Signalling pathways have many different functions during development. Although the cellular consequences of a particular signalling event are usually conserved, the more complex developmental effects vary. These developmental effects result from an interaction between the signalled cell and its neighbours, and might involve other, sometimes reciprocal, signalling events (Freeman,2000; Muskavitch,1994). Furthermore, the combined effects of several signals in developing cells trigger more varied responses than a separate signal does. The cellular-molecular basis for the simultaneous integration of several signalling pathways is found in the cross-talk between pathways, mediated by the regulation of common effector proteins (Freeman,1998; Sharpe et al.,2001). However, a different kind of integration also takes place in the developing cells in the embryo, over time. The cumulative effects of several rounds of signalling produce different outcomes than the original first burst of a signal (Freeman,1996). Prominent for the promiscuity of its interactions, and for the wide range of cellular and developmental outcomes triggered, is the Ras signalling pathway, both in vertebrates and invertebrates alike (Tan and Kim,1999; Simon,2000).
Drosophila is a suitable model in which to study these aspects of signalling and development. In particular, the development of the leg offers a developmental process with a well-described set of patterning genes and patterning events, triggered by well-defined signalling pathways such as those of wingless (wg), decapentaplegic (dpp), EGFR-Ras, and Notch (Couso et al.,1993; Lecuit and Cohen,1997; de Celis et al.,1998; Bishop et al.,1999; Rauskolb and Irvine,1999; reviewed in Couso and Bishop,1998; Kojima,2004). The leg of the adult, or imago, of Drosophila develops over the 10 days of its life cycle. First, a primordium of about 20 cells is segregated in the ectoderm of the embryo (Bate and Arias,1991). This primordium already contains the rudiments of the positional information needed for the development of the future leg, in the form of the expression of several patterning genes such as wg, engrailed, and Distal-less (Dll), encoding dorsal-ventral, anterior-posterior, and proximal-distal cues, respectively (Cohen et al.,1991; Couso et al.,1993). During the first 24 hr of larval life after embryogenesis, the leg primordium physically delaminates from the epidermis and forms a bud, which organises itself as an imaginal disc. Attached to this imaginal disc are adepithelial cells, the precursors of the leg myoblasts, which will form the leg muscles (Bate et al.,1991). The leg imaginal disc resumes cell proliferation at about 48 hr after egg laying (48 hr AEL) (Madhavan and Schneiderman,1977), and continues until the leg reaches approximately 15,000 cells in the early pupa. Although basic patterning cues were present in the embryonic primordium, most leg patterning takes place during this proliferative phase. In particular, from 80 hr AEL (early in the third and final larval instar), a signal encoded by the EGFR ligand Vein (Vn) acts as a morphogenetic signal to establish new distal fates at the tip of the developing leg, which is located at the centre of the imaginal disc (Campbell,2002; Galindo et al.,2002). Towards the end of the larval period, at 120 hr AEL, and during the early pupal hours, the activity of the Notch pathway establishes segment boundaries and allocates presumptive joint regions at appropriate places (de Celis et al.,1998; Bishop et al.,1999; Rauskolb and Irvine,1999). At this time the leg disc, which by now resembles a compressed and folded telescope, evaginates, and the architecture of the final leg becomes clear (Fig. 1; see also Fig. 3A), so that patterns of gene expression can be easily fate-mapped to final structures (Galindo and Couso,2000). The epidermal layer forms a tube with a wave-like pattern of folds that prefigure the position of the final joints. Inside the lumen of this tube, myoblasts and neuronal axons start to assemble the final pattern of muscles and nerves. During the next 100 hr of pupal development, the epidermis differentiates the exoskeleton and its cuticular structures, such as bristles and joints (Fristrom and Fristrom,1993; Mirth and Akam,2002).
We have investigated the different roles of EGFR signalling in each of these developmental processes. In addition to its Vn-mediated role in the patterning of the distal part of the leg during the middle stages of leg development (Campbell,2002; Galindo et al.,2002), EGFR-Ras has also been described to control the development of bristles at the end of leg development (del Alamo et al.,2002), and to be required for leg cell proliferation and survival (Diaz-Benjumea and Garcia-Bellido,1990; Diaz-Benjumea and Hafen,1994). These reports focused on specific functions of EGFR-Ras only, and did not attempt a more comprehensive view of the roles of Ras in leg development. Here we report such a study and show how these functions integrate in space and time. In addition, we have found that signalling through the EGFR-Ras pathway fulfils further functions in Drosophila leg development. Thus EGFR-Ras is active in the developing myoblasts and in the developing tendons. Within the epidermis, late-appearing segmental stripes of EGFR-Ras activity antagonise Notch signalling in the middle of each leg segment, thus confining joint differentiation to its appropriate intersegmental position.
Expression of Components of the EGFR-Ras Pathway
In order to explore the functions of EGFR-Ras signalling during development, we studied the expression of genes encoding for EGFR-Ras ligands: vein (vn, which encodes a secreted protein with homology to neuregulins; Schnepp et al.,1996), and spitz (spi, which codes for a TGFα protein with EGF domains; Golembo et al.,1996a), and also of rhomboid (rho, encoding the transmembrane protein that mediates the proteolytic cleavage and release of Spi and other diffusible EGFR ligands; Urban et al.,2001). We also studied the expression of two effectors: diphospho-MAPK, the active form of the kinase that translocates the signal to the nucleus, and pointedP2 (pntP2), an ETS transcription factor that is post-translationally regulated by Ras signalling and then regulates the expression of transcriptional targets of the pathway (Brunner et al.,1994). The expression patterns of these EGFR-Ras components seem to broadly follow a similar profile, but differences in timing and precise domains of expression can be noticed in each case and are described next, and shown in Figure 1.
vein: vn expression starts early, from 72 hr AEL, in the centre of the disc and is established in a small elliptical central domain by 78 hr AEL (Fig. 1A). This domain occupies the presumptive distal-most region of the leg, the pretarsus, and signalling from it patterns this and adjacent segments (Campbell,2002; Galindo et al.,2002). At 96 hr AEL, this domain starts to evolve with the appearance of two isolated cells expressing vn and a proximal domain in the dorsal pleura (Fig. 1B). In between 96 and 120 hr AEL, the early central domain disappears, whereas the two isolated cells give rise to two symmetrical clusters in the presumptive pretarsus (Fig. 1C). Concentric rings of weak expression in tarsus 1, and near the tibia/femur and proximal femur/trochanter boundaries, are apparent at this stage. Strong sub-epidermal staining in the pretarsus, and weak expression in the tarsal segments, is seen at 128 hr AEL (8 hr after puparium formation) (Fig. 1D).
rhomboid: rho expression does not start until 86 hr AEL, when a central domain of expression begins to be established simultaneously with some concentric rings (Fig. 1E). The expression is very weak until 96 hr AEL. The central domain is then substituted by a small cluster of cells in the ventral pretarsus, in the pit of invaginating cells that give rise to the tendon (Fig. 1F). A similar cluster of invaginating cells is seen in the distal dorsal femur where another tendon and a chordotonal organ form (zur Lage and Jarman,1999). The concentric rings of expression, which are stronger than those of vn, are located in the presumptive tibia, tibia/femur boundary, and trochanter positions. By 120 hr AEL (Fig. 1G), the dorsal femoral cluster has become a twin cluster and there is now a ring in every presumptive leg segment. The pretarsal expression has disappeared. This pattern persists at 128 hr AEL in the pupa, when the rings of expression can be seen as stripes in the everted leg (Fig. 1H) and with the addition of isolated expressing cells in the pretarsus. Each rho stripe straddles the fold between adjacent segments, and has a graded intensity, which is weaker proximally and stronger distally.
spitz: spi expression seems to involve only concentric rings in the absence of a central, distal domain (not shown). Peripheral rings start to appear at 80 hr AEL in the trochanter and distal femur within a general background of expression over the disc. At 96 hr AEL, a tarsus 1/tibia ring and a tarsus 4/5 ring have been added. This expression persists at 128 hr AEL, with the addition of further stripes.
MAPK: the expression of diphospho-MAPK follows a pattern similar to a composition of vn and rho. Expression starts in a central distal domain (Galindo et al.,2002) shortly after 72 hr AEL (Fig. 1I). At 96 hr AEL, a dorsal femoral cluster starts to form, plus a dynamic pattern of cells scattered over the disc (presumably sensory organs, Fig. 1J). By 120 hr AEL, the pattern resembles rho expression with rings in every segment, twin femoral clusters, and no clear pretarsal cluster (Fig. 1K). A closer inspection of the segmental pattern at 128 hr AEL (Fig. 1L) reveals that, in contrast to rho, the diphospho-MAPK domains do not extend across the segmental folds.
pointedP2: pntP2 expression starts at 80 hr AEL in a weak central domain seemingly more extensive than that of vn and MAPK, but much stronger expression is present in the peripodial membrane and adepithelial cells, making assessment difficult (Fig. 1M). By 96 hr AEL, the pattern of pnt has refined into a thick tarsal ring in the approximate position of tarsus 4, a dorsal femoral cluster, and emerging rings in other segments (Fig. 1N). Twin clusters in the pretarsal pits where vn is expressed are also seen. This pattern persists at 120 hr AEL with the addition of more scattered cells similar to those stained with MAPK (Fig. 1O). At 128 hr AEL (Fig. 1P), this pattern persists except that the pretarsal pits are no longer visible; instead, threads of cells inside the leg expressing pnt are seen, presumably tendon cells. Strong expression in the adepithelial myoblasts is noticeable during the whole of leg development (not shown). At this stage, the topology of the epidermal pattern is similar to that of rho (compare with Fig. 1H).
In general, a two-tiered pattern of EGFR-Ras activity can be seen: expression in a presumptive distal domain in the centre of the disc in the early third instar larva (72–88 hr AEL), followed by a periodic pattern of concentric rings, one in every leg segment, plus expression in developing muscle structures, from the mid third instar larva and lasting into the pupa (96 hr AEL to 132 hr AEL). These results are summarised in Figure 2B.
Requirement for EGFR-Ras Function Over Time
To assess the function of these patterns, we used a temperature-sensitive mutant condition for EGFR (EGFR-ts, see Experimental Procedures). Timed cultures laid over 6 to 24 hr (see Experimental Procedures) were raised at the permissive temperature, and then either shifted to the restrictive one at defined points in development (shifts), or exposed to the restrictive temperature for periods of 24 hr only and then transferred back to the permissive temperature (pulses). The results are summarised in Figure 2A. Shifts and pulses before or at 80 hr AEL result in severely truncated legs with loss of pretarsus, tarsal segments 4, 5, and part of 3 (Fig. 3B). Shifts and pulses at 90 hr AEL produce legs with loss of pretarsal components and tendon defects, plus fusion of tarsus four and three (Fig. 3C). Shifts and pulses at 110 hr AEL produce partially undifferentiated claws and tendon defects (Fig. 3D). Less frequently, these mutant legs show ectopic joint tissue development. These ectopic joints are more common in legs from flies mutant for a double-mutant viable combination of hypomorphic alleles of MAPK and EGFR (rl top, see Experimental Procedures; Fig. 3E). They appear proximally to the proper joint and the polarity of the ball and socket structure is reversed.
In summary, again a two-tiered pattern of function appears in correlation with the two-tiered pattern of expression observed above. Appearance of the distal domains of expression correlate in time with a requirement for the development of the pretarsus, tarsus 4, 5, and part of 3, whereas appearance of patterns in rings in every segment and in the tendon pits correlates with a requirement for repression of joint development and promotion of tendon development (Fig. 2).
Ectopic Repression and Activation of EGFR-Ras
In general, directed expression of dominant-negative forms of EGFR-Ras components using the UAS-Gal4 system highlights the same requirements as the EGFR-ts condition, producing phenotypes of loss of pretarsal elements and tarsal segments 3 to 5, and duplicated joints, depending on the Gal4 drivers used. The result of a particular EGFR-ts shift can be mimicked by using an appropriate Gal4 line. For example, the DllGal4 driver is expressed throughout the entire distal half of the leg, including the tarsus and the distal tibia, from the beginning of leg development (Campbell and Tomlinson,1998). DllGal4 UAS-pntP2DN produces a strong phenotype of loss of tarsal segments (Fig. 3J) similar to early shifts of EGFR-ts. This result corroborates that total absence of EGFR function during distal leg development produces loss of pretarsus, tarsus 4 and 5, and even the distal part of tarsus 3. On the other hand, apGal4 UAS-EGFRDN and dlim1Gal4 UAS-EGFRDN mimic an EGFR-ts shift at 96 hr AEL. apGal4 UAS-EGFRDN loses tarsus 4 (Fig. 3I), where apGal4 is expressed, and dlim1Gal4 UAS-EGFRDN loses only the pretarsus (Fig. 3K), where dlim1 is expressed (Pueyo et al.,2000). The similarity between the phenotypes produced by the local absence of EGFR-Ras in these last two Gal4 mis-expression experiments, with the phenotypes following generalised removal of EGFR-Ras throughout the leg produced by EGFR-ts, reveals that the EGFR-Ras phenotypes during distal leg patterning are autonomous, or at least segment-autonomous. This segment-autonomy indicates that most likely the loss of these segments reflects a direct requirement for Ras signalling, and is not due to a secondary signal triggered by an EGFR-Ras transcriptional target.
Late and generalised expression of a dominant-negative form of Ras with the 32BGal4 driver produces ectopic joints with reversed proximal-distal polarity, similar to those observed in rl top flies (Fig. 3E,F). In addition, in all Gal4 mis-expression experiments the bracts are missing in the bristles of surviving tissue expressing Gal4 (not shown), as the bract cell type (which is specific to the leg and the proximal wing) requires an EGFR-Ras signal from the adjacent bristle at 128 hr AEL during pupal development, as previously reported (del Alamo et al.,2002).
Ectopic activation or over-activation of the EGFR-Ras pathway is lethal for the fly in most cases, either during embryogenesis, larval life, or early pupal stages, but again, use of appropriate Gal4 drivers corroborates our loss of function results. First, ectopic overgrowth of distal leg structures can be observed in legs of ap-Gal4 UAS-rafgof flies (Fig. 3L) where the tarsus 4 region is hyperplastic and fused to tarsus 5. This partial non-autonomy is probably due to either invasive over-proliferation of tarsus 4, or to induction of cell death in neighbouring cells (Karim and Rubin,1998). Second, when UAS-rho is driven by bab-Gal4, which is expressed in most of the tarsal region from mid to late third instar, the joints between the tarsal segments were extremely reduced or absent altogether (Fig. 3G and H), reinforcing the hypothesis that EGFR-Ras signalling represses joint development.
Molecular and Genetic Analysis of EGFR-Ras Function in Legs
EGFR-Ras activation in early third instar.
We have explored further some aspects of EGFR-Ras function and expression in legs. Regarding the early domain of expression in the distal leg, we have studied the molecular mechanism that gives rise to vn expression, and the differences between vn and rho function.
The activation of vn expression requires the gene Dll, but also wg and dpp signalling (Galindo et al.,2002; Fig. 4A,B). The close temporal and spatial relationship between the central domains of rho and vn begs an exploration of their functional interactions. Previous works (Campbell,2002; Galindo et al.,2002) differ on the relative contribution of each ligand to distal leg patterning and to their interactions. Thus, Galindo et al. identify Vn as the main ligand triggering EGFR-Ras-dependent distal leg patterning, whereas Campbell considers rho and vn activity to be simultaneous and synergistic for this function. Our detailed study of vn, MAPK, and rho expression (Fig. 1) reveals their temporal displacement and other differences. vn and MAPK appear 8 hr before rho, and rho expression in the distal, central, domain is always weaker, suggesting that vn is the main actor in distal leg patterning (compare Fig. 1A,B with E,F). To resolve this issue, we observed the expression of distal leg markers, whose temporal profile of expression in the wild-type is summarised in Figure 2B, in mutant conditions for vn and EGFR, in parallel with an analysis of the phenotypes produced in the differentiated mutant legs. In general, vn mutants produce weaker phenotypes in the legs, and at a lower frequency than EGFR-ts mutants. A similar situation is observed when molecular markers are studied. Loss of EGFR from 72 hr AEL produces loss of the pretarsal markers al, dlim1, and the tarsus 5 marker Bar (Campbell,2002; Galindo et al.,2002). This correlates with the appearance of the phenotype described above (Fig. 3B) of loss of all of these segments in the final leg. However, loss of vn produces a somewhat different, but interesting, effect. In approximately 15% of vn mutant discs, al expression is not present at its normal activation time of 80 hr AEL, and Bar expression is almost completely lost (Fig. 4D), or else reduced to a few cells present ectopically at the centre of the disc. In a further 35% of discs, these gene products are greatly reduced compared to the wild-type. Remarkably, expression of al and Bar is present, albeit sometimes reduced, in all discs at 96 and 120 hr AEL (Fig. 4E). Since it is not possible to record al and Bar expression during the life of a given mutant disc, two interpretations are possible. Either al and Bar expression recover between 84 and 120 hr AEL, or the individuals with total loss of Bar and al do not survive until 120 hr AEL, leaving only those individuals with partial reduction of al and Bar to be sampled at this time. The vn genotypes used are apparently molecular nulls, and the mutant animals die as larvae smaller than wild-type, presumably due to defects in the larval CNS and muscles, which require vn during embryogenesis (Schnepp et al.,1996). To solve this issue we generated clones of the null allele vnL6 and bypassed larval lethality. Discs with clones covering the vn-expressing region showed similar phenotypes to vn null mutants at 120 hr AEL (Fig. 4F) and in adult legs, i.e., presence of al (Fig. 4F) and rare defects in distal leg (not shown). This suggests that vn phenotypes recover between 80 and 120 hr AEL.
In agreement with this view, expression of dlim1 and ap is not lost and appears with wild-type timing, but in reduced domains in vn mutants (Fig. 4H,I). Interestingly, the reduction of the dlim1 domain is also milder than that suffered by al in rl top mutants, such that a wild-type-like situation of cells co-expressing al and dlim1 only exists in a reduced domain (Fig. 4J,K). It would also appear that most of the remaining Bar cells in a vn− disc express ap (not shown), in contrast to the wild-type where only half of the Bar domain expresses ap (Kojima et al.,2000; Pueyo and Couso,2004). This suggests that tarsus 5 fate (Bar expression alone, as opposed to Bar plus ap for tarsus 4 fate) is lost, and therefore the expression of al, Bar, and ap in the strongest vn mutants mimic the strongest phenotypes observed in the differentiated legs, showing loss of pretarsus and one tarsal segment (Galindo et al.2002). These results suggest that from 72 until 88 hr AEL Vn is the main or sole factor controlling distal leg patterning through the activation of al and Bar. From 88 hr AEL onwards, rho backs-up the Vn-induced activation of EGFR in time for the activation of dlim1 and ap expression.
EGFR-Ras activation in pupal legs: relationship with Notch signalling.
The joint phenotypes observed in the various mutant conditions, in correlation with the periodic pattern of expression in rings in every segment, begs the question of the relationship with Notch signalling. Notch signalling triggers joint development (de Celis et al.,1998; Bishop et al.,1999; Rauskolb and Irvine,1999), and it would appear that EGFR-Ras works antagonistically to Notch in this process. During joint development, stripes of cells expressing the Notch ligands Serrate (Ser) and Delta (Dl) arise adjacent and proximal to the cells that become committed to form joints (Figs. 5A, 6) (de Celis et al.,1998; Bishop et al.,1999; Rauskolb and Irvine,1999). Thus, signalling by Ser- and Dl-expressing cells signals to the cells located distally, but not proximally, and this has been explained by repressory inputs found in the cells proximal to the Ser and Dl stripe (Bishop et al.,1999).
To establish the topological relationship between the repetitive pattern of expression of some of the EGFR-Ras members shown in Figure 1, we double-stained legs for Notch and Ras activity. Activated MAPK is present in domains inside each segment. In a double staining for diphospho-MAPK and the joint marker big brain (bib), it can be appreciated that bib is distally adjacent to the diphospho-MAPK domain, and therefore the activated MAPK domain overlaps with that of Ser (Fig. 5B, compare with 5A; Fig. 6). The rho and pntP2 domains straddle the intersegmental fold and have a graded intensity of expression, but they are absent in a few rows of cells, approximately where Ser is expressed. argos (aos), an inhibitory ligand of EGFR that is also a target of the pathway, and is involved in regulatory feedback loops (Golembo et al.,1996b), shows a similar pattern of expression. Co-immunodetection of an aos reporter and Ser shows that, indeed, the Ser-expressing cells are distal and immediately adjacent to the high levels of aos (Fig. 5C).
Genetic conditions that resulted in a down-regulation of the EGFR-Ras pathway produced ectopic joints with inverted polarity, and over-activation of the pathway resulted in missing or extremely reduced joints (Fig. 3E–H). These results suggest that in the wild-type situation, EGFR-Ras inhibits Notch signalling in the cells proximal to Ser to block joint determination on this side of the stripe. The patterns of expression of Ser in rl−top− and in bab-Gal4 UAS-rho were indistinguishable from the wild-type (Fig. 5D,E), so neither down- nor up-regulation of EGFR-Ras signalling affects Ser expression. Therefore, the EGFR-Ras pathway must act downstream of Ser to down-regulate Notch signalling. We assessed this by looking at the expression of another Notch target and joint marker, the transcription factor AP-2 (Kerber et al.,2001). In the wild-type tarsus, AP-2 expression is similar to bib, in rows of cells corresponding to the joints between the five tarsal segments. In bab-Gal4 UAS-rho legs, the expression of AP-2 is extremely reduced both in extent and intensity (Fig. 5F,G). This result show that, whereas Notch ligand expression (Ser) is not regulated by EGFR-Ras, the outcome of Notch signalling is. An antagonist effect of EGFR-Ras on Notch signalling would explain the phenotypes of ectopic joints observed, and is also found in leg chordotonal organ development, a stretch sensory organ whose function is closely associated with muscles and joints (zur Lage and Jarman,1999). The ligand responsible for this function has to be a rho-dependent one, given that expression of vn is extremely weak in everted legs, compared to rho (Fig. 1D,H) and the fact that vn mutant allelic combinations or clones never show any joint defects (not shown).
The question remains of how the repetitive patterns of EGFR-Ras pathway members are set up. We favour the hypothesis that regulation of the expression of rho, pntP2, aos, and perhaps other members of the EGFR-Ras pathway is mediated by Notch signalling. This may explain both the emergence of the segmental stripes of expression of EGFR-Ras components and their close association to Notchr. The rings of rho and pntP2 expression appear in thin rings, one or two cells wide, in third instar discs, adjacent to stripes of Ser and high levels of Dl, and slightly delayed in time with respect to the emergence of the rings of Dl. The mechanism of Dl ring establishment is different to that of rho and pntP2: the rings of Dl are sculpted in sequence from a high ubiquitous background by a mixture of upregulation and repression in the appropriate places, as if activated by non-periodic signals (Fig. 5H–J, Rauskolb,2001). The rings of rho, however, appear directly and more or less simultaneously, as if activated by a periodic signal already established in rings (Fig. 1E–G). Finally, the expression of rho and pntP2 coincides with Notch pathway activity, which is active all over the segment, but is lowest or non-existent in the stripes of Ser plus high levels of Dl expression, due to an as yet unexplained “autonomous dominant-negative” mechanism noted by several authors in several tissues (de Celis et al.,1997; Micchelli et al.,1997; Bishop et al.,1999, Jacobsen et al.,1998). In this respect, the effect on EGFR-Ras activity can be ascertained using ectopic expression genotypes. Both in dppGal4 UAS-Ser and in apGAl4 UAS-Ser, an upregulation of MAPK is observed (not shown). Several explanations are possible for this, but the simplest is that the activity of an element of the EGFR-Ras pathway is up-regulated by Ser. In ombGal4 UAS-Ser (which is expressed similarly to dppGal4 UAS-Ser) ectopic expression of rho is observed in the regions where MAPK is upregulated (Fig. 5K). Thus, it appears that expression of Ser activates rho expression, and leads to activation of EGFR-Ras signalling.
Recently, a new role has been described for the EGFR-Ras signalling pathway in leg PD patterning (Campbell,2002; Galindo et al.,2002). In this instance, a wave of EGFR-Ras signalling from the distal tip of the leg acts as a morphogen to determine cell fates in the surrounding cells. Although the involvement of the pathway was evident from both reports, some aspects of this process were open to discussion, namely the ligands involved in the activation of EGFR and the extent of the EGFR-Ras-controlled patterning. In order to shed light on these questions, we conducted a thorough analysis of the patterns of expression and developmental genetics of several EGFR-Ras pathway members.
Early PD Patterning Role of EGFR-Ras
In early leg discs, expression of both vn and the ligand activator rho can be detected, although differences between them are evident. First, vn expression is set up earlier, by 78 hr AEL, whereas rho appears at least 8 hr later, and at much lower intensity. Consistent with the activation of these two genes, diphospho-MAPK is detected in a solid circular domain coinciding with the distal leg fates. As a result, when the EGFR-Ras pathway is disrupted during early leg development using an EGFR-ts allele, or by expression of dominant negative forms of pathway members, severe defects in distal patterning are observed. These extend, proximally, from tarsus 3 to the most distal pretarsus. The neuregulin-like ligand Vn is expressed in the centre of the disc, probably as a result of the integration of three developmental inputs: the HOX transcription factor Dll, and the signalling pathways of wg and dpp. These combined requirements explain the timing and positioning of vn expression. Thus, the requirement for wg and dpp signalling allocates vn expression to the centre of the disc, in the presumptive distal-most region where their expression coincides (Diaz-Benjumea et al.,1994). However, wg expression is present and active in this region from 48 hr earlier (Couso et al.,1993; Diaz-Benjumea et al.,1994). dpp expression has been reported in the leg disc from at least 72 hr AEL (Masucci et al.,1990), but it is already active in second instar when it induces the expression of the dachshund gene and maintains Dll expression (Lecuit and Cohen,1997). The requirement for Dll expression might delay the appearance of vn until second instar, when the pattern of Dll is established after disc invagination and separation of Dll-expressing cells of the larval Keilin organ from the Dll-expressing cells of the presumptive disc. Presumably, a requirement for a certain amount of Dll, or a combination of these three inputs in sufficient quantity over time, delays vn expression until the end of second instar at 72 hr AEL.
At 88 hr AEL the ligand activator rho comes into play. The role of rho is the cleavage and release of further EGFR ligands. In imaginal tissues, the two accepted candidates to date are spi and keren. To date, it is not possible to ascertain any role for keren due to the fact that there are no keren mutants available and its expression is usually below the detection level of current techniques.
To assess the respective contributions of vn and rho to the EGFR-Ras-controlled PD pattern, we compared mutant conditions of vn and EGFR, considering that, although the rho phenotypes described are weak or non-existent (Campbell,2002), EGFR must recapitulate the contributions of all the ligands involved. The differences between both types of mutants are evident. In EGFR mutants, there is a total absence of the distal markers al, dlim1, and Bar (Campbell,2002; Galindo et al.,2002). In contrast, vn mutants show an apparent early loss and later recovery of al and Bar, which is corroborated by the fact that we cannot observe lack of al in extensive clones of vn− that span most of the distal domain in a the late disc. Moreover, the later markers dlim1 (pretarsus) and ap (tarsus 4) appear with the correct timing, although their domains are reduced in vn mutants.
Our results suggest that at 72 hr AEL, Vn diffusion starts a gradient of EGFR-Ras signalling in the distal leg primordium. As a result of the activation of the EGFR-Ras pathway, the expression of al and Bar is initiated. From 86 hr AEL, rho backs up the Vn-induced activation of EGFR at the time of dlim1 and ap expression. In vn mutant conditions, in the near-total or total absence of Vn function, proper activation of Bar and al is not produced. However, the subsequent expression of rho is able to rescue these defects to a large extent, such as near-normal expression of al and Bar, and milder leg phenotypes are observed. Reciprocally, absence of rho alone produces little or no effect (Campbell,2002) because vn expression must still remain. This redundancy is highlighted in dlim1Gal4 UAS-EGFRDN flies, in which EGFR function is blocked at the tip of the leg from about the same time as rho activation, and yet produces loss and defects in the pretarsus. These phenotypes are stronger than the loss of rho, and presumably due to the loss of all EGFR-Ras signalling triggered by the combination of rho-dependent and Vn signals. Similarly, clones removing both vn and rho simultaneously produce defects comparable to loss of EGFR (Campbell,2002). From 96 hr AEL, EGFR is no longer required for al and Bar expression (Campbell,2002).
Further factors, other than functional redundancy of vn and rho, may contribute to the resilience of the leg to producing mutant phenotypes when vn or rho are removed independently. Firstly, EGFR-Ras activation levels in the leg (as revealed by phosphorylated MAPK expression) are low in comparison to other tissues like the eye or the CNS, and therefore anything other than total loss of EGFR-Ras signalling might fulfil leg requirements. Secondly, at the tip of the leg a regulatory network of transcription factors reinforces, maintains, and expands the patterning inputs of EGFR-Ras (Kojima et al.,2000; Pueyo et al.,2000; Pueyo and Couso,2004). Firstly, al expression has an activatory input on dlim1 expression, which then feeds back to maintain al expression. Secondly, Bar activates ap expression. Finally, loss of either dlim1 or ap produces only subtle phenotypes, due both to low requirements for total amount of protein being present (ap), and to a requirement in differentiation rather than patterning and proliferation (dlim1) (Pueyo et al.,2000; Pueyo and Couso,2004; Tsuji et al.,2000).
In summary, it appears that, whereas Vn is the main EGFR ligand for distal patterning, rho produces a delayed fail-safe reinforcement of the signal. The situation approaches functional redundancy but is subtly different in that vn is clearly the main partner in the equation.
EGFR-Ras Signalling in Pupal Legs
Our work has revealed a further role of EGFR-Ras signalling in late leg development, which seems dependent on rho-activated ligands only, as shown by vn and rho expression and phenotypes. Some of the mutant conditions employed that affect only late stages of development, or that are mild enough to avoid PD defects, result in defects of the intersegmental leg joints. Down-regulation of the EGFR-Ras pathway produces ectopic joints of inverted polarity similar to those observed in tissue polarity mutants (Bishop et al.,1999), whereas its up-regulation causes loss or reduction of the wild-type joints. The determination of the joints is mediated by Notch signalling (de Celis et al.,1998; Bishop et al.,1999; Rauskolb and Irvine,1999). Joints are induced by Notch activation in rows of cells distally adjacent to domains with high levels of Ser and Dl expression (Fig. 6). We determined the relative positions of the patterns of expression of members of the Notch and EGFR-Ras pathways. Several of the latter, including rho, pntP2, and aos have repetitive domains of graded expression. Each one of these domains starts proximally in a group of cells with a moderate level of expression and extends distally through the intersegmental fold and into the next segment with increasing intensity. For one of these genes, aos, we have determined that the cells with the highest expression are proximally adjacent to the Ser-expressing cells (Figs. 5, 6) Activation of EGFR results in the phosphorylation of MAPK, the main transducer of the canonical Ras pathway, and we have determined that a domain of activated diphospho-MAPK is present inside each segmental unit, overlapping distally with the Ser domain, but it remains proximal to the presumptive joint cells expressing bib (Figs. 5,6). The genetic results, and the topology of the MAPK domains are consistent with a role for the latter in repressing Notch signalling in the cells proximal to Ser (Fig. 6). Moreover, when we up-regulate the EGFR-Ras pathway by over-expression in most of the tarsal region by ectopically driving rho, the joints are reduced or missing. In these legs, the expression of Ser is identical to the wild-type, but expression of the joint marker AP-2 is reduced in number of cells and overall intensity. Our results indicate that Notch targets are down-regulated, and that this attenuation of Notch signalling is not achieved by down-regulation of the Ser ligand, but at some point downstream of the signal transduction process. As we mentioned above, a similar effect is observed in mutants for tissue polarity genes. At this stage, we do not know if either of these two mechanisms is dependent on the other, or if they are redundant.
Interestingly, a requirement for EGFR-Ras in vertebrate joint development has also been reported recently. In vertebrate limbs, EGFR-Ras promotes joint development, as opposed to the repression seen in flies. However, in vertebrates, the action of EGFR-Ras is actually to inhibit chondroplast condensation and subsequent bone differentiation, a cellular function therefore similar to its effect in Drosophila joints, where EGFR-Ras represses the appearance of epidermal constrictions with densely packed epidermal cells and subsequent differentiation of thick cuticle at these sites.
Multiple Roles for EGFR-Ras in Leg Development
Signalling through the EGFR-Ras pathway fulfils multiple functions in Drosophila leg development. It is active in the developing myoblasts and tendons; within the epidermis, it is involved in the early patterning and allocation of proximal-distal fates to the distal part of the leg, as well as being required at a cellular level for the proliferation of the tissue. Correct differentiation of leg sensory organs requires EGFR-Ras to form bract cells and recruit extra neurons, and proper differentiation of joints requires antagonising Notch-mediated signalling in the middle of the segment by EGFR-Ras. However, differences between these functions can be appreciated.
Low levels of EGFR-Ras activity seem to be required autonomously for cell survival and growth in every cell of the leg disc. This requirement is similar to those identified in other Drosophila organs (wing, eye) and in vertebrates, and is revealed by the fact that (1) clones of null alleles of Ras pathway members are cell lethal in the legs but can be partially rescued by giving the cells a growth advantage (Diaz-Benjumea and Garcia-Bellido,1990; Diaz-Benjumea and Hafen,1994; unpublished observations), and (2) extreme EGFR-ts treatments produce cell death (Campbell,2002).This issue is relevant to EGFR-Ras signalling during distal leg patterning, and to the putative range of the gradient of EGFR-Ras activity, which has been described either as affecting up to tarsus 4 (Galindo et al.,2002) or affecting up to tarsus 1 or even more proximally (Campbell,2002). It is likely that some of the proximal phenotypes are either due to cell death or to secondary signalling events. In this regard, we have identified a secondary signalling process that is triggered by Bar expression and that affects the development of more proximal tarsal segments (unpublished observations). The combination of cell survival requirements and secondary signals could explain occasional phenotypes affecting beyond the tarsus 3/4 joint.
The role of the gradient of EGFR-Ras signalling in distal leg patterning is itself also different from the other instances mentioned above. During distal leg patterning, the consequences of EGFR-Ras are more far-reaching. A single signalling event, even though protracted in developmental time over 24 hr, coordinates the development and instructs developmental fates in a population of hundreds of cells. In contrast, all the other instances discussed are local, independent signalling events that affect smaller populations of cells, i.e., every joint and bristle develops independently of the others in populations of tens of cells and individual cells, respectively. Apart from this difference in range of action of the EGFR-Ras signal, another difference is that during distal leg patterning the activity of EGFR-Ras directly instructs the development of at least two (al expression and pretarsal fate and Bar expression and tarsus five fate) and perhaps three (ap expression and tarsus four fate) developmental fates. In the other functions in leg development, a single developmental fate is instructed, in some cases as an extension of an already existing fate such as chordotonal neurons.
Despite these differences between short-range local functions versus long-range organising or morphogen activities, the molecular components of the EGFR-Ras pathway and their molecular mechanism of action seem conserved in the leg and in other cases. The differences might lie in the range of action and intrinsic activity of the ligands or of their antagonists. We observe that the levels of MAPK activation and pnt expression are lower in the leg disc until 96 hr AEL than in other tissues such as CNS and eye, where spi and rho have a more prominent function. It has been noted that Vn is a weaker activator of EGFR signalling and MAPK than Spi (Schnepp et al.,1998). The lower activatory potential of the Vn ligand might be more suitable for producing differences in amount of signalling, and hence differential expression of EGFR-Ras transcriptional targets such as al and Bar, than Spi, whose activity might saturate EGFR-Ras signalling, in effect allowing only an on-off switch for a single cell fate at a time.
Standard fly culture procedures at 25°C were employed except for temperature-sensitive EGFR-ts shifts. For these, flies of a cross topts1a / SM6a TM6b x topCO / SM6aTM6b were allowed to lay eggs for 24 hr at 17°C and the collections kept at this permissive temperature until they were shifted to the restrictive temperature of 29°C, at appropriate times for 24 hr (pulses) or until the end of development (shifts). Larval age was checked at the time of the shifts using morphological markers (Galindo et al.,2002). Other mutant genotypes were: wgIL; Dll3; Dl6B / DlRf; rl1top1 / rl10top7 and rl1top1 / rl10atop4a; null vn mutants were vnL6 / Df(3L)v65c and vnL6 / vnγ3, hypomorphic mutants studied were vnRG / vnγ3 and vnL6 / vnD4. Gal4 lines are described in Pueyo and Couso (2004). bab-Gal4 is described in Cabrera et al. (2002). UAS lines used were UAS-Ser, UAS-DERDN (UAS-EGFRDN); UAS-pntP2, UAS-HrasN17, UAS-Rasv12, UAS-rafgof, and UAS-rlSEM. These genetic variants are described in Flybase (http://fbserver.gen.cam.ac.uk).
Antibody stainings were done as previously described (Bishop et al.,1999; Pueyo et al.,2000; Galindo et al.,2002, Pueyo and Couso,2004). Anti-AP-2 antibody was a gift from Pam Mitchell. β-gal reporter lines used were: pnt1277, rho-lacZ (zur Lage and Jarman,1999), Dl-lacZ (Klein and Martinez Arias,1998), aosW11, spi01068, vnRf264, apUG62. As with EGFR-ts shifts, the age of stained discs and pupal legs were dated according to morphological markers. Full-length DNA probes for rho and vn were used for in situ hybridisation.
We thank A. Simcox for generously sharing unpublished results, E. Martin-Blanco, A. Jarman, and B. Shilo for flies, and I. Pueyo and other members of our lab for discussions, unpublished results, and advice. This work has been funded by The Wellcome Trust through a Senior Research Fellowship to J.P.C. (057730) and a Project Grant 066189.