The shaping of the specific structures making up a complex organ is a key process in the development of multicellular organisms and requires extensive morphogenetic movements, specialization of different cell types and cell-cell signaling. The Drosophila ovary represents an excellent system to study such a dynamic cell reorganization process, because the transition from the embryonic homogenous gonadal mass to the adult organized ovary involves few specific cell types and occurs progressively, beginning in late larval stages and continuing through pupal stages. The Drosophila gonad forms at the end of embryogenesis, by coalescence between primordial germ cells and somatic gonadal precursors (Williamson and Lehmann, 1996). The embryonic ovary, thus, consists of a spherical mass of intermingled mesodermal and germline cells, which grows in size upon active proliferation of both cell types during early larval stages. The morphogenetic program leading to the formation of specific adult ovarian structures starts during the third larval instar with the formation in the apical region of a two-dimensional array of stacks of somatic cells called terminal filaments (Fig. 1A, pink; King, 1970; Sahut-Barnola et al., 1995). During pupariation, a second population of apical somatic cells (Fig. 1A,B, blue) migrate basaUy, first between terminal filaments, and thereafter through central and basal cells (King, 1970; Cohen et al, 2002). As they migrate, these cells secrete a thick basement membrane and thereby separate the central population of cells into approximately 15-20 individual units, called ovarioles, composed of both germline and somatic stem cells and their deriving germline cysts and somatic prefollicular cells, respectively (Fig. 1B,C, green and yellow cells, respectively). Germline stem cells are maintained-in the apical-most part of the ovariole, called the germarium, in direct contact with the somatic terminal filament (Fig. ID, bracket). Germline stem cell division is asymmetric, generating both a daughter stem cell that remains in contact with the basal-most cells of the terminal filament and a differentiated daughter cell, or cystoblast, that looses contact with the terminal filament. Each cystoblast then undergoes four rounds of mitosis with incomplete cytokinesis and produces a cyst of 16 interconnected germline cells (1 oocyte: 15 nurse cells). Each germline cyst then becomes enveloped individually by somatic prefollicular cells in the central region of the germarium to form ovarian follicles (or egg chambers) that will successively bud off from the posterior part of the germarium (Fig. ID, asterisks). In addition, stalks of basal somatic cells (termed basal stalks) assemble in the continuity of the newly-forming ovarioles (Fig. 1B,C, white) and will connect the most basal follicle of each growing ovariole to the oviduct (Fig. ID, arrow; rung, 1970; Spradling, 1993).

To date, few data have accumulated on the nature of genes required for Drosophila ovarian morphogenesis. Convergence and extension movements involved in terminal filament formation have been shown to depend on the nuclear protein Brie a brae, and on the actin-depolymerizing factor Co-filin (Godt and Laski, 1995; Chen et al., 2001). In addition, Cohen et al. have shown that Dwnt4 is required within apical somatic cells to control dynamic accumulation of the focal adhesion kinase (FAK) protein in these cells, thereby facilitating their basal migration during pupal ovarian orga- R nogenesis (Cohen et al., 2002). Little is known, however, about the molecuar mechanisms allowing communication between different cell types and p coordination between their developmental programs.

We have analyzed the ovarian function of the fused (fu) gene, which encodes a serine/threonine kinase best characterized as a positive effector of the Hedgehog (Hh) signal transduction pathway in Drosophila for embryonic and imaginal disc development (Limbourg-Bouchon et al., 1991; Sanchez-Herrero et al., 1996; Alves et al., 1998). In these systems, the Fu protein is part of a cytoplasmic multi-protein complex composed of at least 3 other proteins: the kinesin-related, Costal-2 (Cos2), Suppressor-of-fused (Sufu), and the transcription factor Cubitus interruptus (Ci). Genetic and molecular studies have led to a model in which transduction of Hh signaling is triggered by binding of Hh to its receptor Patched (Ptc) and subsequent release of the inhibition exerted by Ptc on the activity of the transmembrane Smoothened (Smo) protein. This process results in modifications in the properties of the regulatory cytoplasmic complex and, finally, in the c activation of the transcription factor Ci and subsequent transcription of target genes. Members of the Hh family have been shown to control several aspects of growth, patterning, and morphogenesis during development of 3 vertebrates as well as invertebrates ; (Mullor et al., 2002; Lum and Beachy,] 2003). In the imaginal wing disc system in Drosophila, for example, hh controls cell segregation-based definition of anterior and posterior compartments, imaginal disc growth, and definition of different domains, including . vein and intervein domains (Strigini and Cohen, 1997; Dahmann and Basler, 2000; Vervoort, 2000).

Our present study shows that fused is also required for ovarian morphogenesis during pupal stages, in particular for correct packaging of germ cell precursors within individual somatic sheaths to form ovarioles. In addition, we provide evidence that fused function during these stages participates in transduction of a specific somatic Hedgehog signal controlling Drosophila ovarian organogenesis.

The results of our present study indicate that Fused-dependent Hh signaling regulates interactions between somatic and germline cells for Dro* ^ sophila ovary morphogenesis during pupal stages. We have analyzed ovary development in females homozygous for/u and hh strong hypomorphic mutations and find a similar phenotype. For both genes, ovariole formation with respect to the somatic cell population is not perturbed in mutant females. However, a significant number of tfe>e ovarioles ar,e devoid of germ cells. More detailed analysis of fu mutant females indicates that this cell population appears almost normal in number and distribution during late larval stages but becomes progressively disorganized throughout pupal stages. Instead of remaining as a relatively compact group in the central part of the ovary immediately basal to the terminal filaments, germ cells in fu mutant pupal ovaries become progressively separated from the terminal filaments and organize themselves either as several small groups or as larger clusters in the most basal part of the ovary. We show that, by late pupal stages, a large proportion of these cells are not incorporated into growing ovarioles.

Taken together, our data indicate that, in early pupal oyaries, Hh signal emanating from specialized somatic cells is transduced by means of the Fu kinase, possibly in germ cells, leading to proper germ cell inclusion in somatic cell sheaths for ovariole formation. Indeed, (1) hh is expressed specifically in a discrete population of somatic cells (basal terminal filament and cap cells) immediately adjacent to germ cells during pupal ovariole development, (2) fu hypomorphic loss-of-function phenotypes in the pupal ovary are mimicked by hh hypomorphic loss-of-function mutations, and (3) induction of fu germline clones is associated With production of aga-metic ovarioles. It is possible, therefore, that Hh signal transduction in germ cells during pupal stages regulates the expression or activity of one (or several) protein(s) involved in cell adhesion or cell affinity. In the absence of hh in somatic cells or fused in germ cells, germ cells would loose appropriate interactions between themselves or with neighboring somatic cells (terminal filament, cap cells, or somatic germarial precursors; see Fig. 1), which would lead to basal sorting out of germ cells. Importantly, expression of DE-cadherin has been shown to be important for recruitment and maintenance of germ cells in the somatic cell niche composed of basal terminal filament and cap cells during pupal and adult stages (Song et al., 2002). However, we examined the expression of DE-cadherin in fu mutant pupal ovaries and did not observe any obvious difference in abundance or subcellular localization compared to wild-type (F.B. and A.M.P., unpublished results).

Of interest, a role in regulation of cell surface properties and sorting out has already been proposed for Hh in several other systems in Drosophila. In imaginal discs, hh is expressed specifically in posterior compartment cells and is required for cell segregation-based establishment of the anteroposterior compartment boundary. Anterior compartment cells in which smo function (and, hence, the ability to transduce Hh signal) has been suppressed no longer segregate with anterior cells, but instead tend to preferentially associate with posterior cells (Blair and Ralston, 1997; Rodriguez and Basler, 1997). It has been proposed that Hh signaling may modulate the segregation properties of responding cells by means of a transcriptional control of one (or several) cell adhesion molecule(s) (Dahmann and Basler, 2000), which is supported by experiments showing that a difference in the abundance of a single cell adhesion molecule (DE-cadherin) between two cell populations leads to their sorting out (Steinberg and Takeichi, 1994; Dahmann and Basler, 1999, 2000). In addition, specific expression of another cadherin, Cad99C, at the anterior-posterior compartment border in wing imaginal discs has been shown recently to be regulated by Hh signaling (Schlichting et al., 2005).

The Hh protein also controls cell-cell interactions involved in gonad morphogenesis during embryonic development. Previous work has shown that hh is expressed in somatic gonadal precursor cells and controls migration of primordial germline cells toward these cells and coalescence between these two cell types to form the primitive gonad (Deshpande et al., 2001). In this system, cell autonomous components of the Hh pathway are required maternally because germline cell migration defects are observed after induction of smo or fu mutant germline clones and can be rescued by the paternal wild-type copies of the corresponding gene. In addition, such transducers are probably required within germ cells because this paternal rescue is observed only in late embryos (in stage 15 embryos), that is, far later than the beginning of somatic zygotic transcription, but just after the beginning of germline zygotic transcription (Deshpande et al., 2001). The specific targets of Hh signal transduction in embryonic germ cells for gonad morphogenesis have yet to be determined.

In contrast to what is observed for these systems, however, we have shown that the function of the key positive effector of Hh signal transduction, smo, may not be required for pupal ovarian morphogenesis. Indeed, induction of smo mutant germline clones, using an amorphic allele, is not associated with alteration in early ovariole formation or production of agametic ovarioles. Variations in the nature of the components used to transduce Hh signaling have already been described in different systems, including germline cells of the adult ovary (Therond et al., 1999; Suzuki and Saigo, 2000; Vied and Horabin, 2001). Indeed, variations in Hh levels have been shown to affect the dynamic distribution of the Sex lethal (Sxl) protein within anterior germ cells of the adult germarium, in particular the transition between anterior cytoplasmic accumulation in germline stem cells and subsequent protein degradation in differentiating cystoblasts (Vied and Horabin, 2001). Strikingly, regulation of Sxl distribution requires fu and cos2 activity but not that of smo, within germ cells (Vied and Horabin, 2001). This finding suggests, first, that germline cells directly respond to Hh signal, and, second, that cell-specific transducers may be used to transduce Hh signaling in different systems.

Although no alternative to Smo has been clearly characterized, other transmembrane proteins have been recovered by recent RNAi and two-hybrid screens aimed at identifying new components of Hh signaling (Lum et al., 2003; Formstecher et al., 2005). In addition, Smo belongs to the Fz subfamily of G protein-coupled receptors (Alcedo et al., 1996; Hooper, 2003), and a two-hybrid interaction between Ptc and Fz2 has been reported (Formstecher et al., 2005). Of interest, another component of classic Hh signal transduction, the transcription factor Ci has been reported to be absent from germline cells in the adult ovary (Forbes et al., 1996). In addition, Ci-independent Hh signal transduction has been reported in both ovarian germ cells (Vied and Horabin, 2001) and for Boldwig organ formation (Suzuki and Saigo, 2000). Like for Smo, it is possible that alternatives for Ci in Hh signal transduction exist, because other transcription factors have been recovered by means of specific Hh signaling RNAi and two-hybrid screens (Lum et al.,. 2003; Formstecher et al., 2005). Therefore, our results, along with those of others, indicate that the Drosophila germline provides a new system to study variations in the nature of components used to transduce Hh signaling.