Sequential organizing activities of engrailed, hedgehog and decapentaplegic in the Drosophila wing

The adult appendages of Drosophila are each subdivided into precisely defined regions, the anterior and posterior compartments, which derive from adjacent but immiscible cell populations established early in development (Garcia-Bellido et al., 1973, 1976; Morata and Lawrence, 1975). In most regions of the body, these populations are founded around the time that the embryo becomes segmented (Steiner, 1976; Lawrence and Morata, 1977) by a process that involves the heritable activation of the homeodomain protein engrailed (en) in posterior, but not anterior, founder cells (Morata and Lawrence, 1975; Lawrence and Morata, 1976; Kornberg et al., 1985; DiNardo et al., 1985; Hama et al., 1990; Vincent and O’Farrell, 1992). en functions subsequently as a ‘selector’ gene (Garcia-Bellido, 1975), directing posterior cells to form posterior rather than anterior pattern, and to avoid mixing with anterior cells (Morata and Lawrence, 1975; Lawrence and Morata, 1976; Lawrence and Struhl, 1982).

Recently, we and others have found evidence that hedgehog (hh), a protein secreted by posterior cells (Lee et al., 1992; Mohler and Vani, 1992; Tabata et al., 1992; Tashiro et al., 1993) is responsible for organizing wing development, possibly by inducing neighboring anterior cells to express another secreted protein, decapentaplegic (dpp) (Basler and Struhl, 1994; Capdevila et al., 1994; Capdevila and Guerrero, 1994; Tabata and Kornberg, 1994). Here we show (i) that en activity in posterior cells directs them to express hh while at the same time blocking their ability to respond to hh, (ii) that the absence of en activity in anterior cells allows these cells to be induced by hh to express dpp, and (iii) that dpp can exert a long-range organizing influence on growth and patterning in both compartments. These results indicate that the selective deployment of en in posterior cells has two distinct roles. First, it directs the expression of the organizing signals hh and dpp in cells on opposite sides of the compartment boundary. Second, it controls the different ways anterior and posterior cells respond to these signals to generate distinct cell patterns in each compartment.

A plasmid containing the Tubα1>y+>hh gene (Basler and Struhl, 1994) in the Carnegie20 transformation vector (Rubin and Spardling, 1983) was modified as follows: (i) the y⁺ gene in the >y⁺> flp-out cassette was replaced by a 13 kb NotI-SalI fragment containing the f⁺ gene (Petersen et al, 1994), (ii) the ry⁺ gene in the C20 vector was removed, (iii) the hh coding sequence was replaced by the coding sequence for en (position 1-1840, Poole et al., 1985) or dpp (position 1133-3096, Padgett et al., 1987), respectively. Transformants were identified by rescue of the forked mutant phenotype.

Similar to the Tubα1>y+>hh gene (Basler and Struhl, 1994) except (i)the y⁺ gene in the >y⁺> flp-out cassette is replaced by a >CD2, y⁺> flp-out cassette (Jiang and Struhl, 1995) and (ii) the ry⁺ gene in the C20 vector was removed. The >CD2, y⁺> flp-out cassette is identical to the >y⁺> flp-out cassette present in the Tubα1>y+>hh gene except that it contains the coding sequence for the reporter protein rat CD2 (Dunin-Borkowski and Brown, 1995) inserted immediately downstream of the first FRT (>). Transformants were identi- fied by rescue of the yellow mutant phenotype.

Similar to Tubα1>CD2, y+>hh: except that the hh coding sequence is replaced by the en coding sequence.

f^36ahsp70-flp females were crossed to f^36a males carrying a single insert of the Tubα1>f⁺>en or Tubα1>f⁺>dpp gene on the second or third chromosome, and the resulting progeny were subjected to a mild heat shock (34°C or 35°C for 30 minutes) during the first 72 hours of development. Tubα1>f⁺>en or Tubα1>f⁺>dpp adults which emerged from these crosses were collected and their wings dissected in ethanol, mounted in Euparal and screened under the compound microscope for the presence of forked wing hairs formed by f^36a; Tubα1>en or f^36a; Tubα1>dpp cells. Five independent lines of the Tubα1>f⁺>en gene, and two independent lines of the Tubα1>f⁺>dpp gene were used. All of the lines of each transgene behave similarly, so the results in each case have been pooled.

To generate Tubα1>dpp clones in a dpp^d8/dpp^d10 background, a Tubα1>f⁺>dpp transgene on the second chromosome (#27) was recombined onto a dpp^d8 chromosome (St. Johnston et al., 1990) and used to generate f^36ahsp70-flp; dpp^d8Tubα1>f⁺>dpp/dpp^d10 progeny, which were subjected to a mild heat shock (33°C or 34°C for 30 minutes) during the first 48 hours of development. Adults of this genotype were recognized by their reduced eyes and reduced or absent wings.

To assess the size and distribution of ‘neutral’ clones (Results), morphologically normal wings were scored under the compound micro- scope for the presence of f^36aTubα1>en clones. The number of cells in each clone was determined by counting the forked hairs. A total of 215 neutral clones were identified in a sample of approximately 250 wings: 98 of these clones were located in the posterior compartment and the remaining 117 were in the anterior compartment (Table 1A). To assess the size and distribution of ‘reorganizing’ clones (Results), wings with altered venation pattern within the anterior compartment were identified under the dissecting microscope and then mounted and scored for forked hairs under the compound microscope. Clones of marked cells associated with ectopic veins both within and outside of the clone were designated as reorganizing clones (see Table 1B). To investigate the cause of venation defects in the posterior compartment, wings which were normal in shape but contained aberrant posterior venation patterns were identified under the dissecting microscope and scored under the compound microscope for the presence of clones. In 35/40 such wings, the defective venation was associated with clones which cross the line that normally demarcates the A/P-compartment boundary and autonomously fail to differentiate parts of vein 4. In the remaining 5 cases, small ectopic veins or distortions of the normal veins were associated with Tubα1>en clones located entirely within the posterior compartment. To assess further the frequency with which Tubα1>en clones cross the A/P-compartment boundary, a sample of 94 wings collected without reference to their morphology were scored for the presence of clones along the A/P compartment boundary. 16/27 clones found in the vicinity of the compartment boundary appeared to cross the A/P compartment boundary and 15/16 of these crossing clones were associated with venation defects in the posterior compartment.

To assess the distribution of ‘neutral’ Tubα1>dpp clones (Results), approximately 350 wings of normal shape and size were screened for clones under the compound microscope and 238 clones identified. Of these, 89/136 anterior clones and 53/102 posterior clones contributed to only one surface of the wing. Of the remaining clones that contributed to both the dorsal and ventral surfaces, 34/47 anterior clones were located posterior to vein 2 and 47/49 posterior clones were located anterior to vein 5.

To assess the distribution of ‘reorganizing’ clones (Results), wings of abnormal shape or size were mounted and screened for clones. 75 reorganizing clones were identified, all of which contributed to both the dorsal and ventral surfaces of the wing. All 42 of these clones in the anterior compartment were positioned anterior to vein 2; similarly 27/33 of the posterior compartment clones were positioned posterior to vein 5.

To assess the frequency with which Tubα1>dpp clones reorganized pattern in regions anterior to vein 2, or posterior to vein 4, 287 wings obtained without reference to their morphology were scored for the presence of clones that contributed to both surfaces of the wing and were located either anterior to vein 2 or posterior to vein 4. 8/10 such clones located anterior to vein 2 were reorganizing, as were 2/2 clones located posterior to vein 5. In contrast, 16 such clones located between veins 2 and 5 were neutral.

y hsp70-flp larvae, which carried a single copy of the Tubα1>CD2, y+>en gene as well as a single copy of either the dpp-lacZBS3.0 (Blackman et al., 1991), dpp^P10638 (R. Blackman, personal communication), or hh^P30 (Lee et al., 1992) lacZ reporter gene were subjected to a mild heat shock during the first or second larval instar. Their wing discs were then recovered during the mid to late third instar, fixed and stained for lacZ, CD2 and/or en expression by standard immunofluorescence procedures (as in Basler and Struhl, 1994). The mouse monoclonal antibodies OX34 (Serotec) and 4D9 (Patel et al., 1989) were used to detect CD2 and en expression, respectively. A rabbit polyclonal antibody (Cappel) was used to detect lacZ expression. Tubα1>hh clones were generated similarly by using a Tubα1>CD2, y⁺>hh transgene in place of the Tubα1>CD2, y⁺>en transgene.

Marked clones of cells mutant for en and inv were generated by flp- mediated mitotic recombination (Golic, 1991; Xu and Rubin, 1993) by subjecting first instar larvae of the following genotypes to a single severe heat shock (37°C for 1 hour): (i) y hsp70-flp/+; FRT42 πM/FRT42 Df(2R)en^E; hh^P30/+ (to monitor hh-lacZ expression), and (i)y hsp70-flp/+; dpp^P10638FRT42 πM/FRT42 Df(2R)en^E (to monitor dpp-lacZ expression). The resulting third instar larvae were subjected to a second heat shock (37°C for 1 hour), and after a recovery period of 1 hour, their imaginal discs were fixed and stained for immuno- fluorescence as described above. A mouse monoclonal antibody, 9E10 (Evan et al., 1985) was used to detect expression of the πM marker protein.

In a previous study, we analyzed the organizing role of hh protein by ectopically expressing or eliminating hh function in genetically marked cells during imaginal disc development (Basler and Struhl, 1994). Here, we describe the results of similar experiments in which we have ectopically expressed or eliminated en or dpp gene function in the imaginal wing disc.

Clones of cells that express the en coding sequence under the control of the constitutive Tubulinα1 (Tubα1) promoter were generated by heat shocking f^36a larvae carrying the transgenes hsp70-flp and Tubα1>f⁺>en during the first or second larval instar (Materials and Methods). The Tubα1>f⁺>en transgene contains a >f⁺> flp-out cassette which includes the marker gene forked (f⁺) and is flanked by targets (‘>‘) for the site- specific flp recombinase. Hence, flp recombinase expressed under heat-shock control can excise the cassette, thereby eliminating the f⁺ gene and generating marked Tubα1>en cells in which the Tubα1 promoter drives expression of the en coding sequence.

Three classes of Tubα1>en clones were obtained: those that contribute solely to the posterior compartment, those that con- tribute solely to the anterior compartment and a third class containing clones that cross between the two compartments. We describe these classes in turn.

Clones restricted to the posterior compartment develop normally (Fig. 1A), as expected given that the endogenous en gene is normally active in all cells of this compartment.

Clones restricted to the anterior compartment are of two types, which we call ‘neutral’ and ‘reorganizing’ (Table 1). Neutral clones appear morphologically normal and are generally small or found close to the compartment boundary (Table 1A). In contrast, reorganizing clones are found only in more anterior portions of the compartment and are generally larger than neutral clones in the same region of the wing (Table 1B). Moreover, they are associated with abnormal patterns both within and outside of the clone (Fig. 1C-F). Tubα1>en clones in the anterior compartment express only low levels of en protein (below), perhaps accounting for the incomplete pen- etrance of the ‘reorganizing’ phenotype.

The most striking alterations of pattern associated with reorganizing clones are those caused by single clones that con- tribute to both the dorsal and ventral surface of the wing (Fig. 1D), or in rare cases by the conjunction of independent dorsal and ventral clones (Fig. 1C). If the normal pattern is described as 123/45m in which the numbers stand for longitudinal veins, ‘/’ stands for the antero-posterior compartment boundary, and ‘m’ stands for the wing margin posterior to vein 5 (Fig. 1A), then these clones (‘*’) can be described as reorganizing sur- rounding tissue to form double-anterior 23*32 patterns (Fig. 1C,D). Reorganizing clones restricted to only the dorsal or ventral surface do not alter the overall shape and size of the wing. However, they do alter the behavior of surrounding anterior cells on the same surface, typically causing them to form a circular vein 3 (Fig. 1E,F). Thus, reorganizing clones induce neighboring anterior cells to form structures such as vein 3 normally found close to the compartment boundary, and can exert a more long-ranging influence on anterior cells further away, causing them to form more anterior structures such as vein 2. We note that the response of surrounding wild- type cells to Tubα1>en clones appears to depend principally on their distance from the clone. This is apparent both in the mirror-symmetry of the 23*32 patterns associated with clones that mark the dorsal and ventral surfaces of the wing (Fig. 1C,D) and in the formation of circular vein 3’s at a constant distance from the boundary of clones restricted to either of the two surfaces (Fig. 1E,F).

The altered wing patterns formed by wild-type cells sur- rounding reorganizing Tubα1>en clones closely resemble those caused by similarly positioned Tubα1>hh clones (Basler and Struhl, 1994), suggesting that Tubα1>en cells reorganize sur- rounding anterior tissue because they express hh ectopically (see below). However, Tubα1>en cells within these clones differ significantly from their Tubα1>hh counterparts in that they appear to behave like posterior rather than anterior cells. First, they differentiate posterior structures. For example, reorganizing clones that contribute to the wing margin form ‘double row’ bristles characteristic of the posterior wing margin as well as longitudinal veins which meet the wing margin but do not form campaniform sensillae (Fig. 1D). In normal wings, only veins 4 and 5 behave in this way, indicating that the ectopic veins formed by Tubα1>en cells are of posterior type. Because these ectopic veins appear to form very close to the clone border, similar to the proximity of the normal vein 4 to the compartment boundary (Fig. 1A), we infer that they are ectopic vein 4’s, which are specified in response to signals emanating from surrounding anterior tissue. Second, cells within reorganizing clones do not appear to intermix normally with surrounding anterior cells. This is apparent in clones positioned anterior to vein 2, which are generally small and circular (Table 1; Fig. 1E,F) and have an abnormally high cell density (approximately twice that of neighboring wild-type tissue, data not shown).

The last class of Tubα1>en clones consists of clones that typically form a broad swath spanning the normal position of the compartment boundary and sometimes fail to form part or all of longitudinal vein 4 (Fig. 1B). Because Tubα1>en clones restricted to the posterior compartment develop normally whereas sibling clones restricted to the anterior compartment appear to develop like posterior cells, we infer that clones belonging to this third class derive from anterior cells that have acquired ‘posterior’ properties including the ability to assort with normal posterior cells and hence cross into the posterior compartment. The occasional failure of these clones to form normal posterior structures such as vein 4 may be due to the fact that they express only low levels of en protein.

In sum, Tubα1>en clones appear to exert a similar organizing influence on surrounding anterior tissue to that exerted by Tubα1>hh clones. However, they differ from Tubα1>hh clones in that cells within the clone can differentiate posterior structures and assort preferentially with posterior cells.

To examine the consequences of ectopic en expression on hh and dpp expression within the imaginal wing disc, we subjected first and second instar larvae carrying two trans- genes, Tubα1>CD2, y+>en and hs-flp, to a single mild heat shock (Materials and Methods). The Tubα1>CD2, y+>en transgene includes the coding sequence for a reporter protein, rat CD2 (Dunin-Borkowski and Brown, 1995), immediately downstream of the first flp recombination target (Materials and Methods). Hence, excision of the >CD2, y+> flp-out cassette generates Tubα1>en cells, which can be identified in the imaginal wing disc because they do not express CD2.

As shown in Fig. 2A, Tubα1>en clones in the anterior compartment express barely detectable levels of ectopic en protein. Nevertheless, in many clones, this ectopic en expression is sufficient to activate hh transcription in most or all cells of the clone (Fig. 2C), as monitored by the expression of a lacZ enhancer trap insertion into the hh gene (hh^P30; Lee et al., 1992). Further, it correlates with the formation of ectopic ‘doughnuts’ of dpp expression (Fig. 2A,B), as monitored by the expression of either of two dpp-lacZ reporter genes (dpp-lacZBS3.0, Blackman et al., 1991; dpp^P10638; Blackman, personal communication; Materials and Methods). As shown in Fig. 2B, these doughnuts result from expression of the dpp-lacZ gene solely by wild-type cells that surround the Tubα1>en clones. Finally, we note that Tubα1>en clones tend to form circular patches with smooth borders (Fig. 2A-C), providing a further indication that the cells within these clones minimize their contact with surrounding cells.

To compare the behavior of Tubα1>en cells with that of Tubα1>hh cells, we have also examined hh-lacZ and dpp-lacZ expression in wing discs containing marked Tubα1>hh clones. These were generated as described above for Tubα1>en clones, except that a Tubα1>CD2, y⁺>hh transgene was used in place of the Tubα1>CD2, y⁺>en transgene (Materials and Methods). As shown in Fig. 3A, Tubα1>hh clones differ from Tubα1>en clones in that they do not induce ectopic expression of the hh-lacZ gene. Moreover, in contrast to Tubα1>en clones, they are associated with ectopic dpp-lacZ expression in all cells of the clone, as well as in surrounding tissue (Fig. 3B). Finally, these clones form irregularly shaped patches with ragged borders indicating that Tubα1>hh cells are able to interdigitate with surrounding cells (Fig. 3).

We interpret these results as follows. Anterior cells that express en ectopically turn ‘on’ the hh gene and thereby induce neighboring anterior cells to turn ‘on’ the dpp gene. At the same time, they become refractory to the action of hh protein, accounting for the doughnut holes in dpp-lacZ expression. In contrast, anterior cells that express hh ectopically do not lose their competence to respond to hh protein and do not turn ‘on’ the endogenous hh gene. Thus, in terms of their effects on dpp- lacZ and hh-lacZ expression, Tubα1>en cells that arise in the anterior compartment behave as if they are transplanted posterior compartment cells and create an ectopic compartment boundary. In contrast, their Tubα1>hh counterparts behave like normal anterior cells that are positioned close to the compartment boundary and hence exposed to hh protein secreted by neighboring posterior cells.

In wild-type animals, en gene expression is strictly limited to posterior compartment cells until the middle of the third larval instar; however, anterior cells neighboring the compartment boundary subsequently express en or the en-related gene invected (inv), or both (Blair, 1992). This late en-inv expression appears to play a minor role in controlling the differentiation of anterior cells in the immediate vicinity of the compartment boundary (Hidalgo, 1994; our unpublished observations). We note that Tubα1>en gene activity appears to be functionally distinct from this late, endogenous en-inv activity. Specifically, Tubα1>en cells positioned next to the compartment boundary can turn ‘off’ dpp^P10638 gene expression (e.g., Fig. 2A), turn ‘on’ the hh^P30 gene and cause neighboring anterior cells to express dpp^P10638, and hence differ in all three respects from wild-type en-inv-expressing cells in the same position. Conse- quently, we suggest that the late ‘en-inv’ activity in the anterior compartment reflects the expression of only inv and that inv protein differs from en in that it cannot block dpp transcription or activate expression of hh.

Further evidence that the state of en gene activity controls the ability of wing cells to send as well as to receive the hh signal was obtained by examining clones of posterior cells lacking the endogenous en and inv genes. In this case, clones of cells homozygous for Df(2R)en^E, a deletion of en and the adjacent en- related gene invected (inv) (T. Kornberg, personal communica- tion), were induced in wing imaginal discs using flp-mediated mitotic recombination (Golic, 1991). In this experiment, cells within the clone were independently marked either by the loss of en protein expression, or by the simultaneous loss of the marker gene hs-πM (Xu and Rubin, 1993; Materials and Methods).

Clones of Df(2R)en^E cells in the posterior compartment fail to express the hh-lacZ gene, in contrast to surrounding posterior cells (Fig. 4A,B). Moreover, all of the cells within these clones express the dpp-lacZ gene, in contrast to the surrounding posterior cells which do not (Fig. 4C,D). Similar results have also been reported recently by Sanicola et al. (1995). We also find that these Df(2R)en^E clones appear to stimulate excessive proliferation in surrounding wild-type cells. In particular, posterior wing compartments containing several clones appear to be abnormally large and the wild-type tissue neighboring these clones appears to contain extra folds (Fig. 4A,C). Previous analyses of en mutant clones in the adult have also suggested that posterior cells lacking en function stimulate excessive proliferation in surrounding cells (Lawrence and Morata, 1976; Hidalgo, 1994).

We interpret these results as follows. en gene activity normally programs wing cells to be hh signalers: that is, they secrete hh protein, but cannot respond to it by expressing dpp. Conversely, the absence of en activity programs them to be hh responders: they do not express hh, but respond to hh protein by expressing dpp. Consequently, both Tubα1>en anterior clones and Df(2R)en^E posterior clones create ectopic interfaces between signaling and responding cells, and hence lead to ectopic dpp expression.

The results of experiments involving gain or loss of hh or en function reveal a close association between dpp expression and the organizing activities of hh and en (Basler and Struhl, 1994; Capdevila et al., 1994; Capdevila and Guerrero, 1994; Tabata and Kornberg, 1994; here). Hence, dpp protein secreted by anterior cells may organize both anterior and posterior compartment patterns by exerting a spreading influence on cells on both sides of the boundary. If this inference is correct, ectopic dpp expression generated in either compartment should organize symmetric anterior or posterior patterns, depending on the compartmental provenance of the responding cells.

To test this prediction, we have induced clones of Tubα1>dpp cells marked by the loss of the f⁺gene using a Tubα1>f⁺>dpp transgene (Materials and Methods). As described above for the generation of Tubα1>en and Tubα1>hh clones, Tubα1>dpp clones were induced by a single, mild heat pulse during early larval development. We note that we have not been able to detect ectopic dpp protein expression associated with Tubα1>dpp cells with the available antibodies (Panganiban et al., 1990) indicating that the level is low relative to expression of the endogenous gene in anterior cells along the compartment boundary.

Following the induction of Tubα1>dpp clones, we obtained flies with two types of abnormal wings. First, we observed wings that were of normal size and shape, but which contained ectopic vein structures (e.g. Fig. 5A). In virtually all such cases, these ectopic vein structures were formed by Tubα1>dpp cells. Moreover, they could be formed by very small clones consisting of only a few cells. The dpp gene is known to have two distinct functions during wing develop- ment, a disc function, which is required for long-range organization of wing pattern, and a shortvein function, which is necessary for vein differentiation (Spencer et al., 1982; Segal and Gelbart, 1985; Posakony et al., 1990). Hence, we attribute the autonomous venation phenotype exhibited by Tubα1>dpp clones to the shortvein function of dpp and do not consider it further.

Second, we found wings of abnormal size and shape which displayed dramatic alterations of the normal pattern (Fig. 5B- D; see Materials and Methods for quantitative data). These abnormalities were invariably associated with clones of Tubα1>dpp cells that contributed to both the dorsal and ventral surface. However, unlike the autonomous ectopic vein phenotype described above, only a small fraction of the cells that contributed to these altered patterns were genotypically Tubα1>dpp. Hence, Tubα1>dpp cells can exert a long-range organizing influence on the behavior of surrounding wild-type cells. We describe the nature of these reorganized patterns below.

Anterior compartment clones associated with reorganizations of surrounding wild-type tissue were only observed in the more anterior portions of the anterior compartment (anterior to vein 2). In contrast to similarly positioned Tubα1>en and Tubα1>hh clones, these Tubα1>dpp clones appear to organize less extensive patterns, which can be described as 12*21 or just 1*1 patterns, rather than 123*321 or 23*32 patterns (Fig. 5B,C). Thus, Tubα1>dpp cells in clones associated with reorganized anterior patterns appear to behave like wild-type cells positioned close to vein 2 and to organize surround- ing tissue to form symmetric, more anterior patterns on either side.

We also found posterior Tubα1>dpp clones associated with reorganized patterns of posterior wing tissue, including the formation of supernumerary double posterior wings (Fig. 5D). These clones were all positioned posterior to vein 5 and caused patterns that can be described as m5*5m or m54*45m supporting our proposal (Basler and Struhl, 1994) that the normal /45m pattern is organized by dpp protein secreted by anterior cells adjacent to the compartment boundary.

To test further the hypothesis that localized dpp expression is sufficient to organize both anterior and posterior compartment patterns in the wing, we have generated genetically marked Tubα1>dpp clones in dpp^d8/dpp^d10 animals. This combination of dpp mutations reduces normal dpp activity in the wing imaginal disc, blocking proliferation and patterning in both compartments. Consequently, dpp^d8/ dpp^d10 flies form only rudimentary wings (Spencer et al., 1982; Fig. 6A). As shown in Fig. 6B-F, this ‘no wing’ phenotype can be partially rescued by Tubα1>dpp clones. In particular, we find flies that bear ‘winglets’ composed of symmetric double-anterior (1*1) or double-posterior (m*m) patterns and, in every such winglet, we observe a stripe of Tubα1>dpp cells along the plane of symmetry on both the dorsal and ventral surface (Fig. 6C,D). Moreover, we also find flies that form asymmetric anterior-posterior winglets exhibiting a 1*m pattern (Fig. 6B,E,F) as well as flies that form two winglets from a single wing primordium (e.g., Fig. 6F). In all cases in which two winglets are formed by a single primordium, double-anterior winglets arise anteriorly and double-posterior winglets arise posteriorly to the remaining winglet (e.g., Fig. 6F). Thus, ectopic Tubα1>dpp expression can suffice to organize both anterior and posterior wing pattern, even in the absence of endogenous dpp gene function. Moreover, clones arising in the anterior or posterior compartment at a distance from the boundary appear to organize double-anterior or double-posterior winglets, respectively, whereas clones arising in close proximity to the boundary appear to organize asymmetric anterior-posterior winglets composed of cells from both compartments. Thus, we infer that the type of pattern formed depends on the compartmental provenance of the responding cells and hence on the state of activity of the en gene.

We also note that double anterior and double posterior winglets generated by Tubα1>dpp clones in the absence of endogenous dpp activity appear to be less extensive than those organized by Tubα1>dpp clones in wild-type wings. For example, such clones in the anterior compartment form 1*1 rather than 12*21 patterns, while those in the posterior compartment form m*m rather than m5*5m or m54*45m patterns (compare Figs 5C and 6C, and Figs 5D and 6D). In principle this could be because Tubα1>dpp expression in otherwise wild-type wings can induce the expression of the endogenous dpp gene. However, as noted above, we cannot detect any dpp protein expression associated with Tubα1>dpp clones; moreover Tubα1>dpp cells do not appear to induce expression of dpp-lacZ reporter genes (data not shown). Another possi- bility is that in dpp⁺ discs, dpp protein secreted by Tubα1>dpp cells acts additively with endogenous dpp protein secreted by anterior cells along the compartment boundary.

In sum, ectopic expression of dpp can reorganize both anterior and posterior wing pattern, depending on the com- partmental provenance of the responding cells. This result supports the proposal that hh organizes both anterior and posterior wing development by inducing anterior cells along the compartment boundary to secrete dpp. Moreover, it indicates that dpp specifies anterior or posterior wing pattern- ing according to the state of en activity in responding cells. We also find that the level of ectopic dpp expression generated in these experiments is low relative to endogenous dpp expression and suffices to reorganize only those portions of each compartment which are positioned at a distance from the compartment boundary. As we discuss below, these and related findings provide suggestive evidence that dpp exerts its organizing influence by acting as gradient morphogen.

Compartments have been proposed to play at least two roles: (i) to define distinct cell populations within which certain ‘selector’ genes control developmental fate (the compartment hypothesis; Crick and Lawrence, 1975; Garcia-Bellido, 1975; Lawrence and Morata, 1976), and (ii) to define interactive cell populations that generate organizing signals along the boundaries between them (the boundary hypothesis; Meinhardt, 1983; see also Crick and Lawrence, 1975). Our analysis of en, hh and dpp function in the wing provides evidence that anterior and posterior compartments perform both roles during limb development, and that they do so because of the consequences of the different states of en activity in each compartment.

First, our results confirm and extend previous findings (Morata and Lawrence, 1975; Lawrence and Morata, 1976; Lawrence and Struhl, 1982; Morata et al., 1983; Busturia and Morata, 1988) that en activity in posterior cells directs them not to intermix with anterior cells and hence subdivides each limb primordium into adjacent but immiscible cell populations. This is most clearly illustrated by our finding that ectopic en expression in anterior cells causes these cells to form circular patches that minimize contact with surrounding anterior cells or to cross the compartment boundary and mix with posterior cells on the other side (Figs 1, 2; see also Busturia and Morata, 1988). Conversely, we also observe that loss of endogenous en activity in posterior cells can lead to their minimizing contact with surrounding posterior cells, extending previous observations that such clones cross into the anterior compartment or are eliminated from the sur- rounding en-expressing epithelium (Lawrence and Struhl, 1982; Morata et al., 1983).

Second, we show that the state of en activity (whether ‘on’ or ‘off’) also dictates whether wing cells secrete hh protein or can respond to hh protein by secreting dpp (Figs 2, 3). Because en encodes a homeodomain protein implicated as a transcriptional repressor (Jaynes and O’Farrell, 1990; Han and Manley, 1993), en protein could directly bind targets in dpp and block transcription. This possibility is supported by the recent finding of en binding sites in the dpp gene which appear to mediate transcriptional repression of dpp in the posterior compartment (Sanicola et al., 1995). Similarly, it may bind and repress the transcription of other genes expressed only by anterior cells and involved in mediating their ability to respond to secreted hh protein (e.g., patched, cubitus interruptus^D (ci^D); reviewed in Hooper and Scott, 1992). The role of en in driving hh tran- scription is more complex because the initial transcription of hh in posterior cells is not dependent on en (Lee et al., 1992; Mohler and Vani, 1992; Tabata et al., 1992; Tashiro et al., 1993). Furthermore, although hh transcription in embryos sub- sequently becomes en dependent, the prevalent view is that this reflects a non-autonomous requirement for en function mediated by en-dependent wg signaling from neighboring cells (e.g., Perrimon, 1995). For the imaginal wing disc, our data establish that en activity is both necessary and sufficient to drive hh expression (Figs 2, 3) and indicate that the state of en gene activity autonomously dictates the transcriptional activity of the hh gene. Accordingly, we argue that once embryonic cells become heritably commited to either express or not express en, their descendants will be hh signalers or hh responders for the remainder of development.

Third, as a consequence of the ability of en to control tran- scription of downstream genes such as hh, dpp, ptc and ci^D, the selective activity of en in posterior cells creates an inductive interface wherever populations of anterior and posterior cells abut. The importance of this interaction is illustrated by experiments in which en is ectopically activated in anterior compartment cells (Figs 1, 2). Under these circumstances, the creation of an ectopic interface between en- expressing and non-en-expressing cells causes cell pattern to be reorganized on both sides of the interface. This is dramatically illustrated in Fig. 1D by the formation of both posterior (vein 4) and anterior (vein 3) pattern elements on opposite sides of the border of a Tub>en clone. Thus, the different epigenetic states of posterior and anterior cells allows them to interact across the compartment boundary to elicit signals that can organize growth and patterning in both compartments.

Finally, we show that cells expressing dpp can organize both anterior and posterior wing pattern (Figs 4, 5). Although ectopic dpp expression has previously been shown to cause the formation of supernumerary anterior or posterior wing patterns (Capdevila and Guerrero, 1994) these studies did not address the question of whether the ectopic dpp-expressing cells exert a long-range organizing influence on surrounding, non- expressing cells. Here we report that ectopic dpp-expressing cells are indeed able to exert a long-range influence over sur- rounding non-dpp-expressing cells. Moreover, we have obtained evidence that the compartmental provenance of the responding cells determines whether they respond by forming anterior or posterior pattern. We therefore infer that the state of en activity (i.e., whether ‘on’ or ‘off’) governs not only the abilities of wing cells to intermix and to induce organizing signals such as dpp, but also their capacity to respond to these signals in distinct ways, thereby generating different cellular patterns.

In our prior analysis of ectopic hh-expressing cells (Basler and Struhl, 1994), we found that the ability of these cells to organize supernumerary double-anterior wings depends on their contributing to both the dorsal and ventral surfaces of the wing, which are themselves compartments established during larval development (Garcia-Bellido et al., 1973, 1976). In the present study, we have obtained similar findings for clones of ectopic en-expressing and ectopic dpp-expressing cells. These findings reinforce our proposal (Basler and Struhl, 1994) that the organization of cell proliferation and patterning along the proximodistal axis of the wing may require two signals: one, dpp, induced along the anteroposterior compartment boundary; the other induced by interactions between cells on opposite sides of the dorsoventral compartment boundary (Diaz-Benjumea and Cohen, 1993; Blair et al., 1994; Irvine and Wieschaus, 1994; Tabata and Kornberg, 1994; Williams et al., 1994). These two signals would normally be generated together in the vicinity of cells destined to give rise to the distal tip of the adult wing, consistent with the notion (Campbell et al., 1993) that they would define a distal ‘organizer’ governing growth and patterning along the proximodistal axis.

hh and dpp each belong to extensive families of signaling molecules which have potent organizing activities in animal development. Depending on the organism and developmental process examined, they have been described variously as inducers or morphogens. Accordingly, there is considerable controversy about how they exert their organizing influence – in particular, whether they do so by initiating or furthering a chain of inductive interactions, or by a gradient mechanism in which different concentrations dictate different cellular outputs.

In the context of the developing Drosophila wing, our present results provide strong support for the hypothesis that the long-range organizing activity of hh is mediated indirectly, through the short-range induction of dpp (Basler and Struhl, 1994). In the case of the anterior compartment, we previously demonstrated that hh is both necessary and sufficient to induce dpp expression in anterior wing cells and to organize anterior wing pattern. Here we extend these findings by showing that dpp expression can exert a similar long-range organizing influence on anterior pattern. Hence, the long-range organizing activity of hh can be attributed to its ability to induce dpp. The argument that dpp mediates the long-range organizing activity of hh is further strengthened by recent studies of the involvement of protein kinase A (PKA) in hh signal transduction (Jiang and Struhl, 1995; Lepage et al, 1995; Li et al., 1995; Pan and Rubin, 1995). These studies have shown that the loss of PKA activity mimics the reception of the hh signal and causes long-range reorganizations of anterior wing pattern. Moreover, they have estabished that these reorganizations result from ectopic expression of dpp by PKA-deficient cells. The situation in the posterior compartment provides an additional line of evidence that hh organizes wing development by inducing dpp. Our present results show that en gene activity in posterior cells renders them refractory to hh signaling, even though they express the hh gene and are presumably exposed to secreted hh protein. Nevertheless, growth and patterning of this compartment depends critically on hh gene activity (Basler and Struhl, 1994; Hidalgo, 1994) as well as on dpp expression in neighboring anterior cells across the compartment boundary (Posakony et al., 1990). Because en activity blocks the ability of posterior cells to respond to hh protein, it follows that hh must organize posterior compartment development indirectly by inducing dpp expression in neighboring anterior cells. This inference is further supported here by our demonstration that ectopic dpp expression in the posterior compartment can organize symmetric double-posterior wings (Fig. 4D). Thus, as in the anterior compartment, the long-range organizing role of hh in the posterior compartment can be attributed solely to the induction of dpp in anterior cells neighboring the compartment boundary.

Thus, for hh, the available evidence provides a strong argument that most or all of its organizing activities in the developing wing are due to its activity as a short-range inducer. We note that our experiments do not rule out the possibility that cell patterning in the vicinity of the compartment boundary may depend on other short-range responses to hh aside from dpp induction. Indeed, hh appears to trigger several other responses in anterior cells close to the boundary such as the expression of the reporter genes LF06 and P1531 and enhanced transcription of the gene ptc (Tabata and Kornberg, 1994; Capdevila and Guerrero, 1994). However, there is presently no evidence to suggest that different concentrations of hh protein elicit distinct short-range responses involved in organizing cell pattern. Our interpretation of hh as a short-range inducer contrasts with the recent proposal that hh exerts a long-range organizing influence on the dorsal epidermis of the larva by acting as a concentration-dependent morphogen (Heemskirk and DiNardo, 1994). However, we note that dpp is normally expressed in cells contributing to the dorsal larval epidermis (Jackson and Hoffmann, 1994) and, hence, may be mediating the organizing influence of hh in this tissue as well.

In contrast to hh, our results with dpp provide evidence that appears to favor a morphogen interpretation. Specifically, we observe that levels of ectopic dpp expression that are sufficiently low to fall beneath our level of detection are ‘neutral’ when generated close to the compartment boundary, the source of endogenous dpp protein, but can reorganize wing pattern when generated at a distance from the boundary in either compartment. Moreover, in the absence of endogenous dpp activity, these low levels of ectopic dpp expression organize only those portions of anterior or posterior compartment pattern that are normally formed at a distance from the compartment boundary. As noted previously, (Tickle et al., 1975; Tickle, 1981; Struhl and Basler, 1993), the ability of low levels of a putative signaling molecule to organize such subdomains of the normal pattern favors models involving gradient rather than sequential inductive mechanisms.

One criticism of this type of evidence is that hh signaling may induce the expression of molecules other than dpp, and these molecules may act in concert with dpp or modulate its activity to specify other elements of wing pattern closer to the compartment boundary. According to this hypothesis, the restricted organizing activities of ectopic dpp would be due to the absence of these other factors rather than to the limited amounts of ectopic expression that we have been able to generate experimentally. However, we have observed that posterior cells in dpp^d8/dpp^d10 wing discs appear to express hh normally (as assayed by the expression of the hh^P30 reporter gene; data not shown). Hence, other responses to hh signaling aside from the induction of dpp should be normal. Nevertheless, Tubα1>dpp cells appear to have a similarly modest organizing activity in these discs, irrespective of whether they are expressed close to the compartment boundary, or at distance (Fig. 6). Consequently, the restricted organizing ability of Tubα1>dpp cells positioned at a distance from the compartment boundary does not appear to be due to the absence of other factors that are normally induced along the boundary by hh.

Another line of evidence that argues in favor of a gradient mechanism is our observation that the organizing potency of Tubα1>dpp-expressing cells is enhanced by the presence of endogenous dpp gene function (compare Figs 5 and 6). Although we do not know the mechanism by which this enhancement occurs, the fact that it occurs suggests that different concentrations of dpp protein can trigger different responses.

The key attribute of a morphogen, as distinct from an inducer, is that a morphogen organizes pattern by dictating distinct cellular responses at different concentrations. Hence, if dpp functions as a morphogen, we would expect that progressive increases in the level of ectopic dpp expression would correlate with corresponding increases in organizing potency. Such experiments have been performed for the dorsoventral organizing activity of dpp in early embryos and support the conclusion that dpp functions as a morphogen in this context (Ferguson and Anderson, 1992; see also Wharton et al., 1993). It remains to be seen whether equivalent experiments with dpp in the imaginal discs will provide further support for a gradient mechanism during limb development.

We thank T. Kornberg for communicating results and providing the Df(2R)en^E mutation prior to publication, Y. Chen and M. Brunner for help with generating clones of Df(2R)en^E cells, M. Brunner, A. Nakanishi and R. Perez for technical assistance, P. Beachy, R. Blackman, T. Kornberg, T. Xu and the Bloomington stock center for fly stocks, N. Brown for the CD2 coding sequence, N. Patel for the en antibody, V. Corces for the f genomic DNA, and Y. Chen, J. Dubnau, I. Greenwald, E. Hafen, T. Jessell, J. Jiang, P. A. Lawrence and A. Tomlinson for discussion and comments on the manuscript. K. B. is supported by the Swiss National Science Foundation and the Kanton of Zürich, and G. S. is an Investigator of the Howard Hughes Medical Institute.