Zinc-dependent Interaction between Dishevelled and the Drosophila Wnt Antagonist Naked Cuticle*

During Drosophila development, thenaked cuticle (nkd) gene attenuates wingless/Wnt signaling through a negative feedback loop mechanism. Fly and vertebrate Nkd proteins contain a putative calcium-binding EF-hand motif, the EFX domain, that interacts with the basic/PDZ region of the Wnt signal transducer, dishevelled (Dsh). Here we show that Dsh binding by Drosophila Nkd in vitro is mediated by the EFX domain as well as an adjacent C-terminal sequence. In vivo data suggest that both of these regions contribute to the ability of Nkd to antagonize Wnt signaling. Mutations in the Nkd EF-hand designed to eliminate potential ion binding affected Nkd-Dsh interactions in the yeast two-hybrid assay but not in the glutathioneS-transferase pull-down assay. Addition of the chelating agent EDTA abolished the in vitro Nkd-Dsh interaction. Surprisingly zinc, but not calcium, was able to restore Nkd-Dsh binding, suggesting a zinc-mediated interaction. Calcium 45- and zinc 65-blotting experiments show that Nkd is a zinc-binding metalloprotein. The results further clarify how Nkd may antagonize Wnt signaling via interaction with Dsh, and identify a novel zinc-binding domain in Drosophila Nkd that collaborates with the conserved EFX domain to bind Dsh.

The Wg 1 signal transduction pathway regulates pattern formation during Drosophila development by controlling cell fate determination and cell proliferation (reviewed in Ref. 1). Its counterpart in vertebrates, the Wnt/␤-catenin signaling pathway, is critical in many aspects of development. Inappropriate activation of Wnt-induced genes that are normally precisely regulated can lead to diverse types of human cancer (reviewed in Ref. 2). The mechanisms by which Wnt signals are held in check are therefore critical both to normal development and to human disease.
Genetic and molecular analyses have identified evolutionarily conserved proteins that transmit the Wnt/␤-catenin signal. Upon binding of Wnt to frizzled family receptors, Dsh or its mammalian homologs (Dvl) become hyperphosphorylated (3,4). In cells that are not exposed to Wnt, a multiprotein complex promotes phosphorylation and degradation of armadillo (Arm)/ ␤-catenin via the ubiquitin-proteasome pathway (5). The complex includes the proteins Zw3/GSK3, adenomatous polyposis coli, axin, and Arm/␤-catenin. Upon Wnt reception, Dsh transmits a signal that inhibits the activity of the complex, thereby allowing hypophosphorylated Arm/␤-catenin to accumulate by escaping the degradation machinery. Arm/␤-catenin then translocates to the nucleus, where it activates transcription of Wnt target genes through its association with transcription factors of the Lef/Tcf family (6,7).
In addition to Wnt/␤-catenin signaling, Dsh functions in a distinct signal transduction pathway that mediates planar cell polarity (8). This pathway coordinates epithelial morphogenesis in fly and vertebrate development, and involves proteins downstream of Dsh that are distinct from the Wnt/␤-catenin pathway proteins. Several Dsh-binding proteins have been identified. In Wnt/␤-catenin signaling, Dsh/Dvl was proposed to contact the Arm/␤-catenin degradation complex via a direct association with two GSK3-interacting proteins, axin and Frat1 (9). Upon exposure to Wnt, the Dsh⅐axin⅐Frat⅐GSK3 complex dissociates, which may cause Arm/␤-catenin accumulation (9). Biochemical purification of Dsh kinases from Drosophila cell extracts led to the identification of CKII and PAR-1 as additional binding partners of Dsh (10,11). CKII also interacts with the Dvl proteins in mammalian cells (12). CKII and PAR-1 are able to phosphorylate Dsh/Dvl. Studies using Xenopus and Caenorhabditis elegans showed that the CKI family of protein kinases are positive effectors of Wnt signaling (13)(14)(15). They can phosphorylate Dsh and directly interact with the Dsh PDZ domain. Three additional proteins, PP2C␣, Idax, and Stbm, associate with the PDZ domain of Dvl (16 -18). PP2C␣ also interacts with axin and activates Wnt signaling by dephosphorylating axin (16). In contrast, Idax and Stbm are negative regulators of the Wnt/␤-catenin pathway (17,18).
The roles of Wg and its signal transduction system have been particularly well studied in the context of Drosophila embryonic segmentation (1). We previously reported the identification of the segment polarity gene nkd and showed that Nkd acts in a negative feedback loop to restrict Wg activity in the embryo (19). Nkd acts at the level of Dsh through a direct interaction with the central region of Dsh that includes a sequence rich in basic residues and the adjacent PDZ domain (20).
Nkd is a novel protein that includes a single putative Ca 2ϩ binding motif of the EF-hand family. The activities of many EF-hand proteins are sensitive to, and regulated by, changes in intracellular Ca 2ϩ concentration (reviewed in Ref. 21). We and others recently described vertebrate EF-hand-containing proteins related to Drosophila Nkd that also bind Dsh/Dvl proteins and can block Wnt signals (22,23). The conservation of the EF-hand motif in fly and vertebrate Nkd proteins suggests a role for Ca 2ϩ or other divalent cations in Nkd-mediated regulation of Wnt pathway activity. To test this hypothesis, we defined domains in Drosophila Nkd responsible for binding Dsh and antagonizing Wg signaling, and we investigated the possible role of divalent cations in the interaction between the two proteins.
Yeast Two-hybrid Assay-Transformation of the yeast strain PJ69-4A (25) was performed using a variation of the lithium acetate method (Clontech). Yeast growth assay was achieved using the ADE reporter gene present in the strain and medium containing or lacking adenine. In this system, an interaction between two protein domains being tested is indicated by growth in the absence of adenine because of the expression of the reporter gene. As a control, growth was also tested in the presence of adenine. Growth was evaluated between 3 and 4 days at 30°C. ␤-Galactosidase activity (using the LacZ reporter gene also present in the strain) was measured as described by Clontech.
GST Pull-down Assay-Bacterial lysates containing the GST fusion proteins were prepared as described by Amersham Biosciences, except MTPBS buffer (150 mM NaCl, 12.5 mM Na 2 HPO 4 , 2.5 mM KH 2 PO 4 ), containing 1 mM Pefabloc SC, 5 mM dithiothreitol, and an antiprotease mixture, was used instead of phosphate-buffered saline. The lysates were bound to glutathione-Sepharose 4B beads for 30 min at room temperature and washed four times, twice with MTPBS, 1% Triton X-100 buffer and twice with DT80 buffer (20 mM Tris-HCl, pH 8, 80 mM KCl, 0.25% Triton X-100). The beads were then incubated for 2 h at 4°C in the latter buffer (containing 1 mM Pefabloc SC and 1 mM dithiothreitol) with [ 35 S]methionine-labeled Nkd or Dsh proteins, which were produced using the TNT T7 coupled reticulocyte lysate system (Promega). The beads were washed four times with DT buffer (20 mM Tris-HCl, pH 8, 0.25% Triton X-100, containing 80, 150, or 300 mM KCl) and incubated with SDS-PAGE-loading buffer to elute the proteins. Samples were run on SDS protein gels, which were then dried and exposed on BioMax film (Kodak) or a PhosphorScreen (Amersham Biosciences).
For the experiments corresponding to Figs. 3 and 4, GST pull-down assays were performed as described above, with the following modifications. In Fig. 3, EDTA alone, EGTA alone, EDTA ϩ MgCl 2 , or EDTA ϩ LiCl were added to each buffer at the indicated concentrations. For the experiments in Fig. 4, GST-Nkd R1S, once bound to glutathione-Sepharose 4B beads, was washed twice for 5 min with MTPBS, 1% Triton X-100 buffer containing 100 mM EDTA and twice with DT80 buffer containing ZnCl 2 , CaCl 2 , or MgCl 2 at the indicated concentra-tions. The subsequent steps (incubation with Dsh and final washes) were also performed in the presence of ZnCl 2 , CaCl 2 , or MgCl 2 at the same concentrations. Calculation of K d values and free concentration of EDTA was performed using WEBMAXC v2.10 program (www. stanford.edu/ϳcpatton/maxc.html) with the following values: temperature ϭ 4°C, pH 8.0, and ionic strength ϭ 0.1.
nkd Rescue and Overexpression Assays-Fly transformations and culture were performed according to standard methods. For the rescue assay, a third chromosome P(mini wϩ, da-Gal4) insert was recombined onto a nkd 7E89 chromosome and balanced over TM3. For each experiment in Table I, da-Gal4, nkd 7E89 /TM3 females were crossed to UAS-X/UAS-X; nkd 7H16 /TM3 (where X is each mutant Nkd protein) males at 25°C, and embryos were scored according to previously described criteria (19). Constructs NIN3 (aa 1-447), NBam, and EFX were tagged with a C-terminal GFP to determine relative accumulation in tissues. Cuticle preparations were performed as previously described (19) and images were collected by darkfield microscopy on a Zeiss Axioplan 2 microscope. For the overexpression assay, B119-Gal4 (on II) or A8-Gal4 (on X) females were crossed to chromosome II or III UAS-X/UAS-X or balancer males at 29°C, and progeny were scored on a Leica MZ12.5 dissecting scope and photographed using a Nikon Coolpix camera. For sternite bristle counts, all bristles in abdominal segments A2-A6 of gravid females were counted. 45 Ca-blotting Experiments-The 45 Ca-blotting assay was performed as described (26). Five g of GST fusion proteins, purified as specified above, were loaded on a SDS-polyacrylamide gel and transferred without SDS onto a nitrocellulose membrane for 1 h 15 min at 120 V (4°C). As a loading control, an identical gel was run in parallel and stained with Coomassie Blue. The membrane was washed for 1 h at room temperature with KMI buffer (60 mM KCl, 5 mM MgCl 2 , 10 mM imidazole-HCl, pH 6.8). The incubation with 45 CaCl 2 (PerkinElmer Life Sciences) was then performed in the same buffer for 10 min at a concentration of 1 Ci/ml in 10 ml. The membrane was washed for 5 min in distilled water, dried, and exposed on BioMax film (Kodak). For the 45 Ca-blotting assay performed in the presence of Zn 2ϩ (data not shown), 50 or 500 M ZnCl 2 was added to KMI buffer.
Zn-blotting Experiments-We performed the 65 Zn-blotting assay as previously reported (27). Five g of GST fusion proteins were purified using glutathione-Sepharose 4B beads, run on SDS-polyacrylamide gels, and incubated for 1 h at 37°C in transfer buffer (25 mM Tris-HCl, pH 8.3, 192 mM glycine, 0.1% SDS) containing 5% ␤-mercaptoethanol. The reduced proteins were transferred onto a nitrocellulose membrane for 1 h at 100 V (4°C). The blot was washed for 1 h with 10 mM Tris-HCl, pH 7.5, and subsequently incubated for 15 min in 10 ml of TK buffer (10 mM Tris-HCl, pH 7.5, 100 mM KCl) containing 5 Ci of ZnCl 2 . The membrane was then washed twice for 15 min with TK buffer, dried, and exposed on BioMax film (Kodak).

Determination of Dsh-interacting Domains in Drosophila
Nkd-Dsh contains three conserved domains: DIX (also present in axin), PDZ, and DEP. Nkd interacts with the central region of Dsh that includes a basic sequence and the PDZ domain (20). In Nkd, the EFX domain, a 66-amino acid region conserved between fly and vertebrate Nkd is sufficient to bind Dsh in the yeast two-hybrid assay ( Fig. 1A; Ref. 22). To further investigate the role of the EF-hand, we tested Drosophila Nkd deletion mutants in two different Dsh-binding assays, the yeast two-hybrid and the GST pull-down. Mutant Nkd proteins that lack the EF-hand do not interact with Dsh (constructs BamC, NR1, and X), or only interact weakly (construct XS), indicating that the putative Ca 2ϩ binding motif of Nkd is important for the association (Fig. 1). However, the EF-hand alone (construct EF) does not bind Dsh (Fig. 1). In the GST pull-down assay, the interaction between the Nkd EFX domain and Dsh is not as robust as the interaction between full-length Nkd and Dsh (Fig.  1B), indicating that regions in Nkd other than EFX may contribute to the association with Dsh. The XS fragment by itself binds weakly to Dsh, but the interaction becomes as strong as the full-length Nkd protein when fragment XS is adjacent to the EF-hand (R1S construct; Fig. 1B). Thus the EF-hand and the XS fragment cooperate in the full-strength association with Dsh. The R1S domain is also able to bind the central basic/PDZ region of Dsh (data not shown).

The EF and XS Regions Contribute to Nkd Function in
Vivo-To determine which regions of Nkd are important for biological activity, we constructed transgenic flies capable of misexpressing mutant Nkd proteins. The da-Gal4 driver was used to induce ubiquitous expression in strong class nkd 7H16 / nkd 7E89 embryos. Misexpression of full-length nkd cDNA rescued all or nearly all mutant cuticles to wild-type morphology ( Fig. 2A and Table I). Construct NIN3, a C-terminal deletion that includes the EF-hand and the adjacent XS region (see Fig.  1C), partially rescued the cuticle abnormalities in nkd mutant embryos ( Fig. 2B and Table I). In contrast, constructs NBam, R1S, NR1, and EFX had no activity in the rescue assay (Fig. 2, C and D; Table I).
Cuticle rescue is a stringent assay for nkd function that may not reveal partial activities of deleted Nkd proteins, perhaps because of the narrow temporal requirement for nkd activity in the early embryo (19,28). When nkd expression is driven post-embryonically in otherwise wild-type animals, a variety of adult phenotypes that mimic loss of the wg gene arise, including wing to notum transformations and loss of halteres and sternite bristles (19). To further explore the activity of the mutant Nkd proteins, they were produced in the pupal abdomen using B119-Gal4 and the progeny were scored for the loss of sternite bristles. Production of full-length Nkd resulted in near total loss of sternite bristles with occasional residual laterally displaced bristles ( Fig. 2E 2F and Table II). Construct NBam, including the EFX domain but lacking the adjacent Dsh association sequence, produced adults with largely normal bristle patterns and adults with subtle bristle pattern abnormalities, FIG. 1. Dsh interacts with a region of Nkd that includes the EF-hand motif. A, yeast two-hybrid interactions. Various deletion mutants of Nkd, fused to the GAL4 DNA-binding domain (GB), were tested for their ability to interact with Dsh fused to the activation domain of GAL4 (GAD). Yeast colonies were grown on medium containing (ϩ) or lacking (Ϫ) adenine (ADE) to assay the activity of the ADE reporter gene present in the strain. B, in vitro interaction using the GST pull-down assay. GST fusion proteins containing either full-length Nkd (WT) or the indicated mutant forms were incubated with [ 35 S]methionine-labeled Dsh. The eluates from the pellet (P), as well as 1/10 of the supernatant (S), were analyzed by SDS-PAGE and subsequent autoradiography. C, schematic representation of the Nkd deletion mutant constructs used in this paper, summarizing the binding results obtained in yeast two-hybrid (Y2H) and GST pull-down (GST) assays, along with the in vivo data obtained in rescue (RSC) and overexpression (OVE) experiments. The EF-hand motif (EF) and the X sequence are represented as gray boxes. ϩϩ, positive interaction or full effect; ϩ, partial effect; ϩ/Ϫ, weak interaction or weak effect; Ϫ, no interaction or no effect; blank, not tested. A, full-length Nkd protein restores the wild-type segmentally repeated pattern of denticle bands and promotes proper head skeleton involution (arrowhead) and posterior spiracle elongation (arrow). The majority of nkd mutant cuticles rescued by full-length Nkd protein are indistinguishable from wild type. B, construct NIN3 promotes spiracle elongation (arrow) and restores most of the denticles but does not promote head skeleton involution (arrowhead). According to our cuticle scoring scale, this represents a "moderate" class nkd phenotype (cuticle scoring according to Ref. 19; see Table  I for quantitation of cuticle scoring). C and D, constructs NBam (C), R1S, NR1 (D), and EFX did not rescue any aspect of the nkd cuticle phenotype. These embryos resemble "strong" class nkd embryos, possessing none or only partial denticle bands, no spiracle elongation (arrow), and fully exteriorized head skeleton (arrowhead). E-H, ventral view of four segments of adult female abdomen showing segmentally repeated sternite bristle clusters. E, full-length Nkd protein causes loss, lateral displacement, and disorientation of abdominal sternite bristles (see Table II for quantitation of bristle numbers). F, construct NIN3 routinely eliminates about half of the bristles. G, constructs NBam and R1S produced subtle sternite bristle patterning abnormalities, including loss of hemisternite bristle clusters (arrow) and an overall reduction of 8% of bristles (see Table II). H, constructs NR1 and EFX did not routinely give rise to bristle abnormalities and transgenic flies were thus indistinguishable from wild-type flies. most commonly loss of hemisternite bristle clusters (Fig. 2G). Construct NR1 did not result in a loss of bristles; indeed three of five transgenic lines appeared to have increased numbers of bristles (Table II). To confirm the differential activities of NIN3, NBam, and NR1, a more ubiquitous driver, A8-Gal4, was used. Misexpression of NIN3 or NBam gives rise to pharate adults that fail to hatch, with small wings, short legs, and sternite bristle abnormalities, indicating that these constructs can antagonize Wg signaling when expressed at sufficient levels and duration (data not shown). In contrast, NR1 does not result in any wg-like phenotypes or pharate lethality. In a manner similar to NBam, construct R1S induced subtle bristle abnormalities when crossed to B119-Gal4, and pharate lethality when crossed to A8-Gal4, whereas EFX did not produce any adult phenotypes with B119-Gal4 or A8-Gal4 (data not shown). This differential activity between Nkd R1S and Nkd EFX indicates that the region C-terminal of the EF-hand that is necessary for full interaction with Dsh is important for Nkd activity. Taken together, the in vivo data show that the Dsh association sequences deduced from our binding studies contribute to Nkd function, but regions further N-and Cterminal must also be important for full Nkd activity (see summary Table in Fig. 1C).
Role of Divalent Cations in the Interaction between Nkd and Dsh-The ability of EF-hand motifs to bind Ca 2ϩ , and sometimes Mg 2ϩ , is well established (21,29,30). We investigated the potential role of ions in the Nkd-Dsh interaction. Chelators of divalent cations were tested for their effects on the Nkd-Dsh association, using the GST pull-down assay. In the presence of the divalent cation chelator EDTA, full-length GST-Nkd or GST-Nkd R1S fusion proteins lost their ability to bind Dsh (Fig. 3A), indicating that EDTA titrated a cofactor essential for binding. GST-Nkd R1S was chosen for further biochemical study because it binds Dsh as strongly as full-length GST-Nkd (Fig. 1) and its production in bacteria is more efficient (data not shown).
The minimal concentration of EDTA required to completely abolish the Nkd-Dsh interaction was 40 mM (data not shown) and was employed in the following experiments. As a control, increasing concentrations of divalent ions were used to presaturate EDTA and specifically inhibit its effect. The experiment shown in Fig. 3B was performed with MgCl 2 , but similar results were obtained with CaCl 2 or ZnCl 2 (data not shown). 40 mM MgCl 2 restored the interaction to the control level (Fig. 3B). With lower concentrations of MgCl 2 that did not fully saturate EDTA, there was a corresponding decrease in the amount of Dsh bound to Nkd (data not shown). In contrast, 40 mM LiCl, which contributes monovalent ions unable to bind EDTA, did not inhibit the effect of EDTA (Fig. 3B). Thus the action of EDTA is specific and likely because of chelation of divalent ions needed for the Nkd-Dsh interaction.
Another chelator, EGTA, was tested. In contrast to EDTA, EGTA had no effect on the Nkd-Dsh interaction, even at concentrations as high as 100 mM (Fig. 3C). Therefore EGTA cannot titrate away the divalent ion(s) required for the interaction of Nkd with Dsh, possibly because of a weaker affinity for the ion(s). Under our assay conditions, dissociation constants (K d ) of EDTA and EGTA for Ca 2ϩ are quite similar (4 ϫ 10 Ϫ9 and 5 ϫ 10 Ϫ9 ; see "Experimental Procedures" for calculation), suggesting that Ca 2ϩ is not the divalent ion necessary for the interaction. Mg 2ϩ and Zn 2ϩ , however, are good candidates, as EDTA has 4 -5 orders of magnitude higher affinity for Mg 2ϩ and Zn 2ϩ than does EGTA: for Mg 2ϩ , K d (EDTA) is 9 ϫ 10 Ϫ7

TABLE I
Genetic rescue of nkd 7H16 /nkd 7E89 cuticle phenotype by wild-type and mutant Nkd proteins Each UAS line was crossed (with the exception of NBam, which was recombined) onto a nkd 7H16 /TM3 background and crossed to da-Gal4, nkd 7E89 /TM3. Data are presented as percent of cuticles that were scored as wild-type, weak, moderate, or strong phenotype. Cuticle scoring is according to Ref. 19. Briefly, "strong" class cuticles are Ͻ75% of wild-type length, have two or fewer complete denticle bands, and have exteriorized head skeleton and noneverted spiracles. "Moderate" class cuticles have three or more complete denticle bands (usually in odd numbered abdominal segments), a partially internalized head skeleton, and partially to fully everted spiracles. "Weak" class cuticles have focal denticle band loss. Cross number 1 is the control. Each UAS, nkd stock was scored for maintenance of the strong nkd 7H16 /nkd 7H16 phenotype prior to crossing to da-Gal4, nkd 7E89 /TM3 (data not shown). UAS, nkd 7E89 X da-Gal4, nkd 7E89 crosses were performed for each UAS line and give comparable results (data not shown).    -NR1  5  79  7  96  7  143  11  97  6  82  6  UAS-R1S  4  79  4  83  3  86  3  89  2  UAS-EFX  3  99  3  88  3  90  2  UAS-GFP  1  84  3 and K d (EGTA) is 4 ϫ 10 Ϫ3 ; for Zn 2ϩ , K d (EDTA) is 7 ϫ 10 Ϫ15 and K d (EGTA) is 1 ϫ 10 Ϫ10 . Zn 2ϩ , but not Ca 2ϩ or Mg 2ϩ , Is Able to Restore the Interaction between Nkd and Dsh-GST pull-down assays were employed to identify divalent ion(s) necessary for the interaction between Nkd and Dsh. Instead of using EDTA throughout the experiment as above, we used it only to wash GST-Nkd R1S bound to Sepharose beads prior to its incubation with Dsh in the presence of a variety of divalent cations. If no divalent ion is added during the incubation with Dsh, the interaction between the two proteins is greatly diminished (Fig. 4A; compare with Fig.  1A). Dsh is at no point exposed to an ion-free solution, and encounters only residual EDTA if any. This result strongly suggests that EDTA inhibits the interaction by affecting Nkd and not Dsh. The addition of 40 M ZnCl 2 reestablished the association (Fig. 4B), whereas addition of CaCl 2 or MgCl 2 , even at high concentrations, had no effect (Fig. 4, C and D). Therefore Zn 2ϩ , but not Ca 2ϩ or Mg 2ϩ , can stimulate the Nkd-Dsh association in vitro.
Other divalent ions of the transition metal series can substitute for Zn 2ϩ in Zn 2ϩ -binding proteins (31). Cu 2ϩ , Cd 2ϩ , and Co 2ϩ compete for Zn 2ϩ binding to superoxide dismutase or carbonic anhydrase, whereas Fe 2ϩ and Mn 2ϩ are poor competitors (32). In agreement with these data, Cu 2ϩ , Cd 2ϩ , Co 2ϩ , as well as Ni 2ϩ , restored the Nkd-Dsh association, whereas Fe 2ϩ and Mn 2ϩ had no effect (data not shown).
Role of Putative Ca 2ϩ -binding Residues in the Nkd EFhand-The EF-hand motifs adopt a helix-loop-helix structure and bind Ca 2ϩ via consensus amino acid residues present in the loop (21,33). Key residues that coordinate Ca 2ϩ correspond to loop positions 1, 3, 5, 7, and 12 and are often oxygen-donating residues such as Asp, Glu, and Asn. In some EF-hand domains, natural substitutions of loop residues have occurred that preclude ion binding. For example, in recoverin, a protein with four EF-hands, only EF2 and EF3 bind Ca 2ϩ , whereas EF1 and EF4 have loop residue substitutions that prevent Ca 2ϩ binding (33). Both fly and vertebrate Nkd EF-hands possess consensus loop residues compatible with ion binding (for alignment, see Refs. 19 and 22). However, Drosophila Nkd has an unusual pair of histidine residues at the apex of the loop that are not present in other EF-hands or in the vertebrate Nkd proteins.
The roles of ion binding by many EF-hand proteins have been probed by substituting nonoxygen donating residues in loop positions 1 or 12, which eliminates or greatly reduces Ca 2ϩ binding by EF-hand proteins troponin C and calmodulin (34,35). We created two mutations in Drosophila Nkd EF-hand that should prevent ion binding (if any) by individually changing the Asp residues at loop positions 1 and 12 into Ala and Val, respectively. The mutant proteins, Nkd-D201A and Nkd-D213V, can both interact with Dsh in the yeast two-hybrid (Fig.  5A) and GST pull-down assays (Fig. 5B). However, in the yeast two-hybrid assay the mutations decrease ␤-galactosidase reporter activity (Fig. 5A), despite comparable levels of wild-type and mutant proteins (data not shown), suggesting a reduced Nkd-Dsh binding affinity. In the context of the EFX domain alone the point mutations eliminated the Dsh interaction in the yeast two-hybrid assay (data not shown).
The yeast two-hybrid results suggest a role for the EF-hand loop residues in Dsh binding, whereas the GST pull-down assay reveals the contribution of an adjacent Nkd region in binding Dsh as well as the unanticipated dependence of the interaction on ions other than calcium. To investigate functional properties of the EF-hand point mutant Nkd proteins, we misexpressed them in Drosophila. In multiple independent transgenic lines, UAS-Nkd-D213V and UAS-Nkd-D201A induced wg-like phenotypes as strongly as did wild-type Nkd (data not shown). Injection of Nkd-D213V and Nkd-D201A mRNA restored normal denticle belt patterns to nkd mutant embryos (data not shown). The Nkd-D201A result was confirmed using da-Gal4-mediated rescue ( Table I).
The EF-hand of Nkd Does Not Bind Ca 2ϩ in 45 Ca-blotting Assays-The strict conservation of EF-hand loop residues with potential oxygen-donating properties in Nkd proteins, coupled with a requirement of those residues for efficient Dsh binding FIG. 3. EDTA, but not EGTA, abolishes the interaction between Nkd and Dsh in GST pull-down assays. A, effect of EDTA (100 mM) on the interaction between GST-Nkd or GST-Nkd R1S and 35 S-Dsh. As a control, the assay was also performed in the absence of EDTA (Ϫ). B, saturation with MgCl 2 (40 mM) eliminates the effect of EDTA (40 mM), whereas addition of LiCl (40 mM) has no effect. In control assays (Ϫ), GST alone was used instead of GST-Nkd R1S. The free concentration of EDTA in the presence of 40 mM MgCl 2 is 0.18 mM (see "Experimental Procedures" for calculation). This concentration is not sufficient to abolish the Nkd-Dsh interaction (data not shown). C, in contrast to EDTA, EGTA (even at 100 mM) has no effect on the interaction between Nkd and Dsh. P, pellet; S, 1/10th of the supernatant.

FIG. 4. Zn 2؉ -dependent interaction between Nkd and Dsh. A,
EDTA wash of GST-Nkd R1S is sufficient to impair its association with 35 S-Dsh. EDTA was only used in this initial wash, and not during the next steps of the assay (see "Experimental Procedures"). B, addition of 40 M ZnCl 2 during the incubation with Dsh restores the interaction. In contrast, addition of CaCl 2 (C) or MgCl 2 (D) has no effect, even at 50 or 500 M.
in the yeast two-hybrid assay, suggested that Nkd may bind Ca 2ϩ . We used a 45 Ca-blotting technique to directly test this possibility (26). Bacterially expressed proteins, purified using GST affinity columns, were transferred onto a nitrocellulose membrane after SDS-PAGE. The membrane was then incubated with 45 CaCl 2 . The bovine protein recoverin, which contains 4 EF-hand motifs (GST-bRecoverin), as well as the single EF-hand of the yeast protein Pmr1 (GST-Pmr1 EF; Ref. 24), were loaded as positive controls (Fig. 6A, lanes 2 and 3). In contrast to these controls, two different Nkd mutant proteins that contain the EF-hand (GST-Nkd EFX and GST-Nkd R1S) show no 45 Ca 2ϩ binding (Fig. 6A, lanes 4 and 5).
Nkd Is a Metalloprotein That Binds Zn 2ϩ -The zinc that facilitates the Nkd-Dsh interaction is likely to influence Nkd, which suggests that Nkd binds Zn 2ϩ . However, the primary sequence of Nkd does not reveal any known consensus Zn 2ϩbinding motifs. We performed a 65 Zn-blotting technique (27) to show that Nkd was able to directly bind Zn 2ϩ (Fig. 6, C and D). The binding is mediated by the same Nkd R1S fragment (Fig.  6C, lane 2) that was sufficient for robust Dsh binding and important for Nkd activity in vivo. Deletion of the EF-hand (construct XS) does not abolish 65 Zn 2ϩ association (Fig. 6C,  lane 3), so a region distinct from the EF-hand binds Zn 2ϩ . Expression of the GST-Nkd R1S fragment in bacteria gives rise to two major products of 49 and 44 kDa that accumulate to similar levels (Fig. 6D, lane 2). The larger protein has the expected molecular weight of the full-length GST-Nkd R1S fragment, whereas the smaller one is likely to be a partial degradation product. The smaller fragment did not bind Zn 2ϩ (Fig. 6C, lane 2), suggesting that at least one residue important for Zn 2ϩ coordination was missing. We used mass spectrometry to determine the position of the cleavage site, and deduced a loss of 39 amino acids from the C-terminal end of the Nkd R1S fragment (data not shown). Only four of the 20 amino acids found in proteins are known to participate in Zn 2ϩ binding: His, Cys, and to a lesser extent, Glu and Asp (31). Five His and two Glu residues are missing in the smaller peptide, and thus one or more of them may help to coordinate Zn 2ϩ . Taken together, our results indicate that the Nkd XS region, which is located C-terminal of the EF-hand, mediates Zn 2ϩ binding and cooperates with the EF-hand domain to bind Dsh. DISCUSSION During Drosophila embryogenesis, nkd acts as a feedback antagonist to regulate Wg signaling in epidermal segmentation (19,36). Nkd is a cytoplasmic EF-hand protein that binds to Dsh and specifically regulates Dsh activity in vivo (19,20). The EFX domain, the central portion of which is similar to an EF-hand, is the only domain clearly shared by fly and vertebrate Nkd proteins (22). That the fly or vertebrate EFX domain can bind to the Dsh/Dvl proteins suggests that Dsh association is an ancient feature of Nkd activity in regulating Wnt signaling (22). In the present study we show that a robust interaction between Drosophila Nkd and Dsh requires, in addition to the EFX domain, zinc-binding sequences that are located C-terminal of the EFX domain. This result was unexpected as Nkd contains no consensus Zn 2ϩ -binding motif, but instead has a putative Ca 2ϩ -binding EF-hand motif.
Role of the Nkd EF-hand-Nkd is rare among EF-hand proteins because it contains a single EF-hand, whereas most EF-hand-containing proteins have two to four of these motifs side-by-side. Only a handful of the more than 1,000 predicted EF-hands occur singly, including those in the PKD2 and the yeast Ca 2ϩ /Mn 2ϩ -ATPase Pmr1 proteins (24,37). Our data show that the EF-hand sequence is the core of a conserved domain that binds Dsh and contributes to Nkd activity in vivo. Deletion of the EF-hand in mouse Nkd1 inhibits its ability to antagonize Wnt signaling in cell culture (23). However, such a mutant protein still co-immunoprecipitates with Dsh (23), suggesting that in vertebrates either regions distinct from the Nkd1 EF-hand are also able to bind Dsh, or additional proteins in a Nkd-Dsh complex are able to bridge the two proteins in the absence of the Nkd1 EF-hand.
Role of Ca 2ϩ in Nkd Function-Calcium binding to EF-hand proteins can induce dramatic conformational changes, based upon the three-dimensional structures of EF-hand proteins such as recoverin in the presence and absence of Ca 2ϩ (38,39). Mg 2ϩ can substitute for Ca 2ϩ in some EF-hand motifs under physiological conditions, leading to the view that EF-hands are actually Ca 2ϩ /Mg 2ϩ exchange proteins (29,30). EF-hand motifs coordinate Ca 2ϩ by positioning key oxygen donating (or H 2 O binding) loop amino acid residues in a pentagonal bipyramid conformation to match the electronic orbital configuration in the Ca 2ϩ ion. Fly and all four vertebrate Nkd proteins encode amino acids of potential oxygen-donating character in each of the six conserved loop positions.
Mutations in the Drosophila Nkd EF-hand loop residues that were predicted to disrupt ion binding affected Nkd-Dsh interaction in the yeast two-hybrid assay but not in the GST pulldown assay. The yeast two-hybrid assay may be more sensitive in detecting subtle changes in the k a and k d rate constants induced by mutations because the consequence of an interaction is a discrete transcriptional activation event. Distinct results in each binding assay may hint at subtle features of the Nkd-Dsh interaction. Although Nkd proteins harboring mutated EF-hand loop residues did not exhibit altered activity in Drosophila, the assays involve overexpression that could mask alterations in the affinity of a mutant Nkd protein for Dsh in vivo. Indeed, a mouse Nkd1 protein harboring point mutations analogous to ours attenuated, but did not eliminate, the ability of an overproduced protein to antagonize Wnt signaling in cell culture (23). Here three lines of evidence argue against a role for Ca 2ϩ in regulating Drosophila Nkd activity. First, the in vitro Nkd-Dsh interaction, as well as Nkd activity in our in vivo assay, were not disrupted by the putative ion coordinating residue point mutations. Second, when inhibited by ion chelators, the Nkd-Dsh association was restored by Zn 2ϩ but not by Ca 2ϩ . Third, Nkd was unable to bind Ca 2ϩ in a filter-binding assay. Nkd may therefore be free of Ca 2ϩ in vivo. The two adjacent histidines in the Drosophila Nkd EF-hand loop replace a single amino acid typical of the EF-hand consensus motif and might disrupt the Ca 2ϩ binding configuration. Alternatively, the Nkd EF-hand might bind Ca 2ϩ with a low affinity or may require additional proteins or specific modifications to coordinate Ca 2ϩ (or Mg 2ϩ ) efficiently.
Role for Zn 2ϩ in Nkd Function-Our in vivo and in vitro experiments suggest that a Zn 2ϩ -binding region C-terminal of the EFX domain, included in constructs NIN3 and R1S, is important for Nkd-Dsh association and Nkd activity. The in vitro Nkd-Dsh interaction could be restored by zinc, suggesting that zinc may regulate Nkd activity in vivo. How might zinc perform this function? One possibility is that Zn 2ϩ may modify the overall conformation of the R1S region, allowing it to interact with Dsh. In the absence of Zn 2ϩ , the Dsh-binding regions of Nkd may adopt an alternate conformation that prevents Dsh association, or promotes association with other protein(s).
The crystal structure of the psoriasin protein provides a potentially analogous example of Zn 2ϩ -induced conformational modifications (40). Psoriasin is a two EF-hand-containing protein of the S100 family that accumulates in keratinocytes of psoriasis patients (41). The second EF-hand of psoriasin is a canonical motif that strongly binds Ca 2ϩ , whereas the first one lacks three amino acids in its loop compared with the consensus Ca 2ϩ -binding sequence, and cannot accommodate the ion. The crystal structure of psoriasin reveals that Zn 2ϩ binding modifies the structure of the variant EF-hand causing it to adopt the same conformation as the Ca 2ϩ -bound loop (40). In our experiments addition of Zn 2ϩ after EDTA treatment restored the Nkd-Dsh interaction, indicating that zinc binding by Nkd is reversible. The Zn 2ϩ -binding region and the EF-hand motif of Nkd are adjacent and cooperate in binding Dsh. By analogy to psoriasin, Zn 2ϩ binding to Nkd might induce a conformational modification of the EF-hand region that favors interaction with Dsh. This raised the possibility that this Zn 2ϩinduced conformational change could promote Ca 2ϩ coordination by the Nkd EF-hand. However, we were not able to detect any 45 Ca 2ϩ binding in a 45 Ca-blotting assay similar to that in Fig. 6 performed in the presence of ZnCl 2 (data not shown). As in Nkd, residues of psoriasin that coordinate Zn 2ϩ do not form a classical Zn 2ϩ -binding motif. Instead psoriasin binds Zn 2ϩ with a His-X-X-X-His sequence that is located about 50 amino acids downstream of the variant EF-hand. This consensus is not present in the Zn 2ϩ -binding domain of Nkd although it does possess several His-rich sequences.
Much remains to be learned about Zn 2ϩ homeostasis within the cell. Specific transporters enforce strict control of intracellular Zn 2ϩ concentration in eukaryotes (42,43). In the context of the whole organism, zinc accumulation is regulated in flies by the malpighian tubules, an organ somewhat analogous to vertebrate kidneys (44). The concentration of free Zn 2ϩ may be kept very low within the cell (45,46). This has been recently confirmed in Escherichia coli where there is no persistent pool of free Zn 2ϩ in the cytoplasm (47). Metallothionein plays an important role in the regulation of zinc sequestration and distribution (46). At least 17 metallothionein proteins are present in humans, whereas two have been identified in Drosophila. They bind seven Zn 2ϩ atoms in two clusters and are able to supply the cation to target proteins in intermolecular reactions (48). Similarly, delivery of Zn 2ϩ to Nkd by a metallochaperone might regulate the interaction between Nkd and Dsh.
Role of the Dsh PDZ Domain-PDZ domains are modular structures that mediate protein-protein interactions (49). PDZcontaining proteins are often localized at cell junctions or are associated with the inner surface of the cell membrane where they promote the clustering of signal transduction components (50). A four-amino acid consensus motif, usually located at the C terminus of target proteins, binds in a groove present on one face of the PDZ domain. A peptide library screen identified two classes (I and II) of PDZ domains according to the specificity of the peptide ligand-PDZ interaction (51). The PDZ domain of Dsh has been classified as a class II domain that would be predicted to show a preference for X-Phe/Tyr-X-Phe/Ala/Val-COOH peptides. PDZ domains also form dimers with other PDZ domains by binding to non-terminal ␤-hairpin fingers (52,53), or associate with LIM motifs or ankyrin and spectrin repeats (54 -56). In each case, the PDZ domain is sufficient to bind its targets.
The PDZ domain of Dsh mediates interaction with several proteins, including CKI⑀, Frat1, PP2C␣, Idax, and Stbm (9, 13, 16 -18), although it is not known whether any of these proteins bind in the groove. The Dsh PDZ domain is important, but not sufficient, for the association with Nkd; an upstream region in Dsh that contains a basic sequence is also required. Nkd may bind in a unique way to the Dsh PDZ domain, perhaps as an allosteric regulator. To our knowledge this is the first case of an association between a PDZ domain and an EF-hand motif. Structure determination of a Nkd⅐Dsh complex will help reveal the molecular basis of this novel ion-sensitive interaction that is critical for restraining Wnt signal transduction.