bullwinkle and shark regulate dorsal-appendage morphogenesis in Drosophila oogenesis

The folding and remodeling of epithelia into more complex structures is a recurrent phenomenon in metazoan development. Intercellular interactions are important regulatory components of these processes. Adjacent cells typically provide cues that direct morphogenesis or establish an extracellular milieu permissive for cell movements.

In Drosophila melanogaster, remodeling epithelia can interact with an adjacent epithelium. Two well-studied examples include the migration of the embryonic dorsal epithelium over the amnioserosa (reviewed by Jacinto et al., 2002; Knust, 1997), and eversion of leg and wing primordia relative to the peripodial tissue that bounds the imaginal discs (reviewed by Fristrom,1993). These cell layers actively regulate the patterning and movements of neighboring epithelia. Ablation of the peripodial membrane results in growth and patterning defects in the eye and wing discs(Gibson and Schubiger, 2000). In the embryo, the amnioserosa contributes signals(Harden et al., 2002; Reed et al., 2001; Stronach and Perrimon, 2001)and mechanical force (Kiehart et al.,2000) to dorsal closure. During germband retraction, the amnioserosa also signals to (Lamka and Lipshitz, 1999) and extends lamellipodia-like structures towards(Schock and Perrimon, 2002)the retracting germband cells. We elaborate on a novel extracellular pathway defined by bullwinkle (bwk)(Rittenhouse and Berg, 1995)that is essential for proper tubulogenesis of the follicular epithelium during synthesis of the dorsal appendages (DAs), specialized respiratory structures of the eggshell. Additionally, we demonstrate that an adjacent squamous cell layer acts as a substrate for the migrating epithelium and expresses factors required for this morphogenetic process.

DA formation occurs within the context of the Drosophila egg chamber, which consists of ∼650 somatically derived follicle cells(Margolis and Spradling, 1995)surrounding a germline cyst composed of one oocyte and 15 nurse cells(Spradling, 1993). The germ cells are interconnected via cytoplasmic bridges called ring canals, which provide access for the transfer of nurse-cell material into the developing oocyte. At stage 11, the nurse cells transport most of their cytoplasm into the oocyte, and then undergo programmed cell-death(Mahajan-Miklos and Cooley,1994). DA morphogenesis begins at stage 11, coincident with nurse-cell apoptosis.

During DA formation, the somatic layer consists of two major populations with distinctive morphologies, the stretch cells and columnar cells. At the anterior, ∼50 squamous stretch cells cover the nurse cells. These cells provide signals that pattern the anterior eggshell-forming cells and ensure proper nurse-cell cytoplasmic dumping. The columnar cells overlie the oocyte at the posterior and secrete the layers and specialized structures of the eggshell (reviewed by Waring,2000). The anterior-most columnar cells (the centripetal cells)migrate inwards, closing off the anterior end of the oocyte while synthesizing the operculum and micropyle. In addition, two subpopulations of ∼65 dorsoanterior follicle cells form the two dorsal appendages through a complex reshaping and reorganization of a flat epithelium into three-dimensional tubes(Dorman et al., 2004).

These DA-forming cells apically constrict and evert outwards, changing from a flat layer into tubular structures that extend anteriorly. Secretion of chorion proteins into the tube lumens creates the appendages(Fig. 1A). This process occurs during the final stages of oogenesis, downstream of the events that pattern the eggshell and embryonic axes.

Although much is known about the induction and refinement of follicle-cell patterning (Peri and Roth,2000; Schüpbach,1987; Twombly et al.,1996; Wasserman and Freeman,1998), little is known about the factors that govern the cellular movements. One pathway that contributes to the morphogenesis is the Jun-kinase(JNK) pathway. The Drosophila Jun and Fos transcription factors are expressed highly in the stretch cells and in an anterior subset of the two DA-forming cell populations. Loss of JNK-pathway function results in two short paddleless DAs and defective nurse-cell cytoplasmic transport(Dequier et al., 2001; Dobens et al., 2001; Suzanne et al., 2001).

The DA-forming cells require additional extracellular cues for normal tubulogenesis. Mosaic analyses demonstrate that bwk is required in the germline to regulate formation of the dorsal appendages(Rittenhouse and Berg, 1995). bwk encodes several SOX/TCF transcription factors with pleiotropic functions (C.A.B., M. Terayama, D. H. Tran and K. Rittenhouse, unpublished),regulating dorsal follicle-cell migration, anteroposterior (AP) patterning in the embryo, and transport of nurse-cell cytoplasm into the oocyte. In bwk mutants, the DA-forming cells not only fail to migrate anteriorly, but instead extend much more laterally(Dorman et al., 2004), as indicated by the wide DA paddle (Fig. 1B).

To elucidate the role of bwk in DA formation, we set out to identify other components of this germline-to-soma signaling pathway. We screened second-chromosome deficiencies for regions that genetically interact with bwk. Tests of candidate mutations identified shark as a strong Enhancer of bwk. shark encodes an SH2-ankyrin-repeat,tyrosine-kinase protein (Ferrante et al.,1995) that functions upstream of the JNK pathway during dorsal closure of the embryo (Fernandez et al.,2000).

We show here that shark acts downstream of bwk in the squamous stretch cells and mediates the regulation of DA formation by bwk. Furthermore, detailed cellular analyses with stretch-cell markers show that the stretch cells provide a substrate for the DA-forming cells and appear morphogenetically active.

We used Canton S as the wild-type strain. We employed the following mutant Drosophila melanogaster stocks: bwk¹⁵¹ and bwk⁸⁴⁸² P-element insertions; bwk^CT,an EMS mutation (C. Trent and C.A.B., unpublished); bwk^R4,an imprecise excision of the 8482 insertion(Rittenhouse and Berg, 1995); shark¹, an EMS mutation causing a premature stop in the second ankyrin repeat, and UAS-Shark⁺(Fernandez et al., 2000); shark² (an unmapped lethal allele, gift from Rahul Warrior); dpp¹⁰⁶³⁸(Twombly et al., 1996); UAS-bsk⁺ (Boutros et al., 1998); FRT(42B) Ubi-GFP (kindly provided by S. Luschnig and C. Nüsslein-Volhard); Bic-C^AA4 and Bic-C^YC33 (Mahone et al., 1995); Bic-C^WC45(Schüpbach and Wieschaus,1991); GAL4^GR1, UASFLP (kindly provided by T. Schüpbach); GAL4^c415(Gustafson and Boulianne,1996); GAL4^A90(Manseau et al., 1997); GAL4^55B (Brand and Perrimon, 1994); and GAL4^T155 and GAL4^CY2 (Queenan et al., 1997).

The Bloomington stock center provided the second-chromosome deficiency kit, bsk¹ and bsk^J27, UAS-Src42.CA(Tateno et al., 2000), and FRT 42B tubP-GAL80 (Lee et al.,2000).

Second-chromosome deficiency stocks were crossed individually to bwk strains in an F2 screen for dominant modifiers of the bwk DA phenotype. Ten deficiency-bearing females were compared with ten siblings lacking the deficiency in two bwk transheterozygous backgrounds, bwk¹⁵¹/bwk⁸⁴⁸² and bwk¹⁵¹/bwk^CT. Eggshell phenotypes were counted daily for 3 days, without knowing the genotype until the end of the counts.

We developed a numerical scoring system to facilitate identification of dominant modifiers. The bwk mutants used in the screen produced a range of phenotypes, which we sorted into four categories: wild type, short thin, short broad and very short broad. Using a weighted average in which wild type=4, short thin=3, short broad=2 and very short broad=1, we derived a score from 4 to 1 as a composite of the phenotypic classes. Eggs from bwk^151/8482 females averaged a score of 1.43, while eggs from bwk^151/CT females averaged 2.96. The standard deviation for both allelic combinations was ∼0.3. We scored eggs produced by Df/+; bwk/bwk females and compared these values to bwk/bwk siblings without the Df. Scores that differed by more than one standard deviation (0.3) were considered evidence of a significant interaction, while differences greater than 0.6 suggested strong interactions.

We subcloned the Bgl_II insert from a pCaSpeR-hs-sharkplasmid (Fernandez et al.,2000) into pBluescript-SK and made digoxigenin-labeled RNA probes using the Roche DIG-labeling kit. We followed a modified in situ protocol (Tautz and Pfeifle,1989; Wasserman and Freeman,1998).

We used the following primary antibodies: polyclonal rabbit anti-GFP(Clonetech) and monoclonal mouse anti-GFP (Molecular Probes) both at 1/100;rat anti-Fos (Riese et al.,1997) and rabbit anti-c-Jun(Chen et al., 2002) both at 1/100; mouse monoclonal anti-β-gal (Sigma) at 1/500. To detect the primary antibodies we employed secondary antibodies conjugated to Alexafluor488 and Alexafluor568 at 1/500 (Molecular Probes). We followed a modified immunocytochemistry protocol(French et al., 2003).

Clones were induced using the FLP/FRT method(Chou and Perrimon, 1992; Xu and Rubin, 1993). Heat-shock-driven FLP produced both germline and follicle-cell clones(Golic and Lindquist, 1989). GAL4^GR1 (a gift from Trudi Schüpbach), expressed in follicle-cell stem cells and later-stage egg chambers, was used to induce follicle-cell clones. Ubiquitin-GFP was used to mark the clones(Davis et al., 1995).

Positively marked clones were made with a modification of the MARCM method(Lee et al., 2000). Females of genotype hsFLP/+; FRT shark¹/FRT tubP-GAL80; UAS-GFP^S65T/GAL4^T155were heat-shocked for 2 hours, dissected 3-4 days post-heat-shock, and stained with anti-GFP.

UAS-shark⁺, UAS-bsk⁺ and UAS-Src42A.CA⁺ were expressed using GAL4^c415, GAL4^55B or GAL4^CY2 in a bwk^151/8482 background. Eggs laid by 10 females per genotype were examined over 3 days on egg plates. Control sibling flies lacking the GAL4 or UAS elements were also tested.

To identify components of the bwk germline-to-soma signaling pathway, we undertook a deficiency screen looking for dominant genetic interactions with bwk. We examined the effect of heterozygous deficiencies upon the DA morphology of bwk eggs. This approach allowed the identification of genes sensitive to levels of bwkactivity, including those with pleiotropic functions.

We employed two allelic combinations to facilitate isolation of both enhancers and suppressors: bwk^151/8482, a strong loss-of-function combination (Fig. 1B), and bwk^151/CT, a moderate loss-of-function combination. Both combinations produced a range of phenotypes. bwk^151/8482 eggs have mainly short, broad dorsal appendages, facilitating the identification of suppressing mutations. bwk^151/CT eggs manifest an array of DA structures from short, broad to long, tubular appendages. As bwk^151/CTphenotypic profile exhibited a bias towards wild-type length DAs, this combination facilitated the isolation of enhancing interactions.

The Bloomington Stock Center maintains a large collection of deletions that uncover 60-70% of the Drosophila genome. We screened 39 deficiencies that uncover ∼78% of the second chromosome, as determined by polytene-segment coverage (Table 1). We scored F₂ progeny to examine the effect of these deletions on the DA phenotypes of the two bwk-allele combinations. Using a stringent scoring method (see Materials and methods), we identified four deletions that exhibited a strong interaction, two enhancers and two suppressors. Nine deletions exhibited a moderate interaction and seven interacted weakly. We focused our initial efforts on the four strong modifiers.

The Df(2L)J136-H52 and In(2R)pk78s chromosomes(Fig. 1C) strongly suppressed the bwk DA phenotype, while Df(2L)r10 cn and Df(2R)Jp8(Fig. 1D) strongly enhanced it. Extensive analyses using overlapping and smaller deficiencies failed to recapitulate the suppression associated with both Df(2L)J136-H52 and In(2R)pk78s (data not shown). Similar studies did confirm the two strong enhancing interactions and defined the interacting segments as 35D1-35E2 and 52F5-53A. Tests with available mutations in these regions indicated that heterozygous loss of function of either Bicaudal-C(35E2) or shark (52F) enhanced the bwk DA phenotype (data not shown). We focused our efforts on shark due to a previously reported function for this gene in regulating epithelial morphogenesis(Fernandez et al., 2000).

A mutation resulting in a premature stop codon in the shark gene(Ferrante et al., 1995) showed strong enhancement of the bwk phenotype, similar to the original deficiency(data not shown). Germline clones of shark¹, however,failed to produce a detectable phenotype in oogenesis(Fernandez et al., 2000),leading us to investigate a possible somatic function.

First, we examined expression of shark in oogenesis and noticed an unusual pattern (Fig. 2). shark transcript was present in the germline and somatic cells beginning in region 2 of the germarium(Fig. 2A). At the time of dorsal-appendage formation, egg chambers showed a pronounced pattern of discrete spots and tracks near the periphery of the nurse cells(Fig. 2C). These foci were often associated with stretch-cell nuclei (arrowheads, Fig. 2C′), suggesting that shark expression occurs in the thin (<1 μm) stretch-cell layer overlying the nurse cells. After stage 10, during rapid nurse-cell cytoplasmic transport, shark RNA levels increased dramatically in the nurse cells and the unusual foci were no longer readily visible (data not shown).

bwk mutants exhibited an altered pattern of shark mRNA localization (Fig. 2B,D,D′). Although germline staining resembled wild type,the discrete foci were not evident at stage 10 (compare Fig. 2C′ with 2D′). These results suggest that the bwk pathway normally modulates shark expression in the stretch cells, implicating this layer in DA formation. Previous studies, however, had not described a role for stretch cells in this process. We therefore examined the behavior and morphology of the stretch cells during DA morphogenesis.

We examined egg chambers that expressed both stretch-cell-specific and columnar-cell-specific markers. To label stretch cells, we employed the GAL4/UAS system (Brand and Perrimon,1993), driving UAS-GFP^S65T with GAL4^c415 or GAL4^A90(Gustafson and Boulianne,1996; Manseau et al.,1997). c415 drives reporter expression in the stretch cells while A90 labels both the stretch cells and the border cells. To mark the DA-forming cells, we used the P[lacZ; ry⁺]enhancer trap line PZ05650, which expresses highly in the centripetally migrating columnar cells and in the two populations of dorsoanterior follicle cells that synthesize the dorsal appendages(Rittenhouse and Berg,1995).

At stage 10, the stretch cells covered the exterior of the nurse cells(Fig. 3A,B). The thinness of the layer meant that the cells were most visible at the junctions of nurse cells (Fig. 3A′, blue arrowhead) and in regions surrounding the nuclei of the stretch cells(Fig. 3A′, blue triangle). This morphology of the stretch cells, a significant thickening of the layer at discrete locations, could explain the shark RNA foci:localized cell thickening could cause ubiquitously expressed RNA to appear localized. Alternatively, the shark foci could represent actual localization of RNA within the stretch cells.

During centripetal migration at stage 10B, the posterior-most stretch cells also migrated inwards (asterisk, Fig. 3A) accompanying centripetally migrating columnar cells (red nuclei, Fig. 3B). These panels show two egg chambers: the follicle cells in the upper egg chamber have nearly completed centripetal migration, while those in the egg chamber on the lower right, a partial view, have just initiated this process (arrow, Fig. 3B).

At stage 12, the nurse cells were much smaller due to transport of their cytoplasm into the oocyte (compare the nurse cells, labeled NC, in Fig. 3C with those in 3A),while the stretch cell layer had thickened. At this time, the stretch cells exhibited three interesting behaviors. First, in contrast to stage 10, the stretch cells enveloped all nurse cells(Fig. 3C). This envelopment could be due to an active movement or a byproduct of the nurse-cell shrinkage. Second, the migrating DA-forming cells moved over the stretch cells (green arrowhead, Fig. 3C). Finally,the stretch cells occasionally extended small cellular projections towards the DA-forming cells (arrow, Fig. 3E,F). By stage 13, the stretch cells resided between and underneath the two DA cell populations, which have reached the anterior end of the egg (data not shown).

These studies revealed that the stretch cells are a substrate for the migrating DA-forming cells. This result contradicts a previous hypothesis(King and Koch, 1963), who proposed that the DA-forming cells migrated between the stretch and nurse cells in an invasive manner. Thus, the stretch cells form an intervening layer between the germ cells and the migrating DA-forming cells. The stretch cells could express factors that mediate the movement of the DA cells across this layer; shark may be such a factor, as suggested by the genetic interaction and expression data. Because the known shark alleles are lethal, we used mosaic analyses to examine the function of shark in oogenesis.

We induced clones with the shark¹(Fernandez et al., 2000) and shark² (R. Warrior, unpublished) alleles using the FLP/FRT system (Xu and Rubin, 1993). We expressed FLP using either a heat-shock promoter-FLPase transgene or a follicle-cell-specific GAL4 transgene, GR1, driving UAS-FLP. GAL4^GR1 is expressed in the follicle cells continuously from the time of stem-cell division to stage 14 (T. Schüpbach, personal communication). We marked the clonal cells in two fashions: negatively, such that loss of Ubiquitin-GFP (Davis et al., 1995) defined homozygous shark cells, or positively, with a modification of the MARCM/GAL80 method(Lee et al., 2000). In positively marked clones, only homozygous shark cells expressed UAS-GFP^S65T (Amrein and Axel, 1997).

shark clones affected the morphology of the dorsal appendages and the structure of the DA chorion (Fig. 4). In some eggs (Fig. 4A,E), the DA material appeared vacuolated, with gaps interposed with a skeletal network. This phenotype resembled defects seen in mutants affecting eggshell structure (reviewed by Waring, 2000). In contrast to chorion mutants, however, shark mosaic eggs with clones in the main body had no obvious structural defect (data not shown).

Other eggs displayed shortened DAs with normal chorion, suggesting a defect in the anterior migration of the DA-forming cells(Fig. 4B). The shortness of the dorsal appendages varied from the egg in Fig. 4B to the egg in Fig. 4C.

The chorion and short-DA defects were not mutually exclusive; in fact, most chorion-defective appendages were also shortened. These defects were associated with clones in the anterior of the egg; when the entire anterior was clonal, both DAs were short and vacuolated (data not shown).

To establish the precise relationship between clone position and DA defect,we examined small clones and their effect on DA morphology. Scoring small clones by the absence of GFP proved difficult; however, once the DA cells had migrated onto the stretch cells. We used positively marked clones for clarity. In one representative clone, most GFP-positive clonal cells lay between the two DAs (Fig. 4C-G). By position, many of these marked cells should be stretch cells. Several of the GFP-positive cells were also closely associated with the chorion-defective DA,and likely label the DA-forming cells responsible for secreting the appendage chorion in E.

The frequency of these defects was low(Fig. 4H). Clone frequency, as measured by the number of egg chambers with at least one clone, for post-mitotic stages varied from 14.2 to 17.4%. No defects were seen with a wild-type FRT chromosome, while an FRT shark² chromosome recapitulated the shark¹ results. Most clones were made with the shark¹ allele, where 7.5% of all stage-14 egg chambers showed a DA chorion defect and 1.9% had short DAs with normal DA chorion. We attribute this low frequency to several factors: regional specificity within the egg chamber, large clone-size requirement and incomplete penetrance for the short DA defect.

These studies showed that when a significant fraction of DA-cells was clonal, a chorion defect was seen. The large-clone-size requirement implied that neighboring shark⁺ cells could provide cell non-autonomous function for the homozygous shark cells. This effect was limited to the appendage associated with the clone; one DA could be affected while the other was normal.

The short-DA defects were, in turn, associated with large clones encompassing the stretch cells. These short-DA defects exhibited a variety of morphologies, from short and thin to short and broad like bwk mutant DAs. Furthermore, the short-DA defect was not fully penetrant; large clones in the stretch cells could result in mild defects.

To determine the relative contributions of these shark functions in regards to bwk, we asked whether tissue-specific expression of a wild-type shark⁺ transgene could ameliorate the DA defects of a bwk mutant. As the stretch cells do not express or secrete chorion (Margaritis et al.,1980), we could distinguish between the role of shark in chorion production versus DA migration.

We postulated that bwk functioned to regulate sharkexpression and/or activity in the stretch cells. Restoration of sharkexpression in the stretch cells could compensate for loss of bwk, if shark expression in these cells was a key downstream factor. To test this hypothesis, we expressed UAS-shark⁺(Fernandez et al., 2000) in a bwk background using GAL4^c415, which expresses specifically in the stretch cells (Manseau et al., 1997), GAL4^55B, which expresses in both the stretch cells and DA cells (Brand and Perrimon, 1994), and GAL4^CY2, which expresses in all follicle cells (Queenan et al., 1997).

Expression of a wild-type UAS-shark⁺ with the c415 and 55B drivers suppressed the bwk-mutant DA phenotype substantially (Fig. 5B; Table 2A),while CY2-driven expression had little effect(Table 2A). With stretch-cell-specific expression of UAS-shark, we generated a significant shift towards longer and more tubular DAs, a more wild-type-like phenotype. We quantified the suppression using a weighted average of four classes of DA phenotypes, where a difference of greater than or equal to 0.3 between the experimental and control scores indicated a significant interaction (see Materials and methods). The c415-driven suppression was equivalent to the strongest suppression observed in the deficiency-interaction screen. This result indicated that sharkexpression in the stretch cells is a key factor downstream of bwk. Additionally, the chorion function of shark was not crucial to the bwk DA phenotype. To explore shark function in the stretch cells, we assayed candidate factors that might act with shark in this tissue.

As Shark acts in the Jun-kinase pathway during embryonic dorsal closure(Fernandez et al., 2000) and JNK-pathway function is required for DA morphogenesis(Dequier et al., 2001; Dobens et al., 2001; Suzanne et al., 2001), we asked whether the Jun-kinase pathway is a component of the bwk/sharkpathway in oogenesis. We tested whether gain or loss of Jun kinase(basket) affected bwk phenotypes(Table 2B). Basket activates both Jun and Fos in Drosophila(Ciapponi et al., 2001; Riesgo-Escovar et al.,1996).

Expression of UAS-bsk⁺ led to a reduction in the number of eggs laid by bwk mothers but the morphology of the bwkDAs was not modified. Heterozygosity for two strong loss-of-function alleles(basket¹ and basket^J27) also failed to interact with bwk (Table 2B). Additionally, expression and localization of both Jun and Fos were normal in bwk mutants or shark clones (Jun, Fig. 5C,D; Fos, J. Dorman and C.A.B., unpublished).

Another likely candidate gene expressed in the stretch cells is dpp, partial overexpression of which can result in shortened and fringed DAs (Twombly et al.,1996). Expression of both dpp RNA and a dpp-lacZenhancer trap were normal in bwk mutants (data not shown). Heterozygosity of dpp and/or its receptor failed to modify the strong bwk^151/8482 combination(Table 2C), although modest interactions occurred with the moderate bwk^151/CTcombination (data not shown).

In mammalian cells, proteins related to Shark act alongside Src kinases in mediating immunoreceptor signaling (Latour and Veillette, 2001). We tested the Drosophila Src42Agene for interaction with bwk. Src42A, like shark, functions upstream of the JNK pathway in dorsal closure(Tateno et al., 2000). Interestingly, Src42A mutations enhanced the moderate bwk^151/CT DA phenotype (Table 2D). Expression of transgenic UAS-Src42A.CA, an activated form (Tateno et al., 2000),with the stretch-cell-specific GAL4^c415 suppressed the strong bwk DA phenotype (Table 2D). Compared with the UAS-shark suppression, the UAS-Src42A.CA suppression was weaker but still produced a significant shift towards longer DAs.

Previous analyses of bullwinkle revealed the existence of a germline pathway required for DA morphogenesis(Rittenhouse and Berg, 1995). We screened the second-chromosome deficiency kit for modifiers of bwk; follow-up studies identified two Enhancers of bwkencoded by the Bic-C and shark genes. Although Bwk probably acts as a transcription factor, it does not directly regulate either gene. Preliminary studies (not shown) suggest that the interaction between Bic-C and bwk is complex; here we show that sharkacts downstream of bwk, mediating the signal from germline to DA-forming cells.

shark encodes a distinctive multidomain protein that regulates the movements of epithelial cells in the dorsal embryonic epidermis(Fernandez et al., 2000). This non-receptor kinase is conserved, with homologs in Hydra(Chan et al., 1994) and sponge(Suga et al., 1999). The mammalian counterparts contain homologous SH2 and tyrosine-kinase domains but lack the ankyrin repeats (Chan et al.,1991; Taniguchi et al.,1991). These mammalian proteins, Zap70 and Syk, are recruited to immunoreceptor complexes upon ligand binding and regulate immune-cell activation and differentiation, functioning alongside Src kinases (reviewed by Chu et al., 1998). In T-cells,Zap70 also mediates signaling downstream of integrin-receptor complexes that feature in T-cell motility (Bearz et al.,1999; Soede et al.,1998).

We show that shark has two functions in oogenesis and that a bwk/shark pathway could involve the Shark and Src42A kinases in an evolutionarily conserved version of the mammalian signaling pathway.

Mosaic analyses with loss-of-function shark alleles established two somatic functions in DA formation (Fig. 6). First, shark is required in the DA cells for proper DA-chorion deposition, a complex process regulated at many levels (reviewed by Waring, 2000). Mutations that disrupt chorion-gene amplification or chorion-protein synthesis result in thin, collapsed DAs and main-body eggshell(Bauer and Waring, 1987; Landis et al., 1997; Mohler and Carroll, 1984; Nilson and Schüpbach,1998).

Unlike those mutations, loss of shark in the main-body follicle cells does not cause defects in follicular imprints, alter the appearance of the eggshell under darkfield optics, or produce thin chorion and collapsed eggs. Although our methods may miss subtle defects in main-body chorion, shark may play a DA-cell-specific role in the production/formation of chorion. Although regulatory sequences and a putative binding protein drive specific spatial expression of chorion-reporter constructs(Tolias and Kafatos, 1990; Tolias et al., 1993), no reported mutants disrupt DA-specific chorion expression.

The second function of shark lies in the stretch cells and affects the migration of the DA cells. Large stretch-cell clones resulted in shortened DAs that varied in their morphology and penetrance. This variability could result from residual activity of these mutant alleles (see Materials and methods), non-cell autonomy, or functional redundancy. Although no Shark paralogs are encoded in the genome (Adams et al., 2000), several non-receptor tyrosine kinases share homology in the SH2 and kinase domains, including Src42A.

In addition, stretch-cell expression of shark strongly suppressed the bwk-mutant DA phenotype, in concurrence with a direct role for bwk in regulating shark expression in this tissue. These results indicate that shark is key in regulating DA migration downstream of bwk. Full rescue was not likely achieved because of insufficient expression levels, the need to localize shark RNA, or the existence of shark-independent branches downstream of bwk.

These data suggest a model in which BWK regulates factors in the germline that are required for proper shark expression in the stretch cells. Shark then regulates the activity of targets required for DA-cell movement across the stretch-cell layer (Fig. 6A). Another factor that could be regulated by bwk is the Src42A kinase, which behaves similarly to shark(Fig. 6B). Loss of Src42A enhances bwk mutants, while stretch-cell expression of activated Src42A suppresses. Mammalian homologs of Shark function together with Src kinases, suggesting a conserved signaling cascade.

Two other stretch-cell signaling pathways, JNK and DPP, regulate DA morphogenesis. Tests with bwk and shark, however, failed to reveal strong or definitive interactions. Loss of JNK activity in oogenesis results in shortened and paddleless DAs, yet expression of UAS-basket⁺ and reduction of bsk dose did not alter the morphology of bwk eggshells. Furthermore, expression of the AP-1 components was unaffected in bwk mutants and sharkclones. These data support the hypothesis that the bwk/sharkpathway does not primarily act through JNK signaling.

Moderate overexpression of dpp and loss of the type I receptors, tkv and sax, can lead to shortened and somewhat broadened DAs, resembling bwk mutants(Twombly et al., 1996). The expression of dpp RNA and a dpp enhancer trap, however, were unaffected in bwk mutants. Both hypomorphic dpp alleles and loss of type I receptors failed to interact with a strong bwk mutant. Our data suggest that bwk does not directly regulate dppexpression or activity but rather may modulate downstream targets.

DA-cell movement over the stretch cells may require expression of stretch-cell factors that guide or facilitate migration. As noted above,mammalian proteins that share homology with Shark can bind to and regulate integrin complexes. Shark may bind these and/or other adhesion receptors to regulate cell migration either through signaling cues or by modulating the extracellular matrix.

Shark could also regulate stretch cell behaviors, controlling the small cellular projections that extend towards the DA cells during their anterior movement. These extensions may guide or signal the DA-forming cells, as occurs in imaginal discs (Cho et al.,2000; Gibson and Schubiger,2000; Ramirez-Weber and Kornberg, 1999).

Extracellular signals and interactions are key components of morphogenetic processes. We have identified two downstream components of the bwkpathway that act in the stretch-cell layer to relay a novel germline signal required for the movement of a third tissue, the remodeling epithelium of the dorsal appendage cells.

We thank R. Stanley, R. Warrior and P. Lasko for their generosity with the shark and Bic-C reagents. We thank M. Llimargas, S. Luschnig, S. Hou, M. Mlodzik, G. Boulianne, L. Manseau and H. Ruohola-Baker for reagents or equipment. We also thank the members of the Berg laboratory for helpful discussions.