The roles of hedgehog and engrailed in patterning adult abdominal segments of Drosophila

The epidermis of the embryonic segments and imaginal discs of Drosophila is organized into lineage compartments defined by expression of selector genes. Compartments along the anterior-posterior axis are determined by the engrailed (en) gene, which is expressed in the posterior but not the anterior compartment of each segment. Pattern formation within each segment is largely dependent on interactions between anterior and posterior cells at compartment boundaries (reviewed in: Lawrence and Struhl, 1996; Tabata et al., 1995).

An important component of signaling between compart-ments is the secreted protein encoded by hedgehog (hh). hh is expressed in posterior compartment cells under the control of en (Lee et al., 1992; Mohler and Vani, 1992; Tabata et al., 1992; Zecca et al., 1995), and functions as a short-range inducer that causes a stripe of adjacent anterior compartment cells to respond by activating hh target genes, which include the Hh receptor component patched (ptc), the transcription factor cubitus interruptus, and the long-range morphogens wingless (wg) and decapentaplegic (dpp) (Alexandre et al., 1996; Dominguez et al., 1996; Hepker et al., 1997; Ingham, 1993; Johnson et al., 1995).

The long-range morphogens induced by hh serve to pattern both anterior and posterior compartments. wg plays such a role in the embryonic epidermis (Dougan and DiNardo 1992; Lawrence et al., 1996) and dpp has been shown to pattern much of the wing disc (de Celis et al., 1996; Lecuit et al., 1996; Nellen et al., 1996). Both dpp and wg participate in patterning the leg and antennal discs (Brook and Cohen, 1996; Jiang and Struhl, 1996; Penton and Hoffmann, 1996; Theisen et al., 1996). However, some functions of hh in imaginal discs and the embryonic epidermis are independent of dpp and wg (Bokor and DiNardo, 1996; Jiang and Struhl, 1995; Lawrence et al., 1996; Li et al., 1995; Mullor et al., 1997).

In this report, we address the roles of hh and en in the epidermis of the adult abdomen of Drosophila. Transplantation experiments in other insect species suggest that pattern in abdominal segments is established by segmental morphogen gradients (Bhaskaran and Röller, 1980; Campbell and Caveney, 1989; Lawrence, 1966). A major motivation for our work is to determine whether Hh or other signaling molecules are respon-sible for such gradients in the abdomen of Drosophila.

The epidermis of the adult abdomen in Drosophila develops from histoblast nests that are laid down during embryogenesis as derivatives of the embryonic epidermis (Simcox et al., 1991). Each larval abdominal hemisegment contains three major histoblast nests: an anterior dorsal nest, which produces the abdominal tergite, a posterior dorsal nest, which produces the intertergal cuticle, and a ventral nest, which produces the sternite and pleural cuticle (Madhavan and Madhavan, 1980). Unlike imaginal disc cells, abdominal histoblasts do not pro-liferate during the larval stages. Instead, they remain part of the larval epidermis and secrete larval cuticle through each of the larval molts. Histoblast mitoses resume 1-1.5 hours after puparium formation (APF) and continue until approximately 40 hours APF, as the histoblasts migrate and replace the degenerating larval epidermal cells (LEC) (Madhavan and Madhavan, 1980). By 40-41 hours APF, all LEC are replaced by the newly formed imaginal epidermis.

As in other segments, the epidermis of adult abdominal segments is subdivided into anterior and posterior compartments (Hama et al., 1990; Kornberg, 1981). We present evidence that en and hh specify cell fates in the posterior compartment and that hh also functions non-autonomously to induce posterior characteristics within the anterior compartment. Although hh is able to polarize anterior compartment cells, it does not appear to play a primary role in establishing polarity. Rather, we present evidence that the abdominal segment is polarized symmetrically by a mechanism that is independent of hh and that hh causes anterior compartment cells to reverse their interpretation of this underlying polarization.

Enhancer trap lines used were: Xho-25 en-lacZ (Hama et al., 1990), P2023-44 hh-lacZ (Tabata et al., 1992), P{ry=PZ} wg(G. Rubin/ BDGP), P{ry=PZ} dpp(A. Spradling/ BDGP). FLP-out en, hh and dpp constructs are described by Basler and Struhl (1994) and Zecca et al. (1995). The hs-FLP line used was the X-chro-mosomal FLP 122, provided by Konrad Basler. The hs-GAL4 driver used was P{w=GAL4-Hsp70.PB}89-2-1 (Brand and Perrimon, 1993). The hs-hh line is described by Tabata et al. (1992). The UAS-ptc line was a gift from Matt Scott (Johnson et al., 1995).

Adult abdominal cuticles were prepared and mounted as described by Duncan (1982).

Heat-shock pulses were administered in glass vials in an agitated water bath at 37°C. For FLP-out experiments, a single 10-minute heat shock was given.

To prepare the pupal abdominal epidermis for staining, pupae were bisected longitudinally with a razor blade and placed in an Eppendorf tube with 1 ml of ice-cold PBS+0.1% Tween-20 (PBT). The tube was gently agitated to remove most of the gut and fat body. For in situ hybridization, body walls were then fixed for 30 minutes in 5% formaldehyde in PBT. For antibody staining, fixation was for 10 minutes in 1% formaldehyde in PBT, and, for X-Gal staining, 10 minutes in 1% glutaraldehyde in PBT. After fixation, body walls were washed twice for 10 minutes in PBT, stained in 0.5% eosin-Y (Sigma) in PBT and rinsed in PBT. Internal organs and most muscles were then removed with a pair of sharp forceps, leaving the abdominal epidermis exposed.

In situ hybridization was as described by Poeck et al. (1993), with the following modifications. Pupal body walls were digested in 2 μg/ml Proteinase K for 2-3 minutes; stronger treatment destroys the abdominal epidermis. Hybridization and washes were performed at 55°C, and 0.1% Tween-20 was present in all solutions. The hh RNA probe used was prepared as described by Tabata et al. (1992). X-Gal staining was performed as described by Hama et al. (1990). The monoclonal antibody 4D9 of Patel et al. (1989), which recognizes both en and invected proteins, was used for En stainings. Antibody stainings were as described by Kellerman et al. (1990). Stained body walls were washed in ethanol, rehydrated, mounted flat in 80% glycerol and viewed with DIC optics.

Each abdominal segment produces a large dorsal cuticular plate, the tergite, and a smaller ventral plate, the sternite. Tergites of adjacent segments are separated by flexible inter-tergal cuticle, and sternites are separated from one another and from the tergites by pleural cuticle. Each tergite can be divided into three regions: an anterior region (the acrotergite) that contains undecorated sclerotized cuticle, a central region containing an array of microchaetes, and a posterior region that contains a dark pigment band and a row of large macrochaetes at its posterior edge (Fig. 1A). The central and posterior regions, but not the acrotergite, are covered with trichomes. For convenience, we define the posterior boundary of the tergite to be the posterior edge of the pigment band. This does not agree with other authors (Madhavan and Madhavan, 1980; Roseland and Schneiderman, 1979). The intertergal cuticle is unpigmented and composed of an anterior trichome-bearing region (the posterior hairy zone or PHZ) and a posterior region of naked cuticle (the intersegmental membrane or ISM). All trichomes and bristles in the abdomen are oriented from the anterior to the posterior.

The positions of the compartment and segment boundaries relative to cuticular structures in the abdomen can be inferred from expression of en-lacZ in newly emerged adults (Hama et al., 1990, and Fig. 2A). The anterior boundary of en expression, which marks the compartment boundary, is not fixed precisely with respect to cuticular pattern. In some cases it coincides with the tergite-PHZ boundary, while in others it lies 3-4 trichome rows posterior to it (Fig. 2B, C). This indicates that a 3-to 4-cell-wide region of the PHZ can be contributed by either compartment. The posterior edge of each en stripe coincides with the ISM-acrotergite border (Hama et al, 1990, and Fig. 2A).

In larvae, en is expressed throughout the posterior dorsal his-toblast nest, but is not expressed in the anterior dorsal nest. In the single ventral nest, posterior cells express en, whereas anterior cells do not (Hama et al., 1990). We find that this pattern is preserved in prepupae and early pupae (not shown). Following fusion of the dorsal nests at 18-20 hours APF, en protein accumulates in a gradient that is highest at the anterior-posterior compartment boundary and trails off posteriorly (Fig. 3A).

The most posterior row of LEC in each segment is the last to be replaced by the expanding histoblast nests. These ‘border cells’ are easily distinguished from their neighbors by a more convex basal surface (Fig. 3B). The border cells are replaced by en-expressing histoblasts, indicating that adult and larval segment boundaries coincide precisely. The gradient of en protein persists until several hours after elimination of the border cells. Later, en expression in the posterior compartment becomes uniform, as seen in the adult (Fig. 2A).

We have recovered a gain-of-function allele of hh, hh^Mirabile (hh^Mir), that causes a dramatic phenotype in the adult abdomen in heterozygotes: the anterior tergite is absent and replaced by a mirror-image duplication of the posterior tergite and intertergal region (Fig. 1B). The duplicated structures correspond to the posterior compartment and the posterior region of the anterior compartment. Internally, the oenocytes underlying the posterior tergite are also duplicated, and the insertion points of the adult muscles reflect the symmetric epidermal pattern (not shown). The sternites are usually unaffected. With the exception of a mid-dorsal thoracic groove seen in heterozygotes raised at 29°C, the effects of hh^Mir appear to be limited to the adult abdomen.

hh^Mir was isolated following X-ray mutagenesis and is associated with a complex chromosomal rearrangement (new order: 1-20F|81F-64BC|94A1-81F|94A1-100 + 61-64BC|20F). To eliminate involvement of the X chromosome, a derivative of sequence 61-64BC|81F-64BC|94A1-81F|94A1-100 was recovered after further irradiation. By recombining this derivative with other inversions, we mapped the abdominal phenotype of hh^Mir to near the 81F|94A1 breakpoint. Allelism with hh was established by recovering X-ray-induced reversions. Three such revertants had chromosomal breaks at the hh locus (94DE) and behaved as hh nulls or hypomorphs. hh^Mir complements the recessive lethality of hh, suggesting that it does not compromise most normal functions of hh.

In wild type, hh expression in histoblasts closely matches the pattern of en expression. In prepupae and early pupae, hh transcript can be detected throughout the posterior dorsal nest, but is absent from the anterior dorsal nest (Fig. 3C). Following fusion of the two nests, hh is expressed in an anterior-posterior gradient with a sharply defined anterior border, which presumably corresponds to the compartment boundary (Fig. 3D). The enhancer trap P2023-44 (Tabata et al., 1992) is expressed in histoblasts in a pattern identical to the hh transcript (not shown). We also see hh expression in the LEC in a posterior stripe that appears to match the en stripe, with the notable exception that hh is not expressed in the larval border cells (Fig. 3C,D). hh transcript can be detected in the border cells in hs-hh pupae following heat shock (see below), indicating that the failure of these cells to stain in wild type is not due to exclusion of the hh probe.

The basic defect in hh^Mir appears to be that it drives ectopic expression of hh in the anterior compartment in the pupal abdomen. Consistent with the double posterior cuticular pattern seen in the adult, hh is expressed strongly at the anterior and posterior edges of the combined dorsal histoblast nest in hh^Mir heterozygotes (Fig. 3E). hh expression in the ventral histoblasts is usually normal (not shown).

We find that hh^Mir also causes ectopic activation of both the wild-type hh allele and en. hh activation was monitored by the expression of the hh-lacZ enhancer trap P2023-44 (Tabata et al., 1992). When heterozygous with hh^Mir, this reporter was expressed in the ectopic anterior domain, although at only a low level (not shown). We also observe an ectopic en stripe in the middle of the anterior compartment (Fig. 2D). Ectopic expression of hh in the anterior compartment of the wing also causes activation of en and endogenous hh (Guillén et al., 1995).

To test the effects of ubiquitous expression of hh, prepupae and pupae carrying a hs-hh transgene (Tabata et al., 1992) were subjected to 2-hour heat shocks at different stages. Heat-shock treatment during the first 5 hours APF has no effect in the abdominal cuticle. Heat shocks delivered between 5 and 10 hours APF result in the transformation of posterior microchaetes into macrochaetes (Fig. 4A). Flies subjected to single heat shocks between 11 and 17 hours APF are usually unaffected, although weak disturbances in bristle polarity are occasionally observed (not shown). Finally, heat-shock treatment between 18 and 35 hours APF (23-29 hours APF is the most sensitive stage) results in an anterior expansion of the pigment band of the tergite without any effect on the bristle pattern (Fig. 4B).

When flies carrying two copies of the hs-hh construct are subjected to three or four 2-hour heat shocks between 5 and 30 hours APF, tergites are compressed to approximately one half of their normal width, while the ISM is considerably expanded. The polarity of structures in the anterior tergite is sometimes completely reversed (Fig. 4C). We also observed mirror-image posterior duplications in many sternites (Fig. 4E). No com-parable abnormalities were observed in wild-type flies subjected to similar treatment. We observed no stable ectopic activation of either hh or en-lacZ following heat-shock treatment, suggesting that the phenotypes observed were caused by the pulses of uniform hh expression.

The phenotypes produced by ubiquitous hh expression suggest that relatively low levels of ectopic hh are sufficient to promote posterior cell fates in the anterior of the tergite, but higher levels are required to affect polarity. This is supported by the phenotype of a partial reversion of hh^Mir, hh^Mir-RevBx1. Heterozygotes for this revertant show expansion of the posterior pigment band to include most of the tergite and transformation of some microchaetes to macrochaetes (Fig. 1C), but do not show any polarity reversal. Homozygotes die as embryos that show a partial hh loss-of-function phenotype, indicating that hh^MirRevBx1 is associated with a hypomorphic allele of hh. We find that the expression pattern of hh^MirRevBx1 is identical to that of hh^Mir (not shown). Hence, it would appear that hh protein produced by hh^MirRevBx1 is sufficiently active to promote posterior characteristics, but not to cause polarity reversal.

To test the autonomy of hh, we generated ‘FLP-out’ (Basler and Struhl, 1994) hh-expressing clones in the anterior compartment. These produce ectopic ISM, PHZ and posterior tergite structures. Since the cuticular marker used (yellow) cannot be scored in ISM or PHZ, we cannot determine the limits of hh-expressing clones within these cuticle types. Except for an occasional yellow bristle, the posterior tergite structures induced were produced by y⁺ cells, indicating that hh functions non-autonomously.

The effects of hh-expressing clones in the anterior compartment depend strongly on their position in the tergite. Clones in the middle region, anterior to the pigment band, induced con-centric patterns that consisted of a patch of PHZ cuticle sur-rounded by posterior tergite structures (Fig. 5C). Trichomes and bristles in the reorganized areas were oriented radially towards the center of the patch. In contrast, more anterior clones were polarized in their effects, and induced ectopic posterior tergite structures only at their posterior edges. Tergite structures located laterally and anteriorly to such clones were completely unaffected (Fig. 5A). This is not due to differential competence to respond to hh; the same tergite region produces posterior tergite structures when located posterior to hh-expressing clones, but is unaffected when located lateral or anterior to such clones (Compare Fig. 5A and B). These observations suggest that secretion of hh protein may be polarized in these anterior clones. Anterior clones also induced PHZ and ISM; the PHZ always had reversed polarity and was located to the posterior of the ISM. The boundary between the two was straight and perpendicular to the anterior-posterior axis of the segment (Fig. 5A). The pattern produced by large anterior hh-expressing clones is remarkably similar to that of hh^Mir. Identifiable clones in the posterior tergite were rare, but a few patches of PHZ associated with ectopic macrochaetes were found (Fig. 5D). Polarity of the PHZ and macrochaetes was normal.

To remove hh function from the developing abdomen, homozygotes for the temperature-sensitive allele hh^ts2 (Ma et al., 1993) were reared at 18°C up to the late third larval instar and shifted to 29.5°C immediately after pupariation. Among animals surviving to the late pharate adult stage, we found that many hemitergites displayed mirror-image anterior duplications. In such cases, posterior tergite, PHZ and ISM were deleted and replaced with a mirror-image duplication of the anterior tergite, creating a pattern opposite to that seen in hh^Mir(Fig. 6A). Other phenotypes found include complete or partial deletion of PHZ and ISM, missing posterior pigment band and transformation of macrochaetes to microchaetes (Fig. 6B). These weaker phenotypes are reciprocal to the patterns seen in hh^MirRevBx1 heterozygotes and hs-hh animals given single heat shocks at the pupal stage. Ventrally, the sternites were greatly reduced in size and the sternal bristles were usually absent (not shown). No comparable abnormalities were observed in wild-type flies reared at 29.5°C.

The loss-of-function phenotype of hh can be mimicked when hh signaling is blocked by overexpression of ptc. We expressed ptc ubiquitously in the pupal abdomen by subjecting hs-GAL4/UAS-ptc flies (Johnson et al., 1995) to a single 2-hour heat shock at the prepupal or early pupal stage. The phenotypes produced included the loss of the pigment band, transformation of macrochaetes to microchaetes and complete or partial deletion of PHZ and ISM (Fig. 6C).

The function of hh in patterning the posterior portion of the tergite does not appear to be mediated by dpp or wg. dpp is expressed in the pupal abdomen in the pleural portion of the ventral histoblast nest and in a few cells at the dorsal midline in the dorsal nests (Fig. 7A,B). wg is expressed in the sternite primordia and in the medial tergite, but is excluded from cells at the dorsal midline (Shirras and Couso, 1996, and Fig. 7C,D). Neither gene is expressed in the lateral tergite (Fig. 7A-D). The expression of both genes is limited to the posterior region of the anterior compartment (Shirras and Couso, 1996, and Fig. 7) and both are activated anteriorly in hh^Mir (not shown), suggesting that their expression patterns are controlled by hh. However, neither gene is responsible for anterior-posterior patterning. wg functions to promote bristle formation and to dif-ferentiate sternite and tergite primordia from the pleura (Shirras and Couso 1996). The function of dpp is less clear, but we find that large FLP-out dpp-expressing clones develop normally in most of the tergite (not shown).

Two gain-of-function alleles of en, en^Erased (en^Es) (Lindsley et al., 1972) and en^{Apigmented abdomen} (en^Apa) (E. B. Lewis, personal communication) cause ectopic expression of en in the pupal abdomen. Both alleles are associated with rearrangements broken at the en locus (48A). In heterozygotes for these mutants, en is expressed broadly in the anterior compartment. However, expression is always reduced in a 3-to 4-cell-wide stripe of cells anterior to the compartment boundary, and is excluded from apparently randomly distributed patches more anteriorly (Fig. 8C). en expression in the ventral epidermis appears normal. Both en^Es and en^Apa cause weak ectopic acti-vation of hh in the anterior compartment (not shown), indicat-ing that hh is positively regulated by en in the abdomen, as it is in imaginal discs and embryonic epidermis (Guillén et al., 1995; Tabata et al., 1992; Zecca et al, 1995). The en-lacZ enhancer trap element is not activated in the abdomen by en^Es or en^Apa.

The cuticular phenotypes of en^Es and en^Apa are dramatic; with the exception of interspersed islands of apparently normal tergite cuticle, each abdominal segment consists of a broad central region of PHZ cuticle flanked to the anterior and posterior by ISM. Each segment is mirror symmetric, with the PHZ trichomes in the anterior half of the segment directed anteriorly, and the trichomes in the posterior half directed posteriorly (Fig. 8A). However, polarity is always normal within the islands of tergite cuticle; in the anterior of the segment, this leads to a sharp discontinuity in polarity where PHZ and tergite meet. The patches of tergite tissue in the anterior of the segment are lightly pigmented and do not carry macrochaetes (Fig. 8A), indicating that the level of hh produced in the anterior compartment is not sufficient to promote posterior cell fates. en^Es/en^Apa heterozygotes have a similar phenotype, with the exception that no tergite islands are present (Fig. 8B).

To determine whether the tergite islands present in en^Es and en^Apa correspond to the patches in the anterior compartment that do not express en, we stained en^Apa heterozygotes at about 55 hours APF, when cuticular differentiation is well under way, but en can still be detected. We found that en-expressing cells secrete PHZ or ISM cuticle, whereas cells that do not express en produce tergite structures (Fig. 8D). en is also autonomous with respect to polarity reversal; in the anterior of the tergite, trichomes secreted by en-expressing cells are oriented to the anterior, whereas trichomes secreted by adjacent non-express-ing cells are oriented to the posterior.

We have used the FLP-out system (Basler and Struhl, 1994; Zecca et al., 1995) as an additional way to examine the effects of ectopic en expression. The cuticular marker present within the FLP-out construct used (forked) cannot be scored in en-expressing clones in the abdomen, since these produce PHZ and ISM cuticle, and usually lack bristles. However, because the effects of ectopic en expression are autonomous in en^Apa, we assume that the limits of en-expressing clones correspond to the edges of the PHZ or ISM patches seen.

The effects of en-expressing clones depend upon their position within the segment. Clones in the posterior tergite were transformed to PHZ, whereas more anterior clones contained both PHZ and ISM (Fig. 5E-G). Larger clones also caused wild-type cells located to the posterior to produce posterior tergite structures (Fig. 5E). This non-autonomous effect is probably due to activation of hh within the en-express-ing clones. As with hh-expressing clones, ectopic ISM cuticle was always induced anteriorly to the ectopic PHZ, and was never induced posterior to the second (from the anterior) row of microchaetes; tergite structures located laterally and anteri- orly to the clone were always unaffected (Fig. 5E).

Like en^Es and en^Apa, en-expressing clones reveal underlying mirror symmetry to the polarization of abdominal segments. Clones in the anterior tergite produced PHZ with reversed trichome orientation, whereas clones in the posterior tergite produced PHZ of normal polarity (Fig. 5E,G). en-expressing clones that straddled both regions showed an internal line of polarity reversal within the PHZ region (Fig. 5F). By plotting where this line intersects normal tergite tissue for a number of such clones, we were able to define a line of underlying polarity reversal within the tergite. Although not rigidly placed with respect to tergite structures, this line lies within the posterior pigment band in a narrow zone centered on the most posterior row of microchaetes (Fig. 9).

Four lines of evidence are presented that hh functions to pattern the posterior portion of the anterior compartment in the adult abdomen. First, loss of hh function in hh^ts2 pupae causes trans-formation of posterior tergite towards a more anterior fate, as indicated by the loss of the pigment band and transformation of macrochaetes to microchaetes. Second, ubiquitous expression of hh causes reciprocal transformations of anterior to posterior. Third, ‘FLP-out’ clones that express hh in the anterior compartment can induce surrounding wild-type cells to produce posterior tergite characters. Fourth, expression of an ectopic stripe of hh at the anterior edge of the segment in the gain-of-function allele hh^Mir causes double-posterior patterning in the tergite. Most of the anterior tergite develops normally in hh^ts2 homozygotes, indicating that a large part of the abdominal segment is patterned independently of hh.

The nonautonomy of hh, and the graded transformations caused by ectopic expression, suggest that hh may function as a morphogen in the abdomen. The properties that we have defined for hh match closely those of a morphogen postulated by Madhavan and Madhavan (1982) to control posterior pat-terning in the Drosophila tergite. They also correspond closely to a morphogenetic activity that controls posterior patterning in abdominal segments of the moth Galleria (Bhaskaran and Röller, 1980).

In imaginal discs, the long-range patterning effects of hh are mediated by dpp and wg (Brook and Cohen, 1996; de Celis et al., 1996; Jiang and Struhl, 1996; Lecuit et al., 1996; Nellen et al., 1996; Penton and Hoffman, 1996; Zecca et al., 1995). However, the expression patterns of wg and dpp, and direct tests of their function (Shirras and Couso, 1996; this report), indicate that these genes are not involved in anterior-posterior patterning in the abdomen. Whether hh acts directly, or indi-rectly by activation of a secondary morphogen, is not resolved by our work. In the embryonic epidermis, wg and hh mutually activate one another at the compartment boundary (Heemskerk et al., 1991; Ingham, 1993). That a similar reinforcement may take place in the pupal abdomen is suggested by our finding that en and hh expression become graded after fusion of the anterior and posterior histoblast nests, with peak expression occurring at the compartment boundary. It seems unlikely that wg or dpp are involved in this reinforcement, however, since neither is expressed in the lateral tergite region when upregu-lation of en and hh is seen.

The effects of timed ubiquitous expression suggest that hh controls bristle patterning between 5 and 10 hours APF, whereas it controls pigmentation between 18 and 35 hours APF. Bristles appear to be patterned by hh before the anterior and posterior dorsal histoblast nests make contact. This suggests that anterior compartment histoblasts receive hh signal from posterior compartment LEC, which express hh at this time (not shown). Differential control of pigment and bristle patterning is also shown by gain-of-function alleles of optomotor-blind (omb), the main target of hh signaling in the abdomen (Kopp and Duncan, 1997), and is seen after cauterization or transplantation of pieces of abdominal epidermis (Bhaskaran, 1973; Roseland and Schneiderman, 1979). The two main functions of wg in the abdomen, promotion of bristle formation and allocation of the sternite and tergite primordia, also occur at different times (Shirras and Couso, 1996).

In addition to patterning the posterior tergite, hh is required within the posterior compartment. Both PHZ and ISM are often lost in hh^ts2 animals raised at the restrictive temperature during pupal development, and ectopic expression of hh in the anterior compartment causes transformation of anterior tergite to PHZ and ISM. Defining the role of hh in posterior compartment patterning is complicated by our demonstration that hh and en mutually activate one another in the pupal abdomen. This raises the possibility that the effects of hh on posterior com-partment patterning result from alterations in expression of en. However, in hh^Mir, en-lacZ is activated only in the region of ectopic PHZ, and not in the region of ectopic ISM. This suggests that hh can pattern some posterior compartment struc-tures in the abdomen independently of en. hh is also required for normal patterning within the posterior compartment of the dorsal larval epidermis (Heemskerk and DiNardo, 1994).

Segment boundaries have long been postulated to have special insulating properties that allow segments to develop as independent morphogenetic fields (Blennerhassett and Caveney 1984; Campbell and Caveney 1989; Warner and Lawrence 1982). The properties of the LEC at the posterior edge of each abdominal segment suggest they may serve such an insulating role. These ‘border cells’ are morphologically distinct, and are the last LEC to disappear as the histoblast nests expand and fuse. Significantly, although these cells express en, they do not express hh. These properties suggest the border cells may serve as a ‘morphogen seal’, preventing histoblasts of neighboring segments from inter-acting until most patterning has been accomplished.

Consistent with such a role, histoblasts at the anterior edge of the segment appear to be insulated from hh protein produced by posterior compartment histoblasts of the preceding segment. Judging from the effects of timed ectopic expression of hh (this report), omb (Kopp and Duncan, 1997) and wg (Shirras and Couso, 1996), most cell fates and polarity in abdominal segments are already determined by 40-42 hours APF, when the larval border cells are replaced by histoblasts. At this time, ptc is strongly upregulated in histoblasts anterior to the hh stripe, but no ptc expression is seen in histoblasts located posteriorly across the segment border (Fig. 3F). hh also induces the expression of omb anterior, but not posterior, to the hh stripe (Kopp and Duncan, 1997). Ectopic expression of hh in hh^Mir and hs-hh (not shown) induces high levels of ptc and omb in anterior histoblasts, indicating these cells are capable of responding to hh. We suggest they are normally shielded from exposure to hh protein by the larval border cells. Cauterization of the segment boundary often causes mirror-symmetric double posterior hemitergites to develop posterior to the operation (Madhavan and Madhavan, 1982; Roseland and Schneiderman, 1979). Our observations suggest this may occur because damage to the border cells allows hh to influence pattern across the segment boundary.

By analogy with imaginal discs, one might expect abdominal segments to be patterned by a long-range morphogen induced at the compartment boundary by hh (Lawrence and Struhl, 1996). This model makes two clear predictions: first, pattern-ing should be lost when hh function is either removed or provided ubiquitously; second, localized ectopic expression of hh should produce structures organized symmetrically around the source of the hh signal. We find these predictions are not fulfilled, indicating that this ‘compartment boundary’ model is not sufficient to explain pattern formation in the adult abdomen. We present evidence that cuticular pattern in abdominal segments is established through interaction of two distinct patterning systems: a hh-dependent system, responsible for patterning near the compartment boundary, and a more global mirror-symmetric system that is independent of hh.

The underlying symmetry of the abdominal segments is revealed when hh is either inactivated, resulting in a mirror-image double anterior pattern (Fig. 6A), or expressed ubiquitously, creating a reciprocal, double-posterior pattern (Fig. 4C). Additional evidence for a hh-independent mirror-symmetric patterning system is provided by the mirror-symmetric patterning of abdominal segments caused by ectopic expression of en. By examining a number of FLP-out en-expressing clones in the anterior compartment, we were able to determine that a cryptic line of symmetry is present in the posterior tergite; en-express-ing cells anterior to this line always orient toward the anterior, whereas en-expressing cells posterior to this line orient toward the posterior (Figs 5F, 9). Several lines of evidence suggest that this line of symmetry is established independently of hh function. Most convincing, this line appears to be unaffected by hh loss of function, as it coincides with the plane of symmetry in the mirror-image tergites present in hh^ts2 animals raised at the restrictive temperature (Fig. 6A). The line of symmetry also appears to be retained when hh is expressed ectopically, as it coincides with the line of polarity reversal in the double posterior tergites of hh^Mir (Fig. 1B) and hs-hh animals (Fig. 4C), and with the posterior limit of polarity reversals caused by hh-expressing and en-expressing clones (Fig. 5A,E). The line of polarity reversal is also retained in en^Es and en^Apa mutants (Fig. 8A,B), in which ectopic expression of en would be expected to render most or all cells of the anterior compartment insensitive to hh (Guillén et al., 1995; Zecca et al., 1995).

We suggest that hh acts to reverse a ground-state double anterior polarity in the posterior of the segment (see also the model in Fig. 9, Kopp and Duncan, 1997). The phenotypes of hh^ts2 and hs-hh animals suggest that, although the requirement for hh in specifying posterior structures is graded, polarity is reversed in an all- or-nothing fashion by a threshold level of hh.

Interaction between hh and the inferred underlying symmetric patterning system is most clearly seen in the pheno-types induced by hh-expressing clones in the anterior compart-ment. These clones produce the concentric patterns predicted by the ‘compartment boundary’ model only in the central region of the tergite, which we suggest is susceptible to polarization by hh because it is minimally instructed by the hh-independent symmetric polarization (Fig. 5C). Cuticular patterns produced by hh-expressing clones in the anterior of the tergite are dra-matically different: PHZ cuticle has reversed polarity, and always forms posterior to the ISM, and ectopic posterior tergite structures are induced only posterior to the clone (Fig. 5A). This arrangement is a perfect mirror image of the normal pattern seen in the posterior of the segment, suggesting that patterning of posterior structures by hh is superimposed on an underlying mirror-symmetric system. Our results suggest that one effect of this underlying system may be to polarize Hh secretion (see Results). If so, the asymmetric patterning of ectopic ISM and PHZ associated with anterior hh-expressing clones may result from the action of a morphogen induced by hh at the posterior edges of these clones.

The well-characterized limb imaginal discs are derived from only the portion of the segment straddling the compartment boundary, and develop without contact with neighboring segments. This may explain the predominance of the compartment boundary in patterning the imaginal discs. The more global symmetric patterning of the abdominal segments appears to have no equivalent in the imaginal discs and may be directed by the segment boundaries. The observations of Madhavan and Madhavan (1980) suggest that histoblasts may be symmetrically polarized by contact with the larval border cells, and Bhaskaran and Röller (1980) report that cells trans-planted from either the anterior or posterior borders of abdominal segments in the moth Galleria can repolarize centrally located cells.

To summarize, we argue that abdominal segments are patterned by two interacting systems: a global system that specifies mirror-symmetric patterning within the entire segment, and a system dependent on hh that functions only near the compartment boundary. We argue that hh, or a gene product induced by hh, functions as a morphogen that directs anterior compartment cells to produce a range of posterior tergite structures, and may also pattern part of the posterior compartment. hh also determines polarity, apparently by two mechanisms. First, hh has some ability to orient cells toward points of high expression. Second, and perhaps more important, hh expression causes cells to reverse their interpretation of underlying polarity specified by a global mirror-symmetric system.

Coincident with acceptance of this paper, two papers by Struhl et al. (1997a,b) appeared that address the role of hh in patterning the adult abdomen. These authors demonstrate that hh specifies tergite cell fates by functioning directly as a morphogen, but specifies polarity by an indirect mechanism. In addition, these authors show that hh patterns the anterior as well as the posterior edge of the tergite. The anterior region affected appears to be limited to the acrotergite. Consistent with this, we find that the acrotergite is lost in hh^ts2 animals raised at the restrictive tem-perature, while the remainder of the anterior tergite develops normally (Fig. 6A). Struhl et al. (1997a) also show that, in wild type, ptc is upregulated at both the anterior and posterior edges of the tergite. These observations would seem to contradict our proposal that the larval border cells insulate anterior cells from exposure to hh. However, our failure to find activation of ptc at the anterior of the segment up to 40 hours APF indicates that ptc activation here must be delayed until after the disappearance of the border cells, long after ptc is activated at the posterior of the tergite (Fig. 3F). Moreover, by delivering timed pulses of ectopic hh expression, we show that exposure of anterior tergite cells to hh early in pupal development causes them to develop as posterior tergite, PHZ, or ISM, but not acrotergite. These observations indicate that the response of anterior tergite cells to hh signaling is critically dependent on timing, and are consistent with an important role for the border cells in shielding anterior tergite cells from early Hh in pupal development.

We also differ significantly in our interpretation of the role of hh. Struhl et al. (1997a,b) propose that hh is the central determinant of patterning in the tergite, whereas we suggest that the patterning activities of hh in the tergite are superimposed on a global mirror-symmetric pattern. Our view is supported by a number of observations described in this report. Among these are two observations also made by Struhl et al. (1997a). First, both groups find that a large portion of the anterior tergite [regions a2 and a3 of Struhl et al. (1997a)] is patterned independently of hh. We suggest that these regions are patterned by the underlying symmetric system. Second, both groups find that hh-expressing clones in the anterior tergite are internally patterned, with ISM always developing anterior to PHZ. We suggest this patterning is also directed by the mirror-symmetric system.

Finally, our papers differ in the emphasis given to the role of hh in determining polarity. Both papers demonstrate that hh is able to polarize tergite cells. However, both papers also present evidence for a second, hh-independent, polarizing activity. Because of the nature of our data, we stress a hh-independent mechanism, whereas Struhl et al. (1997b) stress the polarization activity of hh itself. Which is more important, and how many polarizing mechanisms exist, remain to be determined.

We are very grateful to Tom Kornberg and Tetsuya Tabata for the en-lacZ, hh-lacZ, ptc-lacZ and hs-hh stocks and for the hh cDNA clone; to Konrad Basler for the FLP-out hh, en and dpp lines and the hs-FLP 122 line; to Ron Blackman and Chris Merli for the dpp cDNA; to Matt Scott for the UAS-ptc line; and to Kathy Matthews at the Bloomington Stock Center and Linda Treeger at the Mid-America Stock Center for various fly stocks. We also thank Dianne Duncan for useful advice, Paula Kiefel for technical help, Richard Emmons and Mark Kankel for support and discussions, and Monica Doran for help in isolating rever-tants of hh. This work was supported by NIH BRSG grant SO7 RR 07031 to M. A. T. M. and NIH grant GM32318 to I. D.