Soma-to-Germline Interactions During Drosophila Oogenesis Are Influenced by Dose-Sensitive Interactions Between cut and the Genes cappuccino, ovarian tumor and agnostic

Corresponding author: Stephen M. Jackson, Department of Genetics, Box 357360, University of Washington, Seattle, WA 98195-7360., sjackson{at}u.washington.edu (E-mail)

Communicating editor: T. SCHÜPBACH

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The cut gene of Drosophila melanogaster encodes a homeodomain protein that regulates a soma-to-germline signaling pathway required for proper morphology of germline cells during oogenesis. cut is required solely in somatic follicle cells, and when cut function is disrupted, membranes separating adjacent nurse cells break down and the structural integrity of the actin cytoskeleton is compromised. To understand the mechanism by which cut expression influences germline cell morphology, we determined whether binucleate cells form by defective cytokinesis or by fusion of adjacent cells. Egg chambers produced by cut, cappuccino, and chickadee mutants contained binucleate cells in which ring canal remnants stained with antibodies against Hu-li tai shao and Kelch, two proteins that are added to ring canals after cytokinesis is complete. In addition, defects in egg chamber morphology were observed only in middle to late stages of oogenesis, suggesting that germline cell cytokineses were normal in these mutants. cut exhibited dose-sensitive genetic interactions with cappuccino but not with chickadee or other genes that regulate cytoskeletal function, including armadillo, spaghetti squash, quail, spire, Src64B, and Tec29A. Genomic regions containing genes that cooperate with cut were identified by performing a second-site noncomplementing screen using a collection of chromosomal deficiencies. Sixteen regions that interact with cut during oogenesis and eight regions that interact during the development of other tissues were identified. Genetic interactions between cut and the ovarian tumor gene were identified as a result of the screen. In addition, the gene agnostic was found to be required during oogenesis, and genetic interactions between cut and agnostic were revealed. These results demonstrate that a signaling pathway regulating the morphology of germline cells is sensitive to genetic doses of cut and the genes cappuccino, ovarian tumor, and agnostic. Since these genes regulate cytoskeletal function and cAMP metabolism, the cut-mediated pathway functionally links these elements to preserve the cytoarchitecture of the germline cells.

INTERACTIONS between different cell types play pivotal roles in developmental programs of tissues and organs. Oogenesis in Drosophila provides an ideal model system for understanding these interactions. The events that occur during oogenesis take place in egg chambers that are composed of somatically derived follicle cells surrounding a syncyctial cyst of germline-derived cells (SPRADLING 1993 ). Each cyst consists of 16 cells; 15 of these are polyploid nurse cells and the sixteenth is the diploid oocyte. Cells in the cyst are connected by cytoplasmic bridges called ring canals, which result from incomplete cytokineses during the earliest stages of oogenesis. After the germline cell cytokineses are complete, the cyst is encapsulated by an epi-thelial monolayer of follicle cells. Interactions between the follicle cells and germ cells are crucial to initiate and coordinate the events that are required for oogenesis (SPRADLING 1993 ; RAY and SCHUPBACH 1996 ).

Genetic analyses have begun to clarify the many signaling pathways operating between the soma and germline that establish the identity and functions of these cells. The small ovaries gene is required in the soma, but influences the differentiation of germline cells (WAYNE et al. 1995 ). Somatic mutations in cut (JACKSON and BLOCHLINGER 1997 ), daughterless (CUMMINGS and CRONMILLER 1994 ), and the neurogenic group of genes (RUOHOLA et al. 1991 ; XU et al. 1992 ; BENDER et al. 1993 ) disrupt the ability of follicle cells to recognize and encapsulate the germline-derived cells. The toucan locus is also required for the follicle cells to encapsulate the germline cyst, but appears to be required in the germline and not the soma (GRAMMONT et al. 1997 ). Germline-to-soma interactions are also important at the posterior of the oocyte, where localized gurken RNA yields a TGF- homologue that signals through the Epidermal growth factor receptor pathway in the follicle cells to establish posterior follicle cell fate. A subsequent signal from these posterior follicle cells back to the oocyte affects Protein kinase A function in the germ cells and initiates a reorganization of the microtubule cytoskeleton that leads to the establishment of the anterior-posterior and dorsal-ventral axes of the egg chamber and embryo (reviewed in RAY and SCHUPBACH 1996 ). Finally, the product of the bullwinkle locus is required in a separate germline-to-soma signaling pathway that influences the migration of the follicle cells as they produce the respiratory appendages on the dorsal side of the eggshell (RITTENHOUSE and BERG 1995 ).

Other genes not directly involved in signaling between soma and germline are important for the regulation of cytoskeleton function. When these genes are mutated, the structural integrity of individual germline-derived cells is compromised and binucleate nurse cells are produced. These binucleate cells are thought to result either from defective cytokineses of the germline cystoblast cells (chickadee, VERHEYEN and COOLEY 1994 ; cappuccino, MANSEAU et al. 1996 ) or from fusion of adjacent nurse cells after cytokineses are complete (spaghetti squash, WHEATLY et al. 1995 ; EDWARDS and KIEHART 1996 ; armadillo, PEIFER et al. 1993 ; rho family of small GTPases, MURPHY and MONTELL 1996 ). Signaling events and the regulation of cytoskeletal function are integrated processes. For example, subcortical actin breaks down at the ring canals when Protein kinase A catalytic subunit (Pka-C1) function is lost in germline clones. This breakdown results in the fusion of adjacent nurse cells into binucleate cells that are morphologi-cally indistinguishable from those seen in the cytoskele-ton-associated mutants described above (LANE and KALDERON 1993 , LANE and KALDERON 1995 ). Thus, cAMP-dependent Protein kinase A regulates multiple important cytoskeleton-depen-dent processes, both actin- and tubulin-based, during oogenesis.

Recently, we described a soma-to-germline signaling pathway that requires the activity of the cut gene (JACKSON and BLOCHLINGER 1997 ). Our analyses showed that in the ovary, cut RNA and protein expression are restricted to the follicle cells; moreover, cut mutant germline clones are phenotypically normal. When cut function is lost in the follicle cells, however, germline-derived cysts are mispackaged into egg chambers with abnormal numbers of cells, and the structural organization of oocyte-nurse cell complexes disintegrates, generating binucleate germline-derived cells similar to those described above. To date, cut is the only gene known to be required in the follicle cells that when mutated results in binucleate cells. The assembly of egg chambers and the maintenance of germline cell morphology therefore requires the activity of the cut gene in the soma, revealing a signaling pathway that influences the morphology and function of the germline-derived cells. In support of this conclusion, cut interacts genetically during oogenesis with two genes that influence intercellular communications, Notch and Pka-C1 (JACKSON and BLOCHLINGER 1997 ).

The cut locus encodes a nuclear homeodomain-containing protein with multiple DNA-binding motifs (BLOCHLINGER et al. 1988 ). Regulation of cut expression is complex and is controlled by a large promoter/enhancer region extending 200 kb from the coding region (JACK 1985 ; JACK and DELOTTO 1995 ). The cut gene is expressed in other tissues besides the ovary, and cut function is required for the development of all the tissues in which it is expressed. The genetic networks that influence the function of cut in these tissues are beginning to be elucidated. Along the wing margin, loss of cut function affects the differentiation of sensory bristles and results in loss of wing blade material (JACK et al. 1991 ; JACK and DELOTTO 1992 ; BLOCHLINGER et al. 1993 ). During development of the wing margin, cut interacts with many genes, including Notch, Serrate, strawberry notch, vestigial, scalloped, mastermind, and Chip (JACK and DELOTTO 1992 ; MORCILLO et al. 1996 ; MAJUMDAR et al. 1997 ). Notch, Delta, and wingless interact with cut by affecting Cut protein expression along the wing margin (MICCHELLI et al. 1997 ). In the embryonic and adult peripheral nervous system, cut acts as a bimodal switch between external sense organ fate and chordotonal organ fate (BODMER et al. 1987 ; BLOCHLINGER et al. 1990 , BLOCHLINGER et al. 1991 ). In developing poly-innervated external sense organs, the pox-neuro gene induces cut expression (VERVOORT et al. 1995 ). cut is also expressed in the developing central nervous system, trachea, and Malpighian tubules and loss of cut function affects differentiation of each of these organs (BLOCHLINGER et al. 1990 , BLOCHLINGER et al. 1993 ; LIU and JACK 1992 ). Although several genes act upstream of cut in developing tissues, downstream targets of cut gene activity are unknown in Drosophila.

Since cut appears to determine cell fate in a cell-autonomous manner in other tissues, we were surprised to find that germline cell morphology was influenced by cut-mediated events in the follicle cells. Virtually nothing is known about this unique interaction; thus, in this work we analyze the basis of the interactions between these two cell types. First, we examined the mechanism responsible for the morphological defects by determining whether binucleate cells form due to defective cytokinesis or by fusion of adjacent cells. We found that in cut mutants, binucleate nurse cells were present only in later stages and contained remnants of ring canals, suggesting that cytokinesis proceeded normally. Surprisingly, and in contrast to current models of cappuccino and chickadee function, we also found evidence of normal cytokinesis in these mutants. Second, we found that cut exhibits dose-sensitive genetic interactions with cappuccino but not with chickadee. Finally, we took advantage of the dose-sensitive nature of cut to identify additional genes in this signaling pathway. We found 16 genomic regions that interact with cut during oogenesis and 8 regions required for the development of other tissues. As a result of this screen, we identified novel genetic interactions between cut and two genes, ovarian tumor and agnostic. We previously hypothesized that this unique signaling pathway regulates the activity of cytoskeleton-associated proteins in the germline cells. This work provides concrete evidence that somatic cut expression cooperates with at least two cytoskeleton-regulating genes in the germline cells, presumably by cAMP-mediated events.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Stocks and crosses:
Flies were raised in vials of standard cornmeal-molasses-agar medium at 25° unless stated otherwise. cut follicle cell clones were generated using FLP-mediated mitotic recombination as described previously (JACKSON and BLOCHLINGER 1997 ). After heat shock-induced generation of ct^C145 null clones, adult females recovered for 4–7 days at room temperature, and egg chamber morphology was examined in dissected ovaries as described below.

Interactions with cut nulls: Genetic interactions between cut null alleles and known autosomal mutations were examined in a doubly heterozygous background by crossing males containing mutations in the gene of interest (geneX) to y w ct^C145/FM3 or y w ct^DB7/FM7c virgins at 25° in vials. Females doubly heterozygous for cut and geneX (e.g., y w ct^C145/+; geneX/+) were harvested and kept with males on fresh yeast for several days before ovaries were dissected. Egg chamber morphology was examined by staining with 4',6-diamidino-2-phenylindole (DAPI) and rhodamine-conjugated phalloidin as described below. First chromosome mutations were examined by crossing y w ct^C145/y⁺ ct⁺ Y males to geneX/Balancer virgins and harvesting double heterozygotes (y w ct^C145/geneX).

Interactions with cut hypomorphs: To examine the ability of second chromosome mutations to interact with cut hypomorphs, males containing the gene of interest (geneX) were crossed to y w ct^L188; Pin/CyO virgins at 22° (room temperature). y w ct^L188/Y; geneX/CyO males were then crossed to y w ct^C145/FM3 or y w ct^DB7/FM7c virgins at 25° in bottles. y w ct^L188/y w ct^C145; geneX/+ females were harvested and kept with males on fresh yeast for several days and scored as described above. Third chromosome mutations were analyzed in a similar manner using appropriate balancers. To examine the ability of a strong armadillo (arm) mutation to interact with cut hypomorphs, arm⁴ was recombined onto the ct^C145 chromosome and balanced using standard genetic techniques. Three independent lines (w arm⁴ ct^C145/FM6, ct) were generated and the presence of arm⁴ in each line was verified by examining cuticle phenotypes. Virgins from these stocks were collected and mated to y w ct^L188 males, and w arm⁴ ct^C145/y w ct^L188 females were collected and assayed as described above. Each independent line gave identical results.

agnostic^ts3 mutants (agn^ts3) were raised at 25°; adult females were placed in vials with fresh yeast at 29° for 5 days to examine the phenotype at the restrictive temperature. Similarly, y w ct^C145/y⁺ ct⁺ Y males were mated to agn^ts3 females at 25°, y w ct^C145/agn^ts3 females were harvested and placed at 29° for 5 days to examine the phenotype. As controls, w¹¹¹⁸ stocks were raised at 25°, placed at 29° for 5 days, and ovaries were dissected and examined at the same time.

Screen for interactions between cut and chromosomal deficiencies:
The stock y w ct^C145/FM3, y⁺ ct⁺ Y was used to screen a collection of chromosomal deficiencies. First chromosome deficiencies were provided by the Bloomington stock center; second and third chromosome deficiencies were also obtained from the Bloomington stock center but maintained separately by S. Parkhurst. For most of the first chromosome deficiencies, Df(1)/Balancer virgins were crossed to y w ct^C145/y⁺ ct⁺ Y males; for some of the deficiencies, it was necessary to use y w ct^C145/FM3 virgins and deficiency chromosome-containing males. For second and third chromosome deficiencies, males containing the deficiency chromosome were crossed to y w ct^C145/FM3 virgins. Flies were crossed in vials at 25°; females doubly heterozygous for y w ct^C145 and the deficiency were placed in fresh vials with yeast and w¹¹¹⁸ or Canton-S males for 2–7 days. The siblings containing cut and nondeficiency chromosomes were examined in an identical manner to rule out the possibility that an interacting mutation mapped outside the deficiency. Ovaries were dissected and stained with DAPI and rhodamine-conjugated phalloidin as described below. A total of at least 500 egg chambers at all stages were examined from each mutant combination for morphological defects. For completeness, we scored egg chambers between stages 2 and 11; if we had counted only stages 9–11 (when the binucleate phenotype was most often observed), the penetrance would have been higher. Deficiencies that displayed a phenotype were checked for specific cut interactions by crossing each deficiency to y w ct^C145/FM3, y w ct^DB7/FM7, or w¹¹¹⁸ stocks. Females were harvested, checked for reproducibility of the phenotype, and ovaries were dissected and stained as described below.

Immunocytochemistry:
Adult females were placed in vials with fresh yeast and males for several days prior to dissection. Ovaries were hand dissected in PBSE (phosphate-buffered saline plus 1 mM EDTA) and fixed in PBTE (PBSE plus 0.2% Tween-20) plus 4% formaldehyde at room temperature for 20 min. Ovaries were rinsed once with PBTE, and then permeablized for at least 1 hr in PBSE plus 1% Triton X-100. The tissue was rinsed once with PBTE, blocked for at least 1 hr in PBTE plus 5% bovine serum albumin and 0.02% sodium azide, and incubated in primary antibodies in blocking solution overnight at 4°. The ovaries were then washed with four 5-min washes with PBTE, incubated in secondary antibody in PBTE for at least 1 hr at room temperature, washed again, and mounted in PBTE plus 35% glycerol and one drop of Vectashield antiphotobleaching agent (Vector Labs, Burlingame, CA). DAPI (0.2 µg/ml) and rhodamine-conjugated phalloidin (2 units/ml, Molecular Probes, Eugene, OR) were included with secondary antibody-staining solution. Antibody concentrations were as follows: anti-c-Myc 9E10 1:20 (Oncogene Science, Manhassett, NY), anti-Hts-RC 1:10, anti-Kelch 1:1 (both from L. Cooley), and Oregon Green-anti-mouse 1:100 (Molecular Probes). Images were examined and photographed on a Nikon Microphot FXA microscope and slides were scanned with a Microtek ScanTek III flatbed scanner or acquired digitally on a Deltavision deconvoluting microscope (Applied Precision, Issaquah, WA) and processed using nearest neighbor algorithms.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Binucleate cells contain remnants of ring canals:
In the germarium, which is found at the anterior tip of the ovary, a germline cystoblast cell divides exactly four times to produce 15 nurse cells and a single oocyte. Cytokinesis is incomplete during these divisions, resulting in a syncytium with channels (ring canals) through which materials are transported or freely pass. Several proteins are added to reinforce the ring canals after the germline cell cytokineses are completed, beginning with a tyrosine-phosphorylated protein(s), followed by filamentous actin and Hu-li tai shao (Hts) protein in region 2 of the germarium and, last, Kelch protein in region 3 of the germarium (ROBINSON et al. 1994 ). Other proteins are also added to ring canals after the germline cell divisions are completed, but these proteins have yet to be ordered with respect to ring canal assembly (Cheerio, ROBINSON et al. 1997 ; Tec29, GUARNIERI et al. 1998 ; ROULIER et al. 1998 ; Src64B, DODSON et al. 1998 ; ROULIER et al. 1998 ). Subcortical filamentous actin marks the boundaries between adjacent cells and in later stages tethers the nucleus to the middle of the cell. These structures can be seen in egg chambers stained with fluorescent phalloidin conjugates (Figure 1). We took advantage of the fact that Hts and Kelch proteins are added to the ring canals after cytokineses are completed to examine the mechanism leading to binucleate cells.

View larger version (57K): In this window In a new window Download PPT slide   Figure 1. Binucleate cells produced by cut follicle cell clones contain remnants of ring canals. (a) Wild-type stage 10 egg chamber. Note the position of the nurse cell nuclei (DAPI stained, blue), subcortical actin marking the boundaries of individualized nurse cells, and nuclear-anchoring actin filaments (rhodamine-conjugated phalloidin stained, red). (b) Stage 10 egg chamber containing homozygous cut follicle cell clones. Two binucleate nurse cells with actin-containing spots are indicated with arrowheads. Homozygous cut clones, marked with a c-Myc epitope tag (green), are over the oocyte and outside the plane of focus. (c and d) Egg chambers with homozygous cut clones have misshapen ring canal remnants that stain with Hu-li tai shao antibodies (green). Two closely apposed nuclei in a binucleate cell can be observed in d. Although the ring canals in unaffected nurse cells are normal and rounded, some of these are outside the plane of focus and appear as diffuse patches of green staining.

In principle, binucleate germline-derived cells could arise either by normal karyokinesis followed by defective cytokinesis, or by normal, incomplete cytokinesis followed by a subsequent breakdown in structural elements of adjacent cells, resulting in cell fusions. The presence of ring canal remnants in binucleate cells would sug-gest that the ring canals were assembled properly after the germline cytokineses and subsequently became detached when membrane integrity became compromised. To determine which of the two possible mechanisms was responsible for the phenotype when cut expression was lost, egg chambers containing cut null (ct^C145) follicle cell clones were generated by FLP-mediated mitotic recombination and stained with antibodies that recognize ring canal markers (Figure 1). In egg chambers containing cut null follicle cells, nurse cells with two nuclei and a spot of filamentous actin-containing material could be found. The binucleate cells could be found at any position and among any of the nurse cells in the egg chamber. A majority of the affected egg chambers contained a single binucleate cell, but occasionally more than one binucleate cell was present (Figure 1B). Most of the time (>90%) these cells were binucleate, but in cut clones and in genetic combinations discussed below, more than two nuclei could occasionally be observed within a single cell. In most cases, it was not possible to correlate the location of the follicle cell clone with the binucleate nurse cell, since the follicle cells migrate over the oocyte during the stages when binucleate cells are most frequently observed (stages 9–11). Nevertheless, a clone in which ~25–30% of follicle cells were null for cut function was sufficient to produce an egg chamber with binucleate cells, suggesting that an important factor is dose sensitive. It is not possible to observe the effect of larger cut clones since larger clones result in cyst-packaging defects (JACKSON and BLOCHLINGER 1997 ).

When mutant egg chambers were stained with antibodies against Hts proteins, the bright actin-containing spot observed in binucleate cells also reacted with Hts antibodies. In these egg chambers, ring canals at various stages of degeneration were observed, from free-floating, detached, round ring canals to dense spots that stained with antibodies against actin and Hts. Identical results were obtained using Kelch antibodies (data not shown). Binucleate cells that lacked Hts or Kelch immunoreactivity or actin-containing spots were never observed. The total number of normal ring canals and ring canal remnants in egg chambers with binucleate nurse cells was 15, suggesting that there was no net change in the number of cystoblast cytokineses. Finally, binucleate cells were not observed prior to stage 5 in cut clones and were observable in stage 10 egg chambers as soon as 4 days after clonal induction. These results are consistent with a model in which the binucleate cells produced by cut follicle cell clones result from the postmitotic fusions of adjacent nurse cells.

We employed a similar analysis of ring canal markers to determine if binucleate cells present in cappuccino (capu) female sterile mutants arise by defective cytokinesis or subsequent nurse cell fusions (Figure 2). In capu^RK12 and capu^G7 homozygous mutant egg chambers, all of the binucleate cells contained fragments of ring canals that reacted with Hts (Figure 2A) and Kelch antibodies (data not shown). As in the cut somatic clones, binucleate cells were not detectable in the germarium or early stages, suggesting that binucleate cells in these mutants also arise by fusion of adjacent cells. Since chickadee (chic) produces similar phenotypes and interacts genetically with capu, we also analyzed the binucleate cell phenotype in chic mutants. Binucleate cells produced by chic⁷⁸⁸⁶ (Figure 2B) or chic¹³²⁰ (data not shown) mutants also contained remnants of ring canals that displayed Hts and Kelch immunoreactivity (Figure 2B and data not shown). Although all of the binucleate cells produced by cut and capu mutants contained ring canal remnants, rare chic binucleate cells (<1%) exhibited no detectable ring canal remnants. chic mutants therefore may produce binucleate cells one of two ways: most commonly, by fusion of adjacent cells, or rarely, by defective cytokinesis. Alternatively, all binucleate cells arise by fusion, but loss of chickadee function accelerates the rate of ring canal degradation. Germline clones of Pka-C1 mutations also undergo fusion of adjacent nurse cells, resulting in binucleate cells with Hts- and Kelch-containing ring canal remnants (LANE and KALDERON 1993 ).

View larger version (94K): In this window In a new window Download PPT slide   Figure 2. Binucleate cells produced by chickadee and cappuccino homozygous females also contain remnants of ring canals. (a) Stage 10 egg chamber from capuRK12 homozygous female. The cell with two closely juxtaposed nurse cell nuclei (DAPI stained, blue) contains a dense spot of ring canal remnant that reacts with Hu-li tai shao antibodies (green staining, arrowhead). (b) Stage 9 egg chamber from chic7886 homozygous female. The ring canal remnant in the cell with two nuclei also stains with Hu-li tai shao antibodies (arrowhead). Again, note that the ring canals in unaffected nurse cells have a normal appearance.

cut interacts with cappuccino, a gene that regulates cytoskeletal function:
Binucleate cells with a phenotype similar to that observed in cut follicle cell clones have been observed in egg chambers produced by flies carrying mutations in other genes, including those that are associated with cytoskeletal, cell adhesion, and cell-signaling functions. To determine whether these phenotypic similarities reflected functional interactions, we determined whether mutations in these genes produced a binucleate cell phenotype in either a doubly heterozygous genetic background, or in a background of hypomorphic combinations of cut alleles (cut hypomorphs; JACKSON and BLOCHLINGER 1997 ). We found that double heterozygotes containing strong alleles of capu (capu^RK12 and capu^G7; MANSEAU and SCHUPBACH 1989 ; EMMONS et al. 1995 ) and null alleles of cut (ct/+; capu/+) produce binucleate cells (Figure 3 and Table 1). We used two different embryonic lethal cut null alleles for our analyses: ct^C145, an EMS-induced allele that produces no detectable Cut protein on immunoblots and ct^DB7, a diepoxibutane-induced allele that makes a truncated, non-nuclear Cut protein (BLOCHLINGER et al. 1990 ). When doubly heterozygous with strong capu alleles, each cut allele produced binucleate cells with a similar frequency, demonstrating that the observed genetic interaction is not restricted to specific alleles. As in the cut clones, a majority (>50%) of affected egg chambers had a single binucleate cell, but occasionally more than one binucleate cell could be observed and these cells could be found anywhere among the nurse cells. The binucleate cells formed by ct/+; capu/+ heterozygotes were not observed prior to stage 5 of oogenesis and contained anti-Hts and anti-Kelch immunoreactive remnants of ring canals (Figure 3), suggesting that, like cut follicle cell clones and capu and chic homozygous mutant ovaries, they arose by fusion of adjacent cells rather than defective cytokineses.

View larger version (75K): In this window In a new window Download PPT slide   Figure 3. cut and cappuccino interact during oogenesis to produce binucleate cells. Stage 10 egg chambers from ctC145/+; capuRK12/+ doubly heterozygous females have binucleate cells (arrowheads) with ring canal remnants that contain filamentous actin (a and c, stained with rhodamine phalloidin) and also stain with Hu-li tai shao antibodies (b) and Kelch antibodies (d). The same egg chambers are shown in a and b and in c and d. The apparent membrane staining in b is bleed-through from the rhodamine staining.

Although a weaker capu allele (capu^HK3) did not interact with null alleles of cut in a doubly heterozygous background, it did enhance the frequency of binucleate cell production in cut hypomorphs (Table 1, compare ct^C145/+; capu^HK3/+ to ct^L188/ct^C145; capu^HK3/+ and compare ct^DB7/+; capu^HK3/+ to ct^L188/ct^DB7; capu^HK3/+). These results demonstrate that the genetic interaction is sensitive to the levels of activities of ct and capu gene products. We also tested chickadee (VERHEYEN and COOLEY 1994 ; MANSEAU et al. 1996 ), spaghetti-squash (WHEATLY et al. 1995 ; EDWARDS and KIEHART 1996 ), armadillo (PEIFER et al. 1993 ), Tec29A (GUARNIERI et al. 1998 ; ROULIER et al. 1998 ) and Src64B (DODSON et al. 1998 ; GUARNIERI et al. 1998 ), each of which produces binucleate cells when homozygous in a wild-type cut background (e.g., +/+; chic/chic). None interacted with cut to produce binucleate cells, either in double heterozygotes or as enhancers of cut hypomorphs (e.g., ct^C145/+; chic/+ or ct^L188/ct^C145; chic/+). In addition, several other genes that either affect cytoskeletal function during oogenesis (quail, MAHAJAN-MIKLOS and COOLEY 1994 ; spire, MANSEAU and SCHUPBACH 1989 ) or signaling between the germline and soma (Egfr^top, PRICE et al. 1989 ) did not interact with cut either in double heterozygotes or as enhancers of cut hypomorphs.

Null alleles of cut exhibit dose-sensitive interactions with specific genomic regions:
The observation that cut and capu exhibit dose-sensitive genetic interactions suggested that other components of the cut-mediated soma-to-germline signaling pathway may also be sensitive to genetic dose. To begin to identify other genes that interact with cut, a collection of Drosophila chromosomal deficiencies was utilized in an F₁ screen for second-site noncomplementing loci that interact with a cut loss-of-function mutation. This collection of 157 genomic deletions with defined chromosomal breakpoints provides an opportunity to assay ~70% of the euchromatic genome for dose-sensitive genetic interactions with cut. Double heterozygotes of the null allele ct^C145 and each deficiency were scored for lethal interactions, defects in adult morphology, and female sterility. Ovaries of double heterozygotes were examined for the presence of binucleate cells after staining with DAPI and rhodamine phalloidin. To verify that observed genetic interactions were specific for the cut mutation, deficiencies that interacted with ct^C145 were also assayed with a different cut null allele (ct^DB7) and with w¹¹¹⁸ as a negative control. Out of 157 first, second, and third chromosome deficiencies screened, 19 deficiencies defining a minimum of 16 regions interacted with cut to give binucleate cells (Figure 4 and Table 2). The penetrance of the binucleate cell phenotype varied from ~1–5% of all egg chambers to ~50% of all egg chambers. Eight deficiencies interacted with cut in other developmental pathways.

View larger version (39K): In this window In a new window Download PPT slide   Figure 4. Genomic map of deficiencies that produce phenotypes when doubly heterozygous with cut. The relative positions of all deficiencies used in the screen are shown. Deficiencies defined by black bars showed no interaction as double heterozygotes with cut, those denoted by red bars interacted to give binucleate cells, those indicated by purple bars were lethal as double heterozygotes, those defined by blue bars interacted with cut in other pathways, and those denoted by green bars uncovered a locus known to interact with cut. Hatching indicates more than one characteristic was associated with the interacting deficiency. Stippling indicates that the breakpoints are not well defined.

Each interacting deficiency was inspected for known mutations that map to the same region (FlyBase, http://flybase.bio.indiana.edu). We reasoned that genes known to cooperate with cut in other tissues or genes involved in cAMP/PKA-mediated signaling, cytoskeletal regulation, or cell adhesion would be good candidates for interacting genes in the ovary. Regions containing genes that met these criteria were subjected to more detailed analyses as described below. Some of the deficiencies that interacted with cut in other pathways in our screen confirmed genetic interactions that have been observed previously. A few regions either did not have obvious candidate genes according to our criteria or did not have mutations available in candidate genes. Although these genomic regions are of interest, our initial efforts concentrated on regions for which mutations were available and that exhibited a high penetrance of the binucleate cell phenotype.

Regions that are known to interact with cut:
Two deficiencies were identified that exhibited loss of wing margin material in double heterozygotes. Df(1)N-8 removes the Notch locus and, although small wing margin gaps were present in all flies that carried this deficiency, the wing margin notches appeared to be deeper and more frequent in flies doubly heterozygous for cut and Df(1)N-8. Similarly, Df(1)sd72b removes the scalloped gene, and double heterozygotes of this deficiency and cut exhibited wing notches as well, but to a smaller extent than observed with Df(1)N-8. Neither of these deficiencies interacted with cut to produce binucleate cells in ovaries. Previous studies showed that both N and sd interact with cut in wing margin development (JACK and DELOTTO 1992 ), confirming that known genetic interactions can be detected in this screen. Finally, Df(2L)30C deletes the Pka-C1 locus, but binucleate cells were infrequent (~1%) in double heterozygotes of cut and this deficiency. These observations are consistent with our previous results, however, since genetic interactions between a lethal P-element insertion in Pka-C1 (Pka-C1^l(2)01272) and cut were not observed in double heterozygotes. Rather, Pka-C1^l(2)01272/+ dominantly enhanced the frequency of binucleate cell formation in cut hypomorphs (JACKSON and BLOCHLINGER 1997 ).

Several regions when heterozygous with cut are haploinsufficient during embryogenesis:
Three regions were lethal when doubly heterozygous with cut. A fourth deficiency, which deleted cut, was also lethal. The 11D1;11D2 region demonstrated a lethality that was highly penetrant (3% of expected + Df/ct + females survived), and this lethality was genetically separable from the agnostic locus (Figure 6 and discussed below). Df(3R)P14 (90C2;91A1-2) exhibited a lethal interaction with cut (6% of expected ct/+; Df/+ flies survived). The 97A;98A2 region uncovered by Df(3R)Tl-P was partially lethal; 38% of expected females of the double heterozygous class eclosed. Although the developmental defects of the lethal classes were not characterized further, no dead larvae or pharate adults were observed, suggesting that each interaction resulted in embryonic lethality. Adult escapers of each of the lethal classes produced egg chambers with binucleate cells, but no other defects in adult morphology or fertility were observed.

View larger version (61K): In this window In a new window Download PPT slide   Figure 5. cut interacts with ovarian tumor to produce binucleate nurse cells. Stage 10 egg chamber from y w ctC145/y otuP1 female stained with DAPI (blue) and rhodamine-phalloidin (red). Two binucleate cells with actin-containing ring canal remnants are indicated with arrowheads.

View larger version (60K): In this window In a new window Download PPT slide   Figure 6. Positions of deficiencies used to identify cut interactions with agnostic. Deficiencies that produced a binucleate cell phenotype are indicated with gray bars; noninteracting deficiencies are shown in black. Stippled regions of Df(1)JA26 and Df(1)C246 indicate that the breakpoints of these deficiencies are not well defined. The relative position of the agnostic gene (11A8) is shown. A second gene, distinct from agnostic, is lethal when heterozygous with cut and resides in the 11D1-2 region.

Two regions failed to give a reproducible phenotype when uncovered by different stocks. Df(1)A209 (20A;20F) produced shortened and tapered bristles when heterozygous with cut. Two other deficiencies that uncover the same region (Df(1)HF396, 18E1;20 and Df(1)DCB1-35b, 19F1;20E-F) failed to produce the same phenotype. Since the proximal breakpoints of these deficiencies are not well defined, the interacting gene may have resided in a unique region of 20F. Alternatively, another mutation existed elsewhere on the X chromosome that may have interacted with cut to produce this phenotype. Df(2L)GpdhA (25D7;26A9) produced binucleate cells when heterozygous with cut, but Df(2L)cl-h3 (25D2;26B2-5) deleted a larger region and failed to produce binucleate cells. It is possible that one of the assigned breakpoints of these deficiencies is inaccurate. Alternatively, an interacting mutation residing elsewhere on the chromosome might account for the observed interaction of cut with Df(2L)GpdhA. For these reasons we conclude that cut does not interact specifically with Df(2L)GpdhA. Of note, chickadee is contained in the region uncovered by Df(2L)GpdhA.

Finally, a deficiency uncovering the cappuccino gene failed to interact with cut in our screen. No binucleate cells were observed in egg chambers from females doubly heterozygous for a cut null allele and Df(2L)ed1, a deficiency that fails to complement the female sterility defects associated with several capu alleles (MANSEAU and SCHUPBACH 1989 ). Each of the capu alleles used in our studies are EMS-induced missense mutations that map to different domains of the capu coding sequence (EMMONS et al. 1995 ), raising the possibility that they are gain-of-function rather than loss-of-function alleles. No null mutations in capu have been identified. Our conclusion is consistent with the dose-sensitive nature of the ct^null/+; capu^strong/+ vs. cut hypomorphs; capu^weak/+ interactions.

cut does not interact with Src64B during oogenesis:
Two overlapping deficiencies on the third chromosome, Df(3L)GN50 and Df(3L)GN24, interacted with cut to produce binucleate cells. Although Src64B mutations produce binucleate cells (DODSON et al. 1998 ), and this locus is uncovered by the overlap of these two deficiencies, we found no interactions between cut and a Src64B mutation, either as double heterozygotes or enhancers of cut hypomorphs (Table 1). In addition, no binucleate cells were seen in egg chambers produced by flies doubly heterozygous for cut and Df(3L)10H (64B10-64B12; 64C5-64C9), a smaller deficiency that removes the Src64B locus. Therefore, a locus residing in 63F4-64B10, included in the overlap in the interacting deficiencies but excluded from Df(3L)10H, interacts with cut to produce a binucleate cell phenotype.

cut interacts with the ovarian tumor gene to produce binucleate cells:
Df(1)RA2 and Df(1)KA14 overlap in the 7F1-8A5 region of the first chromosome and both deficiencies produced binucleate cells when doubly heterozygous with cut. One potential interacting gene of the minimum of 112 genes contained within this overlap (FlyBase) is ovarian tumor (otu, 7F1). otu mutations produce a range of phenotypes during oogenesis (KING et al. 1986 ; KING and STORTO 1988 ), including agametic ovaries (quiescent alleles), tumorous egg chambers (oncogenic alleles), and nurse cell dumping defects or arrest at late stages (differentiating alleles). Recent studies suggest that these phenotypes result from misregulation of actin cytoskeleton function at multiple stages of oogenesis (RODESCH et al. 1997 ). otu therefore met our criteria as a potential interacting gene. To determine whether otu interacts with cut, females doubly heterozygous for cut and the strong alleles otu^P1 (a P-element excision allele that is null for otu; GEYER et al. 1993 ) or otu¹¹ (an allele of the tumorous class; STEINHAUER and KALFAYAN 1992 ) were examined for ovariole morphology. Binucleate cells were observed when either allele was doubly heterozygous with cut (Figure 5). As in other mutant combinations discussed above, a majority of the affected egg chambers contained a single binucleate nurse cell positioned anywhere in the chamber, but occasionally two or sometimes three binucleate cells could be observed in a single egg chamber. Binucleate cells were not observed, however, when otu-mutant chromosomes were placed in trans with the w¹¹¹⁸ chromosome, indicating that heterozygosity in otu alone is insufficient to produce binucleate cells. Other oogenic defects associated with loss of otu function, such as tumorous or dumpless egg chambers, were not observed in double heterozygotes. These results demonstrate that cut and otu cooperate during oogenesis to affect membrane integrity of the germ cells. Current models of otu function suggest that this interaction may influence the function of elements associated with the actin cytoskeleton.

cut interacts with the agnostic gene to produce binucleate cells:
Df(1)N105, uncovering the region from 10F7 to 11D1, interacted with cut to produce binucleate cells. A neighboring deficiency, Df(1)JA26 (11A1;11D-E), was lethal when doubly heterozygous with cut. The basis for the lethality has not yet been explored. Nevertheless, rare escapers of the Df(1)JA26/cut class that survived the lethal phase produced egg chambers that contained binucleate cells. This result suggests that an interacting locus is contained within the overlapping region of these two deficiencies. Smaller deficiencies in this region were viable in combination with cut and resulted in binucleate cells (Figure 6). The proximal breakpoint of Df(1)JA26 is not well defined, but Df(1)N12 (11D1-2;11F1-2) did not interact with cut at all. This region therefore appears to contain two interacting loci. One produces binucleate cells and is defined by the overlap of all the deficiencies tested (11A2;11A8), and a second is in the 11D1;11D2 region that is lethal in combination with null alleles of cut.

A scan of the 74 known mutations that map to the 11A2;11A8 region (FlyBase) for loci that met our criteria for cut-interacting genes suggested the agnostic gene (agn, 11A8) as a likely candidate. agnostic is thought to regulate the activity of adenylyl cyclases and phosphodiesterases indirectly by encoding a calmodulin-inhibiting protein (PERESLENI et al. 1997 ). A temperature sensitive-lethal allele of agn was obtained (agn^ts3), and the requirement for this gene during oogenesis was characterized at the restrictive temperature. In addition, we asked whether the agn mutation produced binucleate cells at the restrictive temperature when doubly heterozygous with cut.

After 5 days at 29°, ovaries from agn^ts3 homozygous females exhibited striking defects that were completely penetrant (Figure 7). In stage 5–6 egg chambers, the follicle cell epithelium became gapped. The follicle cells at this stage also appeared more elongated and less cuboidal. No squamous follicle cells (stretch cells) could be found associated with the gaps in the epithelium. In later stages, the follicle cell epithelium was absent, and binucleate germline-derived cells could be observed. Rarely, single egg chambers contained large nurse cells in which three or more nuclei were present. The morphology of these late-stage egg chambers was similar to that observed in cut hypomorphs (JACKSON and BLOCHLINGER 1997 ). Although w¹¹¹⁸ females were unaffected at 29° for 5 days, agn^ts3 females stopped laying eggs under these conditions. Defects in oogenesis were observable after only 2 days at the restrictive temperature.

View larger version (72K): In this window In a new window Download PPT slide   Figure 7. agnostic function is required during oogenesis. Egg chambers from agnosticts3 females placed at the restrictive temperature for 5 days, stained with DAPI (blue) and rhodamine phalloidin (red). (a) Germarium (far left) and early-staged egg chambers. The stage 5–6 egg chamber at the center is elongated, and the oocyte is displaced to the side of the chamber. In the posteriormost egg chamber (right), the follicle cell epithelium is gapped (bracket), and remaining follicle cells on the other side of the chamber appear elongated. The oocyte is indicated with an arrow. (b) Later-staged egg chamber; the exact stage could not be determined. Note the lack of a follicle cell epithelium and disorganization of the nurse cells. Two binucleate nurse cells with ring canal remnants are indicated with arrowheads. The oocyte, which has failed to take up any material from the nurse cells, is indicated with an arrow.

When ct^C145/agn^ts3 adult females were placed at 29° for 5 days, egg chambers were produced with binucleate cells (1–5% of all egg chambers). These egg chambers were indistinguishable in morphology from those produced by cut clones and in other double heterozygous combinations described above. The frequency at which binucleate cells were observed in ct^C145/agn^ts3 females was less than that observed with the deficiencies (20–50% of all egg chambers contained at least one binucleate cell). This difference in expressivity probably reflects the fact that agn^ts3 is not a null allele. Alternatively, two interacting genes could reside in the region uncovered by the overlapping deficiencies. Notably, no defects were observed when either agn^ts3 homozygotes or ct^C145/agn^ts3 double heterozygotes were maintained at the permissive temperature of 25°; ovaries were mostly wild type with the exception of rare binucleate cells observed in agn^ts3 homozygous females. These results demonstrate that the agnostic gene product is required during oogenesis and that agnostic cooperates with cut during oogenesis to affect egg chamber morphology.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

In this study, we used two approaches to understand the mechanism by which somatic expression of the homeodomain protein cut affects the morphology of the germline cells. First, using a cytological approach, we found that the binucleate nurse cells produced by cut, cappuccino, and chickadee mutants probably arise by fusion of adjacent cells, rather than by defective cytokineses of the cystoblast cells. Second, using a genetic approach, we found that cut exhibits dose-sensitive interactions with specific genes and regions of the genome, some of which include genes involved in cAMP-mediated signaling or cytoskeletal function. These results are consistent with our hypothesis that cut affects cAMP-mediated events that ultimately influence the function of the cytoskeleton of the germline-derived cells.

Binucleate cells result from fusions of adjacent germline-derived cells:
The maintenance of germline cytoskeletal membrane integrity is a complex process, since mutations in several genes produce a binucleate cell phenotype. Many of these genes encode cytoskeleton-associated proteins and their regulation is therefore important for preventing binucleate cell formation. The molecular mechanism by which disrupting cytoskeletal function results in binucleate cells, however, is not yet clear. Work from other investigators suggested that binucleate cells observed in chic and capu mutants resulted from cytokinesis defects. For example, chickadee encodes a Drosophila homolog of Profilin (VERHEYEN and COOLEY 1994 ), which regulates cytokinesis in a number of systems (EDMATSU et al. 1992 ; BALASUBRIMANIAN et al. 1994 ; HAUGWITZ et al. 1994 ; VERHEYEN and COOLEY 1994 ; GIANSANTI et al. 1998 ). In addition, genetic and biochemical interactions have been detected between Profilins and Formin homology domain-containing proteins (such as Capu) in yeasts and in flies, implicating the Formin homology domain proteins in regulating actin cytoskeleton function (MANSEAU et al. 1996 ; EVANGELISTA et al. 1997 ); moreover, some Formin proteins are required for cytokinesis (CASTRILLON and WASSERMAN 1994 ; IMAMURA et al. 1997 ). Thus, the connection between these proteins, the actin cytoskeleton, and cytokinesis appears to be reasonable. This hypothesis is complicated, however, by the observation that microtubule function is also affected by mutations in the capu gene (MANSEAU et al. 1996 ). This result raises the possibility that misregulation of microtubule function may contribute to binucleate cell production in capu-mutant egg chambers.

Our results suggest that defective cytokineses are not responsible for the binucleate cell phenotype in cut, capu, and chic mutants. First, binucleate cells were not present in any mutant combination prior to stage 5 of oogenesis. Second, stage 10 egg chambers with cut follicle cell clones exhibited binucleate nurse cells as soon as 4 days after clonal induction. Since 16-cell cysts require 5 days to mature to stage 10, and single cystoblast cells require 7 days (KING 1970 ), binucleate cells in stage 10 egg chambers must have arisen after the divisions were completed. Third, remnants of ring canals stain with Hts and Kelch antibodies, two proteins that are added to ring canals after the germline cell divisions are completed. Finally, the total number of normal ring canals and ring canal remnants was 15, suggesting that cytokineses proceeded normally. Taken together, these results are consistent with a model in which binucleate cells form by fusion of adjacent cells after cytokineses are complete. Presumably, these fusions result from the disruption of cytoskeletal functions other than those necessary for cytokinesis. Alternatively, sufficient product is available to ensure proper cytokinesis, but this level of activity may be insufficient to maintain the integrity of the membrane/cytoskeletal junctions during the considerable growth of the maturing egg chamber. In addition, our results do not rule out the formal albeit unlikely possibility that ring canal-like structures, containing filamentous actin, Hts, and Kelch proteins, form spontaneously when cytokinesis is defective. Finally, we have not observed a noticeable loss in fecundity in binucleate cell-producing mutant combinations. This result suggests that loss of membrane integrity between adjacent cells does not dramatically affect the functions of the nurse cells in the syncytium, provided the disruptions are not severe (e.g., as in agnostic mutants).

Other investigators have also hypothesized that binucleate cells arise by fusion of adjacent cells. Disrupting actin cytoskeleton function by expressing dominant negative RhoL or dominant negative and constitutively active Cdc42 mutant proteins in the germline during oogenesis also resulted in binucleate nurse cells (MURPHY and MONTELL 1996 ). The authors hypothesized that these binucleate cells resulted from fusions rather than cytokinesis defects. Although our screen detected interactions between cut and a deficiency that deletes the RhoL gene (Df(3L)AC1), the lack of RhoL mutations limits further analyses. A deficiency that deletes the Cdc42 gene was not present in our collection and mutations in this locus (FEHON et al. 1997 ) have not been tested for genetic interactions. Loss of armadillo (ß-catenin) or spaghetti squash (regulatory light chain of nonmuscle myosin) function during oogenesis also produces binucleate cells (PEIFER et al. 1993 ; WHEATLY et al. 1995 ; EDWARDS and KIEHART 1996 ). In arm and sqh mutants, binucleate cells were observed only by constructing germline clones (arm) or using heat shock-induced rescue of loss-of-function alleles (sqh). Although it was hypothesized that binucleate cells were produced by nurse cell fusions, these investigators did not examine ring canal markers and so it remains a possibility that defective cytokinesis is responsible for the phenotype in these mutants. Nevertheless, we failed to detect interactions between these genes either by looking specifically with known alleles or in the screen. Similarly, mutations in cheerio (a ring canal component) also produce binucleate cells (ROBINSON et al. 1997 ), but we did not detect an interaction with a deficiency that removes the cheerio gene. It is possible that all these genes function in the process that regulates nurse cell/oocyte structural integrity during the middle stages of oogenesis, but their products are too abundant to reveal a role in this process using our methods.

cut interacts with the cappuccino and ovarian tumor genes during oogenesis:
Because cut is expressed and required only in the follicle cells, its influence on germline cell morphology must be mediated across the soma-germline boundary. Our results demonstrate that capu and otu, which are both required in the germline (WIESCHAUS et al. 1981 ; PERRIMON and GANS 1983 ; MANSEAU and SCHUPBACH 1989 ; STEINHAUER et al. 1989 ; EMMONS et al. 1995 ; SASS et al. 1995 ; TIRRONEN et al. 1995 ), interact genetically with cut and may facilitate cut-mediated events originating in the soma. Although cut is a transcription factor, the clear separation of cell types in which these genes are expressed suggests that cut does not regulate the transcription of capu or otu directly by binding to their promoters and/or enhancers. Rather, we propose a multistep model in which cut activity in the follicle cells first directs expression of a gene or set of genes that regulates adhesion or signaling between the somatic and germline cells. This soma-to-germline interaction then influences cAMP-dependent function in the germline cells. The activity of Capu and Otu is in turn regulated by these cAMP-mediated events, perhaps by post-translational modifications or by alterations in the subcellular localization of one or both of these proteins. Finally, the regulation of Capu and Otu by cAMP results in altered cytoskeletal function. This hypothesis makes several testable predictions that are currently under investigation. Since it is not yet known if agn is required in the germline cells or follicle cells, we cannot rule out the possibility that cut influences agn levels directly by regulating agn transcription in the follicle cells. A less favorable model is that capu, otu, and/or agn function genetically upstream of cut, and loss of germline function of these genes influences cut activity in the follicle cells. At some point in this model, however, cut activity in the follicle cells must influence the function of the germline cytoskeleton, since loss of cut function in the follicle cells is sufficient to produce binucleate germline cells.

The observed genetic interactions between cut and otu are consistent with our model that cut-mediated events disrupt the function of the germline cytoskeleton. Recently, RODESCH et al. 1997 showed that actin cytoskeleton function was disrupted in otu mutants; they hypothesized that this defect was the underlying cause for the various otu phenotypes. Although otu has been cloned and antibodies have been raised against the protein, the gene's sequence (STEINHAUER et al. 1989 ) and the uniform distribution of Otu protein within the germplasm (STEINHAUER and KALFAYAN 1992 ; SASS et al. 1995 ) give no clues as to how Otu is affecting the function of the cytoskeleton. Nevertheless, our findings suggest that otu regulates cytoskeleton function in response to signaling events that occur after the cystoblast cell divisions are completed and egg chambers leave the germarium. Finally, one of the otu mutant phenotypes is the production of tumorous egg chambers filled with extra germline-derived cells. We did not observe egg chambers that were tumorous or that contained extra germline cell nuclei in cut/otu double heterozygotes, suggesting that cut and otu do not interact in the germarium to regulate germline cell divisions.

agnostic is required during oogenesis:
We have provided clear evidence that agnostic is required during oogenesis. Loss of agnostic function affects the morphology of the follicle cell epithelium and, because follicle cells are missing in late-stage egg chambers, may influence the survival of the follicle cells. In addition, loss of agnostic affects the morphology of the germline-derived cells. It is not known whether agnostic is required in the follicle cells, germline cells, or both. Signaling between these two cell layers occurs throughout oogenesis and models can be hypothesized in which loss of agnostic function in one cell type affects the function and morphology of the other cell type. Nevertheless, agnostic is thought to be involved in cAMP metabolism (SHARAGINA et al. 1997 ). Since the requirement for Protein kinase A is restricted to the germline cells (LANE and KALDERON 1993 , LANE and KALDERON 1995 ), it is tempting to speculate that minimally, agnostic is required in the germline cells.

Irrespective of the cell type in which agnostic is required, both adenylyl cyclase and phosphodiesterase enzymatic activity are altered in agnostic mutants (SAVVATEEVA et al. 1985 ; SHARAGINA et al. 1997 ). These results raise the possibility that cut function impinges on the activity of either or both of these enzymes. In Drosophila, mutations in a 3',5'-cyclic nucleotide phosphodiesterase gene (dunce) and an adenylyl cyclase gene (rutabaga) have been isolated (DUDAI et al. 1976 ; ACEVES-PINA et al. 1983 ). Interestingly, some dunce alleles are female sterile, revealing a role for dunce during oogenesis (BELLEN et al. 1987 ). Deficiencies uncovering dunce and rutabaga failed to interact with cut in our screen, however, and no morphological defects were observed in the ovaries of double heterozygotes of cut and specific mutant alleles of dunce or rutabaga. Thus, cut does not appear to interact individually with either dunce or rutabaga in the same dose-sensitive manner as agnostic. There are three other adenylyl cyclase genes identified in Drosophila that may be regulated by agnostic and/or cut (LEVIN et al. 1992 ). Finally, it is interesting to note that both adenylyl cyclase and phosphodiesterase activity are increased in agnostic mutants (SAVVATEEVA et al. 1985 ), suggesting that the role of this gene in regulating cAMP levels is complex.

At least 24 genes show genetic dose-sensitive interactions with cut:
A strategy in which the genetic dose of cut and genes contained in genomic deficiencies were halved resulted in the identification of 16 regions that produced binucleate cells, 3 that were lethal, and 5 that produced other defects when heterozygous with cut. Some of these regions produced multiple phenotypes, suggesting a minimum of 24 genes that are sensitive to genetic dose in a cut heterozygous background. There are likely to be more cut-interacting loci, however, than those identified by our screen. First, cut-interacting genes that are abundant or stable would be difficult to detect by our strategy. For example, cut and wingless are known to interact during wing development (MICCHELLI et al. 1997 ); nevertheless, we did not see wing margin defects when cut was heterozygous with a deficiency uncovering wingless. Second, the phenotypes that we examined were limited to lethality, visible defects in adult cuticle morphology, and oogenesis. Examination of other tissues or stages of development using more stringent criteria may have revealed other interacting regions. For example, pox-neuro interacts with cut during establishment of certain external sense (es) organ neurons (VERVOORT et al. 1995 ), but in our screen, no embryonic lethality or visible adult peripheral nervous system defects were observed in double heterozygotes of cut and a deficiency that removes pox-neuro (Df(2R)Jp1). Interactions between these genes and others in es organ lineage establishment may have been detected if we had specifically examined embryos for es organ morphology. Finally, the chromosomal deficiency kit does not cover the entire genome, and regions not present in the collection may contain additional interacting genes. Although our strategy may not have uncovered all cut-inter-acting loci, it was successful in identifying a number of regions that are important in the soma-to-germline signaling pathway essential for germline membrane integrity and ring canal maintenance. Since very little is known about this pathway, these regions will be the basis for identifying genes essential for this process and for characterizing interactions between pathway members.

Our previous results led us to propose a model in which Cut function in the follicle cells during oogenesis directs expression of a gene or set of genes whose products transduce a signal across the soma-germline boundary and ultimately influence the morphology or function of the germline cytoskeleton (JACKSON and BLOCHLINGER 1997 ). Results presented in this study provide further insights into this process by demonstrating that this signaling pathway is sensitive to genetic doses of cut and the genes cappuccino, ovarian tumor, and agnostic. The fact that these genes regulate cytoskeletal function and cAMP metabolism allows us to formulate hypotheses that are starting points for further molecular studies. Moreover, other genomic regions contain genes presumably also involved in regulating or responding to this pathway, including genes that may be direct targets of cut during oogenesis. By pinpointing and analyzing genes contained in these regions, we can understand better how signaling pathways influence the morphology and function of cytoskeletal elements.

	ACKNOWLEDGMENTS

We thank the Bloomington stock center for first chromosome deficiencies, Susan Parkhurst for second and third chromosome deficiencies, Rod Nagoshi for otu^P1, Georgette Sass for otu¹¹, Elena Savvateeva for agnostic^ts3, Doug Guarnieri for Src64B¹⁵ and Df(3L)10H, and Dan Kiehart for sqh¹ and sqh² stocks. We thank Lynn Cooley for providing the Hts and Kelch monoclonal antibodies. We thank Barbara Wakimoto, Hannele Ruohola-Baker, and members of the Berg lab for critically reading drafts of the manuscript. This work was supported by National Institutes of Health (NIH) grant GM-45248 to C.A.B. and NIH grant 1F32HD08254 to S.M.J.

Manuscript received January 18, 1999; Accepted for publication May 21, 1999.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ACEVES-PINA, E. O., R. BOOKER, J. S. DUERR, M. S. LIVINGSTONE, and W. G. QUINN et al., 1983  Learning and memory in Drosophila, studied with mutants. Cold Spring Harbor Symp. Quant. Biol. 48:831-840.

BALASUBRIMANIAN, M. K., B. R. HIRANI, J. D. BURKE, and K. L. GOULD, 1994  The Schizosaccharomyces pombe cdc3+ gene encodes a profilin essential for cytokinesis. J. Cell Biol. 125:1289-1301[Abstract/Free Full Text].

BELLEN, H. J., B. K. GREGORY, C. L. OLSSON, and J. A. KIGER, JR., 1987  Two Drosophila learning mutants, dunce and rutabaga, provide evidence of a maternal role for cAMP on embryogenesis. Dev. Biol. 121:432-444[Medline].

BENDER, L. B., P. J. KOOH, and M. A. T. MUSKAVITCH, 1993  Complex function and expression of Delta during Drosophila oogenesis. Genetics 133:967-978[Abstract].

BLOCHLINGER, K., R. BODMER, J. JACK, L. Y. JAN, and Y. N. JAN, 1988  Primary structure and expression of a product from cut, a locus involved in specifying sensory organ identity in Drosophila. Nature 333:629-635[Medline].

BLOCHLI