Hox genes differentially regulate Serrate to generate segment-specific structures

In Drosophila, the Hox family of homeotic genes diversifies the morphology of larval and adult segments along the anterior/posterior (A/P) axis (reviewed in McGinnis and Krumlauf, 1992; Lawrence and Morata, 1994). At the molecular level, the Hox genes encode homeodomain-containing transcription factors that regulate target genes in segment-specific patterns (reviewed in Botas, 1993; Morata, 1993; Graba et al., 1997). Many of the known Hox target genes encode either transcription factors (e.g. Distal-less, empty spiracles, forkhead and teashirt), or signaling factors (e.g. decapentaplegic (dpp), wingless (wg) and scabrous) (reviewed in Graba et al., 1997), suggesting that regulation of cell-specific morphological diversification is in part indirect. The structural mediators of Hox morphogenetic functions, called realizator proteins, have been proposed to generate differences in cellular and segmental morphology through their effects on the rate and orientation of mitosis, on cell-cell adhesion and on cell shape (Garcia-Bellido, 1977). A few Hox downstream genes that apparently perform realizator functions have been identified, including β3-tubulin (β-Tub60D), connectin (con) and centrosomin (cnn) (Gould and White, 1992; Hinz et al., 1992; Heuer et al., 1995).

Some intermediary genes in the presumptive pathways from Hox protein to realizator function have been shown to activate discrete subprograms under homeotic control. For example, Ultrabithorax (Ubx) and abdominal-A (abd-A) functions are required to activate the expression of two signaling factors, decapentaplegic (dpp) and wingless (wg), in adjacent groups of visceral mesoderm cells (Immerglück et al., 1990; Panganiban et al., 1990; Reuter et al., 1990). The localized function of both signaling molecules is required to induce the central midgut constriction in Drosophila larvae (Immerglück et al., 1990; Panganiban et al., 1990; Reuter et al., 1990). Dpp and Wg also act in neighboring endodermal cells to activate expression of labial, which is responsible for the differentiation of a specific endodermal cell type, the copper cell (Bienz, 1994; Hoppler and Bienz, 1994). In another region of the gut, pointed and odd-paired are regulated by a combination of abd-A and wg, and are required for generation of the posterior midgut constriction (Bilder et al., 1998). These are examples of regulatory pathways required for development of Hox-dependent cell fates. However, there is still little understanding of the extent of and connections among Hox-regulated pathways that result in morphological diversification.

Through a screen for Hox genetic modifiers (Harding et al., 1995), we identified Serrate (Ser) as a potential component of Hox-dependent pathways. Ser encodes a transmembrane signaling protein which interacts through its extracellular domain with the receptor protein Notch (Fleming et al., 1990; Rebay et al., 1991; Thomas et al., 1991). Models of Ser function depend substantially on studies of its role in wing margin development in Drosophila (reviewed in Fleming et al., 1997b). Ser is expressed throughout the dorsal half of the wing disc and its function is required to induce the wing margin (Couso et al., 1995; Diaz-Benjumea and Cohen, 1995; de Celis et al., 1996). Studies of Ser loss-of-function mutants and Ser ectopic expression experiments suggest that Ser is necessary and sufficient to promote wing outgrowth (Speicher et al., 1994; Jönsson and Knust, 1996; de Celis and Bray, 1997). It does so by activating Notch protein in the adjacent ventral cells at the dorsal/ventral boundary, which results in the localized expression of vestigial, wg and cut at the wing margin (Couso et al., 1995; Diaz-Benjumea and Cohen, 1995; Kim et al., 1995; de Celis et al., 1996; Micchelli et al., 1997). Although Notch is expressed throughout the wing pouch, Ser is prevented from activating Notch in dorsal cells by the localized expression of the fringe (fng) gene (Kim et al., 1995; Fleming et al., 1997a; Panin et al., 1997). While much has been learned about the function of Ser in wing patterning, little is known about Ser function in the developing embryo.

To better understand how cells acquire location-specific identities via Hox-dependent pathways, we have studied the embryonic regulation of Ser and the role of Ser in larval cuticle patterning. We show that both Hox and segment polarity genes are required to localize Ser transcription in stripes in the abdominal epidermis, and that this pattern of Ser expression is required to fully diversify denticle belts. Ser regulation of denticle row identity requires the Ser-dependent activation of rhomboid (rho) expression in a single row of cells. Thus Ser expression is an intermediate step in a Hox-regulated pathway leading to morphological diversification at the cellular level. Ser is also required for mouth hook development and overexpression of Ser produces dorsal pouch sclerotization in the head region. However, Ser does not regulate rho in the mouth hook primordia, suggesting that the outcome of Ser expression is context-dependent.

The following alleles were used: Ser^5A29 (Harding et al., 1995), Ser^rx106 and Ser^rx82 (Thomas et al., 1991), Ser^rev6-1 (Fleming et al., 1990), Dfd¹³, Dfd³, Dfd¹⁶ (Merrill et al., 1987), cnc^VL110 (Mohler et al., 1995), Scr⁴, Ubx¹abd-A^D24Abd-B^D18 (Hopmann et al., 1995), Scr⁻Antp⁻BX-C⁻ (Struhl, 1983), wg^l-17, hh¹¹, ptc⁹, Df(2R)en^SFX31, ve⁶ (rho^del1) (Bier et al., 1990), fng¹³ and fng⁸⁰ (Irvine and Wieschaus, 1994), UAS-Ser (Gu et al., 1995), arm-GAL4 (Sanson et al., 1996), 69B-GAL4 (Brand and Perrimon, 1993), UAS-Ser;Df(3R)Ser^+82f24Tb/TM3SbSer (kindly provided by R. Fleming), hs-Dfd (Kuziora and McGinnis, 1988), hs-stg (Edgar and O’Farrell, 1990) and rho-lacZ(X81) (Freeman et al., 1992).

Each Ser allele tested was crossed onto a chromosome containing the hypomorphic allele Dfd¹³(Harding et al., 1995). Males carrying a single copy of this doubly mutant chromosome were then crossed to females heterozygous for the temperature-sensitive Dfd³. Their progeny were raised at 29°C and adults were scored for genotype. Reduced viability was determined by dividing the number of Dfd¹³Ser/Dfd³ adults by the number of siblings (Dfd¹³Ser/balancer and Dfd³/balancer). Controls were performed with a Dfd¹³Ser⁺ chromosome. The interaction strength of Ser with Dfd is defined as the viability of the Ser allele in the Dfd mutant background divided by the viability of the control chromosome in the same background. Thus an interaction strength near zero is very strong, while an interaction strength of 100% indicates no interaction.

Mutant chromosomes were outcrossed to eliminate balancer chromosomes from the stocks before mutant cuticles were collected. Embryos were collected for approximately 12 hours and aged for more than 24 hours before preparing cuticles by standard techniques. Cuticle preparations of overexpressed Ser in a Ser null background were from cages of UAS-Ser;Df(3R)SerTb/TM3SbSer virgin females and Ser^rx106arm-GAL4/TM3SbUbx-lacZ males. In this case, hatched larvae were collected and, after washing in water, treated with 1:4 glycerol:acetic acid and mounted as above.

Embryos for SEM analysis were collected and aged as for cuticle preparations, dechorionated in bleach, washed into scintillation vials and devitellinized in methanol. Embryos were then rinsed in 50:50 methanol:isoamyl-acetate and left in 100% isoamyl-acetate overnight for complete dehydration. The embryos were transferred in isoamyl-acetate to double-sided carbon tape on a specimen mount. After the isoamyl-acetate had completely evaporated, the larvae were coated with gold and photographs were taken by standard methods.

Embryos were fixed and hybridized with DIG-labeled RNA probes using a variation of the protocol described in Tautz and Pfeifle (1989). The Ser probe (Fleming et al., 1990) and stg probe (Edgar and O’Farrell, 1989) were generated from full-length cDNA clones. The reaper probe was made from a clone (White et al., 1994) from which a 350 bp PstI fragment was deleted at the 3′ end. Embryonic stages were defined as in Campos-Ortega and Hartenstein (1997).

Embryos were fixed and hybridized with DIG-labeled probes following the in situ hybridization protocol, with the exception that a 5- to 10-fold excess of probe was used. Following post-hybridization washes, the embryos were incubated with antibodies (biotinylated secondary) by standard methods with the exception that a 5-fold excess of primary antibody was used. Horseradish peroxidase-biotin-conjugated avidin (standard dilution, Vectastain) and alkaline phosphatase-anti-DIG (1:1000) were bound in a combined incubation.

Incorporation and detection of BrdU in embryos was performed as described in Edgar and O’Farrell (1990). Staining with acridine orange was performed as described in Abrams et al. (1993).

We isolated a single allele of the Serrate gene (Ser^5A29) in a screen for mutations that enhanced hypomorphic phenotypes of mutations in the Hox gene Dfd (Harding et al., 1995). The Ser^5A29 allele reduced the survival of Dfd hypomorphs to 40% of normal levels. To test whether this interaction with Dfd was allele-specific, we tested null mutants of Ser. The interaction strength of two such alleles (Ser^rx106, Thomas et al., 1991 and Ser^rev6-1, Fleming et al., 1990) with Dfd was 20-30% of the control chromosomes. Therefore, we concluded that the function of Dfd is sensitive to the dose of wild-type Ser activity.

To explore the connection between Ser and Dfd functions, we reexamined the embryonic/larval phenotypes of Ser mutants. Speicher et al. (1994) previously described the Ser larval phenotype as blunt mouth hooks, anterior spiracle malformations and small imaginal discs in larvae of indeterminate age. We find that some animals lacking Ser die as embryos without emerging from the chorion, while others survive for variable periods as larvae. In both classes, the mouth hooks are blunt, lacking the sclerotic material that forms a curved hook in the normal structure (Fig. 1A,B). Since the mouth hook is a Dfd-dependent structure that is formed principally from maxillary cells (Jürgens et al., 1986; Merrill et al., 1987; Regulski et al., 1987), the requirement for Ser in normal mouth hook development is consistent with the genetic interaction between Ser and Dfd.

Additionally, we have discovered that Ser mutant larvae have abnormally patterned abdominal denticle belts. Within each wild-type denticle belt from A2 to A8, the row 3 and 4 denticles have similar sizes and shapes but the hooks of row 3 point to the posterior, and those of row 4 point to the anterior (Fig. 1C). In Ser mutants, these two rows of denticles are combined into one row, the 3/4 row; Ser^5A29 and Ser^rev6-1 homozygous and transheterozygous mutant combinations show row fusion in almost all larvae, while Ser^rx106 homozygotes show complete fusion in about half of the larvae. The denticles in the 3/4 row have no regular polarity and about half of the embryos develop small denticles in this row (Fig. 1D). To determine whether the 3/4 row of denticles in Ser mutant larvae is a fusion of the two rows or a loss of one, we counted denticles, using the fourth abdominal segment (A4) as an example. In wild-type denticle belts, row 3 has 20(±2) denticles, while row 4 has 19(±2) denticles. In Ser mutant animals, there are 31(±5) (Ser^5A29) to 33(±4) (Ser^rx106) denticles between rows 2 and 5, a result that is inconsistent with complete loss of either row 3 or row 4.

Previous experiments have failed to detect any gross morphological defects in the cuticular features of embryos when Ser is ubiquitously expressed throughout Drosophila embryos (Speicher et al., 1994). Using a Ser cDNA under Gal4-UAS control (Gu et al., 1995), we also assayed the phenotypic effects of ectopic Ser, focusing on denticle morphology. Ubiquitous expression of Ser driven by arm-GAL4 or 69B-GAL4 driver constructs in a Ser⁺ genetic background induced no detectable changes in the shape, size or pattern of denticles (data not shown). Since Notch receptor activation can be dependent on relative levels of ligand (Muskavitch, 1994; Simpson, 1997), we theorized that embryos with ubiquitous Ser expression superimposed on the native Ser expression might still have differential levels of expression sufficient to activate Notch. Consistent with this, when Ser is ubiquitously expressed in a Ser heterozygous or null background, significant morphological defects are observed. The mouth hooks develop additional sclerotic material in the middle and base of the structure (Fig. 2C, compare with Fig. 2A,B), and about a quarter of the embryos show duplication of mouth hook tips (Fig. 2C, insets). Excessive sclerotic material also develops in the dorsal pouch. The expressivity of this phenotype varies from an extended and fragmented dorsal bridge to extreme sclerotization of the dorsal pouch and shortening of the lateralgräten (Fig. 2C). About half of the denticle belts in embryos of this genotype display the wild-type pattern but only rarely do embryos develop extra denticles around row 3 (data not shown). An additional result of ectopic Ser expression in a reduced-dosage background is excessive sclerotization of the proventriculus and gut (Fig. 2F, compare with Fig. 2D,E). Both the dorsal pouch and the proventriculus are responsible for secretion of specialized cuticle to make head sclerites and gut lining, respectively (Jürgens et al., 1986; Skaer, 1993).

Ser expression begins during embryonic stage 11, and eventually is transcribed in regions of the epidermis, tracheal trunks, foregut and hindgut, central nervous system and salivary ducts (Fleming et al., 1990; Thomas et al., 1991). Because of the phenotypic effect on mouth hooks, we examined expression of Ser in the embryonic head. The transcript pattern in the head region at stage 12 includes the mandibular segment, the anterior and posterior lateral borders of the maxillary segment, and the anterior lateral and posterior ventral borders of the labial segment (Fig. 3A). A subset of cells in the clypeolabrum and dorsal head also accumulate Ser transcripts.

We tested the dependence of Ser expression on Dfd function and vice versa and found an epistatic relationship consistent with the genetic interaction. While there is no detectable change of Dfd expression in Ser mutants (data not shown), Dfd and other head homeotic genes are required for the normal pattern of Ser transcription in the head. In Dfd mutant embryos, Ser transcripts are not expressed in cells in the anterior maxillary segment, which will eventually secrete part of the mouth hook (Turner and Mahowald, 1979; Jürgens et al., 1986). The posterior maxillary pattern is unchanged (Fig. 3B). When ectopic expression of Dfd protein is generated by heat shock (hs-Dfd) (Fig. 3D), ectopic mouth hooks form in the labial and thoracic segments (Kuziora and McGinnis, 1988). We find that these segments also express ectopic Ser on the lateral anterior and posterior borders (Fig. 3C). Thus, regulation of Ser by Dfd correlates with the segments where both normal and ectopic mouth hooks develop, and we conclude that Ser is one of the Dfd target genes that mediate mouth hook development. In Sex combs reduced mutant embryos, Ser transcripts are not expressed in the posterior maxillary segment or the anterior labial segment (Fig. 3E). cap’n’collar mutant embryos, which develop ectopic mouth hooks from the mandibular segment (Mohler et al., 1995), express Ser in the mandibular segment in a pattern similar to that of the maxillary segment (Fig. 3F).

Ser transcripts in the trunk are first detected at the extended germband stage in ventral patches in the middle of abdominal segments A2-A8 and in offset lateral patches (Fig. 4A; Fleming et al., 1990; Thomas et al., 1991). The ventral regions of thoracic segments do not exhibit Ser expression at this stage. As the germband retracts, the abdominal stripes intensify and develop sharp anterior borders (Fig. 4B,C). The first abdominal segment (A1) is unique: Ser expression begins later than in the other abdominal segments and forms a narrower stripe after germband retraction (Fig. 4A-C). After germband retraction, Ser transcripts can also be detected in the ventral regions of thoracic segments in broad, faint patches (Fig. 4B and Fleming et al., 1990; Thomas et al., 1991).

Embryos mutant in all genes of the Bithorax Complex (BX-C), Ubx, abd-A and Abdominal-B (Abd-B), develop thoracic-type denticles throughout the trunk region. Consistent with this transformation, stage 11 and 12 BX-C mutant embryos have no Ser expression in ventral regions (Fig. 4D). Ventral Ser expression does begin in BX-C mutants after germband retraction, but the location and level of expression matches that of the thoracic segments (Fig. 4E). As expected, Ubx mutant embryos show a transformation of abdominal- to thoracic-type Ser expression only in A1; abd-A mutants show Ser transcript stripes in A2-A8 that are similar to the wild-type A1 pattern, and Abd-B mutants display no change in A1-A8 ventral Ser transcription (data not shown). Thus Ubx function is sufficient to activate some Ser expression in the center of each segment, but abd-A function is required for the earlier, broader pattern of Ser transcription in A2-A8, a transcript pattern that correlates with complete diversification of denticle belts. Embryos lacking all trunk Hox functions (Scr⁻Antp⁻BX-C⁻; Struhl, 1983; Macias and Morata, 1996) express Ser at the margins of the anterior part of each trunk segment and at lower levels in the center of this region (Fig. 4F), a pattern almost the inverse of that seen in wild type. Transcription of Ser in the posterior-most region of each segment, probably corresponding to the posterior compartment, is completely suppressed (arrows, inset, Fig. 4F).

The delimitation of Ser expression to reiterated subsegmental stripes in the embryonic metameres suggested that segment polarity genes also regulate the Ser transcript pattern. ptc mutants lack ventral abdominal Ser transcripts (Fig. 5B, compare with Fig. 5A), correlating with the loss of denticle diversity and number in ptc denticle belts (Bejsovec and Wieschaus, 1993). Ser transcription in wingless (wg) mutants appears in broad stripes (Fig. 5C), while hedgehog (hh) and engrailed (en) mutant embryos exhibit Ser transcription throughout almost the entire ventral epidermis of the abdominal segments (Fig. 5D and data not shown). Broadened patterns of Ser transcription in these segment polarity mutants correspond to expanded fields of denticles that lack significant diversity of denticle type (Bejsovec and Wieschaus, 1993; reviewed in Peifer and Bejsovec, 1992).

The segmental boundary is persistently identified by en expression in the posterior compartment of each segment (reviewed in: DiNardo et al., 1994), and, after germband retraction, by rho in the two most anterior cell rows of A2-A8 (O’Keefe et al., 1997; Szüts et al., 1997). By stage 14, Ser transcripts are expressed in two to three rows of cells directly posterior to rho-lacZ expression (Fig. 6A,B). The final position of Ser relative to En and rho-lacZ suggests that Ser is expressed in the cells that will produce denticle row 4 and those to the posterior. The anterior boundary of Ser expression appears to lie between the two rows of denticles that Ser affects phenotypically (Fig. 6C), suggesting that the function of Ser is limited to cells near its anterior boundary of expression. We find no changes of denticle pattern in fringe mutant embryos, indicating that fringe is not required to restrict Ser morphological function in the abdominal segments (E. L. W. and W. McG., unpublished data).

Loss of a row of denticles in Ser mutant larvae could be the result of insufficient embryonic cell division or of excessive cell death. Ser is required for outgrowth of the wing (Speicher et al., 1994; Jönsson and Knust, 1996; de Celis and Bray, 1997), and a similar role in proliferation and growth might be conserved in the embryonic epidermis. To determine whether the fused row of denticles is associated with a failure of cell proliferation in Ser mutants, two methods of identifying dividing cells in the embryo were used. First, we analyzed the pattern of incorporation of the nucleoside analog bromodeoxy-uridine (BrdU) during S phase of cell division and, second, we examined the transcription pattern of string (stg). Wild-type and Ser mutant embryos display comparable BrdU incorporation during the final round of ventral epidermal cell division (Fig. 7), which occurs just after the beginning of Ser transcription (Fig. 7; Campos-Ortega and Hartenstein, 1997). If Ser were required to drive division of some cells in this region, cells across the middle of the mitotic domain would lack incorporation of BrdU in the mutant epidermis. stg is expressed at the transition from G₂ to M phases of cell division, and therefore serves as a marker for cells entering mitosis (Edgar and O’Farrell, 1990). The pattern of stg expression in Ser mutants is indistinguishable from that seen in wild type, and exogenously driven stg expression does not suppress the Ser mutant phenotype, indicating that Ser does not regulate mitosis upstream of stg (data not shown). Cell death also appears to be unaffected by loss of Ser function. Comparison of wild-type and Ser embryos stained with acridine orange (Abrams et al., 1993) or reaper antisense probe (White et al., 1994), both of which mark dying cells, suggests that cell death is unaffected by loss of Ser function (data not shown). Thus, in the embryonic epidermis, Ser function is not required for normal cell division, nor does it inhibit normal apoptosis.

To understand how Ser affects denticle belt patterning at the single cell level, we investigated interactions of Ser with other genes that affect denticle diversity. In particular, the expression pattern of rho at the anterior of the Ser expression pattern made it a candidate for genetic interaction with Ser. The spitz-group gene rho is required for normal development of abdominal denticle rows 1 and 4 (Mayer and Nüsslein-Volhard, 1988). rho transcription first appears in a segmental pattern during stage 12, when it is activated in the anterior cells of each segment (Bier et al., 1990). Thoracic segments express rho in one row of cells at the anterior border of each segment, while ventral regions of segments A2-A8 express rho in two rows of cells, the posterior of which is dependent on Ubx/abd-A function (Szüts et al., 1997). A1 develops rho expression in an intermediate pattern one to two cells wide (Szüts et al., 1997). In a Ser null background, the abdomen-specific row of rho expression is missing and rho is transformed to the thoracic pattern in all abdominal segments (Fig. 8A,B). Stage 13 and early stage 14 Ser mutants show a row of unlabeled cells between Ser expression and the single row of rho expression (Fig. 8B), demonstrating that it is the posterior row of rho expression that is dependent on Ser. Lower levels of rho expression in Ser mutants (compare to underlying ventral midline staining, Fig. 8A,B) suggest that anterior rho is partially dependent on Ser function, perhaps through the intervening posterior rho row. By late stage 14 in Ser^5A29 mutant embryos, Ser transcripts have expanded anteriorly and are juxtaposed to the single row of rho (data not shown). Consistent with this, Ser transcription is detected in an additional anterior row of cells in rho mutant embryos (Fig. 8C,D). Thus Ser is required for activation of rho in the posterior row of cells in the abdominal epidermis. Subsequently, rho is required to repress Ser within these same cells.

Denticle phenotypes show similarities between Ser and rho consistent with the observed regulatory interactions. A wild-type embryo produces six rows of denticles with each row identifiable by polarity and/or size (Fig. 8E). As shown in Fig. 1, Ser mutants fail to separate rows 3 and 4, leaving five rows of denticles overall (Fig. 8F). In rho mutants, individual denticle rows are not as well defined, but it is possible to identify row 5 denticles in the middle of mutant denticle belts (Fig. 8G). Anterior to row 5 is a row that consists of small, stubby denticles (Fig. 8G) (called row 3 type in Mayer and Nüsslein-Volhard, 1988). The most anterior row of denticles in the rho mutants contains sparse row 2 denticles (Fig. 8G). In rho,Ser double mutants, all of the denticles are similar to each other, and resemble a composite of type 5 and 2 denticles (Fig. 8H). The denticles are also disorganized, so that separate rows are not distinguishable (Fig. 8H). The difference between rho and rho,Ser double mutant phenotypes shows that the two anterior denticle rows in a rho mutant are still dependent on Ser for their diverse structures. This suggests that Ser function partially determines anterior denticle identity both within and anterior to its expression domain independently of rho function. Additionally, the differences between Ser and rho,Ser double mutants show that a single row of rho expression is only partially sufficient to generate denticle diversity to its posterior. Thus the diversity of denticle rows 3 and 4 is dependent both on Ser regulation of rho and on independent Ser functions in denticle rows 3 and 4. Transcription of rho in maxillary segment cells is unchanged in Ser mutants, a result that correlates with wild-type mouth hooks in rho mutants and with our observation that rho,Ser and Ser mutants have identical mouth hook defects (data not shown).

Our results indicate that the abd-A-dependent activation of Ser in a subset of trunk segments produces more structural complexity in the abdominal segments than in the thoracic segments. As outlined in Fig. 9, early stripes of Ser expression specific to A2-A8 are dependent on abd-A function. A principal role of Ser expression in the abdominal segments is to activate rho transcription anterior to the Ser pattern, thereby generating identity for a single row of denticle-producing cells. Ser-dependent rho expression thus leads to denticle row diversity in the abdomen, presumably by rho-dependent activation of EGFR. Ser is also regulated in a Hox-dependent manner in the gnathal segments, including activation of Ser transcription by Dfd in the maxillary segment. While mouth hook development is dependent on Ser, it is independent of rho function; it is likely that regulation of different genes by Ser in the maxillary segment leads to mouth hook development. In both maxillary and abdominal segments, the localized expression of Ser contributes to diversification of Hox-dependent morphological features.

The selector genes of the trunk, Sex combs reduced, Antp, Ubx, abd-A and the segment polarity genes are responsible for establishing segment-specific transcription patterns of Ser in the ventral thorax and abdomen. During stages 11 and 12, only the abdominal segments show ventral stripes of Ser transcription. The significance of the difference between thoracic and abdominal Ser transcription is underscored by the transformation of Ser expression from an abdominal- to a thoracic-type pattern in embryos mutant for Ubx and abd-A. The difference in timing of expression in abdominal and thoracic segments is probably important to the function of Ser in patterning abdominal denticle belts, since the segmental pattern of rho, which partially mediates the effect of Ser on denticle belts, is activated during stage 12 (Bier et al., 1990). At this stage, Ser is transcribed in the abdominal segments, but is not yet transcribed in thoracic segments.

Localization of Ser expression within segments is critical, since the placement of the anterior border of Ser is required for the positioning of the abdomen-specific rho stripe. This localization is achieved through the functions of the segment polarity genes. Ser expression is absent in ptc mutants, and this could be due to lack of activation by ptc or to repression by ectopic en/hh and wg (Bejsovec and Wieschaus, 1993). Conversely, en, hh and wg mutants all show expanded Ser expression, but the maintenance of ptc function and expression is dependent on all of these genes (Hidalgo and Ingham, 1990). Since Ser is repressed at the anterior of each segment at a time when Wg protein is no longer detected in or near these cells (Gonzalez et al., 1991), we believe that wg signal is indirectly involved in anterior Ser repression. A model consistent with these and other findings is that wg represses Ser expression by maintaining en/hh expression, En/Hh represses Ser expression by repressing Ptc function, and Ptc activates Ser expression by inhibiting Smoothened function (Alcedo et al., 1996; Chen and Struhl, 1996). Indeed, Szüts et al. (1997) proposed positioning of segmental rho by Hh signal as the simplest way to explain their observations about the dependence of EGFR activity on segment polarity genes; this regulatory effect of Hh appears to depend on positioning of the Ser stripe.

The location of Ser expression in Hox and segment polarity mutants correlates with diverse denticle row phenotypes, and with an expansion in the breadth of the denticle belt. In segment polarity mutants that have excessive denticle development, such as wg, en and hh, the Ser transcription pattern is comparably expanded. Conversely, in mutant embryos that show fewer, underdeveloped denticles, such as Ubx,abd-A or ptc mutants, Ser expression is lacking. These data, combined with the embryonic Ser mutant phenotype, suggest that Ser is an integral part of the Hox- and segment polarity-regulated denticle patterning pathway, required for the segregation and identity of denticle rows 3 and 4.

In Scr⁻Antp⁻BX-C⁻ embryos, Ser expression extends into regions where it is normally repressed by hh and wg. This loss of repression is probably not due to changes in segment polarity function, since segment polarity genes are thought to function independently of Hox genes (see Castelli-Gair, 1998, and references therein). Segment polarity genes are therefore not sufficient, although required, for Ser repression within the anterior compartment. Thus coordinated regulation by Hox and segment polarity genes is required for proper Ser transcription in the trunk. Hox proteins can function as interpreters of incoming signals (Grieder et al., 1997; Maloof and Kenyon, 1998), so it is possible that Hox genes integrate segment polarity signals with their own activities in the regulation of Ser transcription.

Expression of Ser in the head region is also regulated by Hox genes, but apparently by different regulatory mechanisms. While Ser is expressed in the center of the abdominal segments, it is expressed at the borders of the maxillary and labial segments. Expression of Ser in head regions of ptc mutants is comparable to wild type (data not shown), demonstrating that ptc is not required for Ser expression in the head. Ectopic expression of Dfd activates ectopic Ser only in limited regions of each segment. Thus it appears that different factors restrict Hox-dependent Ser transcription in the head.

Expression of Ser in the embryonic epidermis results in context-dependent responses, including rho expression, denticle belt patterning and normal development of the mouth hooks. These embryonic roles of Ser are apparently different from its roles in wing margin determination and wing outgrowth. One similarity is the short range over which Ser function is exerted, either at the anterior border of its ventral A2-A8 expression pattern, or at the dorsal/ventral margin of its expression boundary in the wing pouch.

The spitz-group gene rho can potentiate EGFR activation via the spitz (spi) ligand (reviewed in: Wasserman and Freeman, 1997; Bier, 1998). EGFR activation is required from late stage 11 to early stage 13 for patterning of the denticle belts (Clifford and Schüpbach, 1992), and rho, unlike spi and Egfr, has a spatially and temporally regulated expression pattern (Bier et al., 1990). Abdomen-specific rho expression is required for patterning of abdominal denticle rows 1 through 4, probably by allowing secretion of Spitz protein from denticle row 2 and 3 cells, which activates EGFR in neighboring cells (Szüts et al., 1997; O’Keefe et al., 1997). We have shown that Ser function is required for activation of the abdomen-specific posterior row of rho transcription, expression of which is also dependent on Ubx/abd-A (Szüts et al., 1997). The evidence presented in this paper suggests that Ser provides a critical intermediate which translates broad Hox and segment polarity domains into narrow stripes of rho expression, which specify diversification at the single cell level. Ser and rho mutants each show only a single row of denticles between rows 2 and 5 of A2-A8, indicating that Ser and rho are both required for normal development of rows 3 and 4. rho,Ser double mutants develop row 5-like denticle identities throughout the denticle belt. Thus, either gene alone provides some A/P denticle diversity, while the double mutant lacks any diversity. If the only role of Ser were regulation of rho in denticle row 3 cells, then rho,Ser mutants should develop the same phenotype as rho mutants. Since this is not observed, we conclude that Ser has identity functions independent of rho regulation. Ser function is required in the cells immediately to its anterior expression boundary and within the most anterior row of Ser-expressing cells; the effect within its own domain of expression may be a result of signaling from cells within the same row, or from those to the posterior.

A model for the row identity function of Ser in A2-A8 is proposed in Fig. 9. Ser function, presumably signaling through Notch, is required to determine the identity of the denticle row 3 cells, including activation of rho expression. Segmental rho, in turn, is one of the gene products required for localized EGFR activation in denticle row 4 cells (Szüts et al., 1997). Thus denticle row 3 identity is dependent on Ser signaling and Rho function, while denticle row 4 identity depends on Ser function and on feedback from rho-expressing cells. This series of events suggests that abdominal Hox functions direct cellular diversification through the establishment of signaling centers. The limitation of Ser function to its anterior border generates a novel boundary within each segment. Activation of a single row of rho expression at this boundary then creates an additional signaling center in A2-A8, which controls the greater morphological diversity in A2-A8 epidermis. This use of a Notch ligand to produce a single cell signaling stripe may be a common theme: in dorsal/ventral wing margin patterning, Notch is activated in a single row of cells on either side of the margin in response to Delta and Ser boundaries, and Notch activation leads to a narrow, Wg-expressing, signaling center (de Celis et al., 1996; Neumann and Cohen, 1997).

Regulation of rho by the Notch pathway has been demonstrated in Drosophila wing vein patterning (de Celis et al., 1997). However, in the larval and pupal wing primordia, Delta-activated Notch signaling results in repression of rho expression outside the forming vein (de Celis et al., 1997). The difference in responsiveness of cells in the wing and the embryonic epidermis could be due to the identity of the signal (Ser versus Delta protein), or to additional identity factors present in the cell. rho, in turn, is required to maintain Delta expression in the provein region (de Celis et al., 1997). Regulation of Notch function by EGFR activity has also been described by Levitan and Greenwald (1998), who have shown that translation of a C. elegans Notch homologue is downregulated in response to EGF signaling. Similar regulation may result from rho expression in the Drosophila embryonic epidermis, generating a feedback loop between Ser and rho as has been observed for Dl and rho in the forming wing.

We thank R. Fleming, B. Edgar, A. Bejsovec, K. Irvine, E. Knust, G. Morata, E. Bier and the Bloomington stock center for providing fly stocks; R. Fleming, B. Edgar and H. Stellar for plasmids. Anti-En antibody developed by C. Goodman was obtained from the Developmental Studies Hybridoma Bank maintained at Johns Hopkins University and the University of Iowa under contract number NO1-HD-2-3144 from the NICHD. We also thank L. Washington and C. Warble for help with electron microscopy and B. Edgar and D. Lehmann for advice on BrdU incorporation. We thank R. Fleming, A. Bejsovec, N. McGinnis, B. Florence, X. Li, A. Veraksa and K. Mace for comments on the manuscript. This work was supported by grant HD28315 from the NIH.