland,branchial arch, dorsal otic vesicle, fin mesenchyme and pronephric duct(Fig. 3M).

krm1 is first detected by in situ hybridisation at neurula stages,when it shows a similar expression pattern to krm2(Fig. 3C-E). Staining in sections of early neurulae are similar to those shown for krm2(Fig. 3I,J) and therefore not shown. At tailbud stage, as for krm2, expression is seen in hatching gland and fin mesenchyme, but krm1 shows additionally expression in notochord and weakly in somites (Fig. 3F).

We previously showed that dkk1 and Krms synergise to inhibit Wnt signalling in HEK 293 cells and in the Drosophila wing(Mao et al., 2002). To test if they also functionally interact in Xenopus embryos, we carried out axis duplication assays with dkk1 and krm. In these assays,Wnt signalling is read during a period when both endogenous Krm and LRP6 are present (because of maternal contribution), but when Dkk1 is absent.Xwnt8 mRNA injection induces about 60% secondary embryonic axes and this is effectively inhibited by co-injection of dkk1 mRNA[Fig. 4A,B,G (columns 1 and 2)], but not krm1 or krm2 (data not shown). It is thus unlikely that Krm can function without Dkk1. In contrast to its inhibition ofWnt8-induced axis duplication, dkk1 fails to inhibitXwnt8/Lrp6 induced axis duplication[Fig. 4C,D,G (column 3)]. However, co-injection of krm2 and dkk1 mRNAs, but notkrm2 alone, leads to complete inhibition ofWnt8/Lrp6-induced axis duplication[Fig. 4E,F,G (columns 4 and 5)]. We conclude that Dkk1 and Krm proteins can functionally synergise during inhibition of Wnt signalling in Xenopus embryos.

A hallmark of Dkk1 is its ability to induce enlarged head structures inXenopus embryos when overexpressed. This is because Dkk1 functions to induce head formation by interfering with posteriorising Wnt signals during gastrulation (Niehrs, 1999). Thus, if Dkk1 acts via Krm to affect Wnt signalling, then Krm overexpression itself may mimic the effects of Dkk1. To test this, krm2 mRNA was microinjected into four to eight-cell stage embryos. This resulted in anteriorised embryos, with large heads and cement glands, similar to what is observed after dkk1 overexpression(Fig. 4H-J). Consistent with the anteriorised phenotype, animal caps from krm2 mRNA injected embryos, like those injected with dkk1 mRNA, show upregulation of the anterior neural markers otx2(Blitz and Cho, 1995) andXAG1 (Sive et al.,1989), and the pan neural marker NCAM(Tonissen and Krieg, 1993)(Fig. 4K). To test if this anteriorisation is due to inhibition of posteriorising Wnt signalling, embryos were microinjected with pCSKA-Xwnt8 DNA, which induces microcephalic embryos, lacking eyes and cement gland(Christian and Moon, 1993)(Fig. 4L,M). When pCSKA-Xwnt8 is co-injected with krm2, normal head formation is restored (Fig. 4N). Thus,similar to dkk1, krm2 overexpression dorsoanteriorises earlyXenopus embryos and it does so by inhibiting posteriorising Wnt signals.

If Krm proteins act as receptors for Dkk1 to inhibit Wnt/LRP signalling inXenopus, one would expect excess Krm to compensate for a reduction in Dkk1 activity. This is indeed the case, as shown by the ability of injectedkrm2 mRNA to rescue embryos posteriorised by inhibitory anti-Dkk1 antibody (Fig. 4O-Q).krm1 mRNA also rescues this Dkk1 loss-of-function phenotype (data not shown). In the reverse situation, dkk1 overexpression shows partial rescue of the phenotype elicited by krm1/2 antisense morpholino (Mo)(see below) injected embryos (data not shown).

Dkk1 is essential for formation of the anterior CNS, both inXenopus and mouse (Glinka et al.,1998; Mukhopadhyay et al.,2001). To test if Krms are likewise required for Xenopusanterior CNS development we first injected mRNA encoding a soluble form of Krm2, containing all extracellular domains but lacking transmembrane and intracellular regions, as we reasoned it might function as a dominant negative. However, this was not the case (data not shown), suggesting that membrane attachment of Krm proteins is important for mediating Wnt/LRP inhibition by Dkk1.

We therefore injected morpholino-antisense oligonucleotides (Mo), which function as specific translational inhibitors in both Xenopus and zebrafish embryos (Heasman et al.,2000; Nasevicius and Ekker,2000; Heasman,2002). When co-injected into Xenopus embryos,krm1 and krm2 Mo specifically inhibited translation of their cognate mRNA target without affecting translation of the respective orthologue(Fig. 5A). Phenotypically,krm1-Mo injection has no effect, and krm2-Mo injection yields mild anterior defects (not shown). However, as krm1 andkrm2 are co-expressed during early Xenopus embryogenesis,they may function redundantly. Indeed, co-injection of krm1 andkrm2-Mo (krm1/2-Mo) results in microcephaly(Fig. 5B,C). In addition, axial malformations such as bent and shortened trunks were observed. These defects could be partially rescued by co-injection of plasmid DNA encoding N-terminally modified krm2 DNA (lacking the antisense-Mo target sequence; Fig. 5D), indicating that the phenotype was specific. Krm1/2-Mo injected embryos showed reduced expression of the forebrain marker bf1(Bourguignon et al., 1998)(Fig. 5E-H), but the midbrain marker en2 appeared normal (Fig. 5G,H).

If Krms function downstream of Dkk1 to inhibit Wnt signalling during head induction, krm1/2-Mo should enhance the phenotype produced by a reduction of Dkk1 activity. This is indeed the case(Fig. 5I-P). Co-injection of limiting amounts of an inhibitory anti-Dkk1 antibody(Glinka et al., 1998) withkrm1/2-Mo results in embryos with head defects far more severe than seen in either krm1/2-Mo or anti-Dkk1 antibody-injected embryos. In situ hybridisation for bf1 shows that reduction of prospective forebrain territory in these embryos at neurula stage parallels the phenotypic deficiencies (Fig. 5Q-T). Together, these data indicate that Krm proteins functionally interact with Dkk1 to inhibit posteriorising Wnts during embryonic head development.

The head inducer Dkk1 functions as a secreted Wnt inhibitor that binds to and blocks signalling by the Wnt receptor LRP5/6(Bafico et al., 2001;Mao et al., 2001;Semenov et al., 2001), thereby antagonising posteriorising Wnts during patterning of the vertebrate AP axis. However, the recent finding that the transmembrane proteins Krm1 and 2 can bind to Dkk1 (Mao et al.,2002) raised the possibility that Dkk1 alone is not sufficient to inhibit Wnt/LRP signalling. Although their functional interaction in cell culture, as well as in Drosophila, is consistent with Krm proteins enhancing the Wnt inhibitory effect of Dkk1, it is not known if they are physiologically required for the action of Dkk1 and, if so, how they contribute to AP patterning.

To address these questions, we have isolated and characterisedXenopus Krm genes, and have shown that they functionally interact with dkk1 in vivo. Furthermore we provide evidence that this interaction is required for the formation of the anterior CNS. In axis duplication assays krm2 synergises with dkk1 in inhibiting Wnt/LRP6 signalling. In these experiments, the inability of dkk1 to inhibit Wnt/LRP6 signalling is overcome by co-expression of krm2,suggesting that endogenous Krm proteins become limiting. By themselves, Krm1 and Krm2 are unable to inhibit Wnt signalling in these assays, when no endogenous Dkk1 is present. This suggests that Krm and Dkk1 are required equally to block Wnt/LRP signalling. This conclusion is also supported by the findings that both the Dkk1 antibody and krm1/2 Mo phenotypes can be rescued by overexpressed krm and dkk1, respectively. Furthermore, the synergistic effect of combined Krm and Dkk1 loss-of-function indicate that Krm proteins are physiologically relevant receptors that mediate Dkk1 inhibition of Wnt/β-catenin signalling.

Krm1 and Krm2 are thus transmembrane inhibitors of a Wnt/β-catenin signalling pathway, the significance of which in vertebrate AP patterning is emerging. Highlighting the role of Wnt/β-catenin signalling in posterior specification, inactivation of members of the Wnt1 class of ligands, both in the mouse (Takada et al.,1994) and zebrafish (Erter et al., 2001; Levken et al.,2001), as well as of the murine Wnt/β-catenin specific transducer Lrp6 (Pinson et al.,2000), leads to posterior defects. In addition, combined loss of the murine Tcf1 and Lef1 transcription factor genes, which mediate downstream Wnt/β-catenin target gene activation, results in similar posterior deficiencies (Galceran et al., 1999). In contrast to mutations in stimulatory components of the Wnt/β-catenin pathway, mutations in both zebrafish axin(masterblind) and tcf3 (headless) genes, which encode intracellular inhibitory components of the Wnt/β-catenin pathway,result in anterior neural deficiencies limited to the forebrain and its derivatives (Kim et al., 2000;Heisenberg et al., 2001;van de Water et al., 2001). The latter mutant phenotypes are similar to the loss-of-function of Dkk1 in frog and mouse (Glinka et al.,1998; Mukhopadhyay et al.,2001). Thus, studies from loss-of-function mutations provides substantial genetic evidence to support a conserved and essential role for Wnt/β-catenin signalling in vertebrate AP patterning and inhibition of this pathway during anterior CNS formation. The data presented here are consistent with the proposal that Krm proteins are required co-receptors for Dkk1 to inhibit Wnt/LRP signalling and apparently independent of Dishevelled(Dsh) (Li et al., 2002),thereby promoting anterior CNS formation(Fig. 6).

krm1 is expressed maternally, and in both mouse and frog it is expressed in early anterior neural folds. Furthermore, both Xenopus krm1 and krm2 are co-expressed with dkk1 in the prechordal plate underlying the anterior neurectoderm. These expression domains are consistent with a role of Krm proteins during early anterior development. Maternal krm1 mRNA suggests that there is also maternal protein which would not be affected by morpholino knockdown. Hence, the observed phenotype may be hypomorphic. Similar to the loss-of-function of Dkk1 in frog and mouse (Glinka et al.,1998; Mukhopadhyay et al.,2001), the observed neural deficiencies are limited to the forebrain and its derivatives, as in cases of mutations of intracellular Wnt inhibitors (Kim et al., 2000;Heisenberg et al., 2001;van de Water et al., 2001). Conversely, zebrafish wnt8 mutants show expanded forebrain(Levken et al., 2001),indicating that this region of the CNS is most sensitive to Wnt/β-catenin signalling, while more posterior CNS structures may become affected only in compound mutants of Wnt inhibitors.

At tadpole stages, dkk1, krm1 and krm2 show complex expression patterns, with co-expression observed in the otic vesicle, fin mesenchyme and proctodeum, where the genes may interact. During mouse organogenesis, dkk1, krm1 and krm2 are co-expressed in the apical ectodermal ridge of mouse limb buds(Monaghan et al., 1999;Nakamura et al., 2001) (and results presented here). As Dkk1 is required for limb formation(Mukhopadhyay et al., 2001),it may interact with Krm proteins in this context to inhibit the Wnt receptor LRP6, which is ubiquitously expressed during embryogenesis(Pinson et al., 2000). However, although there are several sites of dkk1/krm co-expression,it cannot be excluded that Dkk1 can also function independently of Krm1 and Krm2, e.g. by recruiting yet unknown co-receptors. Likewise, their multidomain ECDs and the intracellular domain raise the likely possibility that Krm proteins have functions in addition to mediating Dkk1 action. The prominent expression, e.g. in trunk mesoderm, where Dkk genes are not expressed and mild trunk defects observed following morpholino knockdown would be consistent with such additional functions. One other potentially relevant, high-affinity ligand for Krm1 and Krm2 is Dkk2, which is co-expressed with krm1 in branchial arch, otic and optic vesicles and limb bud(Monaghan et al., 1999;Wu et al., 2000;Nakamura et al., 2001).

We thank Dana Hoppe for technical assistance, Marlene Rau for help with in situ hybridisation and Ralf Klaeren for help with sequencing.