Mutual antagonism between signals secreted by adjacent Wingless and Engrailed cells leads to specification of complementary regions of the Drosophila parasegment

Organizers specify pattern across cellular fields (Spemann, 1938). Work has focussed on elucidating the properties of organizers because, aside from revealing how the specific tissue is patterned, principles uncovered for one organizer will apply to others. In Drosophila, cellular fields are established with the subdivision of the body plan into parasegments (Martinez-Arias and Lawrence, 1985). Signals emanating from cells at the boundary between adjacent parasegments guide patterning across each parasegment in the epidermis (Baker, 1988; Martinez-Arias et al., 1988; Bejsovec and Martinez-Arias, 1991; Heemskerk and DiNardo, 1994; Bokor and DiNardo, 1996). Similar mechanisms act in patterning the imaginal discs, as signals emanate from the compartment boundaries that are inherited from embryonic parasegment boundaries (Garcia-Bellido, 1975; Basler and Struhl, 1994; Capdevila and Guerrero, 1994; Tabata and Kornberg, 1994; reviewed in Lawrence and Struhl, 1996). Since these signals act to establish cell type diversity across the field, the parasegment and compartment boundary are analogous to classical pattern organizers. Thus, identifying signals from parasegment or compartment boundaries and understanding their control will provide general insight into organizer function.

Screens have identified mutations in the segment polarity genes involved in parasegmental patterning (Nusslein-Volhard and Wieschaus, 1980), and their molecular analyses is providing a paradigm for organizer function. One of the signals secreted by this organizer is the Wnt gene family member, Wingless (Wg) (Cabrera et al., 1987; Rijsewijk et al., 1987; Baker, 1988). It is expressed anterior to cells expressing the homeoprotein Engrailed (En; Ingham et al., 1988). The Enexpressing cells co-express Hedgehog (Hh), member of another signaling family (Lee et al., 1992; Mohler and Vani, 1992; Tabata et al., 1992; Krauss et al., 1993; Riddle et al., 1993). There are two phases to patterning the epidermis. First, reciprocal signaling occurs between Wg- and En/Hh-expressing cells, serving to consolidate parasegmental subdivision of the body plan (DiNardo et al., 1988; Martinez-Arias et al., 1988; Bejsovec and Martinez-Arias, 1991; Heemskerk et al., 1991; Cumberledge and Krasnow, 1993; Ingham, 1993). Subsequently, cells fates are established (Bejsovec and Martinez-Arias, 1991; Dougan and DiNardo, 1992).

Ventrally, Wg signaling is required to establish fate of half of the parasegment, specifically the region that differentiates into smooth cuticle (Baker, 1988; Bejsovec and Martinez-Arias, 1991; Dougan and DiNardo, 1992; Noordermeer et al., 1992; Pai et al., 1997). The signals specifying remaining fates within each parasegment, those that differentiate into diverse denticle types in abdominal segments 2-7, are less clear, although Wg signaling has been implicated (Bejsovec and Martinez-Arias, 1991; Bejsovec and Wieschaus, 1993; Yoffe et al., 1995; Lawrence et al., 1996). One difficulty in establishing a direct role for Wg in generating denticle diversity is to separate that contribution from its early role in maintaining En/Hh expression; especially since we previously found that Hh plays a significant, Wg-independent role in organizing dorsal epidermal pattern (Heemskerk and DiNardo, 1994). That work coupled with two studies on ventral pattern (Bejsovec and Wieschaus, 1993; Sampedro et al., 1993) lead us to consider the contributions of the En- and Hh-expressing cells in generating denticle diversity. We recently showed that the ligands Spitz and Wingless emanate from distinct borders of the En domain, and proper patterning across the En domain is guided by the balance between these two antagonistic signals (O’Keefe et al., 1997). Further results suggested that Spitz was also required for some denticle cell fates just posterior to the En domain (Szüts et al., 1997).

We now bypass the need for Wg in the stabilization of En/Hh expression to investigate the role of En and Hh in generating denticle diversity. This establishes definitive evidence for Wg-independent specification of anterior denticle fates. In addition, we uncover evidence that Wg and Hh activity each ensures that signaling from the organizer is asymmetric, thereby generating proper intrasegmental pattern.

smo³, hh^13C, wg^cx4 and wg^IL (wg^ts) were from Bloomington; Df(3R)GR2 was from J. Mohler; Df(2L)NL Df (2R) en^SF31 was from A. Bejsovec. A smo³ FRT 40A and a wg^cx4smo³ FRT 40A chromosome was used to generate germline clones (Chou et al., 1993), and the resulting females crossed to wg^cx4smo³ FRT 40A / CyO males. UAS-Arm-S10 is described in Pai et al. (1997); UAS-dTCF delta N dominant negative and En-GAL4 were from M. Peifer (Cavallo et al., 1998) and A. Brand. Ptc-GAL4 is described in Speicher et al. (1994). To bypass the requirement for Wg in maintaining En expression, we constructed and crossed wg^cx4 En-Gal4 / CyO × wg^cx4 UAS-ArmS-10/CyO.

Cuticles were prepared as usual (van der Meer, 1977). β-galactosidase activity stains on cuticle preparations (Heemskerk and DiNardo, 1994) were performed using Wg-lacZ (wg1-en11; Kassis et al., 1992) or En-lacZ (Xho-25; Hama et al., 1990) reporters.

RNA in situ antibody double labeling was as in Dougan and DiNardo (1992). Digoxigenin-labeled RNA probes were from cDNAs for Wg, Rhomboid and Serrate (Baker, 1987; Bier et al., 1990; Thomas et al., 1991). Anti-mAb4D9 (En/Inv) was used undiluted (Patel et al., 1989); anti-Patched at 1:200 (Capdevila et al., 1994); anti-β-galactosidase (Cappel) at 1:1,000; biotinylated antibodies (Vector), Cy2- and Cy3-labeled secondary antibodies (Jackson), and HRP-conjugated streptavidin (Chemicon) at 1:400. Embryos were filletted in 80% glycerol; some data were collected on a Zeiss LSM 510 confocal microscope; images were processed in Adobe Photoshop.

Ventrally across each segment of the second through seventh abdominal segments six rows of epidermal cells choose to differentiate smooth cuticle, while the next six rows secrete cuticle that has protrusions called denticles (Fig. 1A; Lohs-Schardin et al., 1979). These denticle-secreting cells form a belt, wherein denticle rows can usually be distinguished from each other based on size and polarity. Row five denticles are largest, with tapered, posteriorly pointing tips. Row six has tiny denticles of indeterminate polarity. Denticles of the first four rows are all smaller than row five, and their tips are hooked more than tapered. Rows one and four point anteriorly, while two and three point posteriorly.

Wingless and Engrailed are essential in organizing this pattern. Wg-expressing cells lie anteriorly adjacent to the En cells. While the Wg-expressing cells differentiate into smooth cuticle (Fig. 2A; O’Keefe et al., 1997), the posterior-most En-expressing cells differentiate into row 1 denticles (Fig. 2B; Dougan and DiNardo, 1992). Embryos null for both wg and en have no smooth cuticle, and the remaining denticles are row 5-like, mostly large with tapered tips (Figs 1B, 2C; Bejsovec and Martinez-Arias, 1991). While this demonstrates that Wg and En together are essential for normal pattern, the individual contribution made by the En-expressing cells is not clear.

In contrast, Wg activity is necessary and sufficient for smooth fates (Wieschaus and Riggleman, 1987; Baker, 1988; Lawrence et al., 1996; Hays et al., 1997; Pai et al., 1997). Smooth fates are specified late, after the Wg-dependent stabilization of en gene expression. Removing Wg function after about 5 hours after egg laying (AEL) by wg^ts upshifts, or using Ptc-GAL4 to express a dominant negative form of dTCF (Molenaar et al., 1996; Brunner et al., 1997; van de Wetering et al., 1997), thereby inhibiting Wg transduction in cells outside the En domain, leads to loss of smooth cell fates (Fig. 2D). The pattern within the denticle belt is quite normal in such embryos (Bejsovec and Martinez-Arias, 1991).

Furthermore, in place of smooth cuticle, there is an approximate mirror-image duplication of denticle pattern (Fig. 2D). This suggests that, in the absence of late Wg function, whatever signal(s) organizes denticle pattern now acts bi-directionally, generating a mirror-image pattern. It has been suggested that early Wg function specifies denticles diversity. However, even in wg null mutants, there is remaining mirror-pattern (Fig. 1C), as large row 5-type denticles (Fig. 2E, large arrowheads) alternate with small row 2-4 type denticles (Fig. 2E, small arrowheads). Thus, in the absence of Wg, any remaining signal(s) now acts symmetrically, generating a roughly mirror-image pattern.

A signal made by the En cells could be responsible for production of the now symmetrically acting signal. In wg mutants, this signal would persist only transiently, as En expression decays early. If true, then maintaining En expression should generate more organized pattern. We therefore examined the pattern generated when En expression is maintained in the absence of Wg function. Previously, we accomplished this in dorsal epidermis by precociously activating En autoregulation using Hs-En in wg mutants (Heemskerk and DiNardo, 1994). Here we use En-GAL4 to express an activated form of the Wg signal transducer Armadillo in the En-expressing cells in wg nulls. This bypasses the transient requirement for Wg in maintaining En expression, leading to significant stabilization of En-expressing cells (see Fig. 3D). By cell intrinsically activating the Wg pathway in En cells, this maintains Hedgehog and other Wg-responsive genes in the En cells, while removing any contribution of Wg to patterning any other cells.

Maintaining En-expressing cells leads to obvious differences comparing the pattern of wg^cx4 En-Gal4 UAS-ArmS10 embryos with wg mutants. First, strips of segmentally repeated smooth cuticle are often found (Fig. 1D). These regions of smooth cuticle mark the cell-autonomous differentiation of En cells expressing activated Armadillo (O’Keefe et al., 1997; Pai et al., 1997; Sanson et al., 1999). Second, there are fewer large, tapered row 5-type denticles; nor are there any tiny row 6-type denticles. This suggests a loss of posterior belt identities (Fig. 2F). Instead, the denticles formed are small and hooked, row 2-or 3-type denticles (Fig. 2F, arrows).

Significantly, the denticle pattern exhibits some mirror symmetry. An axis of symmetry lies within each strip of En-expressing smooth cells as denticles point away from the nearest region of smooth cuticle (Fig. 2F). Thus, maintaining En in the absence of Wg leads to loss of posterior fates and duplication of anterior denticle identities. Given that En maintenance by activated Armadillo is cell autonomous, the changes in denticle patterning outside the En domain suggest strongly that En-expressing cells have a non-autonomous effect in generating this symmetrical pattern.

Molecular markers confirm our interpretation of the cuticle phenotype. Since there is a decrease in row 5 cell types in wg^cx4 En-Gal4 UAS-ArmS10 embryos, we first examined the expression of Serrate. At 8 hours AEL Serrate is activated in about 3 rows of cells (Fig. 3A; Thomas et al., 1991), the anterior of which will correspond to the row 4/5 border. Serrate normally acts as a Notch ligand and, although the role of Notch in specifying denticle diversity has not been investigated, Serrate is required for posterior belt identities, such as row 4 (Wiellette and McGinnis, 1999). In wg mutants, where there is an expansion in posterior row cell types, Serrate expression is also expanded (Fig. 3B; Wiellette and McGinnis, 1999). Expression does not fill each segment, reflecting the remaining mirror-image pattern in wg mutants (Fig. 1C). Indeed, in wg en double mutants, where only row 5 denticles are specified, Serrate expression virtually fills the parasegment (Fig. 3C). This further confirms that the remaining pattern in wg mutants is due to En (and Hh) function. In wg^cx4 En-Gal4 UAS-ArmS10 embryos, Serrate expression is greatly reduced (Fig. 3D), consistent with a loss in posterior denticle rows.

We next examined Rhomboid, which at 7 hours AEL is expressed in a row of cells posterior to the En domain (Fig. 4A; Bier et al., 1990). Rhomboid induces EGF-Receptor signaling, which is required for row 1, as well as row 2-3 denticle cell types (O’Keefe et al., 1997; Szüts et al., 1997), the rows that we suspect to be duplicated in the wg^cx4 En-Gal4 UAS-ArmS10 embryos. In the wg^cx4 En-Gal4 UAS-ArmS10 embryos, En-expressing cells are now flanked by Rho stripes anteriorly (Fig. 4B), as well as posteriorly. Thus, the mirrorimage cuticle pattern is paralleled molecularly by symmetrical expression of Rho. Occasionally Rho-expressing cells surround En-expressing cells (Fig. 4C), a topology suggesting that induction of Rho is En-dependent and short range. These data suggest that Wg acts to ensure that signaling from the En domain is asymmetric, leading to the induction of Rhomboid only posteriorly.

Since Hh is the known ligand secreted by En cells, we examined the effects of removing Wg function on symmetry of Hh signaling. The transmembrane protein Patched binds Hh (Hooper and Scott, 1989; Nakano et al., 1989; Stone et al., 1996). The incorporation of Patched into responding cells in multivesicular bodies is a reflection of the strength of Hh signaling (Capdevila et al., 1994). In wild type, Patched protein is most strongly expressed in cells adjacent to En-expressing cells (Fig. 4D). From 6 hours AEL, there is normally asymmetry in Hh signaling as revealed by the increased number of dots in cells just posterior compared to anterior to En cells (Fig. 4D). However, in wg^cx4 En-Gal4 UAS-ArmS10 embryos, dots are symmetrically distributed (Fig. 4E), suggesting that, in the absence of Wg function, Hh now signals equally strongly anterior to En cells. We conclude that Wg plays an important role in maintaining (or establishing) asymmetry in signaling from En cells.

Our results suggest that a locally acting signal expressed from En cells controls part of the cuticle pattern by inducing Rhomboid only in cells posterior to the En domain. To test whether this signal was Hh, we first asked whether increasing Hh expression in En cells would affect Rho expression. In embryos carrying En-GAL4 and UAS-Hh, Rho expression expands posteriorly to 2-to 3-cell rows (Fig. 5B). Fate specification was also affected as we observed an excess of small denticles, apparently of row 2-4 type at the expense of row 5 and 6 (Fig. 5C; Lee et al., 1994; Fietz et al., 1995). So excess Hh can broaden Rho expression, and this has a consequence on fate selection. Thus, the level of Hh signaling from En cells can limit the Rho expression domain.

Reciprocally, in hh mutants, Rho expression is reduced (Fig. 5D). Thus, both loss- and gain-of-function experiments implicate hh as a positive regulator of Rho. The hh mutant cuticle pattern is consistent with this since, in hh or smo mutants, mostly type 5 denticles (Fig. 5E) are observed and Serrate expression is broadened encompassing all but the residual En-expressing cells (Fig. 5F; Wiellette and McGinnis, 1999).

Interestingly, in hh mutants, we still observed residual Rhoexpressing cells, even in embryos homozygous for a deficiency deleting hh, or embryos deficient for smoothened function, a receptor that transduces the Hh signal (Alcedo et al., 1996; Chen and Struhl, 1996; van den Heuvel and Ingham, 1996). Thus, residual Rho expression cannot be due to any residual Hh signaling. Therefore, Hh is important but is not absolutely required for the induction of Rho expression, implicating an unidentified signal partially redundant with Hh. This second signal may also emanate from the En cells. In hh mutants, En expression does not completely decay and the residual patches of Rho expression are often found in close proximity to Enexpressing cells (Fig. 5D, arrows). Thus, Hh as well as a second, likely En-dependent signal positively regulate Rho expression, thereby contributing to the correct specification of particular denticle identities.

Further evidence for a second En-dependent signal derives from analyzing Wg control of Rho expression. In wg null mutants, we observed an ectopic stripe of Rho in each segment (Fig. 6B). Thus, Wg negatively regulates Rho expression. The induction of an ectopic Rhomboid stripe is consistent with the wg mutant mirror-symmetric cuticle pattern.

Since Hh can positively regulate Rho, the ectopic Rho stripe in wg mutants could be the result of the now symmetrical signaling by Hh in the absence of Wg. To test this we removed both Hh and Wg function. We and others have difficulty constructing wg;hh doubly mutant stocks. To circumvent this, we removed all Hh-dependent signaling by examining smoothened (smo) mutants. We first made germline clones, removing both maternal and zygotic contribution. Resulting embryos show reduced Rho expression (Fig. 6C), identical to hh mutants (Fig. 5D). We next examined embryos lacking wg function as well as both maternal and zygotic smo activity (functionally wg; hh double mutants). Rho is expressed in duplicate stripes in these embryos (Fig. 6D), as in wg single mutants. Thus, although hh plays a role in Rho regulation (Fig. 5), the robust Rho expression in wg mutants cannot be due to symmetrical Hh signaling.

Note also that Rhomboid expression is still spatially restricted in embryos lacking both Wg and Hh signaling, just as in wg single mutants (Fig. 6B,D). Thus, some other localized factor must be required for the spatially restricted activation of Rhomboid. We could not use En as a registration marker to determine the position of ectopic Rho expression, since En expression is lost in wg null mutants. To circumvent this, we either inactivated wg in later stage embryos, after early Wg signaling has stabilized En expression, using wg^ts upshifts, or examined Ptc-GAL4 UAS-dominant negative dTCF embryos wherein Wg signal transduction is inhibited outside of the En domain. This also resulted in an ectopic stripe of Rho (Fig. 6E), positioned just anterior to now stabilized En-expressing cells. Thus, in each case where Rho has been mapped, it is expressed adjacent to En cells (Figs 4B,C, 6E). This, plus the observation that Rhomboid is still spatially restricted in embryos lacking both Wg and Hh signaling, strongly suggests that there exists an En-dependent inducer besides Hh. This factor must be made early by En cells. In the absence of Wg function, this inductive signal now acts symmetrically, on both sides of the En domain, just as we have shown that Hh can act symmetrically in the absence of Wg function (Fig. 4E).

The repression of Rhomboid by Wg also suggested why most Rhomboid expression is lost in smo or hh mutants. When wg function is removed from smo mutants, Rho expression is restored (Fig. 6D). Thus, in the absence of Hh signaling, Wg activity now represses Rho expression in the domain posterior to En cells. This suggests that the asymmetric induction of Rho normally comes about because Hh blocks Wg from signaling to cells posterior to the En domain. To test this, we asked whether activation of the Wg pathway in cells posterior to the En domain would be sufficient to block Rho induction. In Ptc-GAL4 UAS-ArmS10 embryos, Rho is repressed (Fig. 6F). Thus, during normal development, it is important to block activation of the Wg pathway in cells posterior to the En domain. We infer from this that one role for Hh is to establish or maintain asymmetry of signaling during patterning. Thus, wg and hh confer directionality upon each other’s signaling from the segment organizer.

Signaling from parasegment boundaries represents one paradigm for organizer function. This segmental organizer is bipartite, composed of adjacent rows of cells, one expressing Wingless and the next co-expressing Engrailed and Hedgehog. We now assign a specific, Wg-independent role to Hh, as well as to an unidentified second signal, also likely expressed from En cells. Coupled with prior analysis for the role of Wingless, we conclude that these two rows of cells emit signals that each specify a particular portion of the pattern. We also find that each signal prevents the other one from acting bi-directionally in order to create zones that are predominantly patterned by either Wg or by signals from the En cells.

The segmentation gene hierarchy establishes the asymmetric composition of the organizer, with Wg-expressing cells anteriorly adjacent to En/Hh-expressing cells. Sampedro and Lawrence (1993) suggested that the strict spatial order of these two signals did not matter for intrasegmental patterning, but simply defined the position of the parasegment boundary. However, since we find that Wg and En/Hh specify distinct fates, proper ordering of their expression is essential to ensure correct polarity of pattern across the segment.

Reciprocal signaling between Wg- and En/Hh-expressing cells stabilizes one another’s expression, consolidating the parasegmental body plan (DiNardo et al., 1988; Martinez-Arias et al., 1988; Bejsovec and Martinez-Arias, 1991; Heemskerk et al., 1991; Cumberledge and Krasnow, 1993). At this time, signaling is effective only locally, thereby restricting the expression of the other signal to a narrow strip of cells (Martinez Arias et al., 1988; Vincent and O’Farrell, 1992; Ingham, 1993). This ensures that the bipartite organizer remains a line source rather than broadening during patterning. Ventrally, the organizer specifies half the fates as smooth cell types and compelling evidence demonstrates that Wg specifies these (Wieschaus and Riggleman, 1987; Baker, 1988; Bejsovec and Martinez-Arias, 1991; Dougan and DiNardo, 1992; Noordermeer et al., 1992; Lawrence et al., 1996; Pai et al., 1997). This step occurs after Wg stabilizes En/Hh expression. In contrast, the identity of the signals responsible for specifying the diverse denticle cell types has been less clear. By bypassing the need for Wg input to En cells, we find row 2- to 4-type denticles specified in the absence of Wg. This conclusion was confirmed by analyzing the control of Rhomboid expression in cells flanking the En domain, as Rhomboid is essential for the proper differentiation of row 1-fates (O’Keefe et al., 1997; Szüts et al., 1997).

Previous studies indicated that denticle diversity arises early, around the time Wg stabilizes En/Hh expression (Bejsovec and Martinez-Arias, 1991; Sampedro et al., 1993). Our data suggest that the stabilization of En/Hh expression establishes the conditions to generate denticle diversity, but that diversity is not specified until later as reflected in the induction of Serrate and Rhomboid (∼7-8 hours AEL). Indeed, excess Hh delivered at late times can broaden Rho stripes and this still affects denticle diversity.

Due to a lack of molecular markers, the exact posterior boundary of the En/Hh-dependent domain is unclear. However, since row 5 cell types are specified in embryos lacking all Hh/Smo signaling, En/Hh influence can only extend up to row 4. Consistent with this, Hh signaling sets the anterior Serrate expression boundary. Since Wg signaling sets Serrate’s posterior boundary, the Serrate domain defines a region of positional values within the segment where Hh and Wg cooperate in patterning. In fact, Serrate expression is perhaps a molecular marker for a default state, as its expression is almost global in cases where only row 5 cell types are specified, such as in wg en doubly mutant embryos. We presume it would indeed be globally expressed in a wg en; hh triple mutant, which we have not been able to construct and test.

Although previous work suggested that Wg signaling generates denticle diversity, several observations made us question whether this role is direct. First, Sampedro et al. (1993) found that removing Hh function blocks the ability of Wg to generate a “mirror phenotype”, a reflection of denticle diversity. This Hh requirement supports the role that we find for En/Hh cells in defining row 2-4 denticles. Second, although cells can respond to different levels of Wg, the values specified are limited to midregions of the segment, just anterior to and within the Wg-expressing cells (Lawrence et al., 1996). This is equivalent to the smooth region of the segment in A2-A7 and does not bear on denticle diversity. Third, and most compelling, is the isolation and characterization of wg alleles that distinguish the specification of smooth cell fates from denticle diversity (Bejsovec and Wieschaus, 1993; Hays et al., 1997). Since the authors showed that generating denticle diversity correlates with the ability to sustain En expression, we interpret their data to suggest that Wg acts as a relay: the stabilization of En/Hh allows, as we show, for the production of signals necessary to specify some denticle diversity.

Bejsovec and Wieschaus (1993), in their comprehensive analysis of segment polarity mutant phenotypes, implicated several genes in specifying denticle diversity. Their evidence came from modifications of denticle type in various double mutants with wg. Three such mutants directly implicate En or Hh: (1) ptc, which would lead to excess Hh pathway activity, (2) en, which would lead to loss of the En-derived signals that we identify, and (3) hh. Thus, our observations confirm their conclusions, as we demonstrate that a Wg-independent En/Hh signal is involved in generating denticle diversity. They also implicate nkd, but we find two limitations in their interpretation. First, they define the wg phenotype to be of single, row 5 denticle type, while we present evidence for alternating intervals of row 5 with row 2-4 type. Second, since En/Hh expression is maintained for a longer period in wg; nkd doubly mutants (Bejsovec and Wieschaus, 1993), we would predict excess denticle diversity due to persistence of the En/Hh-dependent signals defined here. This is indeed the case. Thus, the contribution of nkd to denticle diversity is likely indirect, relayed through the effect on En/Hh stabilization.

Together, Hh and a second En-dependent signal lead to Rho induction, which, in turn activates EGF-R signaling. Since EGF-R signaling is essential for row 1-4 denticles, this supports a relay mechanism for patterning. However, our data suggest that Hh also plays a direct role. Specifically, wg mutants have duplicate stripes of Rho, and duplicate domains of row 2-to 4-type denticles. However, when smoothened is also removed from the wg mutant embryos, thereby removing Hh signaling, there are still duplicate stripes of Rho, but no small denticles. Thus, Rho is not sufficient to specify row 2-4 fates without Hh having signaled these cells also. Perhaps Hh acts early to induce Rho expression and separately acts to confer competence to a domain of cells that will respond to EGF-R pathway activation later. These data distinguish three distinct functions of Hh in epidermal patterning: Wg maintenance, Rho induction and a Rho-independent role in the specification of rows 2-4. This is reminiscent of wing patterning where Hh has been shown to have several functions as well: Dpp induction (Zecca et al., 1995) and a Dpp-independent role in the specification of cell types in the vicinity of the compartment boundary (Mullor et al., 1997; Strigini and Cohen, 1997).

Our data suggest that both Wg-and the En/Hh-expressing cells establish a block so that each signal operates largely unidirectionally. Rho is repressed by Wg signaling and it is important to block activation of the Wg pathway from cells posterior to the En/Hh domain. This block seems to be Hh-dependent as Rho expression is greatly reduced in hh mutants but maintained in wg smo double mutants. Elegant evidence for Hh restricting the range of Wg has been presented by Sanson et al. (1999). Our data also shows that Wg imparts asymmetry to signals from En cells. Without Wg function, Hh can signal more strongly to the anterior compared to wild type. Importantly, the cuticle pattern generated is also now symmetric relative to the En/Hh cells, strongly suggesting that a normally biased signal is now sent or received bidirectionally. Whereas we can show this directly for Hh signaling by examining Patched protein distribution, proof for asymmetric functioning of the unknown En-dependent signal awaits its identification.

The signals expressed from an organizer are developmentally potent, as they confer pattern over a large cellular field. Thus, once the appropriate expression of these signals is established, an important facet to organizer function is the temporal and spatial restriction of signaling. In some cases, activation of a signaling pathway induces an inhibitor of that same pathway. For example, we previously showed that inhibition of signaling was crucial for proper fate specification by this parasegment organizer (O’Keefe et al., 1997). In examining fate across the En domain, Rho-dependent activation of the EGF-R pathway posterior to the En cells leads to the induction of the diffusible inhibitor, Argos. Argos attenuates EGF-R activation in anterior En cells, allowing Wg signaling to win out, leading to proper fate specification of anterior En cells. In this paper, our data do not address the possible mechanism(s) that account for the bias in Wg or En/Hh signaling. The investigation of other pattern organizers has revealed several mechanisms that constrain signaling function, some of which apply to Wg or to Hh, such as inhibitor induction (Leyns et al., 1997), receptor sequestration (Chen and Struhl, 1996; Cadigan et al., 1998) or directed transcytosis (Bejsovec and Wieschaus, 1993; Dierick and Bejsovec, 1998). Most of these rely on the cognate signal affecting its own signaling properties or responses. The parasegment organizer is different in one regard, however, as one signal restricts the function of another.

In discs, signals emanate from compartment boundaries, which are inherited from the embryonic parasegment boundaries (Basler and Struhl, 1994; Capdevila and Guerrero, 1994; Tabata and Kornberg, 1994; Zecca et al., 1995; reviewed in Lawrence and Struhl, 1996). For the compartment organizer, Hh locally induces a line source for a long-range morphogen, either Wg or Dpp, which each act symmetrically. Cells exposed to the same ligand concentration on opposite sides of the source, adopt the same positional value (Fig. 7A, asterisks). However, the anterior compartment cell will select a different fate from the posterior compartment cell at the same positional value. This is because the posterior compartment expresses a unique transcription factor, En, and therefore is programmed intrinsically with a different response repertoire to the morphogen (Fig. 7A).

In parasegments, 10 of 12 rows of cells are intrinsically equivalent anterior compartment cells, while the posterior Enexpressing compartment only accounts for 2 cells. Thus, compartmental organization, with each compartment sporting unique transcription factors, can only make a small contribution toward distinguishing cell fate selection. Perhaps for this reason patterning cannot rely on induction of one longer-range morphogen. Instead, a bipartite organizer is used with each signal acting essentially unidirectionally. Equivalent cells to each side of the parasegment boundary develop differently because they are exposed to different signals (Fig. 7b).

We thank Mark Peifer and Bloomington stock center for flies. J. P. Vincent and E. Wiellette for communicating unpublished results. Supported by NIH GM45747. U. G. is a German National Scholar and V. H. is supported by the Cancer Research Fund of the Damon Runyon-Walter Winchell Foundation Fellowship, DRG-1442.