Wingless signaling generates pattern through two distinct mechanisms

In Drosophila embryos, the wingless signaling pathway is required for correct specification of cuticular pattern elements secreted by individual epidermal cells (reviewed in Bejsovec and Peifer, 1996). The wild-type cuticle pattern consists of 6 rows of uniquely oriented, morphologically distinct denticles in the anterior of each segment, forming segmental denticle belts that are interspersed with an expanse of naked cuticle covering the posterior of each segment. In the absence of Wg signaling, both aspects of the wild-type pattern become defective: wg mutants show reduced denticle diversity and no naked cuticle specification. This indicates that, in a wild-type embryo, Wg signal produced by a single row of cells in each segment can direct two distinct cellular responses in different groups of cells within the 10- to 12-cell-wide segment. Wg activity generates diversity of cell fates in the anterior cells that secrete the diverse denticle types and it instructs cells in the posterior of the segment to secrete cuticle that is devoid of denticles.

Phenotypic effects of Wg signaling can also be detected at the molecular level. Wg has an autocrine function: its activity is required for its own continued expression (Bejsovec and Martinez-Arias, 1991; Li and Noll, 1993) and unrestricted Wg signaling throughout the segment can induce wg expression in inappropriate rows of cells (Noordermeer et al., 1992; Bejsovec and Wieschaus, 1993; Hooper, 1994; Yoffe et al., 1995). Wg activity is also required for stabilizing the expression of engrailed (en), a segment polarity gene expressed in the row of cells immediately posterior to the wg-expressing row of cells (DiNardo et al., 1988; Martinez Arias et al., 1988). Unre-stricted Wg signaling can induce en expression in inappropriate rows of cells, expanding the en expression domain in a posterior direction (Noordermeer et al., 1992; Bejsovec and Wieschaus, 1993; Yoffe et al., 1995). Wg activity also regulates Armadillo protein levels within the cell (Riggleman et al., 1990), stabilizing the cytoplasmic pool of Arm protein, which then acts as an effector of the Wg signaling pathway (Peifer et al., 1994; Pai et al., 1997).

Recent work has identified several new components of the Wingless signaling pathway, including a cell surface receptor for Wg, Dfrizzled 2 (Dfz2) (Bhanot et al., 1996), and a transcription factor, dTCF (Brunner et al., 1997; van de Wetering et al., 1997), which when bound to stabilized Armadillo is able to activate target gene expression (Riese et al., 1997; van de Wetering et al., 1997). However, it is still unclear how Wg signaling differentially specifies the cuticular cell fates. Previous work identified two wg mutant alleles that show non-overlapping subsets of the wild-type Wg signaling activities (Bejsovec and Wieschaus, 1995). wg^PE2 mutants secrete diverse denticle types but no naked cuticle, while wg^PE4 mutants secrete patches of naked cuticle in an expanse of denticles showing little diversity. Curiously, both mutant lines show similar molecular responses to Wg signaling: both stabilize wg and en expression, but neither shows significant stabilization of Arm protein.

These mutations partially complement each other, together producing an almost wild-type cuticle pattern, indicating that they disrupt different signaling activities of Wg independently. These signaling activities, however, do not map to specific functional domains within the molecule. The wg^PE2 lesion causes a missense change near the carboxy terminus, initially suggesting that this region might be required for naked cuticle specification. However, this region is deleted in wg^PE4 mutants, which retain the ability to specify naked cuticle. The wg^PE4 mutation introduces a stop codon at position 250, producing a molecule that is truncated severely but is partially functional in patterning the cuticle. Thus the overall conformation of Wg appears to determine which cellular response is elicited.

One can imagine two general mechanisms by which Wg conformation could influence cell response: (1) The distinct responses may be transduced through different pathways, utilizing at least two different classes of receptor for Wg. One class of receptor, when activated by Wg binding, transduces naked cuticle cell fate while the other, when activated, induces cell fate diversity. The mutant Wg molecules may be altered in their binding affinity for one set of receptors versus the other, and thus each would activate only one of the cellular response pathways. (2) Differing Wg levels across the segment may direct the distinct responses. The wg-expressing row of cells is roughly centered in the naked cuticle-secreting region of the segment (Dougan and DiNardo, 1992; Peifer and Bejsovec, 1992). This suggests that naked cuticle specification may be the cellular response to the very high levels of Wg protein found in the wg-expressing row of cells and its immediate neighbors. Ectopic expression of wild-type Wg causes all cells of the segment to secrete naked cuticle when Wg levels are uniformly high (Noordermeer et al., 1992; Lawrence et al., 1996). Lower levels of Wg moving out across the segments, below the threshold level for specifying naked cuticle, may promote cell fate diversity. The wg^PE2 and wg^PE4 mutant phenotypes might then be explained by altered protein distribution rather than by disrupted signal transduction. However, wg^PE2 mutant protein shows a distribution similar to wild-type Wg protein, and wg^PE4 mutant protein is not detected with existing Wg polyclonal antisera.

To determine the mechanism by which Wg acts to specify unique cuticular cell fates and to address questions about how Wg structure relates to its signaling function, we have engineered wild-type and mutant wg transgenes. We have expressed these molecules ubiquitously or in defined domains within the segment and assessed their ability to pattern the cuticle and to influence target gene expression. These transgenic molecules were also tagged with an epitope, inserted at a site that does not interfere with Wg function, to provide an unbiased means of determining transgenic protein distribution. We find that, within the Wg protein, an 85 amino acid region that is not conserved in other members of the Wg/Wnt family (Rijsewijk et al., 1987) is dispensable for Wg signaling activity, and that a 32 amino acid region amino-terminal to the nonconserved region inhibits function of truncated molecules. Furthermore, we find that the signaling activities of Wg are mediated by two distinct mechanisms. wg^PE2 protein fails to direct naked cuticle specification under all experimental conditions examined. This supports the idea that the mutant protein activates only one of two independent signaling pathways, possibly involving two discrete classes of receptor. In contrast, wg^PE4 protein can promote either cellular response when provided ectopically, demonstrating that its failure to promote denticle diversity results from defective protein transport. Thus both models stated above are tenable: multiple signaling receptors and proper Wg protein distribution both may be required for correct epidermal patterning in the wild-type embryo.

All constructs were derived from the wild-type wg cDNA, subcloned in a pSP65 vector (Rijsewijk et al., 1987). The wgmutant lesion, changing valine 453 to glutamic acid, was introduced by in vitro mutagenesis using a Unique Site Elimination (U.S.E.) Mutagenesis kit from Pharmacia. The wgmutant lesion was engineered by a PCR-based strategy, introducing a nonsense mutation at amino acid 250 followed by a restriction site to allow insertion of the PCR fragment into the 3′ polylinker. A similar strategy was used to generate wgΔ, an in-frame deletion that removes the nonconserved 85 amino acid region, taking advantage of the EcoRV site at the 3′ end of the non-conserved region. During the course of this subcloning, a second transgene, called wg, was generated. This transgene encodes a molecule truncated at amino acid 282, where the nonconserved region begins (see Fig. 1G).

The 10-amino-acid hemagglutinin epitope was inserted in frame at amino acid 111, the position of a unique XhoI site in the wg cDNA sequence. All experiments were performed with both tagged and untagged wild-type wg transgene controls and identical phenotypic effects were observed. Constructs were subcloned into the Casper transformation vectors pCasper-hs-act (Thummel et al., 1988), which drives transgene expression with the hsp70 promoter, and pUAST (Brand and Perrimon, 1993), which allows transgene expression to be directed by a variety of GAL4 lines. In addition, a polyadenylation sequence flanked by FLP recombinase target sites (Golic and Lindquist, 1989), designed by R. Holmgren (Buenzow and Holmgren, 1995), was inserted between the promoter and the wg coding sequences to prevent ectopic expression in injected embryos. All constructs were verified by sequence analysis. Plasmids were purified by Qiagen column preparation and injected into embryos to obtain germline transformants. When transgenic lines were established, a source of FLP recombinase was introduced into each line to excise the polyadenylation sequence and render the transgene functional. For each construct, at least 10 independent transformant lines were generated and tested for expression level. Lines showing comparable levels of expression by western blot and by RNA in situ hybridization were selected for the experiments described in this paper. Stability of the mutant proteins was comparable to the wild-type protein: this was assessed by inducing expression of the heat-shock transgenes with a single heat pulse, fixing embryos at defined time points following induction and detecting the transgenic protein in situ with antibody staining. For functional analysis in the absence of endogenous wg activity, transgenes were crossed into the wgmutant background. This mutant allele produces no wg RNA or protein (Baker, 1987).

The following stocks were used to drive expression of the UAS-wg transgenes: wg-GAL4 from J. Pradel (personal communication), prd-GAL4 (Yoffe et al., 1995), and E22C-GAL4, a ubiquitous constitutive driver line from the Bloomington stock center. For heat-shock induction of the hsp70 promoter-driven transgenes, embryos were collected at 1 hour time intervals, aged to appropriate developmental stage, dechorionated and placed in a 36°C water bath for 30 minutes. To achieve maximal induction, a multiple heat-shock regimen was followed: three consecutive 30 minute heat shocks beginning at 3 to 4 hours of development were performed, with 1 hour of recovery at 18°C in between heat shocks. The 18°C recovery temperature was necessary because the wgprotein is temperature sensitive.

For cuticle preparations, embryos were allowed to age at 18°C and then mounted in Hoyer’s medium (van der Meer, 1977). For antibody staining, embryos were fixed in 4% formaldehyde 30 minutes after the final heat shock and processed according to DiNardo et al. (1985). Anti-hemagglutinin monoclonal antibody, a gift from R. Lamb, was used at 1:10,000 dilution. Two different anti-Wg rabbit polyclonal antisera, gifts from S. Cumberledge and R. Nusse (van den Heuvel et al., 1989), were used at 1:1,000 dilution. Anti-En monoclonal antibody was used at 1:50 dilution. Endogenous wg expression was assayed using a lacZ enhancer trap inserted into the wg locus. Anti-β-galactosidase antibody was used at 1:2,000.

Transgenic lines expressing at similar levels were assayed for function in both wild-type and wg mutant backgrounds. In a wild-type background, three of the mutant wg transgenes produce phenotypic effects similar to the wild-type wg (wg⁺) transgene. This is particularly clear in their effects on expression of engrailed, a target gene activated in response to Wg signaling. Ubiquitous wild-type wg expression causes an expansion in the en expression domain, increasing its width several cell diameters posterior to its normal, 1- to 3-cell-wide, stripe of expression (Fig. 1A,B, see also Noordermeer et al., 1992). When expressed with either the heat-shock promoter or the ubiquitous E22C-GAL4 driver, wg^PE2 (Fig. 1C), wg^PE4 (Fig. 1E) and wg^Δ85 (Fig. 1F) also expand the engrailed expression domain. Thus these three transgenes are functional and provoke a wild-type Wg signaling response with regard to en regulation. Other phenotypic effects of Wg signaling are also observed, such as increased expression of the endogenous wg gene (wg^PE2 and wg^PE4 not shown, wg^Δ85 see Fig. 2E) and widespread specification of naked cuticle (wg^PE2 see Fig. 4D, wg^PE4 and wg^Δ85, not shown).

The wg²⁸² transgene, however, shows none of these effects of normal Wg signaling activity, even though it is expressed at a level equivalent to that of the other transgenes (see Materials and Methods). In fact, ubiquitous wg²⁸² expression causes a slight dominant negative effect, narrowing the expression domain of engrailed (Fig. 1D). It is unclear whether this disruption is due to a passive effect on general Wg metabolism, such as overloading the secretory apparatus, or to an active competition for cell surface receptors. However, the reduction in en expression does not correlate with any significant deviation from the wild-type cuticle pattern. The observation that this truncated molecule lacks signaling activity, while the shorter wg^PE4 mutant protein shows activity, suggests that the longer molecule includes a sequence that exerts a negative effect on Wg signaling.

The Wingless protein has an insert of 85 amino acids relative to its closest vertebrate homologue, Wnt-1 (Rijsewijk et al., 1987). This nonconserved insert is hydrophilic and contains no cysteine residues, suggesting that it may loop out from the folded structure of the Wg molecule. To test its contribution to Wg signaling, we generated an in-frame deletion of the region (Fig. 1G). This deleted molecule, wg^Δ85, shows normal protein distribution when detected using the hemagglutinin (HA) epitope tag (Fig. 2A,D), but is undetectable using several different Wg polyclonal antisera (Fig. 2B,E). Thus the nonconserved region appears to serve as the primary epitope recognized by antibodies against Wg. Indeed, we find that all Wg molecules missing this 85 amino acid region, including the wg^PE4 protein, are not detected in situ with existing Wg polyclonal antisera. They are, however, detectable using western blotting techniques, suggesting that other epitopes are revealed when the protein is denatured (unpublished).

wg^Δ85 shows normal signaling activity. When expressed in alternate segments using a prd-GAL4 driver, the UAS-wg^Δ85 promotes increased expression of the endogenous wg gene, as does the wild-type UAS-wg⁺ line (Fig. 2C,E). Furthermore, it is able to rescue the wg null mutant cuticle pattern to the same extent as does the wild-type wg transgene (Fig. 2F, see also Yoffe et al., 1995). This rescue is diminished at 18°C whereas rescue with the full-length wild-type Wg protein is not, indicating that the wg^Δ85 protein is cold-sensitive relative to the wild-type protein.

Uniform overexpression of wg⁺, wg^PE2, wg^PE4 or wg^Δ85 transgenic protein in a wild-type background results in specification of excess naked cuticle. This is due in part to their ability to increase expression of the endogenous, wild-type wg gene product. When the contribution from endogenous wg is removed by assaying the transgenes in a wg null mutant background, the transgenes show differences in phenotypic effect.

hs-wg^PE2 expression, induced by three consecutive heat shocks, restores denticle diversity to a wg null mutant embryo (Fig. 3A,B). Under identical conditions, the hs-wg^PE4 (Fig. 3C) and wild-type hs-wg⁺ (Fig. 3D) transgenes restore both denticle diversity and naked cuticle, in an array that approximates the wild-type cuticle pattern (Fig. 3E). Higher uniform expression levels are produced with the E22C-GAL4 driver line. Under these conditions, UAS-wg^PE2 (Fig. 3F) still rescues only denticle diversity while the UAS-wg^PE4 (Fig. 3G) and wild-type UAS-wg⁺ (Fig. 3H) transgenes promote specification of naked cuticle in all cells of the segment. Thus the wg^PE2 mutant protein is unable under any circumstances to promote the naked cuticle cell fate in ventral epidermal cells.

Since denticle diversity and naked cuticle specification are temporally distinct events (Bejsovec and Martinez-Arias, 1991), it is possible that wg^PE2 mutant protein interacts with a Wg receptor that is expressed only at early times and simply is not present at the time epidermal cells are instructed to secrete naked cuticle. To explore this possibility, we induced expression of the hsp70-driven transgenes by a single heat pulse at defined developmental time points. First we assessed the effects in a wild-type background. When induced with a single heat shock at 4 hours of development, both the wild-type hs-wg⁺ transgene (Fig. 4A) and the hs-wg^PE2 transgene (Fig. 4D) induce an ectopic stripe of endogenous wg expression and produce an excess of naked cuticle. For either transgene, a single heat shock at 6 hours of development does not induce an ectopic stripe of wg expression (not shown). While the wild-type hs-wg⁺ (Fig. 4B) directly instructs cells to secrete naked cuticle under these conditions, hs-wg^PE2 subtly changes the denticle pattern but makes no naked cuticle (Fig. 4E). Thus, the autoregulatory function of wg requires early wg input. wg^PE2 is able to promote this function and activation of the endogenous wg gene is responsible for the excess naked cuticle observed. When expression is induced at later times, auto-activation of the endogenous wg gene does not occur, and only the wild-type wg transgenic protein is able to promote naked cuticle secretion. However, expression of the hs-wg^PE2 protein at 6 hours of development does produce an effect on the denticle pattern. A partial loss or rearrangement of denticles in the second and third row of the denticle belt was observed reproducibly in these embryos. This indicates that late expression of wg^PE2 is able to produce phenotypic effects and so the mutant protein presumably interacts with receptor molecules expressed at later times.

These effects were confirmed by repeating the early and late heat shocks in a wg null mutant background. A single early heat shock caused both hs-wg⁺ and hs-wg^PE2 to generate denticle diversity (not shown), whereas a single late heat shock resulted in naked cuticle specification by the hs-wg⁺transgene (Fig. 4C), but none for the hs-wg^PE2 transgene (Fig. 4F). Both transgenes, however, produce dramatic effects on denticle orientation at this time. When induced at 6 hours, both hs-wg⁺ and hs-wg^PE2 redirect denticles to point toward the ventral midline. It is not clear why the normal anterior-posterior denticle polarity is disrupted by late uniform expression of wg. However, this alteration of the pattern demonstrates that hs-wg^PE2 still promotes some signaling function in epidermal cells at late stages, suggesting that its receptor is still expressed at these times. These data suggest that the wg^PE2 mutant protein can influence denticle type or orientation at all times tested, but may be unable to bind and/or activate a particular class of Wg receptor specific to a naked cuticle cell fate pathway.

Uniform expression of either hs-wg^PE4 (Fig. 3C) or hs-wg⁺ (Fig. 3D) produces diverse denticle types interspersed with expanses of naked cuticle in a pattern that approximates the wild type (Fig. 3E). This pattern is somewhat different from the symmetrical ‘mirror-image’ duplication of denticles that has been reported previously for overexpression of a wild-type wg transgene in a wg mutant background (Sampedro et al., 1993). This ‘mirror-image’ effect is observed when our transgenes are induced with a single heat shock at early stages. Continuous uniform expression induced by multiple heat shocks generates a more substantial rescue, restoring the normal asymmetry of the pattern. The denticle belts produced show correct denticle types and polarity, suggesting that, in the wg mutant background, some patterning information already distinguishes epidermal cell populations within the segment. Cells receiving the same level of Wg signaling respond differently depending on their position within the segment: some cells are directed to secrete naked cuticle, while others secrete diverse denticle types. This observation hints at a differential distribution of molecules that interact with Wg to effect its signaling activity.

All cells within the segment, however, are competent to secrete naked cuticle even in a wg null mutant background. When expression levels of either wg^PE4 (Fig. 3G) or wg⁺ (Fig. 3H) are driven at higher levels with the E22C-GAL4 ubiquitous promoter, all cells in the ventral epidermis produce naked cuticle with few or no denticles. This indicates that all cells in the segment can be directed into the naked cuticle cell fate, and therefore presumably would express a naked cuticle specifying Wg receptor. Furthermore, this cell fate pathway appears to be triggered when Wg activity exceeds a certain threshold level. Dose sensitivity for naked cuticle specification is also observed under a variety of other circumstances. When wg⁻ embryos carrying the hs-wg⁺ transgene are treated with single or double heat shocks, no naked cuticle is produced, although these embryos do show denticle diversity, en stabilization and wg autoregulation as measured by wg-lacZ expression (not shown). Furthermore, previous work has demonstrated that the excess naked cuticle produced in a naked mutant embryo can be suppressed by reducing the gene dosage of wg by half (Bejsovec and Wieschaus, 1993).

When expressed uniformly, wg^PE4 transgenes (Fig. 3C,G) rescue the wg mutant phenotype in a manner similar to the wg⁺ transgenes (Fig. 3D,H), promoting both denticle diversity and naked cuticle specification. This contrasts with the original wg^PE4 mutant cuticle pattern, which shows little denticle diversity. One potential explanation for these disparate results is that the wg^PE4 mutant protein may have a limited range of action. To test this possibility, we expressed the transgenic protein in the native wg domain, using a wg-GAL4 line, and examined its distribution by detecting the hemagglutinin epitope tag. We find that, while UAS-wg⁺ (Fig. 5A) and UAS-wg^PE2 (Fig. 5B) transgenic protein can be detected over several cell diameters surrounding the wg-expressing row of cells, UAS-wg^PE4 (Fig. 5C) cannot. Rather, this transgenic protein shows limited movement away from the wg-expressing row of cells. The transgenic protein also appears to accumulate to higher levels within the wg-expressing domain, suggesting that it is not exported as efficiently as the full-length Wg molecule. Reduced secretion would contribute to the limited distribution of the mutant protein, but functional analyses (see below) indicate that at least some wg^PE4 mutant protein reaches neighboring cell populations.

When expressed in this restricted domain, wg^PE4 (Fig. 5F) cannot rescue the wg mutant pattern to the same extent as the wg⁺ transgene (Fig. 5D). UAS-wg⁺ expression fully rescues the ventral pattern (Fig. 5D), but the embryos are slightly distorted because dorsal pattern elements are not rescued. Dorsal cell fates require early input from Wg signaling (Bejsovec and Martinez-Arias, 1991) and wg-GAL4-driven transgene expression is slightly delayed with respect to the native wg promoter. Unlike UAS-wg⁺, UAS-wg^PE4 expression does not fully rescue the ventral surface; rather, these embryos show a substantial expanse of naked cuticle but little denticle diversity (Fig. 5F). Thus, the ventral cuticle pattern of the original EMS-induced allele is recapitulated when transgene expression is driven by wg-GAL4 in a wg mutant background. Likewise, under these conditions, UAS-wg^PE2 expression generates denticle diversity but no naked cuticle (Fig. 5E).

Expressing UAS-wg^PE4 in a restricted domain diminished, but did not eliminate, its ability to regulate en expression in the neighboring row of cells. The engrailed-expressing cells, just posterior to the wg-expressing row of cells, require Wg signaling activity for continued expression of en (DiNardo et al., 1988; Martinez Arias et al., 1988). In a wg mutant embryo, en expression decays by stage 9. When either the UAS-wg⁺ or UAS-wg^PE2 transgenes are expressed with the wg-GAL4 driver, we see en expression stably maintained even in quite late stage embryos (Fig. 6A,B). Both UAS-wg⁺ and UAS-wg^PE2 rescue the ventral portion of the en stripe to the same extent; again, dorsal cells do not show stabilized en expression because they require early input not provided by the GAL4 system (Bejsovec and Martinez-Arias, 1991;Heemskerk et al., 1991). UAS-wg^PE4 (Fig. 6C) also shows stabilization of ventral en expression, but to a lesser extent. This demonstrates that the mutant protein is exported since it can trigger response in a neighboring cell population. However, it does not appear to accumulate as efficiently in this cell population; fewer cells receive levels of Wg signaling sufficient to trigger en maintenance.

Wg protein transport can be tested even more rigorously by expressing the transgenes under the control of the prd-GAL4 line. prd-GAL4 drives expression in a 5-cell-wide domain encompassing both the wg- and en-expressing rows of cells, in odd-numbered abdominal segments (Yoffe et al., 1995). In a wg mutant background, we can assess Wg protein transport away from the prd expression domain by measuring rescue of en expression in the even-numbered segments. UAS-wg⁺ (Fig. 6D) or UAS-wg^PE2 (Fig. 6E) stabilize, and slightly expand, the en expression domain in the prd stripes. In addition, en expression is stabilized and expanded in a ventral portion of the alternate segments (arrows). This pattern may reflect how long it takes for sufficient Wg to move to this distant cell population. Ventral cells are the last to be specified to maintain en expression in response to wg function (Bejsovec and Martinez-Arias, 1991), and may be the only cells still competent to stabilize en expression by the time sufficient Wg protein has accumulated. Alternatively, dorsoventral differences in transport of Wg protein might reduce movement through the dorsal portions of the epidermis and/or enhance movement through the ventral midline cells. In either case, both UAS-wg⁺ or UAS-wg^PE2 show the same degree of rescue in this distant ventral cell population, indicating that both molecules are equally able to move across many cell diameters. In contrast, UAS-wg^PE4 (Fig. 6F) shows no stabilization of en in this domain, although it stabilizes and expands en expression within the prd domain to the same extent as do UAS-wg⁺ and UAS-wg^PE2. This is consistent with the observed restriction in distribution of the wg^PE4 transgenic protein.

The manipulated expression of wg^PE2 and wg^PE4 has revealed different biological functions for the Wingless protein. wg^PE2 is able to auto-activate the endogenous wg and to activate and stabilize en expression in adjacent cell populations, but it is unable to direct the specification of naked cuticle. Because target gene response is identical to that of the wild-type protein, we believe that specification of naked cuticle may occur through a distinct signaling pathway, perhaps utilizing a discrete class of cell surface receptor molecules. The idea of distinct classes of receptor molecule, which each transduce a different cellular response, is further supported by the observation that different populations of cells within the segment respond differently when provided with the same low level of ubiquitous wg expression. It is unlikely to result exclusively from different expression domains for each receptor class, since all cells of the segment can be induced to secrete naked cuticle when Wg levels are sufficiently high. Rather, this effect could result from different cell surface densities of specific receptor classes, or from differential distribution of receptors with higher and lower affinities for Wg binding. In this regard, it is intriguing to note that the Dfrizzled2 receptor shows a segmentally repeating pattern of elevated expression during developmental stages 9 through 12 (Bhanot et al., 1996), when Wg acts to specify the epidermal pattern elements (Bejsovec and Martinez-Arias, 1991).

In contrast, analysis of the wg^PE4 mutant reveals that generation of cell fate diversity depends critically on proper distribution of Wg protein across the segment. This supports the idea that Wg can act as a long-range signaling molecule, since the denticles are secreted by cells that lie at a distance from the wg-expressing row of cells (Fig. 7A). Our work argues against the possibility of patterning through short-range relay mechanisms. When expressed in a restricted domain using either the wg-GAL4 or the prd-GAL4 driver lines, transgenic wg^PE4 mutant protein is able to direct the stabilization of en expression in adjacent epidermal cells. These responding cells apparently do not initiate further signaling events. None of the long-range effects on en expression or on cuticular pattern element specification produced by the wild-type wg transgene are observed. Restriction of Wg signaling activity was also observed when Wg transport was blocked by the shibire mutation (Bejsovec and Wieschaus, 1995). These data suggest that the wild-type Wg molecule promotes response in distant cell populations directly. Analogous long-range effects of Wg in imaginal discs have been inferred from experiments using a transgenic Wg protein that is artificially tethered to the cell membrane (Zecca et al., 1996).

The basis for denticle diversity is still mysterious. Denticle type is not specified simply by level of Wg activity received, since uniform low levels of Wg can promote correct diverse denticle identities. Previous work has indicated that other segment polarity genes, such as patched and engrailed, play a role in determining denticle identity (Bejsovec and Wieschaus, 1993). These gene activities may be involved in dictating the ‘pre-pattern’ observed when uniform wg expression is provided to a wg mutant embryo. Indeed, when ectopic wild-type Wg is provided in a wg en double mutant background, naked cuticle is rescued but very little denticle diversity is observed (Lawrence et al., 1996).

The wg^PE4 mutant protein is competent to generate denticle diversity, as well as naked cuticle cell fate, when it is provided ectopically. Therefore, it appears to be specifically defective in export and transport by epidermal cells. Wg protein transport requires active cellular processes related to endocytosis, since blocking endocytosis restricts Wg protein distribution and function (Bejsovec and Wieschaus, 1995). This suggests that wg^PE4 protein may be defective in its interaction with a class of cell surface receptor that is uniquely involved in the transport process. Further support for a separate class of transport receptor comes from analysis of the longer truncated molecule encoded by the wg^CE7 allele (Bejsovec and Wieschaus, 1995). wg^CE7 mutant protein is distributed over many cell diameters on either side of the wg-expressing row of cells and is internalized into cells without transducing any of the known signaling activities of Wg (Bejsovec and Wieschaus, 1995). wg^CE7mutants secrete a cuticle pattern indistinguishable from that of null mutants; they show no stabilization of cyto-plasmic Armadillo protein, and no stabilization of wg and en expression. Thus intercellular movement of Wg protein, to produce a broad distribution across the segment, is independent of Wg signaling activity.

Analysis of the wg^CE7 mutant protein also suggested that a portion of the Wg molecule may have an inhibitory effect on Wg function. This molecule is truncated at amino acid 367, at the end of the nonconserved region, yet shows no activity whereas the shorter wg^PE4 molecule does show activity. Removing the nonconserved region does not disrupt Wg activity; the wg^Δ85 transgenic protein is distributed properly and shows wild-type Wg signaling in both pattern rescue and molecular response. Therefore this nonconserved region apparently contributes neither positive nor negative effects. The negative effects appear to depend on the 32 amino acids lying between the position of the wg^PE4 truncation and the position of the wg²⁸² truncation. The wg²⁸² transgene shows no evidence of Wg signaling activity and, indeed, produces a slight dominant negative effect. We suspect that the presence of the 32 amino acid region may alter protein conformation in a way that prevents the mutant molecule from interacting with its receptor(s), and that this conformational constraint is regulated in the full-length Wg protein by interaction with the region carboxy terminal to the nonconserved region.

Our current model for how Wg signaling generates pattern is depicted in Fig. 7B. We propose that three classes of receptor for Wg protein may mediate the distinct aspects of Wg biology revealed by our transgene analysis. Two separate signaling receptors may be responsible for the different cellular responses apparent in the cuticle pattern: diverse denticle types and naked cuticle. The putative naked cuticle-specifying receptor must be present in all ventral epidermal cells (although not necessarily at the same level) and its activation is sensitive to Wg dose, since all cells can be directed to secrete naked cuticle if they receive sufficiently high levels of Wg signaling activity. The wg^PE2 mutant protein appears to be specifically unable to activate this class of receptor. It can, however, activate a separate class of denticle-diversity generating receptor. This, or a separate receptor, may also be responsible for transducing the signal that stabilizes en expression in adjacent cell populations, since wg^PE2 can perform this aspect of wild-type Wg function. Alternatively, it is possible that a single signaling receptor triggers both cellular pathways by some unknown means and that the wg^PE2 mutant protein is unable to produce maximal activation of this receptor, which is required for naked cuticle cell fate specification. In any case, we propose a distinct class of cell surface receptor for Wg that is devoted exclusively to its proper transcytotic movement. wg^PE4 mutant protein has normal signaling activity but is not distributed properly, while the previously characterized wg^CE7 protein shows no signaling activity but normal distribution. Thus transport and signaling can be completely uncoupled, suggesting that separate pathways are responsible. It remains to be determined which of these receptor classes is related to Dfz2, a 7-membrane-spanning protein that clearly functions as a receptor for Wg in cell culture (Bhanot et al., 1996).

We are deeply indebted to R. Holmgren for sharing his FRT technology along with many other reagents and ideas. We thank R. Lamb, S. Cumberledge, R. Nusse, and M. Peifer for antibodies, J. Pradel, H. Krause and A. Brand for fly stocks, and H. Dierick, R. Holmgren, and E. Goodwin for comments on the manuscript. We also thank F. W. Outten, T. Sproul and S. Li for assistance during their laboratory rotations. This work is supported by Basil O’Connor Starter Scholar Research Award FY95-1107 from the March of Dimes Birth Defects Foundation and by the National Science Foundation under Grant No. 96-00539.