Evidence that the cysteine-rich domain of Drosophila Frizzled family receptors is dispensable for transducing Wingless

Abstract

Members of the Frizzled family of serpentine transmembrane receptors are required to transduce Wingless/Int (Wnt) signals and contain in their N-terminal regions a conserved Wnt-binding cysteine-rich domain (CRD). Each CRD has specific affinities for particular Wnts, and it is generally believed that signal transduction depends on the strength of this interaction. Here, we report in vivo evidence that the CRD is dispensable for Frizzled family receptors to transduce Wingless (Wg), the primary Wnt signal in Drosophila. Thus, we infer that signal transduction does not require binding of Wg to the CRD, but instead depends on interactions between Wg and other portions of the receptor, or other proteins of the receptor complex.

Wingless/Ints (Wnts) are secreted glycoproteins that play profound and pervasive roles in animal development (1). Most Wnts are transduced by the Frizzled (Fz) family of serpentine transmembrane receptors (2) in conjunction with the Arrow/low-density lipoprotein receptor-related protein (LRP) family of coreceptors (3). Some, but not all, Fz family proteins transduce Wnts through the “canonical” Armadillo/β-catenin pathway (1). However, Fz proteins also can transduce Wnts, or other classes of ligands, through noncanonical pathways that are independent of Armadillo/β-catenin (4).

Structure/function analysis is essential to understand the mechanism by which Fz proteins are activated by Wnts. All Fz proteins contain a large extracellular N-terminal motif with 10 conserved cysteines, known as the cysteine-rich domain (CRD; Fig. 1A). In general, intact Fz proteins and their isolated CRDs can bind Wnts in cell culture assays, but different Fz CRDs have different binding affinities for specific Wnts (2, 5–8). The structures of two CRDs recently have been determined, and the surfaces that are likely to interact with Wnts were identified by mutational analysis of Wnt/CRD-binding interactions in cell culture assays (9). These and other data (10) have led to the view that Wnt binding to the CRD is essential for activating signal transduction.

Fig. 1.

Rescue of fz^P21 Dfz2^– homozygotes by ΔCRD forms of Fz and Dfz2. (A and B) Full-length and ΔCRD forms of Dfz2, with the positions of the Dfz2^C1,2,3 and ^-4 mutations, as well as the corresponding position of the fz^P21 mutation, indicated (see Materials and Methods). (C and D) Ventral larval cuticles formed by wg^CX4 and fz^P21 Dfz2^– homozygotes: Both form lawns of ventral hairs, the hallmark phenotype corresponding to lack of Wg signal transduction. (E and F) Ventral larval cuticles formed by fz^P21 Dfz2^– homozygotes carrying the Tubα1-fz^ΔCRD (E) or Tubα1-Dfz2^ΔCRD (F) transgene: Both show normal, segmented patterns of ventral hairs.

Inferences about Fz protein structure and function from crystallographic analysis and cell culture experiments can be tested in vivo in Drosophila. The Drosophila genome encodes seven Wnts and four proteins of the Fz family (11). Of these, Wingless (Wg) is the predominant Wnt signal in Drosophila and is transduced by two of the four Fz proteins, Fz and Dfz2, through activation of the canonical Armadillo/β-catenin pathway (1). Fz and Dfz2 are fully redundant for Wg transduction, because loss of either has no detectable effect on Wg signaling at any stage of development, whereas loss of both abrogates virtually all known responses to Wg (12, 13). A third protein, Dfz3, can bind Wg in cell culture assays but has little or no transducing activity and instead likely functions as an antagonist of Wg signaling (6, 8). Dfz4 is a divergent Fz protein that lacks the capacity to bind Wg (8); furthermore, its overexpression or down-regulation by RNA interference appears to have no effect on Wg transduction (14).

Here, we use in vivo assays to test the hypothesis that the CRD is essential for the activation of Fz and Dfz2 by Wg. Surprisingly, we find that deletion of the CRD does not block the capacity of either receptor to transduce Wg. These results suggest that Wg interacts with both receptors at sites other than the CRD, or with other proteins of the receptor complex, to activate the canonical Armadillo/β-catenin transduction pathway.

Materials and Methods

Mutations. fz^H51, fz^P21 (15), Dfz2^C1 (13), Df(3L)469-2 (12) (referred to as Dfz2^–), and Dfz3^G10 (6) have been described. Dfz2^C2,^3,4 were isolated as described for Dfz2^C1 (13, 14). Dfz2^C2 and Dfz2^C4 have stop codons after amino acids 515 and 604, respectively; Dfz2^C3 is a substitution of the fourth conserved cysteine by serine (14).

Transgenes. vg^Q1206-Gal4, MS1096-Gal4, UAS-flp, Tubα1-fz, Tubα1-Dfz2, and hsp70-GFPnls have been described (13) (see also Flybase, http://flybase.bio.indiana.edu). The Tubα1 promoter drives low to moderate levels of transcription in most or all cells (e.g., ref. 16). hsp70-CD2⁷⁶, y⁺ is inserted within 1% recombination distance of the Dfz2 locus. The Tubα1-fz^ΔCRD and Tubα1-Dfz2^ΔCRD transgenes were derived from Tubα1-fz and Tubα1-Dfz2 transgenes (13) by replacing the N-terminal domains including the endogenous signal peptide and the CRD (amino acids 1–156 for Fz and 1–166 for Dfz2) with the Wg signal peptide and three Flu tags (16); Tubα1-fz^GSG#15, Tubα1-fz^GSG#18, Tubα1-Dfz2^GSG#15, and Tubα1-Dfz2^GSG#18 were derived similarly, except that the deleted CRDs were replaced by the appropriate mutant CRDs (5). Flu- and GFP-tagged forms of Dfz2⁺, Dfz2^C2, and Dfz2^C3 were generated by replacing the Dfz2 signal peptide with the Wg signal peptide plus three Flu tags and by inserting the GFP-coding sequence just upstream of the last amino acid of Dfz2⁺ and Dfz2^C3 or by fusing it to the C terminus of Dfz2^C2; all three forms were expressed in vivo under Gal4/UAS control.

Rescue of fz^P21 Dfz2^– and Dfz3^G10; fz^P21 Dfz2^– Embryos. The y; fz^P21 Df(3L)469-2 zygotes were generated by crossing y; fz^P21 Df(3L)469-2/fz^P21 hs-CD2⁷⁶, y⁺ females to y; fz^P21 Df(3L)469-2/TM6B, y⁺ males. Tubα1-fz, Tubα1-fz^ΔCRD, Tubα1-Dfz2, and Tubα1-Dfz2^ΔCRD transgenes were introduced on the second chromosome from the female, male, or both, either homozygous or in trans to a CyO, y⁺ balancer. First instar larvae derived from y; fz^P21 Df(3L)469-2 zygotes carrying the transgene were identified unequivocally by the yellow cuticle phenotype (Fig. 1 D and E). The y; fz^P21 Df(3L)469-2 embryos could not be similarly identified before cuticle formation, owing to the lack of an appropriate genetic marker. Hence, the rescuing transgene was introduced from either homozygous males or females. Under these conditions, 100% of the progeny should develop normally, if the transgene rescues the mutant condition; otherwise, 25% should show loss of Wg transduction phenotypes. At least 100 embryos were assayed for each rescue experiment; in all cases (except for the negative controls lacking transgenes), >95% of the progeny developed normally, the remainder showing only random defects. Rescue of Dfz3^G10; fz^P21 Dfz2^– embryos was performed in the same way. Embryos were assayed for cuticle phenotype, En, Eve, and Lab expression as described (13).

Rescue of fz Dfz2 Mutant Clones in the Imaginal Wing Disk. Clones of fz^P21 Dfz2^C1,2,3, or ^-4 cells, labeled by the absence of the hsp70-GFP, ubiquitin-GFP or y⁺ marker genes, were generated as described (13, 14). The standard genotype used was: y; fz^P21 Dfz2^C1,2,3, or ^-4 ri FRT2A/marker ri FRT2A [or marker M(3)ⁱ⁵⁵ ri FRT2A] with a single Tubα1-fz⁺, Tubα1-fz^ΔCRD, Tubα1-Dfz2⁺, or Tubα1-Dfz2^ΔCRD transgene (or no transgene) and either hsp70-flp or vg-Gal4 plus UAS-flp. The same protocols were followed for assaying rescue in Dfz3^G10 larvae and adults. For rescue of Dfz3^G10;fz^P21 Dfz2^C2/fz^P21 Dfz2^– larvae by the Tubα1-fz^ΔCRD and Tubα1-Dfz2^ΔCRD transgenes (see Fig. 4 E and F), DNA was extracted from larval carcasses from which stained wing imaginal discs were derived and analyzed by PCR and sequencing to confirm the fz, Dfz2, Dfz3, and transgene genotypes.

Fig. 4.

Rescue of fz^P21 Dfz2^C2 cells in Dfz3^G10 wing imaginal discs by ΔCRD forms of Fz and Dfz2. (A and B) Rescue of Wg transduction in fz^P21 Dfz2^C2 clones in Dfz3^G10 larvae by Tubα1-fz^ΔCRD or Tubα1-Dfz2^ΔCRD (analyzed as twin spots, as in Fig. 3 G–J). (A–D) Clones are marked black by the absence of GFP; Dll is stained in red. (C and D) Rescue of Wg transduction in fz^P21 Dfz2^C2 clones in Dfz3^G10 larvae by Tubα1-fz^ΔCRD or Tubα1-Dfz2^ΔCRD by using the Minute technique: Mutant clones have a competitive growth advantage and generally populate most of the developmental compartment in which they reside. (E and F) Wing discs from Dfz3^G10; fz^P21 Dfz2^C2/fz^P21 Dfz2^– larvae rescued to the late third instar by the presence of a single Tubα1-fz^ΔCRD or Tubα1-Dfz2^ΔCRD transgene. Dll (red) is expressed normally in a broad stripe flanking Wg-secreting cells (blue) along the dorsoventral compartment boundary [discs are of normal size but are shown at lower magnification than in A–D (see Materials and Methods)].

Ab Staining. Standard protocols were followed by using mouse αWg, rabbit αDll, rabbit αEve, rabbit αLab, mouse αEn antisera, and mouse αFlu (12CA5, BAbCo, Richmond, CA), as in ref. 13. For Ab staining of live cells, imaginal discs were incubated at 4°C in M3 media with antisera for 10 min, rinsed, and then processed by the standard fixation and immunofluorescence protocol. For staining fixed, but nonpermeabilized, cells, discs were fixed and processed by the standard protocol except that detergent was withheld.

Results

Assessing the Role of the CRD of Fz Proteins in Wg Signal Transduction. As an initial test of the role of the CRD in Wg transduction, we asked whether Tubα1-fz^ΔCRD or Tubα1-Dfz2^ΔCRD transgenes (Fig. 1B and Materials and Methods) can rescue Wg transduction in fz^P21 Dfz2^C1 mutant embryos and cell clones. Such mutant embryos and clones are normally devoid of Wg-transducing activity (12, 13); however, Wg transduction is fully rescued by the presence of either transgene (data not shown). These findings indicate either that the Fz^ΔCRD and Dfz2^ΔCRD proteins have the intrinsic capacity to transduce Wg, or alternatively, that they can act in conjunction with endogenous Fz family proteins, particularly the products encoded by the fz^P21 or Dfz2^C1 mutant alleles, to confer Wg-transducing activity.

The fz^P21 allele is a molecular null that truncates the Fz protein 67 aa downstream of the initiator ATG and yields no detectable protein expression (15). The Dfz2^C1 allele is a nonsense mutation that terminates the coding sequence of the receptor just before the first transmembrane domain (13) (Fig. 1 A). Hence, the Dfz2^C1 locus encodes a potentially secreted form of Dfz2 that consists of just the CRD-containing extracellular domain. This truncated endogenous protein might plausibly complement the missing CRD domains of the Fz^ΔCRD and Dfz2^ΔCRD proteins and thereby restore their signal-transducing activity. In subsequent sections, we test the intrinsic Wg-transducing activity of the Fz^ΔCRD and Dfz2^ΔCRD proteins by using new Dfz2 alleles and other genetic backgrounds to reduce greatly, or abolish, the possible contribution of endogenous Dfz2 protein.

Rescue of Wg Signal Transduction in Embryos Null for fz and Dfz2 by ΔCRD Forms of Fz and Dfz2. In the first set of experiments designed to remove endogenous Fz and Dfz2, we assayed the ability of ΔCRD forms of Fz and Dfz2 to rescue Wg signal transduction in fz^P21 Df(3L)469-2 homozygotes derived from fz^P21/fz^P21 Df(3L)469-2 transheterozygous females. Df(3L)469-2 is a deletion that removes the entire Dfz2-coding sequence (12). Such embryos (referred to subsequently as fz^P21 Dfz2^– homozygotes) are devoid of Fz and have been reported to derive no functionally detectable Dfz2 from maternal Dfz2⁺ transcripts (the only possible source of native Dfz2) as they appear indistinguishable from wg^– embryos (12).

We assayed several well characterized Wg signaling events that occur during embryogenesis including (i) the secretion of bands of naked ventral cuticle separating the bands of ventral hairs decorating each larval segment, (ii) the maintenance of segmentally reiterated stripes of Engrailed expression in the ectoderm, (iii) the activation of Even-skipped expression in presumptive motor neurons and heart precursor cells, and (iv) the up-regulation of Labial in the endoderm (refs. 12 and 13 and references therein).

As described (12), we find that fz^P21 Dfz2^– homozygotes generally behave like wg^– embryos. However, we can detect weak rescue of the cuticle and En maintenance phenotypes. First, although fz^P21 Dfz2^– homozygotes secrete a continuous “lawn” of ventral hairs typical of wg^– embryos (Fig. 1C), some have 10–20% longer bodies than wg^– embryos, an indicator of residual Wg signaling (17). Second, although fz^P21 Dfz2^– homozygotes generally fail to maintain En expression in the ectoderm, we detect occasional small clusters of ectodermal cells that show persistent En expression, a phenotype not observed in wg^– embryos.

In contrast, fz^P21 Dfz2^– homozygotes carrying a single copy of either the Tubα1-fz^ΔCRD or Tubα1-Dfz2^ΔCRD transgene introduced via either the male or female parent showed complete rescue for all of the Wg signaling outputs described above (Fig. 1 E and F, Materials and Methods, and data not shown).

Rescue of Wg Signal Transduction by ΔCRD Forms of Fz and Dfz2 in fz Dfz2 Dfz3 Triple Null Embryos. Dfz3, like Fz and Dfz2, contains a CRD that can bind Wg (6, 8). Hence, even though Dfz3 appears to have little, if any, ability to transduce Wg under normal conditions (6), it might nevertheless be able to combine with Fz^ΔCRD and Dfz2^ΔCRD to form a functional receptor, e.g., as a heterodimer (18). To assess this possibility, we asked whether the presence of either the Tubα1-fz^ΔCRD or Tubα1-Dfz2^ΔCRD transgene could rescue Wg transduction in Dfz3^G10; fz^P21 Dfz2^– homozygotes derived from Dfz3^G10; fz^P21/fz^P21 Dfz2^– females [Dfz3^G10 is a deletion that removes the ATG and the first exon (6)]. We observed that the presence of a single copy of either the Tubα1-fz^ΔCRD or Tubα1-Dfz2^ΔCRD transgene introduced via the male rescues Wg transduction in the triple mutant embryos using each of the assays described above (data not shown and Materials and Methods).

Thus, the Fz^ΔCRD and Dfz2^ΔCRD proteins both appear capable of transducing Wg in multiple contexts in embryos that have severely reduced levels of endogenous Dfz2 activity and are devoid of endogenous Fz and Dfz3 protein.

Rescue of Wg Signal Transduction in Clones of fz Dfz2 Double Mutant Cells by ΔCRD Forms of Fz and Dfz2. The Dfz2^– deletion removes additional loci, including one or more required for cell viability, precluding our assaying rescue of fz^P21 Dfz2^– cells by Tubα1-fz^ΔCRD or Tubα1-Dfz2^ΔCRD transgenes during subsequent development. We therefore sought to obtain a null mutation of Dfz2 to be used in cis with the fz^P21 null allele to test whether ΔCRD forms of Fz and Dfz2 can rescue Wg transduction in imaginal cells devoid of endogenous Fz and Dfz2.

Three Dfz2 mutations, Dfz2^C2, Dfz2^C3, and Dfz2^C4, were obtained in genetic screens for second site mutations that block Wg signal transduction in fz^P21 cells (see Materials and Methods). Dfz2^C2 and Dfz2^C4 are stop codons, which truncate the coding sequence after the fifth and last transmembrane domains, respectively; Dfz2^C3 is a missense mutation that substitutes the fourth conserved cysteine in the CRD for a serine (Fig. 1 A and Materials and Methods). Clones of cells homozygous for each of these mutations in cis with fz^P21 behave indistinguishably from clones of fz^P21 Dfz2^C1 cells, indicating that all three alleles encode Dfz2 proteins devoid of Wg-transducing activity. We focused on Dfz2^C2 because mutant forms of other serpentine receptors that are similarly truncated in the middle of the multiple pass transmembrane domain stack fail to reach the cell surface and are unable to transduce extracellular signals (19–21).

To test whether the Dfz2^C2 mutant protein is excluded from reaching the cell surface, we expressed doubly tagged forms of Dfz2⁺ and Dfz2^C2 bearing Flu epitopes at the N terminus and GFP at the C terminus (Fig. 2 A and B) and assayed for the appearance of the Flu tag on the surface of living, or fixed but nonpermeabilized, imaginal disk cells (see Materials and Methods). For both assay methods, the Flu epitope was readily detected on cells expressing doubly tagged Dfz2⁺; however, we failed to detect the epitope on cells expressing doubly tagged Dfz2^C2 (Fig. 2 C–F and data not shown). Thus, it appears that little or no Dfz2^C2 is able to reach the cell surface, making it unlikely to contribute to Wg transduction, whether alone or in conjunction with Fz^ΔCRD or Dfz2^ΔCRD.

Fig. 2.

Lack of cell surface accumulation of Dfz2^C2 in vivo.(A and B) Doubly tagged forms of Dfz2⁺ and Dfz2^C2 carrying three copies of the Flu-epitope at the N terminus and GFP at the C terminus. (C–J) Flu-epitope staining (red) and GFP fluorescence (green) associated with doubly tagged Dfz2⁺ and Dfz2^C2 proteins expressed in the “pouch” of the wing imaginal disk [the MS1096-Gal4 transgene drives stronger expression dorsally (up) than ventrally]. In living discs (data not shown) and fixed but nonpermeabilized discs (C–F), Flu-epitope staining was observed only for doubly tagged Dfz2⁺ and only at the cell surface (apical plane, C). In discs fixed in the presence of Triton X-100, Flu-epitope staining is observed for doubly tagged forms of both Dfz2⁺ and Dfz2^C2 at both the apical and subapical planes of focus (G–J). GFP fluorescence was readily detected in all conditions.

We next tested whether the presence of either the Tubα1-fz^ΔCRD or Tubα1-Dfz2^ΔCRD transgene can rescue Wg transduction in fz^P21 Dfz2^C2 cells during imaginal disk development. We used mitotic recombination to generate “twin” spots composed of marked fz^P21 Dfz2^C2 and WT sibling clones derived from the same mother cell, as well as single clones of fast-growing fz^P21 Dfz2^C2 cells [by using the “Minute technique” (22)] that compete out surrounding WT cells in the same developmental compartment (see Materials and Methods). We also obtained entirely mutant discs from fz^P21 Dfz2^C2/fz^P21 Dfz2^– zygotes rescued by the presence of a single Tubα1-fz^ΔCRD or Tubα1-Dfz2^ΔCRD transgene. In all three contexts, we find that the presence of either the Tubα1-fz^ΔCRD or Tubα1-Dfz2^ΔCRD transgene rescues Wg transduction in a manner indistinguishable from that of their Tubα1-fz and Tubα1-Dfz2 counterparts (Fig. 3 and data not shown).

Fig. 3.

Rescue of fz^P21 Dfz2^C2 double mutant clones in the wing imaginal disk by ΔCRD forms of Fz and Dfz2. (A–E) Anterior edge of wings carrying multiple clones of homozygous fz^P21 Dfz2^C2 cells (see Materials and Methods). In the absence of rescuing transgenes, loss of endogenous fz and Dfz2 gene function blocks the formation of wing margin bristles by mutant cells and causes wing notching, indicating a failure to transduce Wg (A). Cells within the clone are marked by the yellow cuticle mutation, not readily visible at this magnification. (B–E) These phenotypes are fully rescued by single transgenes expressing Fz⁺, Dfz2⁺, Fz^ΔCRD, or Dfz2^ΔCRD under the control of the Tubα1 promoter. (F–J) Wing discs containing “twin spots” composed of fz^P21 Dfz2^C2 double mutant clones (marked black by the absence of GFP expression) and their WT sibling clones (marked bright green by the presence of two copies of the GFP marker gene) induced during the first or second instar. In the absence of rescuing transgenes, double mutant clones survive only briefly in the wing blade primordium (marked by the central domain of Dll expression; red) and stop expressing Dll before being lost from the epithelium, indicating a failure to transduce Wg (F, boxed portion of the GFP image is shown at higher magnification in the Dll and merged images). In the presence of any one of the Tubα1-fz⁺, Tubα1-fz^ΔCRD, Tubα1-Dfz2⁺, or Tubα1-Dfz2^ΔCRD transgenes, such double mutant clones survive in the presumptive wing blade and appear to proliferate equally well, relative to their WT sibling clones (G–J). Further, the pattern of Dll expression appears normal in the rescued clones.

In the wing imaginal disk, Wg acts at short range to induce the formation of wing margin bristles, and at longer range to define the presumptive wing primordium through the activation of target genes such as Distalless (Dll) (16, 23). Clones of fz^P21 Dfz2^C2 cells behave like fz^H51 Dfz2^C1 clones (13): They cease to express Dll and either die or sort out after three to five cell divisions (Fig. 3F and data not shown), resulting in a loss of wing blade tissue and margin bristles (Fig. 3A). All of these phenotypes are fully rescued by the presence of a single copy of either the Tubα1-fz^ΔCRD or Tubα1-Dfz2^ΔCRD transgene (Fig. 3 D, E, I, and J and data not shown). Full rescue was also observed for Wg transduction in the other imaginal discs, as assayed by Dll expression in the distal antennal and leg discs, as well as the ability of mutant clones to contribute to adult derivatives, such as the eye, head capsule, antennae, and legs (data not shown). All of the adult derivates formed by rescued fz^P21 Dfz2^C2 cells appeared phenotypically normal, except for defects in planar cell polarity caused by the fz^P21 mutation, which are not rescued by Tubα1-Dfz2^ΔCRD and poorly rescued by Tubα1-fz^ΔCRD.

We note that Dll is normally expressed in a graded fashion in the wing primordium in response to the amount of Wg, which declines as a function of distance from its source at the prospective wing margin (e.g., as in Fig. 4 E and F). Hence the pattern of Dll expression provides an assay for the capacity ΔCRD forms of Fz and Dfz2 to transduce Wg relative to the native proteins. In the presence of either a single Tubα1-fz^ΔCRD or Tubα1-Dfz2^ΔCRD transgene, or their WT Tubα1-fz⁺ or Tubα1-Dfz2⁺ counterparts, we find that the graded pattern of Dll expression appears normal, irrespective of the presence of large clones of fz^P21 Dfz2^C2 mutant cells (Fig. 3 G–J). Further, in all four cases, the relative sizes of sibling WT and fz^P21 Dfz2^C2 twin spots were similar and contributed to normally proportioned wing blade primordia. Hence, when expressed under Tubα1 control, Fz, Fz^ΔCRD, Dfz2, and Dfz2^ΔCRD all appear to have similar capacities to transduce Wg.

We also performed rescue experiments of the fz Dfz2 double mutant condition by using each of the other three Dfz2 alleles, Dfz2^C1, Dfz2^C3, and Dfz2^C4, in cis with fz^P21. In all cases, we found that a single copy of either the Tubα1-fz^ΔCRD or Tubα1-Dfz2^ΔCRD transgene fully rescued Wg signal transduction in double mutant imaginal disk cells (data not shown). We note that the Dfz2^C3 protein shows a severely reduced capacity to reach the cell surface (assayed as in Fig. 2; data not shown), suggesting that improper folding caused by the cysteine-to-serine substitution in the CRD is sufficient to account for the failure of the mutant receptor to transduce Wg.

Finally, systematic scanning of Dfz2 with glycine–serine–glycine (GSG) tripeptide insertions has identified two insertions, nos. 15 and 18, that block intrinsic Wnt binding by the CRD, but cause only a modest (no. 15), or no (no. 18), reduction in binding by the intact receptor in tissue culture assays (5). We therefore repeated our rescue experiments, using Tubα1-fz and Tubα1-Dfz2 transgenes, which carry these insertion mutations. We find that Tubα1-fz^GSG15, Tubα1-Dfz2^GSG15, Tubα1-fz^GSG18, and Tubα1-Dfz2^GSG18 all fully rescue Wg transduction in clones of fz^P21 Dfz2^C2 imaginal disk cells (data not shown).

Collectively, these data indicate that forms of Fz and Dfz2 that lack the CRD, or bear mutant CRDs that appear to lack intrinsic Wnt binding activity, nevertheless retain the capacity to transduce Wg in vivo.

ΔCRD Forms of Fz and Dfz2 Rescue Wg Signal Transduction in Clones of fz Dfz2 Dfz3 Triple Mutant Cells. To determine whether the presence of Dfz3 might contribute to the rescuing activity of either Fz^ΔCRD or Dfz2^ΔCRD in the fz Dfz2 double mutant condition, we examined the capacity of both the Tubα1-fz^ΔCRD and Tubα1-Dfz2^ΔCRD transgenes to rescue Wg transduction in fz^P21 Dfz2^C1, fz^P21 Dfz2^C2, fz^P21 Dfz2^C3, and fz^P21 Dfz2^C4 clones induced in the imaginal discs of Dfz3^G10 larvae. In all cases, we observed full rescue (Fig. 4 A–D and data not shown). In addition, we assayed Wg signal transduction in the imaginal discs of entirely mutant Dfz3^G10; fz^P21 Dfz2^C2/fz^P21 Dfz2^– larvae rescued by the presence of the Tubα1-fz^ΔCRD or Tubα1-Dfz2^ΔCRD transgene. As in triple mutant clones, Wg transduction appears to be fully rescued by each transgene (Fig. 4 E and F).

Internalization of Wg by Clones of fz Dfz2 Dfz3 Triple Mutant Cells in Wing Disks That Express ΔCRD Forms of Fz and Dfz2. Wg secreted by prospective wing margin cells is endocytosed by neighboring, nonsecreting cells over a range of several cell diameters and accumulates in cytosolic puncta in the receiving cells. To determine whether internalization in these cells is mediated by binding to the CRD of Fz, Dfz2, or Dfz3, we assayed Wg staining in clones of fz^P21 Dfz2^C2 cells in wing discs of Dfz3^G10 larvae carrying a single copy of the Tubα1-Dfz2^ΔCRD transgene. As shown in Fig. 5, we can detect a graded distribution of Wg staining puncta extending over a range of several cells in rescued, triple mutant tissue, as in neighboring nonmutant tissue. Hence, Wg uptake, like Wg transduction, does not appear to depend on the presence of Fz family proteins carrying CRDs.

Fig. 5.

Distribution of secreted Wg in Dfz3^G10; fz^P21 Dfz2^C2 mutant wing disk cells rescued by expression of Dfz2^ΔCRD. (A and B) Wing disk carrying large fz^P21 Dfz2^C2 clones obtained from a Dfz3^G10 larvae carrying a single Tubα1-Dfz2^ΔCRD transgene (Minute technique; the clones are marked black by the absence of GFP). Wg (white in A and red in B) is secreted by a thin line of cells along the dorsoventral compartment boundary and accumulates in cytosolic puncta in surrounding cells over a range of several cell diameters even in the triple mutant cells rescued by the presence of Dfz2^ΔCRD.

Discussion

The CRDs of most Fz family receptors have intrinsic Wnt-binding activity; moreover, the affinity of Wnt/CRD binding for any given Wnt/Fz pair has been suggested to correlate with the capacity of the receptor to transduce the ligand in cell culture and in vivo (7, 8). These observations have led to the view that Wnts bind and activate Fz family receptors via direct interactions with the CRDs.

In Drosophila, the primary Wnt signal Wg can be transduced by either Fz or Dfz2. Surprisingly, both receptors appear to have similar capacities to transduce Wg throughout development (13), even though the CRD of Dfz2 has a 10-fold higher binding affinity to Wg than the CRD of Fz (7). This difference suggests that the binding affinities measured between particular Wnts and Fz family CRDs in vitro (7, 8) may not reflect ligand specificity of the intact receptors in vivo and raises the more radical possibility that Wnt/CRD-binding interactions may not be essential for signal transduction. Other results are consistent with this possibility (5, 24), most notably, the identification of mutant forms of the CRD that appear to block its ability to bind Wnts in cell culture, but have little or no effect on Wnt binding by the intact receptor (5).

Here, we have tested the role of the CRD by examining the intrinsic ability of Drosophila Fz and Dfz2 proteins that lack the CRD to transduce Wg in vivo. Our results argue that the CRD is dispensable for normal Wg transduction by both receptors.

A key question in assessing our results is whether the ability of the ΔCRD forms of Fz and Dfz2 to transduce Wg depends on the presence of the CRD contributed by any of the endogenous Fz family proteins that might remain in the fz Dfz2 Dfz3 triple mutant backgrounds we have used. We have eliminated possible contributions from Fz and Dfz3 by using molecular null alleles of the fz and Dfz3 loci. In addition, we consider a contribution by Dfz4 unlikely given that the Dfz4 CRD lacks Wg-binding activity (8) and that Dfz4 does not normally appear to be involved in Wg transduction (12–14). For Dfz2, we depend on genetic backgrounds that approach the molecular null condition. Specifically, we assayed rescue of triple null (fz^– Dfz2^– Dfz3^–) embryos derived from double null (fz^– Dfz3^–) females; such embryos derive minimal Dfz2 activity from maternal Dfz2⁺ transcripts and are virtually devoid of endogenous Wg-transducing activity. In addition, we have assayed rescue of putative, triple null cells in the developing imaginal discs using functionally null Dfz2 alleles, including Dfz2^C2, which encodes a truncated protein that lacks the last two transmembrane domains and appears unable to reach the cell surface. We find that expression of either Fz^ΔCRD or Dfz2^ΔCRD fully rescues the loss of Wg transduction in all of these mutant conditions. Although it is possible to envisage scenarios in which expression of ΔCRD forms of Fz or Dfz2 might facilitate some form of complementation by the CRDs from either endogenous Dfz2 or Dfz4 in these experiments, the most plausible interpretation of our results is that the rescuing activities of the ΔCRD forms of Fz and Dfz2 reflect their intrinsic capacity to transduce Wg.

Thus, we infer that the key interactions between Wg and Drosophila Fz family proteins that activate the canonical Armadillo/β-catenin transduction pathway do not require the CRD. An analogous situation has been reported for the human IL-8 receptor. This receptor has an N-terminal domain, which binds ligand with high affinity, but is dispensable: It is the low-affinity ligand-binding sites in the extracellular loops linking the transmembrane domains that appear to be critically required for receptor activation (25). Thus, with Fz family receptors, there may be Wg-binding domains elsewhere in the protein, such as the “nonconserved extracellular region” of the receptor or the extracellular surface of the seven transmembrane domain stack. Alternatively, Wg may activate Fz through direct interactions with closely associated proteins, such as the transmembrane protein Arrow/LRP6, a coreceptor that is essential for Wg transduction (3).

Our results also bear on recent findings that the CRDs of Fz family proteins mediate dimerization of the receptors (9, 18) and that dimerization may be a prerequisite for signaling through the canonical Armadillo/β-catenin transduction pathway (18). Our results suggest that if receptor dimerization is necessary it can occur without any direct requirement for the CRD.

What then might the primary role of the CRD be? One possible answer is that the CRD functions to bind and concentrate Wnts in the vicinity of other binding sites within the receptor complex that mediate signal transduction. Accordingly, deletion of the CRD should lower the affinity of Fz proteins for Wnts without abolishing their intrinsic signal-transducing activity. One result which argues against this explanation is that the Tubα1-fz^ΔCRD and Tubα1-Dfz2^ΔCRD transgenes fully rescue Wg transduction in the fz Dfz2 Dfz3 mutant condition, a finding exemplified by our observation that rescued cells in the developing wing imaginal disk appear to show a complete and normally patterned response to the Wg morphogen gradient. However, we cannot presently exclude the possibility that these transgenes generate abnormally high levels of Fz^ΔCRD and Dfz2^ΔCRD expression and thus compensate for a reduced capacity of the truncated proteins to transduce Wg.

A second possible answer is that CRD/Wg interactions involving Fz and Dfz2 are required to modulate the tissue distribution or accessibility of extracellular Wg. This possibility is supported by prior evidence that the CRD mediates binding interactions between Wg and Dfz2 that regulate the extracellular distribution of Wg (26) and is also suggested by the existence of endogenous Fz family proteins such as DFz3 and vertebrate secreted Fz-related proteins that may function solely as Wnt binding, but nontransducing, antagonists of Wnt signaling (6, 10). According to this view, one might expect to find evidence for inappropriate Wg movement or signaling in fz Dfz2 Dfz3 mutant tissue rescued by ΔCRD forms of Fz and Dfz2. However, we failed to detect changes in the pattern of Dll expression or the punctate distribution of internalized Wg. It is possible that our current assays are not sensitive enough to reveal such differences, especially given that other Wg-binding proteins may act redundantly with Fz family members to modulate the extracellular distribution of Wg (27).

A third possible answer is that the CRD is essential for binding and transducing particular extracellular ligands, including some Wnts, but does not have a general or obligate role in Wnt signal transduction per se. It is notable that Fz family receptors respond not only to Wnts, but also to other types of ligands, with some receptors capable of triggering two or more distinct intracellular pathways depending on the specific activating ligand and the developmental context (1, 28, 4, 29). Thus, the CRD may be one of several binding modules responsible for linking specific ligands, either Wnts or others, to the activation of particular transduction pathways by Fz family receptors. In the case of Wg, we suggest that the key binding domain required to trigger the canonical Armadillo/β-catenin transduction pathway is located elsewhere in Fz and Dfz2 or on a coreceptor such as Arrow/LRP.