Dpp receptor levels contribute to shaping the Dpp morphogen gradient in the Drosophila wing imaginal disc

Secreted signaling proteins of the Wnt, TGF-β and Hedgehog families have been implicated in providing positional information in a variety of developmental systems. Localized sources of these signaling molecules have been shown to organize spatial pattern in a manner that suggests that they function as concentration-dependent morphogens (reviewed in Lawrence and Struhl, 1996; Neumann and Cohen, 1997b). There is compelling evidence from genetic mosaic analyses of developing Drosophila limbs that Wingless, Dpp and Hedgehog can form activity gradients in situ, and that these activity gradients act directly to instruct cells about their prospective fate as a function of the level of signal that they receive (Capdevila and Guerrero, 1994; Diaz-Benjumea et al., 1994; Zecca et al., 1995, 1996; Lecuit et al., 1996; Nellen et al., 1996; Lecuit and Cohen, 1997; Neumann and Cohen, 1997a; Strigini and Cohen, 1997).

If ligand gradients are used to specify positional information in developing systems, it becomes important to understand how such gradients are formed. The simplest models involve passive diffusion of ligand through the extracellular space (to avoid saturation this also requires a mechanism to clear ligand from the system). Gradients generated in this way have been predicted, on the basis of physical parameters, to form rapidly and to work over relatively long distances (Crick, 1970; Slack, 1991). The best documented case for a diffusion-generated ligand gradient is that of the Xenopus TGF-β family member, activin. Activin can induce different cell fates at different concentrations (Green and Smith, 1990, 1992; Gurdon et al., 1994, 1995). Activin can be shown to form a local concentration gradient that defines the spatial domains of target gene expression when a localized source is provided in the form of an activin-coated bead (Gurdon et al., 1996). Although this is an artificial situation, the evidence suggests that the activin gradient is formed by diffusion under these conditions and not by relay of other signals (McDowell et al., 1997). The activin gradient forms over a range of 250-300 μm in a few hours. Gradient formation is rapid, as would be expected for unhindered diffusion.

Recent studies suggest that the Hedgehog and Wingless gradients do not form by unhindered diffusion. Hedgehog induces elevated expression of its receptor Patched, which in turn limits the range of Hedgehog movement (Chen and Struhl, 1996). Wingless also regulates its receptor, but in the opposite way, to reduce its expression in the region of the disc where ligand levels are highest (Cadigan et al., 1998). Dfz2 is expressed at higher levels in cells at a distance from the source of Wingless and stabilizes low levels of Wingless by binding. These results have been interpreted as evidence for a biphasic Wingless gradient: steep near the source and shallow in cells at a distance. Dpp also forms a long-range activity gradient. Dpp has been shown to act directly over a range of up to 60 μm to regulate its target genes; and to regulate different target genes at different levels of activity (Lecuit et al., 1996; Nellen et al., 1996). Here we report that Dpp reduces expression of its receptor Thick veins in the wing pouch. High levels of Thick veins outside the wing pouch appear to limit the spread of Dpp and thereby modulate the shape of the ligand gradient. In addition, we show that the level of Tkv expression modulates sensitivity of cells to Dpp. Thus regulation of receptor levels by Dpp modulates the shape of the Dpp gradient.

Antibodies used were: anti-Dpp (Panganiban et al., 1990), anti-Spalt (Kuhnlein et al., 1994), anti-Omb (Grimm and Pflugfelder, 1996) and mouse anti-Patched (kindly provided by Phil Ingham). For double labeling with antibody, in situ hybridization was carried out essentially as described (Milan et al., 1996) using a digoxigenin-labeled tkv cDNA probe. Discs were dissected in PBS, fixed for 20 minutes in 4% formaldehyde/PBS-0.1% Triton X-100 (PBX), dehydrated through an ethanol series and stored for 24 hours in methanol at −20°C. After 5 minutes incubation at room temperature in ethanol and ethanol/xylene (1/1), discs were incubated for 1.5 hours in xylene and washed again in ethanol/xylene, ethanol and methanol. After rehydration in PBX, discs were fixed again for 20 minutes in 4% formaldehyde and rinsed, followed by treatment with Proteinase K and post-fixation for 20 minutes in 4% formaldehyde. In situ hybridization with the DIG-labeled probes was visualized with alkaline phosphatase anti-DIG. Discs were then washed in PBX and PBX with 0.1% BSA for 1 hour, followed by labeling with anti-Spalt or anti-Omb using biotinylated secondary antibody and streptavidin-HRP (Lecuit et al., 1996).

Clones expressing low levels of Dpp were produced using a tub>CD2, y+>dpp. Clones expressing high levels of Dpp were produced using a UAS>CD2, y+>dpp construct, combined with C765Gal4 (Nellen et al., 1996). Overexpression of Dpp in its own domain using dpp-Gal4; UAS-dpp is described in (Lecuit et al., 1996).

The thick veins cDNA was cloned into pUAST (Brand and Perrimon, 1993) to generate UAS-tkv. UAS-tkv was misexpressed in the central region of the wing using 71BGal4 (Brand and Perrimon, 1993). For clonal analysis, the flip-out cassette (FRT-CD2, y+-FRT) was introduced between the promoter and tkv coding sequences in UAS-tkv as an NheI fragment at a unique SpeI site introduced as part of the polylinker that was cloned with tkv in pUAST. Several independent transformant lines carrying UAS>>CD2, y+>>tkv were obtained and gave similar results. tkv was misexpressed in clones using a combination of the Gal4 and FLP out systems as described (Nellen et al., 1996). Larvae were heat-shocked for 25 minutes at 36.5°C at 60 hours AEL (eggs were collected for 24 hours, so the age range was 12 hours of development). Late third instar yellow⁺ male larvae were dissected and stained for CD2 and Spalt, or Omb.

The type I Dpp receptor thick veins (tkv) is expressed at low levels in the center of the disc and at higher levels toward the edges of the disc (Brummel et al., 1994; de Celis, 1997). Although tkv levels are low in the center of the disc, clonal analysis has shown that tkv activity is stringently required in this region for growth and for target gene expression (Burke and Basler, 1996; Nellen et al., 1996). For example, tkv mutant clones grow poorly in the center of the disc where receptor levels are low and grow better in lateral positions (Burke and Basler, 1996).

Double labeling for Omb protein and tkv transcript shows that the transition from low to high receptor levels is at the edge of the Omb expression domain in the mid-third instar disc (Fig. 1A). At later stages tkv is expressed at intermediate levels in the posterior compartment, so that the expression domains overlap more extensively (Fig. 1B). The pattern of tkv expression in mid-third instar discs suggested that tkv might be a target for regulation by Dpp. We asked whether Dpp signaling regulates tkv expression by examining the effects of clones of cells expressing Dpp at lateral positions in the disc where the level of tkv is normally high. Dpp-expressing clones (C765Gal4; UAS>>Dpp) were marked indirectly by their ability to induce ectopic Spalt expression. tkv transcript levels are reduced where Spalt is misexpressed, suggesting that Dpp can act at a distance to repress tkv expression (Fig. 1C,D, arrows; Spalt protein, brown; tkv transcript, blue). C765Gal4; UAS>>Dpp clones express higher levels of Dpp than are found in the endogenous stripe (Fig. 1E). The Dpp-expressing clones remain very small, in the range of 5-10 cells, but have long-range effects on Spalt and Tkv expression (see Fig. 5 and Nellen et al., 1996). These results suggest that the reduced levels of tkv transcript in the center of the disc are due to downregulation by Dpp acting at a distance. We observed ectopic expression of tkv in discs carrying clones mutant for Mad; however, this must be interpreted with caution because the clones were not directly marked (data not shown). During much of the third instar the domain of low tkv expression corresponds to the presumptive wing pouch.

We next asked whether the reduced levels of the Tkv expression are important for the formation of the Dpp activity gradient. To address this we examined the effects of clones of cells that overexpress wild-type Tkv on expression of the Dpp-target genes Spalt and Omb. Tkv-expressing clones well inside the endogenous domains show little effect on Spalt or Omb expression (arrowheads, Fig. 2A,C). Clones near the edge of the endogenous Spalt domain show increased Spalt expression and those near the edge of the Omb domain show elevated Omb expression (arrows). Tkv-expressing clones located outside but near the endogenous Spalt domain show ectopic induction of Spalt (box, Fig. 2A). Clones located farther from the Spalt domain do not show ectopic activation of Spalt (data not shown), but can show ectopic Omb (Fig. 2C,D). Together, these observations suggest that overexpressing Tkv can increase the sensitivity of cells to low levels of Dpp.

The clone in Fig. 2B is located in a region where the level of Dpp activity falls below that normally required to induce Spalt. Spalt expression is high on the side of the clone near the source of Dpp and decreases across the clone. The graded activation of Spalt in the clone shows that the overexpressed wild-type Dpp receptor depends on exogenous ligand, otherwise we would expect uniform activation of Spalt, as described previously for clones expressing the activated ligand-independent form of the Dpp receptor (Lecuit et al., 1996; Nellen et al., 1996). The graded expression of Spalt in the clone may reflect uniform sensitization of cells to the underlying Dpp gradient.

We next expressed Tkv using the 71BGAL4 driver to assess the consequences of broadly elevating Tkv expression levels in the central region of the disc. 71B GAL4 directs target gene expression in a domain resembling that of Spalt (Brand and Perrimon, 1993). 71BGal4 UAS-tkv wings are reduced in size (Fig. 3A,B). The effect is stronger in the posterior compartment, with the region between veins 4 and 5 being more reduced than the region between veins 2 and 3. The region between veins 3 and 4 is relatively normal, possibly because the size of this intervein region is specified directly by Hedgehog, not by Dpp (Mullor et al., 1997; Strigini and Cohen, 1997). These observations suggest that the long-range activity of Dpp in the vein 2-3 and 4-5 regions is compromised by overexpression of the Dpp receptor.

The effects of overexpressing Tkv in the center of the disc could be due to reduced expression of Dpp. However, we did not observe any obvious change in Dpp protein levels under these conditions (data not shown). Nonetheless, overexpressing Tkv strongly reduces the size of the Spalt domain in the Posterior (P) compartment (Fig. 3D,F). The stripe of Patched expression in the Anterior (A) compartment coincides with the source of Dpp at the AP boundary. We note that the effect on Spalt expression is much stronger in the P compartment than in the A compartment and that the Spalt domains in both compartments appear to be less graded at their edges than in wild type. The effective range of the Dpp activity gradient appears to be limited to a few cells in the P compartment in mid- and in late third instar discs (compare Fig. 3C with D; E with F). This suggests that overexpression of receptor can limit the spread of Dpp in the P compartment. Antibody to Dpp is only able to detect the high levels of Dpp protein in the endogenous expression domain (e.g. Fig. 1E), so it has not been possible to directly assess the effects of overexpressing Tkv on movement of Dpp. With this limitation in mind, these observations suggest that high levels of the receptor might sequester ligand and limit its movement across the wing disc. Possible reasons for the AP asymmetry will be discussed below.

These observations suggested that the higher levels of Tkv at the edge of the wing pouch might serve to limit the effective range of Dpp in the developing wing disc. In wild-type discs the Omb domain fills most of the wing pouch and Spalt is expressed in a narrower region (Fig. 4A). The lateral edge of the Omb domain corresponds roughly to where the levels of Tkv expression increase at the edge of the wing pouch (Fig. 1). In discs where Dpp is overexpressed in its normal pattern (dpp-Gal4; UAS-Dpp), we observed previously that the Spalt and Omb domains were of approximately the same size (Lecuit et al., 1996). If Dpp movement is slower at the edge of the wing pouch, Dpp might accumulate in the wing pouch if it is produced more rapidly than it can be removed by degradation and by spreading laterally. We examined the time course of formation of the Spalt and Omb domains in discs overexpressing Dpp to ask whether their expression patterns are compatible with this model.

In wild-type discs Omb is expressed earlier than Spalt. Spalt begins to accumulate in the center of the Omb domain in early third instar (de Celis et al., 1996) and spreads laterally thereafter. In dpp-Gal4; UAS-Dpp discs the Omb domain is broader than normal in early third instar (data not shown) and in mid-third instar discs (Fig. 4B). At this stage there is an anterior-posterior asymmetry in Spalt and Omb expression, which may be due to altered behavior of the dpp-Gal4-expressing anterior cells. We limit our analysis to the posterior compartment where the effects of Dpp are exclusively non-autonomous. In mid-third instar Omb is expressed in an enlarged posterior domain (double-headed arrow, Fig. 4B), while Spalt is still expressed at a very low level in these cells. As the disc matures the Omb domain does not increase much in width, suggesting that the Dpp activity gradient does not extend further laterally (Fig. 4B,C; the double-headed arrows are the same size in B and C). In the same time period Spalt is induced in a domain that almost completely overlaps the Omb domain. This suggests that Dpp levels increase above the threshold required for induction of Spalt, without concomitant broadening of the Omb domain. Thus overexpression of Dpp increases the size of the wing pouch significantly in early to mid-third instar, but only very little at later stages. We have not attempted to determine why the expansion of the wing pouch slows as the disc matures, but note that this correlates with an increase in the level of Tkv expression in lateral regions (data not shown).

This observation is consistent with the model that high levels of Tkv limit the spread of the Dpp gradient at the edge of the wing pouch. An alternative interpretation might be that Dpp is unable to induce Omb in more lateral regions. This seems unlikely because ectopic expression of Dpp in lateral regions can induce Omb (Lecuit et al., 1996; Fig. 5). We propose that the elevated receptor levels function as a barrier to limit movement of Dpp. Thus the shape of the gradient would be determined by the rate of ligand production, the rate of ligand degradation and by the limit to ligand spreading due the receptor distribution profile.

We next compared the time course of Dpp gradient formation in the lateral part of the disc when ligand is produced at different rates. Dpp gradient formation was monitored using Spalt and Omb as reporter genes activated at different threshold levels of Dpp activity. Clones were examined outside the endogenous Spalt and Omb domain where Tkv levels are high, so that de novo induction of both target genes could be compared.

Clones expressing low levels of Dpp (tub>>dpp) induce Omb within 24 hours but require 72 hours to induce Spalt (Fig. 5A-C). These clones are associated with significant overgrowth and repatterning in the lateral region of the disc. By contrast clones expressing high levels of Dpp (C765Gal4; UAS>>dpp) induce Spalt in the lateral part of the wing pouch in as little as 10 hours (Fig. 5E, arrows). By 24 hours Spalt and Omb are induced in nearly identical domains by clones outside the endogenous omb domain (Fig. 5F). This suggests that Dpp accumulates locally to levels sufficient to induce both target genes before it can spread far enough to form a spatially resolved activity gradient (see Fig. 5E). By 36 hours Omb expressed in a slightly larger domain than Spalt (Fig. 5G). The slope of the activity gradient appears to be quite steep compared to the endogenous expression domains (e.g. Fig. 5D). By 50 hours the ectopic Omb domain is broader than that of Spalt, suggesting that a shallower gradient has formed (Fig. 5H, the edge of the endogenous domain is visible at the left of the panel). The ectopic Omb domain can reach a diameter of 50-60 μm, approximately half the size of the endogenous Omb domain, in about 2 days (approximately the duration of third instar). Note that the clones of C765Gal4; UAS>>Dpp consist of only a few cells after 2 days (Fig. 1E). Thus formation of the Dpp activity gradient must be due to spread of ligand through the region around the clone in the 2-day time course of the experiment.

Previous studies have shown that tub>>dpp or Ubx>>dpp clones grow extensively, but exert relatively short-range effects on gene expression in the surrounding region (Lecuit et al., 1996; Nellen et al., 1996). Thus expansion of the Omb expression domain is largely due to growth in and around the tub>>dpp clone. The delay in Spalt induction under these conditions may reflect the time required to accumulate sufficient Dpp. In contrast clones expressing high levels of Dpp do not themselves grow large but induce Spalt and Omb in a large surrounding area. Initially the Dpp gradient appears to be very steep, as reflected in the coincident expression of Spalt and Omb. The expression domains gradually become nested, suggesting that the Dpp gradient becomes shallower with time. This correlates with overgrowth of the surrounding tissue and with downregulation of Tkv expression (Fig. 1C,D). We propose that the high levels of receptor in the lateral region may initially limit the spread of Dpp, resulting in local accumulation of Dpp and a steep gradient, and that formation of a shallow long-range Dpp gradient requires both downregulation of the receptor and growth in the surrounding tissue.

Several factors might affect formation of a morphogen gradient by a secreted signaling protein. Factors that affect the shape of the ligand gradient include the rate of ligand production relative to the rate of its removal and the rate of ligand movement. Other factors could modulate the responsiveness of cells without directly affecting ligand distribution. For example, differential expression of receptor or other downstream factors could affect the sensitivity of a cell to a given amount of ligand, without modulating the shape of the ligand gradient per se. In this report we have presented evidence that the level of Dpp receptor expression influences the sensitivity of cells to Dpp and that Dpp signaling regulates receptor levels in the wing pouch. We also report that high levels of receptor appear to limit the ability of Dpp to move through the disc. In both cases elevated levels of receptor may limit the effective range of Dpp by sequestering it. We note that it has not yet been possible to directly visualize the Dpp protein gradient, so the shape of the gradient can only be inferred from the expression of Dpp target genes. Our results suggest that the level of Dpp activity may help to shape both the ligand gradient and the activity gradient.

Experiments using activin beads in Xenopus explants suggest that gradient formation occurs significantly faster than we observe for Dpp in the wing disc. Activin gradients form over 250-300 μm in a few hours (Gurdon et al., 1995, 1996; McDowell et al., 1997). Recent studies suggest that the relatively little of the exogenously supplied activin is bound to its receptor and suggests that activin diffusion is relatively unhindered in Xenopus animal cap assays (Dyson and Gurdon, 1998).

Our results indicate that the Dpp gradient appears to form much more slowly in the wing disc. The endogenous gradient forms over 2 days, concomitant with growth of the wing pouch. The time required to make a long-range Dpp gradient in the lateral region of the disc is similar to the time required to make the endogenous gradient in the wing pouch and, surprisingly, appears to be relatively independent of the level of Dpp expression. Even when Dpp is expressed at higher than normal levels, it spreads slowly in the lateral region. Our results suggest that ligand movement is slowed by binding to its cell surface receptor and that formation of a spatially resolved activity gradient requires down-regulation of the Dpp receptor. There may also be a role for extracellular matrix components in limiting ligand movement (Jackson et al., 1997). Thus in contrast to the Activity bead experiments, there appears to be relatively little free ligand and thus limited diffusion, at least in the lateral regions where Tkv levels are high.

The level of Tkv expression also affects the sensitivity of cells to Dpp. Clones of Tkv-expressing cells induce Spalt expression at positions in the Dpp gradient where cells expressing normal levels of receptor cannot. The effect is ligand-dependent, and suggests that cells respond as though they were located farther up the gradient. Dyson and Gurdon (1998) have shown that increased activin receptor expression increases ligand binding on a per cell basis and that a particular threshold response occurs at a constant absolute level of receptor occupancy. Consequently cells expressing more receptor show a given threshold response at a lower concentration of ligand than cells expressing less receptor. Thus the level of receptor expression can modulate the apparent shape of the activity gradient by making cells sensitive to lower levels of ligand than would normally be required.

Hedgehog and Wingless also regulate expression of their receptors. Hedgehog induces Patched expression, which in turn serves to limit Hedgehog movement to a narrow stripe in the center of the disc (Chen and Struhl, 1996). Wingless represses its receptor Dfz2 (Cadigan et al., 1998). High levels of Dfz2 help to stabilize low levels of Wingless to form a broad shallow tail to the gradient, but do not appear to limit movement of Wingless (Cadigan et al., 1998). As for Tkv, elevated expression of Dfz2 sensitizes cells to Wg. Dpp signaling controls receptor expression, suggesting that a complex interplay between ligand and receptor levels shapes the gradient. A similar relationship between Dpp and tkv expression has been observed for the vein primordia of the pupal wing (de Celis, 1997). tkv levels are elevated adjacent to the veins, adjacent to where Dpp is expressed. Dpp activity appears to be very short range in this context. Thus each of the three ligands regulates its own receptor to modulate the shape of both the activity gradient and of the ligand gradient

When Tkv was overexpressed using 71BGal4 the Spalt domain expanded much more in the A compartment than in the P. This AP asymmetry is visible in wild-type discs, but is less pronounced. One explanation for the asymmetry may be that growth can help to spread Dpp in the anterior compartment. If we assume that Dpp is relatively stable in the extracellular space, adding new cells could dilute and spread the sequestered Dpp, thereby expanding the gradient. Dpp could be sequestered by binding to its receptor or to the extracellular matrix. The difference that we observe between A and P compartments when Tkv is overexpressed probably reflects the fact that cells originating in the Dpp expression domain can contribute to formation of a large part of the anterior compartment, but not to the posterior compartment. Thus cells originating in the Dpp domain could ‘carry’ Dpp protein away from the source as they and their progeny are displaced by addition of new cells (the displacement process can be directly visualized by lineage tracing cells originating the in the dpp-expression domain, Katrin Weigmann, unpublished data). This raises the intriguing possibility that growth of the field could directly contribute to gradient formation by dilution of stably sequestered ligand.

We thank our colleagues and Pernille Rorth for helpful discussions. We also thank M. Hoffmann, P. Ingham, R. Schuh and G. Pfugfelder for antibodies, K. Basler for DNA and flies and Ann-Mari Voie for technical assistance.