Drosophila. Cubitus interruptus-independent transduction of the Hedgehog signal in

Transcription

1 Development 127, (2000) Printed in Great Britain The Company of Biologists Limited 2000 DEV Cubitus interruptus-independent transduction of the Hedgehog signal in Drosophila Armel Gallet 1, Corinne Angelats 2, Stephen Kerridge 2 and Pascal P. Thérond 1, * 1 Institute of Signaling, Developmental Biology and Cancer Research, CNRS UMR 6543, Centre de Biochimie, Parc Valrose, NICE Cedex 2 France 2 Laboratoire de Génétique et de Physiologie du Développement, UMR 9943 CNRS-Université, IBDM CNRS-INSERM-Université de la Méditerranée, Campus de Luminy Case 907, F Marseille Cedex 09, France *Author for correspondence ( therond@unice.fr) Accepted 9 October; published on WWW 14 November 2000 SUMMARY The Hedgehog (Hh) family of secreted proteins are key factors that control pattern formation in invertebrates and vertebrates. The manner in which Hh molecules regulate a target cell remains poorly understood. In the Drosophila embryo, Hh is produced in identical stripes of cells in the posterior compartment of each segment. From these cells a Hh signal acts in both anterior and posterior directions. In the anterior cells, the target genes wingless and patched are activated whereas posterior cells respond to Hh by expressing rhomboid and patched. Here, we have examined the role of the transcription factor Cubitus interruptus (Ci) in this process. So far, Ci has been thought to be the most downstream component of the Hh pathway capable of activating all Hh functions. However, our current study of a null ci allele, indicates that it is actually not required for all Hh functions. Whereas Hh and Ci are both required for patched expression, the target genes wingless and rhomboid have unequal requirements for Hh and Ci activity. Hh is required for the maintenance of wingless expression before embryonic stage 11 whereas Ci is necessary only later during stage 11. For rhomboid expression Hh is required positively whereas Ci exhibits negative input. These results indicate that factors other than Ci are necessary for Hh target gene regulation. We present evidence that the zinc-finger protein Teashirt is one candidate for this activity. We show that it is required positively for rhomboid expression and that Teashirt and Ci act in a partially redundant manner before stage 11 to maintain wingless expression in the trunk. Key words: Hedgehog, Cubitus interruptus, Teashirt, Signal transduction, Patterning, Drosophila INTRODUCTION Vertebrate and invertebrate development requires numerous cellular interactions and communications to specify normal patterning of the embryo. Cell communication is mediated by signalling molecules that instruct cells in specific ways. The ventral ectoderm of the Drosophila embryo is a good model for such studies. At the end of embryogenesis, the trunk (thorax and abdomen) of the larva exhibits a clear segmental pattern. The anterior part of each abdominal segment is covered by small hooks or denticles that are required for the larval movement, while the posterior of each of these segments is covered by naked cuticle. Screening for mutations that disrupt the cuticle pattern allowed the identification of mutations in the segment polarity class of genes, which are required for establishing the segmental borders and the polarity of the segment (Nusslein- Volhard et al., 1980, 1984; Wieschaus et al., 1984; Jurgens et al., 1984). Interestingly most of these genes have been conserved in vertebrates. Among these genes, some encode secreted signalling molecules such as Hedgehog (Hh) or Wingless (Wg)/Wnt, while others encode proteins required for the secretion or the transmission of the Hh and Wg signals (Hammerschmidt et al., 1997; Wodarz and Nusse, 1998). Because they influence cells within a large area, the expression of these signalling molecules must be tightly regulated. Even minor mis-expression of one of these genes leads to dramatic consequences during embryonic development. The Hh family of molecules has a central role for such regulation since some of its target genes encode long range signalling molecules including the Wnt/Wg and TGFβ protein families in vertebrate and invertebrate species (reviewed by Hammerschmidt et al., 1997). This precise regulation is well illustrated during Drosophila embryonic development where Hh is expressed and secreted from two rows of cells in each segment, corresponding to the posterior compartment (Mohler and Vani, 1992; Tabata et al., 1992; Lee et al., 1992). Anterior to its source of production, Hh is required to maintain the expression of wg in single row of cells (Ingham and Hidalgo, 1993; Hidalgo and Ingham, 1990). Stabilisation of wg expression allows the maintenance of hh transcription in the adjacent cell row, and thus Hh and Wg

2 5510 A. Gallet and others reinforce each other s expression. In a second, later step these secreted molecules will specify cell fate choices. Wg is directly responsible for the naked cell fate (Bejsovec and Martinez- Arias, 1991; Noordermeer et al., 1992; Lawrence at al., 1996), whereas Hh will indirectly govern some of the denticle identities by activating rhomboid (rho) transcription in rows of cells posterior to the Hh-secreting cells (Sanson et al., 1999; Gritzan et al., 1999). rho encodes a transmembrane protein required for EGF signal activation required for denticle diversity (Szuts et al., 1997; O Keefe et al., 1997). Hh maintains, on both side of its expression domain, a third target gene encoding the Hh receptor Patched (Ptc) (Nakano et al., 1989; Hooper and Scott, 1989; Hidalgo and Ingham, 1990). In addition to transducing the Hh signal, Ptc limits the range of Hh action by restricting its diffusion (Chen and Struhl, 1996). Genetic analysis has shown that the loss of hh function abolishes wg, rho and ptc expression and gives rise to larvae without any naked cuticle and with no denticle diversity. By contrast, hh overexpression promotes the expansion of wg, rho and ptc expression and the larvae display mirror image duplication of their denticle belts (Ingham, 1993; Tabata and Kornberg, 1994; Alexandre et al., 1999, this study). In conclusion, the alternate expression of wg, hh, ptc and rho is responsible for cellular identity and polarity of the embryonic segments (Alexandre et al., 1999; Gritzan et al., 1999). To control its target genes, Hh needs to antagonise Ptc activity (Hooper and Scott, 1989; Nakano et al., 1989). Upon binding of Hh, Ptc releases its inhibition on the serpentine protein Smoothened (Alcedo et al., 1996; van den Heuvel and Ingham, 1996). The Hh signal is then transmitted within the cell through the activation of a multiprotein complex, containing Cubitus interruptus (Ci) bound to microtubules (Robbins et al., 1997; Sisson et al., 1997). Although the regulation of this complex is not yet understood, Hh activity releases the complex from the microtubules. This protein complex may facilitate Hh signalling by governing the stability and access of Ci to the nucleus. Ci is the most downstream component so far identified in the Hh pathway (reviewed by Aza-Blanc and Kornberg, 1999). It encodes a zinc-finger transcription factor that mediates all known Hh cellular responses in the imaginal discs and in the embryo. Ci is required for both activation and repression of Hh target genes. In the presence of Hh, Ci is converted into an active form (Ci act ), while the absence of Hh promotes the cleavage of Ci into a repressive form (Ci rep ; Alexandre et al., 1996a; Dominguez et al., 1996; Méthot and Basler, 1999; Aza-Blanc et al., 1997; Hepker et al., 1997; Von Ohlen et al., 1997; Ohlmeyer and Kalderon, 1998; Wang and Holmgren, 1999). Here, we ask whether all cells responding to Hh also invariably use Ci for target gene regulation. We have surprisingly found that Ci is not required for all Hh functions during embryogenesis: it is only required for ptc regulation, for late, but not early, wg maintenance and not for rho expression. It has been shown that the zinc-finger protein Teashirt (Tsh) is required for the maintenance of wg expression (Gallet et al., 1998) in a similar manner to Ci (this work). Tsh is expressed exclusively in the trunk (or thorax and abdomen) where it is required for patterning in combination with specific Hox proteins (Fasano et al., 1991; Röder et al., 1992). We show that Tsh and Ci are required in a redundant manner for the Hhdependent regulation of wg before stage 10. Finally, we show that Tsh is required for the expression of the Hh target gene rho. Our results suggest that Hh target gene regulation requires the activity of two zinc-finger proteins: Ci and Tsh. MATERIALS AND METHODS Fly stocks ptc 9, hh AC, hh 15, ci 94 and tsh 8 are null alleles (Hooper, 1994; Lee et al., 1992; Molher, 1988; Méthot and Basler, 1999; Fasano et al., 1991). Ci cell2 is a dominant negative allele of ci (Méthot and Basler, 1999). ptc 9, UASHh strains were gift from P. Ingham. UASCi rep strain was a gift from T. Kornberg. The Ci rep protein that is expressed is similar to the endogenous Ci75 molecule (Aza-Blanc et al., 1997). prdgal4 was a gift from L. Fasano. UASTsh#20 is an puastsh insertion on the second chromosome that allows the expression Tsh at a moderate level. All the other stocks were obtained from the Bloomington stock centre. The UAS-Gal4 system and the pattern of expression of the Gal4 drivers used are described in Brand and Perrimon (1993). All experiments described in this paper were carried out with both null alleles of hh (hh AC, hh 15 ); photographs show only hh AC. Genetic experiments and statistics In order to simplify the statistical analyses only larvae displaying mutant phenotypes were counted. Flies were allowed to lay on grape juice plates for hours. Plates were removed and incubated until the hatching of wild-type larvae were removed. ptc 9 ; ci 94 double mutant embryos were obtained from ptc 9 /+; ci 94 /+ parents. 106 mutant larvae were analysed with 12 (11.3% in place of 14.3%) presenting a ptc; ci 94 double mutant phenotype. prdgal4 UASHh, ci 94 embryos were obtained by crossing prdgal4/+; ci 94 /+ to UASHh/+; ci 94 /+ parents. 182 mutant larvae were analysed where 31 (17% in place of 14.3%) presented a prdgal4uashh; ci double mutant phenotype. hh, ci 94 double mutant embryos were obtained from hh/+; ci 94 /+ parents. 97 mutant larvae were analysed with 13 (13.4% in place of 14.3%) presenting a hh; ci 94 double mutant phenotype. ptcgal4 UASCi rep, ci 94 embryos were obtained by crossing ptcgal4/+; ci 94 /+ to UASCi rep /+; ci 94 /+ parents. In this experiment, ptcgal4uasci rep larvae do not die during embryogenesis and do not display any mutant phenotype, so they were not counted. 152 mutant larvae were analysed with 35 (23% in place of 25%) presenting a ptcgal4uasci rep ; ci 94 double mutant phenotype. tsh 8, ci 94 embryos were obtained from the stock tsh 8 /SM6ß; ci 94 /ey D. tsh 8, 69BGal4 UASHh embryos were obtained by crossing tsh 8 /CyO; UASHh/ UASHh to tsh 8 /CyO; 69BGal4 parents. tsh 8, prdgal4 UASHh were obtained by crossing tsh 8 /CyO; UASHh/UASHh to tsh 8 /+; prdgal4/+ parents. For ectopic expression of Hh in tsh ci embryos, UASHh tsh 8 /CyO; ci 94 /M(4)101 females were crossed to tsh 8 /CyO; prdgal4/ prdgal4; ci 94 /ey D males. For ectopic expression of Tsh in hh AC homozygotes, UASTsh20/UASTsh20; hh AC /TM3 females were crossed to armgal4/armgal4; hh AC /TM3 males. Cuticle preparation, in situ hybridisation and immunostaining of embryos Cuticles and embryo preparations were described in Gallet et al. (1998). RNA in situ hybridisation is described in Francois et al. (1994). Immunostaining of embryos is described in Gallet et al. (1998). ptc antisense RNA probe was synthesised from the ptc cdna cloned into pnb40 plasmid using T7 polymerase. rho antisense RNA probe was synthesised from the rho cdna cloned into the pbluescript plasmid (gift from J.-P. Vincent) using T7 polymerase. wg antisense RNA probe was made from wg-pbluescript plasmid using T7 polymerase. Mouse polyclonal anti-engrailed (at 1/1000) antibodies were a gift from T. Kornberg. After in situ hybridisation, embryos of the correct genotype were identified using secondary fluorescent

3 Cubitus interruptus-independent Hedgehog signalling 5511 labelling after the following primary antibodies: anti-en to recognise ptc or hh embryos; anti-ci to recognise ci 94 embryos; and anti-tsh to recognise tsh embryos. Rat monoclonal anti-ci 2A1 (at 1/10) were made from hybridoma-expressing cells (gift from R. Holmgren). Rat anti-tsh antibodies (at 1/500) were described by Gallet et al. (1998). Fluorescent secondary antibodies used were anti-rat Cy 3 and antimouse FITC at a working dilution of 1/200 (Jackson Laboratory). RESULTS Ci is required to control ptc and late wg expression but not rho expression The second to seventh abdominal segments of wild-type larvae are covered by two types of cuticle: in the anterior part, six rows of pigmented, cytoskeletal extrusions formed the denticle belts, whereas the posterior part was covered by naked cuticle (Fig. 1A; Lohs-Schardin et al., 1979). Denticle rows were characterised by their size, shape and polarity (type 1 to 6, Fig. 1A ). The first four rows are smaller than the fifth one and were hooked. While the first and fourth rows pointed anteriorly, rows 2, 3 and 5 point posteriorly. Finally, the most posterior, type 6, row almost always displayed doubled, small denticles pointing posteriorly. According to Alexandre et al. (1999), row 1 is only made at the interface between the naked and denticled cuticle and row 4 is only made when adjacent to row 5. Recently, a null allele of the ci gene was identified (Slusarski et al., 1995; Méthot and Basler, 1999). Although numerous studies (reviewed by Aza-Blanc and Kornberg, 1999) have shown that Ci is required to transduce all Hh signalling instructions to the nucleus in the imaginal discs, we decided to re-examine Ci activity during embryogenesis. Surprisingly, we found that the zygotic ci null phenotype was weaker than that of the hh mutant (compare Fig. 1B with 1C). First, no segmentation was seen in an hh null mutant (Fig. 1B), while obvious segmentation was apparent in ci 94 homozygous embryos (Fig. 1C). Second, loss of Hh caused the absence, whereas loss of Ci caused only a reduction, of the naked cuticle type. Third, whereas in hh null mutants only type 5 denticles differentiate (Gritzan et al., 1999; Fig. 1B ), ci 94 mutant cuticles showed mirror image duplications of denticle types 2-5 (Fig. 1C ), with type 1 denticles occasionally developing in segments where naked cuticle persists (thin arrow in Fig. 1C ). This difference in phenotype could be explained by a maternal contribution of the Ci product. However no maternal contribution of Ci has been found (N. Méthot and K. Basler, personal communication; Ramirez-Weber et al., 2000) showing that Ci only gives zygotic input for larval patterning. To confirm our cuticle analyses we compared the expression of three hh target genes (ptc, wg and rho) in ci and hh null mutations. Although during early embryogenesis ptc was ubiquitously transcribed in the anterior compartment (corresponding to the non-engrailed (En)-expressing cells) of each wild-type trunk segment, its expression was then upregulated around stage 10 in two stripes of cells, one stripe each side of the En/Hhexpressing cells (Fig. 2A; Nakano et al., 1989; Hopper and Scott, 1989). In stage 11 embryos, most ptc expression disappears from the epidermis in the absence of Hh activity (Fig. 2C; Hidalgo and Ingham, 1990) and is expressed in more cells following Hh overexpression (see Fig. 6F; Ingham, 1993). In ci 94 embryos, ptc mrna was no longer upregulated. It disappeared from the epidermis, as in hh null mutants, retaining only a weak basal level (Fig. 2B-C). wg is expressed in a stripe anteriorly to the En/Hh expressing cells in each segment. During the stabilisation phase (between stages 8 to 10), wg is a target of Hh signalling while later, during the specification phase, wg transcription becomes independent of Hh signalling but is a target of its own signalling (Ingham and Hidalgo; 1993; Bejsovec and Martinez- Arias, 1991; Dougan and DiNardo, 1992, Li and Noll, 1993; Gallet et al., 1998). Hence, in hh null embryos wg was downregulated during stage 9 (compare Fig. 2F with 2D; Ingham and Hidalgo, 1993) and wg transcripts were no longer detected in later stages (compare Fig. 2I,L with 2G,J) in the epidermis. In contrast, in ci 94 embryos, wg transcripts were still detectable until stage 11 (Fig. 2E,H) and only disappeared completely in ventral positions during stage 12 (Fig. 2K); in the dorsal domains wg was detectable throughout embryogenesis. rho expression was detected in a three-cell wide stripe posteriorly to the En/Hh-expressing cells (Fig. 2M; Alexandre et al., 1999). The pattern of rho expression is complex and is controlled by Hh and Serrate (Ser) signals and also by Hox genes expressed in the trunk (Alexandre et al., 1999; Gritzan et al., 1999; Wiellette and McGinnis; 1999; Sanson et al., 1999; Szuts et al., 1997). In the abdominal segments of wild-type embryos, the most anterior row of rho-positive cells displayed intermediate levels of expression, while the most posterior one showed the highest level of expression (Fig. 2M,M ; Alexandre et al., 1999). These different levels of expression are due to Hh-mediated regulation on the anterior rho-expressing cells and to Ser-mediated regulation of the posterior rho cells (Alexandre et al., 1999). In absence of hh function, rho expression is dramatically reduced but is still detectable in some cells (Fig. 2O,O ; Gritzan et al., 1999; Sanson et al., 1999). Because rho transcription is completely abolished in hh Ser double mutants (Alexandre et al., 1999), it has been proposed that residual rho expression in hh mutant is due to Ser activity. Strikingly, the rho expression domain in ci 94 embryos expanded to 4 or 5 cells wide posteriorly to the En cells and was not reduced as expected (Fig. 2N,N compare with wild type in Fig. 2M ). Close analysis of rho expression revealed that the modulation of its expression level is similar in ci 94 and wild-type embryos. This wider domain of rho expression correlates with the expansion of rows of denticles types 2-4 in the ci 94 cuticular phenotype (Fig. 1C ). These results show that the regulation of Hh target genes is different in ci and hh null mutant backgrounds. In summary, Ci appears to be sufficient to relay Hh signal to regulate ptc expression, insufficient for early wg regulation and acts as a negative, rather than a positive, regulator for rho expression. Ci is epistatic to Hh for ptc and late wg regulation but not for rho and early wg expression To analyse further the hypothesis that Ci is insufficient to transmit all Hh signalling to the nucleus, we performed epistasis experiments using either ptc loss-of-function or ectopic Hh expression in a ci 94 background. Because Hh and Ptc are antagonistic, the loss of ptc function mimics the effects of ubiquitous expression of Hh (Sampedro and Guerrero, 1991; Ingham, 1993; Ingham et al., 1991). We reasoned that if all Hh signalling was mediated by Ci, the cuticle phenotype of ptc,

4 5512 A. Gallet and others Fig. 1. The embryonic phenotype of the ci 94 null allele is weaker than the hh null phenotype. Anterior is towards the left. (A) Cuticle of a wildtype larva. Note the invariable pattern of denticle belts from abdominal segments 2 (A2) to 7 (A7). (A ) Enlargement of wild type 3 rd and 4 th abdominal segments. Each denticle in rows 1 to 6 is characterised by its size, shape and orientation. Each denticle belt is separated by cells secreting naked cuticle (N). (B) hh AC /hh AC mutant cuticle. No segmentation or denticle diversity is detectable. Only type 5 denticles are made (B ). (C) ci 94 /ci 94 mutant cuticle. Some segmentation and denticle diversity are visible compared with hh AC. (C ) The denticle belts are duplicated in mirror image. Note that there are fewer type 4 denticles than in wild type. Denticle type 1 is also absent except in unfused segment (arrow). ci 94 double mutant embryos should resemble the ci 94 one. ptc mutant larvae displayed mirror image duplication (Fig. 3A) of denticle rows 1 and 2 with absence of rows 3, 4 and 5 (Fig. 3A ). This phenotype resulted first from the expression of the wg gene in more cells than usual (Fig. 3B and C; Martinez- Arias et al., 1988). The expansion of the wg-expression domain promoted de novo en expression anteriorly to the wide wg domain. Consequently, an ectopic groove formed in each segment (arrows in 3C) due to the formation of new segmental frontier between Wg- and En-expressing cells (Hooper and Scott, 1989; Nakano et al., 1989). Second, the constitutive activation of the Hh pathway, caused by the loss of ptc function, caused ectopic expression of the rho gene between the two en stripes (Fig. 3D; Alexandre et al., 1999). Thus, a new ectopic rho/en interface was formed and was responsible for the ectopic 1/2 type denticles present at the posterior of each belt (Fig. 3A ). In ptc, ci 94 double mutant embryos the cuticular phenotype displayed an intermediate phenotype, resembling neither the ptc nor the ci single mutant (Fig. 3E). While Ci appears to act downstream of Ptc for denticle identities, the formation of naked cuticle does not require the presence of Ci (Fig. 3E ). However thoracic segments never produced any naked cuticle and some abdominal segments were often fused laterally (Tx and arrows in Fig. 3E). The cuticle phenotype of the ptc, ci 94 double mutant embryos could be correlated with the expression of wg and rho. In such embryos, wg expression was expanded anteriorly during germ-band extension (Fig. 3F) as in the ptc single mutant (Fig. 3B), but was no longer maintained during later stages and had almost completely disappeared by stage 12 (Fig. 3G), as in ci 94 embryos (Fig. 2K). Wg activity is required for naked cuticle formation from stage 8 to 11 (Noordermeer et al., 1992; Gallet et al., 1998) and, thus, the persistence of wide wg expression until stage 11 in ptc, ci 94 double mutant embryos could be sufficient to promote the naked cell fate choice in the rescued segments. In these embryos the ectopic groove that is usually formed anteriorly to the ectopic wg expression domain is not present (arrows in Fig. 3C,G). This correlates with the absence of ectopic en expression (not shown). Thus, no ectopic en/wg and rho/en interfaces are formed contrary to the situation in ptc embryos. Subsequently, duplication of rows 1 and 2 is not observed in ptc, ci 94 double mutant embryos. However, the strong expression of rho at stage 13 in ptc, ci 94 embryos (Fig. 3H) still promotes the expansion of denticle types 2 and 3 (Fig. 3E ). In addition, type 4 and 5 denticles developed laterally in the double mutant. These were not visible in the ptc single mutant (compare Fig. 3A with 3E ). To further characterise Ci-independent Hh signalling activity, we removed Ci activity in embryos overexpressing a UASHh transgene under the control of the prdgal4 driver. Using prdgal4 as a driver provides an internal control because of its expression in a pair rule pattern (Brand and Perrimon, 1993). Cuticles from prdgal4-uashh embryos displayed mutant phenotypes in every other segment, consistent with the domains where Hh was overexpressed (Fig. 4A,A ). While thoracic and first abdominal segments were principally covered by naked cuticle, abdominal segments 3, 5 and 7 presented mirror-image duplication of their denticle belts, similar to that observed in ptc embryos. These phenotypes are in accord with the wg and rho expression domains in this genotype (Fig. 4B-D). wg was expressed in a 4-5 cell-wide domain and rho was transcribed uniformly in its expression domain in odd-numbered

5 Cubitus interruptus-independent Hedgehog signalling 5513 Fig. 2. Ci does not regulate all Hh target genes. Expression of ptc (A-C) and wg (D-L) mrnas. Double staining for rho expression (blue)/en protein (brown) (M-O). (M -O ) Enlargement of abdominal segments for M-O embryos but simply stained for rho mrna. (A,D,G,J,M,M ) Wild-type embryos; (B,E,H,K,N,N ) ci 94 embryos; (C,F,I,L,O,O ) hh AC embryos. Ci and Hh are required for ptc transcription (compare A, B and C). wg is detected at wild-type levels in ci 94 embryos at stage 9 (E, compare the hh-independent wg level of expression in the head with the hh-dependent wg level in the trunk) but levels decrease at stage 11 (H, compare with wild-type embryo in G) and wg transcripts have faded completely by stage 12 (K compare with J). wg transcripts are at lower levels in hh AC embryos earlier at stage 9 (compare D with F) and are completely absent by stage 11 in the trunk (I). No more wg transcripts are detected during later stages in hh AC embryos (stage 12 in L). (M) In stage 13 wild-type embryos, rho (blue) is transcribed in three rows of cells posterior to En (brown) expressing cells. (M ) Enlargement of adjacent abdominal segments showing the complex rho expression pattern: rho is stronger in the most posterior row of cells and weakest in the middle row. (N and enlargement in N ). In ci 94 embryos, the rho expression domain is wider (4 to 5 cell wide); the arrows indicate the segmental groove. (O and enlargement in O ) rho expression in hh AC embryos is strongly reduced. abdominal segments (Fig. 4B-D). As in ptc embryos, ectopic wg expression induced a new 2/1 type at the posterior interface of odd-numbered abdominal belts (clearly visible here in the A5 segment in Fig. 4A ). The removal of Ci activity in prdgal4-uashh embryos did not affect rho overexpression induced by UASHh (Fig. 4H). However loss of Ci strongly reduced endogenous and ectopic wg expression induced by UASHh from stage 12 onwards (Fig. 4G). Consequently, the mirror-image duplication of odd-numbered abdominal denticle belts was rescued (Fig. 4E,E ). Note that naked cuticle was more often present in abdominal segments 3 and 5, and was almost always absent in abdominal segments 1 and 7, and in all thoracic segments, as is the case in ptc, ci 94 double mutant embryos (Fig. 3E,E ). The absence of

6 5514 A. Gallet and others Fig. 3. Ci is not fully epistatic to Ptc. Anterior is towards the left. (A-D) ptc 9 embryos; (E-H) ptc 9 ; ci 94 double mutant embryos. (B,C,F,G) wg RNA in situ; (D,H) rho RNA in situ. Abdominal denticle belts of ptc mutant embryos exhibit mirror image symmetry (A); only denticles types 1 and 2 are made (arrows in A ). (B,C) In these embryos, wg expression is wider compared with expression in wild-type embryos: three cells wide at stage 9 (B, st9 compare with Fig. 2D) and four to five cells wide at stage 12 (C, st12 compare with Fig. 2J). (D) rho expression is strong in the three rho-expressing cells, independent of their position (compare with Fig. 2M ). (E) In ptc 9, ci 94 double mutant embryos, naked cuticle is observed but is more limited (compare with A). Note that some segments are fused (white arrows), and that thoracic segments do not produce any naked cuticle (Tx). (E ) However, abdominal denticle diversity is restored and type 5 denticles are principally present laterally (arrow). (F) In ptc 9, ci 94 double mutant embryos, wg expression is wider (two to three cells wide) than in wild type at stage 9; most importantly, its expression disappears at later stages (G). rho expression (H) is similar to that observed in ptc 9 single mutants (D). Note that an ectopic groove is present in ptc 9 mutant embryo (left-hand arrow in C; the posterior arrow indicates wild-type groove) that is not made in ptc 9, ci 94 double mutant embryos (arrow in G). naked cuticle correlates with the lack of overexpression of wg in the thorax (Tx in Fig. 4G). We also checked ptc expression in ci 94, prdgal4-uashh embryos and found that ptc upregulation failed to be detected in odd-numbered abdominal segments (data not shown) since Ci activity is required downstream of Hh for this function (Fig. 2B). Similar results were obtained in ci 94 embryos overexpressing UASHh under the control of the ubiquitous driver 69BGal4 (data not shown). Taken together our results show that while Ci is strictly required downstream of Hh to regulate ptc expression and wg expression after stage 10, it is not necessary for either early wg regulation or for rho activation by Hh. Ci rep activity can repress rho expression Recent studies have shown that Ci protein co-exists in multiple forms. Upon Hh signalling, full length or Ci 155 is converted into an activating form (Ci act ) while in absence of Hh, Ci 155 is cleaved to yield a smaller form of Ci possessing repressive properties (Ci rep ) (Aza-Blanc et al., 1997; Hepker et al., 1997). Nevertheless, several studies in cell culture or in wing imaginal discs have shown that in absence of Hh, Ci 155 is still present, while Ci rep concentration is increased (Aza-Blanc et al., 1997; Ohlmeyer and Kalderon, 1998; Chen et al., 1999). Our results do not yet rule out the possibility that the phenotypes caused by the total lack of ci function (e.g. Ci act and Ci rep ) in ci 94, are weaker than in hh embryos owing to the absence of the Ci rep form. Indeed, Méthot and Basler (1999) have recently shown that in the anterior cells of ci 94 wing imaginal disc, the Hh target gene decapentaplegic (dpp) is still transcribed, albeit at a low level. The authors have shown that this basal level of dpp expression is due to the lack of Ci rep activity that is required to fully repress the dpp transcription far from the Hh source. In order to know what the contribution of Ci rep is during embryogenesis, we performed epistasis experiments between hh and ci. The cuticular phenotype of hh, ci 94 double mutant embryos was weaker than the hh single mutant, but stronger than the ci single mutant (Fig. 5A, compare to 1B and C). Detailed examination of the cuticle showed that denticle types 4 and 5 were present (Fig. 5A ), whereas hh embryos were only covered by type 5 denticles (Fig. 1B ). This result suggests that in absence of Ci rep activity, denticle diversity is partially restored. This correlates with the broader domain of rho expression in hh, ci 94 embryos compared with a hh single mutant, confirming the Ci rep contribution in this phenotype (compare Fig. 5B with Fig. 2O ). The analysis of wg expression in hh, ci 94 embryos revealed that even in absence of Ci rep activity no upregulation of wg transcription was detected. Early wg expression decayed during stage 9 (compare Fig. 5C with Fig. 2F), as in a hh null single mutant, confirming that a Ci-independent, positive input from Hh is required for the maintenance of early wg expression.

7 Cubitus interruptus-independent Hedgehog signalling 5515 Fig. 4. Ci is not necessary for the early wg and rho Hh-dependent expression. Anterior is towards the left. (A-D) prdgal4 UASHh embryos; (E-H) prdgal4 UASHh; ci 94 embryos. (B,C,F,G) wg mrna in situ; (D,H) rho mrna in situ. (A,A ) Ectopic Hh expression under prdgal4 driver control promotes the duplication of denticle rows 1 and 2 in odd-numbered abdominal segments (A3 to A7). (E,E ) Removing ci function in such embryos abolishes the denticle duplication phenotype and restore correct denticles identity and polarity. Note that type 5 denticles (arrows in E ) are present only laterally and naked cuticle is still made in most odd-numbered segments. During early stages, prdgal4-induced Hh promotes ectopic expression of wg in odd-numbered abdominal segments independently of the presence of Ci (black circles in B,F). Ectopic wg expression however is less robust in prdgal4uashh; ci 94 embryos (F). During later stages, wg expression fades in the absence of Ci during stage 11 (compare dots in C with G). Note that thoracic and A1 segments are more sensitive to the lack of Ci than other abdominal segments (Tx and A1 in G). (D) rho expression is also affected in abdominal odd-numbered abdominal segments when Hh is under prdgal4 control: rho expression does not display anymore modulation; rho mrna is expressed at high equivalent levels in all three rows of rhoexpressing cells (compare A5 with A4 segment in D). Loss of ci function does not affect this rho overexpression (segment A5 in H). lb, labial segment; mx, maxillary segment; Tx, thoracic segment. To further address the role of Ci rep we decided to overexpress a UAS Ci rep construct driven by ptcgal4 in a ci 94 background. Because the basal level of ptc that starts at gastrulation is independent of the Hh pathway, we used the ptcgal4 driver to allow the expression of UAS Ci rep in a domain complementary to that of Hh. If the hh phenotype is the result of the loss of positive inputs from ci ACT and of the gain of Ci rep activity, then ci 94 ; ptcgal4-uasci rep embryos should phenocopy hh null embryos. In these embryos the cuticular phenotype was weaker than the loss of function of hh but stronger than the ci 94 cuticle (compare Fig. 5D,D with 1B- C ). Type 4 denticles were still present although sometimes reduced in some segments (arrow in Fig. 5D ). In these embryos, wg was lost early during embryogenesis (Fig. 5F) in a manner similar to hh mutant embryos, showing that Ci rep exerts a repressive function on wg transcription. Interestingly, rho expression was interrupted in some segments (arrow in Fig. 5E), suggesting that Ci rep could act as a repressor of rho expression. The ci Ce2 allele that encodes a mutant form of Ci, producing only the repressor form (Méthot and Basler, 1999), displayed a similar phenotype (compare Fig. 6G with 6H) to ci 94, ptcgal4-uasci rep embryos, with interruption of the rho expressing stripes (arrows in Fig. 5H). In conclusion, our results show that in a hh single mutant, rho expression is restricted by both loss of positive input from Hh signal and by a repressive activity of the Ci rep form consistent with the loss of denticle identities. Thus, the expansion of rho expression observed in ci 94 (Fig. 2N ) must therefore be the consequence of the absence of Ci rep. Moreover, even if Ci rep is able to repress wg transcription, the absence of Ci rep activities in hh, ci 94 double mutant embryos does not compensate for the absence of the positive input from Hh that regulates early wg expression. This shows that in ci 94 embryos the persistence of wg expression until stage 11 is not due to a basal transcriptional activity of the wg promoter. Ci and Tsh act together for normal Hh signal transduction in the trunk Together our results suggest that during embryogenesis Ci is not the sole transcription factor employed to regulate Hh target genes. Other or another factor seems to be responsible for rho regulation and for early wg expression. A possible candidate for such functions is the zinc-finger transcription factor Tsh (Fasano et al., 1991), which, like Ci (Fig. 6C), is required for the maintenance of wg expression after stage 10 (Gallet et al., 1998). Previous studies have shown that Tsh is exclusively expressed in the trunk (Alexandre et al., 1996b) and plays homeotic functions required in cooperation with Hox genes to

8 5516 A. Gallet and others Fig. 5. Ci rep activity contributes to the hh loss of function phenotype. Anterior is towards the left. (A-C) hh AC ; ci 94 double mutant embryos; (D-F) ci 94 ; ptcgal4 UASCi rep embryos; (G,H) Ci cell2 embryo; (B,E,H) rho mrna in situ; (C, F) wg mrna in situ. (A,A ) Double mutant embryos for hh AC ; ci 94 display a cuticular phenotype that is weaker than hh AC (compare with Fig. 1B,B ) but stronger than ci 94 (compare with Fig. 1C,C ). Type 4 denticles are present (A ) correlating with the stronger rho expression (B) than in hh AC single mutant (see Fig. 2O ). (C) However, wg transcript is no longer present after stage 10 in hh AC ; ci 94 embryos, as in hh AC single mutant (see Fig. 2F,I). Embryos expressing a Ci repressor form (Ci rep ) in a ci null background (D and enlargement in D ) display a cuticular phenotype close to Ci cell2 embryos (G and enlargement in G ). Note that both genotypes still display denticle diversity unlike hh null mutant larvae (Fig. 1B). This diversity correlates with rho expression (E,H). However, because of Ci rep activity, rho expression is often interrupted (arrows in E,H) correlating with segments showing reduced denticle diversity (arrows in D,G ). Moreover, Ci rep activity is able to repress wg expression since wg mrna is disappearing during stage 9 in ci 94 ; ptcgal4 UASCi rep embryos (F), while wg transcript is lost during stage 11 in ci 94 single mutant embryos (Fig. 2H). specify trunk identities (Röder et al., 1992; de Zulueta et al., 1994). Tsh also displays a weak segment polarity phenotype with disruption of both naked cuticle and denticle belt organisation (Fig. 6A). Detailed examination of the denticle belts from tsh null embryos show that only denticle types 5 and 6 are made (Fig. 6A ), similar to embryos overexpressing a dominant negative form of EGFR (Szuts et al., 1997). To investigate a possible role of Tsh downstream of Hh we analysed the regulation of rho and ptc in tsh null embryos. While the transcription of ptc was unaltered (Fig. 6B), rho transcription was not detected in the epidermis of tsh embryos (Fig. 6D). We then performed epistasis experiments by over expressing Hh in a tsh null background. As expected, UASHh under the control of the prdgal4 driver even in absence of tsh gene, promoted the spreading of ptc expression (compare Fig. 6B with 6F). However, wg was ectopically expressed anteriorly to its source during the stabilisation phase (data not shown), but this expression decayed during later stages (Fig. 6G). This result is comparable with that observed in ci 94, prdgal4- UASHh embryos (Fig. 4G). Finally, we found that rho expression was not initiated in tsh 8 ; prdgal4-uashh embryos, showing that Tsh is necessary for normal rho regulation (Fig. 6H). To determine whether Tsh could be a direct regulator of rho expression, we decided to look at rho expression in embryos overexpressing Tsh. We have previously shown that ubiquitous overexpression of Tsh using the UAS system promotes the suppression of denticles leaving only naked cuticle (Gallet et al., 1998). In order to circumvent this difficulty, we used an UASTsh insertion that displays a moderate level of Tsh expression without affecting the denticle formation at 25 C (Fig. 6I). Under these conditions, we observed that the anterior row of rho-expressing cells was changed, being at a higher level comparable with the most posterior rho (compare Fig. 6J with Fig. 2M ). We obtained similar result when the experiment was performed at 29 C (Fig. 6K). Nevertheless expansion of rho expression was never observed. Furthermore, we found that overexpression of Tsh in hh mutant could not rescue denticle diversity (data not shown). The fact that wg expression disappeared during stage 11 in tsh or ci 94 mutants raises the possibility that Ci and Tsh act in a redundant manner for early wg regulation prior to stage 11. In this model Hh could employ either Tsh or Ci for wg regulation. In order to test this hypothesis we analysed wg and En expression in tsh; ci 94 double mutant embryos. In wild-type embryos, wg was expressed in a one-cell-wide stripe, whereas En was expressed in a two to three-cell-wide stripe (Fig. 7F). If both Ci and Tsh are required redundantly to maintain wg expression during the stabilisation phase, En expression should

9 Cubitus interruptus-independent Hedgehog signalling 5517 Fig. 6. Tsh is necessary for rho and late wg expression. Anterior is towards the left. (A-D) tsh 8 embryos. (E-H) tsh 8, prdgal4 UASHh embryos. (I-K) 69BGal4 UASTsh embryos raised at 25 C (I,J) or 29 C (K). (A) Cuticle from tsh 8 null embryo displays small naked areas in each segment and weakly pigmented denticles type 5 and 6. (B-D) Tsh does not regulate ptc transcription (B) but is required for late wg transcription maintenance (C) and rho transcription (D). Note that some rho expression is still present around the ventral midline of the embryos (arrow in D). (E) Cuticle from tsh null embryo overexpressing UASHh under the control of the pair rule prdgal4 driver. Denticles types are unchanged compare to tsh null embryos (e.g. type 5 and 6 denticles). However, in some segments overexpressing UASHh, denticle belts are reduced (arrow in E). Note also the modified shape of the denticle belt. (F) In tsh 8, prdgal4 UASHh embryos, ptc expression is expanded under the action of ectopic Hh in odd numbered abdominal segments (circles) and lack of tsh function does not affect this expansion. (G) However, Tsh is required, as Ci, for late ectopic wg expression induced by prdgal4-uashh (compare with Fig. 4C). Note that wg expression is more intense in the head (arrow in G) compared with the disappearing expression in the trunk (black circles in G). (H) rho expression is not upregulated by ectopic Hh in absence of tsh function. Note that some rho expression persists in the ventral midline (arrow in H). (I) Cuticle from 69BGal4 UASTsh embryos grown at 25 C does not display any disruption of its denticlar pattern. (J) The anterior row of rho-expressing cells exhibit a higher level of rho (compare with wild type in Fig. 2M ). (K) rho expression in 69BGal4 UASTsh embryos grown at 29 C display a similar expression pattern to that obtained at 25 C. also be affected by the lack of Wg activity during this period. In the trunk of tsh ; ci embryos, wg expression was lost during stage 9-10 (Fig. 7D,G), a result similar to that observed in hh mutant embryos (Figs 2F, 7H). Moreover En expression was also at a lower level in stage 9-10 tsh, ci 94 double mutant embryos compared with wild type, reflecting the disappearance of wg expression (Fig. 7G). Most of the remaining En signal came from neuronal cells underneath the ectoderm. Comparable En disappearance was observed in wg (Fig. 7I) null embryos but not in single ci (Fig. 2N) or tsh (Gallet et al., 1998) homozygotes. In single ci or tsh homozygotes, ptc and rho, respectively, disappeared; both target gene expression patterns disappeared in the double mutant (Fig. 7A,B). Finally, cuticle analysis of tsh; ci 94 double mutant embryos showed that only type 5 denticles were made without production of naked cuticle, correlating with the expression of the target genes (Fig. 7J). The analysis of ptc, rho and wg in tsh; ci 94 ; prdgal4 UASHh triple mutant embryos confirmed the above results. As expected, we did not detect any ptc or rho expression in such embryos (not shown); however, while prdgal4 UASHh promoted the expansion of wg expression in gnathal segments (compare arrow heads in Fig. 7D and 7E) wg expression disappeared during stages 9-10 in the trunk (Fig. 7E). In conclusion, Ci and Tsh are both required for normal Hh target genes expression. Ci and Tsh act independently to regulate respectively ptc and rho expression, while both zincfinger transcription factors act first in a redundant manner to regulate wg transcription until stage 10 but then are required together to maintain wg expression later during the cell specification phase. DISCUSSION In the ventral ectoderm, Hh-receiving cells displayed differential responses as a function of position with respect to the En/Hh-producing cells. We show that, depending on the position of the receiving cells, Hh requires different regulators in order to control its target genes and subsequent developmental cell behaviour. On both sides of the Hhproducing domain, a Ci-dependent pathway is employed to

10 5518 A. Gallet and others Fig. 7. Ci and Tsh activities are redundant for early wg expression control. Anterior is towards the left. (A,B,D,G,J) tsh 8, ci 94 double mutant embryos; (C) ci 94 single mutant; (E) tsh 8, ci 94 ; prdgal4 UAShh embryo; (F) wild-type embryo; (H) hh AC embryo; (I) wg CX4 embryo. In situ for ptc (A), rho (B) and wg (C-E), and double staining for En protein (brown) and wg mrna (blue) (F-H). Single En staining (brown in I). In tsh 8, ci 94 double mutant embryos, ptc (A) and rho (B) expression disappears. (D,E) wg expression is lost at an earlier stage in the trunk compared with ci or tsh single mutants (compare with C and Fig. 6C) and at a comparable stage to that observed in hh mutant embryos (G, with H and Fig. 2F). Hh over expression does not affect wg expression in the trunk of tsh 8, ci 94 double mutant embryos (E) except in the gnathal segments where wider wg expression is detected during stage 10 (arrowheads, compare with C,D). In tsh 8, ci 94 embryos, En expression is also altered (G) in a similar way to wg (I) or to hh null embryos (H). Note that the remaining En staining in each genotype correspond to neural cells. (J) cuticle from a double mutant tsh 8, ci 94 embryo produces only weakly pigmented, type 5 denticles and no naked cuticle. regulate ptc expression. Anteriorly to the En/Hh cells Hh used Ci and Tsh zinc-finger factors to control wg expression in the trunk precisely. Finally, the Hh pathway required Tsh in the trunk, independently of Ci, to regulate rho posteriorly to the Hh source in each trunk segment (Fig. 8). Ci is only required for a subset of Hh functions during embryogenesis Ci is required to transduce Hh signal in order to activate its target genes. In cells that do not receive Hh, Ci is cleaved and represses Hh target genes (reviewed by Aza-Blanc et Kornberg, 1999). However, compelling results point out a more complex role for Ci activity during embryonic development of Drosophila. We show that the embryonic phenotype resulting from the complete loss of Ci function was weaker than the complete loss of Hh function. The phenotypic differences observed between hh and ci null mutations resides in the following observations: in ci 94 embryos one observes (1) the presence of segmentation due to maintenance of wg expression until stage 11 and (2) the presence of denticle diversities due to an expansion of EGF signalling illustrated by an expansion of rho expression. We note that Ci does not have a maternal contribution, as ci 94 homozygotes issuing from germ-line clones homozygous for ci 94, did not show a stronger phenotype than embryos lacking only zygotic Ci product (N. Méthot and K. Basler, personal communication), and also that embryos Fig. 8. Branch points in the Hh pathway that control gene expression in different positions within the segment. Schematic view of Hh-receiving cells anterior and posterior to its source. In both cell types Smo is required to transduce Hh pathway. Anteriorly to its source of secretion, Hh regulates wg and ptc expression through the cytoplasmic kinase Fu and the transcription factor Ci. Ci act would be required for the maintenance of wg transcription under the influence of Fu. However, Tsh acts downstream of Wg and is required redundantly with Ci for early wg transcription maintenance by Hh. Posteriorly to its source, Hh is required for ptc transcription but also for the regulation of rho, all this is independent of Fu. Ci 155 corresponds to the Ci isoform that is detected posteriorly. While Hh acts through Ci for ptc regulation, it does not require Ci activity to control rho expression. Here, Tsh is required for rho expression downstream of Hh and possibly downstream of Ser. SIGNAL FACTORS TARGET GENES Anterior to En/Hh expressing cells Tsh and/orci act wg Fu ptc Smo Hh Smo ptc Posterior to En/Hh expressing cells Ci 155 Tsh? rho Ser

11 Cubitus interruptus-independent Hedgehog signalling 5519 hatching after ci RNA interference experiments showed similar phenotypes to ci 94 embryos (Ramirez-Weber et al., 2000). Furthermore if rho expression present in ci mutants was due to maternal production of Ci, one has to explain how two Ci targets would behave differently in absence of zygotic Ci contribution; wg expression disappeared, whereas rho was expressed in more cells. For all these reasons we are confident that Ci has no maternal contribution. Consequently, other factors are substituting for Ci activity in the transduction of Hh signalling. The ci 94 phenotype could be due to the fact that the ci 94 mutation disrupts both activator and repressive (Ci rep ) functions of Ci (Méthot and Basler, 1999) and that the resulting weak phenotype could thus be due in part to the loss of Ci rep activity. Indeed, in the wing disc loss of ci gene alleviates the repressive function of Ci rep on dpp transcription that is then transcribed at a basal level (Méthot and Basler, 1999). Although we show that Ci rep could repress wg transcription, its absence in the hh AC, ci 94 mutant embryo was not sufficient to induce an upregulation of wg transcription (Fig. 5). Hence the maintenance of wg expression until stage 11 in ci 94 is not due to the loss of Ci rep activity but is controlled by Hh input through Tsh activity (see below). Loss of ci induced an expansion of rho expression (Fig. 2N) instead of a reduction, as seen in a hh loss of function (Fig. 2O), showing that Ci is not involved in the activation of rho expression. The fact that rho disappears in tsh mutant embryos strongly suggests that the Tsh zinc-finger protein regulates rho expression or is at least necessary for instructing cells to respond to Hh for rho expression (Fig. 6 and see below). Nevertheless, one has to explain why rho expression is expanded in ci 94. Loss of Ci rep activity could be responsible for this effect. Indeed, overexpression of Ci rep in a ci null background or analyses of the ci Ce2 mutant, which ectopically expresses Ci rep, reveals a repressive effect of Ci rep on rho expression (Fig. 5). Therefore, Ci rep could be used as a gatekeeper in order to repress hh target genes tightly where they should not be expressed, and thus to overcome misregulation of key genes such as rho or wg. Nevertheless, these observations contradict previous analyses showing that Ci rep is not required for correct embryogenesis, since loss of ci function is rescued by a ci transgene lacking the Ci 75, repressor form of Ci (Méthot and Basler, 1999; von Mering and Basler, 1999). An alternative explanation can be gleaned from the fact that ci 94 cuticle phenotypes resemble those lacking Wg activity during the cell specification stage (Bejsovec and Martinez- Arias, 1991). Because it has been shown that Wg exerts a repressive role on rho expression (since absence of Wg activity promotes ectopic expression of rho; Sanson et al., 1999; Gritzan et al., 1999), rho expansion in ci 94 could be an indirect consequence of the late disappearance of wg expression during stage Redundant activities of Tsh and Ci for wg regulation Here we have shown that before stage 11, either Tsh or Ci is sufficient for wg regulation because only the loss of both gene activities results in a downregulation of wg, a situation comparable with that observed in hh mutants (Fig. 7). It is interesting to note that Ci seems to display differential requirements for wg maintenance and naked cuticle differentiation in the abdomen versus the thorax. While Ci is dispensable until stage 11 for wg expression and naked cuticle differentiation in the abdomen (Figs 3, 4), its presence in the thorax is required. We are currently studying this specific Ci function in the thorax. Both Ci and Tsh transcription factors, when overexpressed can induce ectopic wg expression (Gallet et al., 1998; Alexandre et al., 1996a). The two factors do not display the same features: Tsh has three atypical, widely spaced, zincfinger motifs, whereas Ci has conserved spacer regions between its five zinc fingers; the binding sites identified so far for these two proteins are different. It would be interesting to know whether Tsh can bind directly to the wg promoter and to identify its binding sites. It is also noteworthy that between stage 8 and 10 wg requires, in parallel to Hh, its own activity for the maintenance of its transcription (Yoffe et al., 1995). It has previously been shown that Tsh is a modulator of Wg signalling (Gallet et al., 1998, 1999). It becomes phosphorylated and accumulates at a higher level in the nucleus in Wg-receiving cells compared with cells lacking Wg signal. Hence, in the trunk Tsh could be employed both by Wg and Hh signalling in order to maintain wg transcription. The redundancy exhibited between Tsh and Ci for wg regulation changes after stage 10, as loss of Ci or Tsh results in the downregulation of wg transcripts (Figs 2H,K, 6C; Gallet et al., 1998). We do not know if this observation is the result of a cooperation between Tsh and Ci. At least one other gene, gooseberry, is required for the maintenance of wg transcription at this stage (Li and Noll, 1993), indicating that multiple inputs for the maintenance of wg expression are necessary for normal embryonic development. Bifurcations in the Hh pathway are responsible for regulating distinct target genes and cell fate choices Studies on the developing wing blade show that Ci transduces all Hh-delivered information. However, this and other studies on the Hh pathway support the idea that Ci is not always involved in Hh signalling showing that branchpoints are common for distinct Hh signalling steps. First, it has been shown that neither Ci nor Fused (Fu) are involved in the Hhdependent formation of Bolwig s organ in Drosophila (Suzuki and Saigo, 2000). Second, a Hh-responsive wg reporter gene with no Ci-binding sites does not require Ci activity for its regulation until stage 11 (Lessing and Nusse, 1998). Third, studies on the talpid 3 gene in chicken suggest that Gli proteins, the vertebrate homologues of Ci, regulate only a subset of Hh target genes, the others being regulated by an unidentified transcription factor (Lewis et al., 1999). Fourth, a Sonic Hedgehog response element on the COUP-TFII promoter binds to a factor distinct from Gli (Krishnan et al., 1997). Fifth, this study shows that Hh signalling does not require Ci activity to regulate rho. Although we favour the idea that Tsh regulates rho expression directly in response to Hh signal we cannot exclude the hypothesis that Tsh plays a more permissive role allowing Hh to regulate rho via another factor apart from Ci (Fig. 8). In conclusion, Hh requires at least two different transcription factors during Drosophila embryogenesis to regulate its multiple target genes and to instruct cells with precise behaviours. The transcription factors may act independently (e.g. Ci for ptc; Tsh for rho), cooperatively (e.g. Ci and Tsh