The genesis of shape, or morphogenesis, is one of the most fascinating problems of biology. Following fertilization, cells adopt specific fates that determine their identities. Subsequently, groups of cells with similar fates undergo complex cellular rearrangements that shape the various tissues that form the animal. In the past 10 years much information has been obtained toward an understanding of the mechanisms underlying cell fate determination. It is now clear that a complex interplay of transcription factors determines the destiny of cells. This program of transcription factors integrates input from extracellular signals that play critical roles in coordinating and modulating developmental decisions. Many of the routes that integrate signals have now been well characterized, in particular the receptor tyrosine kinases (RTKs), TGF-§, Wnt, Hedgehog (Hh), Notch, JAK/STAT and NFkb signaling pathways. The general interest of our laboratory is to understand the mechanisms by which cells talk to each other to coordinate the formation of specific structures during development. Most of the work that we have conducted in the past has focused on the identification of molecules that relay signals from the membrane to the nucleus. Most of the studies that we have done in this area have involved the characterization of two Drosophila receptor tyrosine kinases (Torso (Tor) and Epidermal Growth Factor Receptor (EGFR)), Wingless (Wg/Wnt) and JAK/STAT signaling pathways. Our contributions to the field have been to identify, using genetics, molecules involved in transducing the signal from the membrane to the nucleus, and to describe their functions. In addition we have developed a number of methods that facilitate gene manipulations.

We have now become more interested in addressing more cell biological issues of signal transduction.
The ongoing projects in the lab are focused on three basic questions of cell signaling and morphogenesis.

1: Role of the extracellular matrix in cell signaling and morphogenesis
2: Role of cell polarity in the organization of signaling complexes
3. Characterization of signaling pathways that regulate the ability of cells to adopt a variety of shapes and
to carry out coordinated and directed movements.

A brief summary of our past and current work is provided below.

1. Gene manipulation methods

To facilitate our analyses of signal transduction pathways, we have developed a number of methodologies that allow us to explore the functions of specific genes in developmental and cellular processes. The first method is a directed mosaic technology whereby the FLP recombinase is expressed in a tissue specific manner (Duffy et al., 1998). Tissue specific expression of FLP is achieved using the Gal4/UAS system (Brand and Perrimon, 1993). This very efficient method allows the induction of mutant clones only in a tissue of interest. The second method is the "Multiple Germ Line Clone" technology (E. Selva and N. Perrimon, in preparation). Previously, with the exception of mutations that are located on the same chromosomal arm, it was very difficult to examine the phenotype of embryos derived from females that lack multiple gene activities. We have generated FRT-containing chromosomes that carry multiple dominant female sterile mutations (DFS; as a DFS we used a combination of nanos-Gal4 and UAS-ovoD1). The DFS mutations are located on different chromosomal arms such that two chromosome arms can be recombined efficiently at the same time in the female germ line. With this system, the only females that lay eggs are those that have lost both DFS mutations. This method allows the analysis of embryos derived from germ lines that lack multiple gene activities and provides various applications in the study of signaling pathways. The third method is the "Positively Marked Mutant Lineages" technique (E. Spana and N. Perrimon, in preparation). The goal of this method, which is a more sophisticated version than a previous similar technology (Harrison and Perrimon , 1993), is to generate clones of mutant cells that express either GFP or LacZ. Our newest method is based on the reconstitution of an actin-Gal4 gene following an FRT-mediated recombination event between two homologous chromosomes. Our method differs from a related technology recently developed by Luo and Lee (1999) which is based on the loss of the Gal80 repressor. The main advantage of our system is that it is not associated with perdurance of the Gal80 repressor. Since the only marked cells are homozygous for the mutation of interest, our method facilitates the detection and analysis of small clones of homozygous mutant cells in a heterozygous background. This technology is especially powerful for the identification of mutant clones in complex tissues such as the nervous system. Also, it allows expression of specific genes only in mutant tissues via a UAS construct.

2. Genetic screens

Using the "FLP-DFS" technique (Chou and Perrimon, 1996), we have completed several large -scale mutageneses for essential genes with distinct maternal effect phenotypes (Perrimon et al., 1996; L. Cunningham, E. Noll, C. Arnold and N. Perrimon, unpublished). We have generated germ line clones of more than 7,500 EMS induced mutations and recovered mutations in 654 loci with specific mutant phenotypes. Our mutageneses have covered more than 90% of the genome and represent the first systematic analysis of the contribution of essential genes during oogenesis and embryonic development. Further, it identifies for us a collection of mutations in interesting essential genes for further study. Indeed, most of the genes that have been analyzed in our laboratory originate from this collection.

3. Studies on Receptor Tyrosine Kinase pathways

The RTK Tor utilizes the Ras/Raf/MAPK signaling cassette to specify cell fates at the embryonic termini. In collaboration with Drs. V. Cleghon and D. Morrison (NCI, Frederick), we have conducted a structure/function analysis of the tyrosine residues on Tor that become phosphorylated upon activation of the RTK (Cleghon et al., 1996; Cleghon et al., 1998; Gayko et al., 1999). This work has clarified the mechanism by which Tor regulates the activity of downstream transducers and demonstrated the redundancy between some of the phosphotyrosine residues. As part of this analysis, we have documented a novel antagonistic interaction between the non-receptor tyrosine phosphatase Corkscrew and RasGAP. We have also made some progress in understanding the mechanism of Raf activation following RTK activation. Specifically, we have characterized the role of a 14-3-3 protein following RTK activation (Li et al., 1997), illustrated a dual function for Ras in Raf activation (Li et al., 1998), and completed a genetic screen for additional candidate genes involved in Raf activation (Li et al., 2000). Modulation of the activity of an RTK appears to be critical for its biological functions. From our studies of the Tor and EGFR RTK pathways, we have been able to demonstrate the developmental significance of the temporal and quantitative regulation of RTK activities. For example, in the Tor system, we found that quantitative variations in the level of MAPK activity controls patterning of the embryonic termini (Ghiglione et al., 1999a), as determined using both expression of the downstream component tailless (tll) and differentiation of cuticular markers. Further, in the wing imaginal disc, we found that a temporal switch in EGFR signaling controls the specification and differentiation of veins and interveins (Martin Blanco et al., 1999). Negative feed back loops have emerged as an important mechanism to regulate signaling pathways (see review by Perrimon and McMahon, 1999). In the case of the EGFR we have identified the transmembrane molecule Kekkon 1 (Kek1) as such a molecule (Musacchio and Perrimon, 1996; Ghiglione et al., 1999b). Kek1 expression is positively regulated by the EGFR pathway, and although at physiological levels Kek1 is a modest negative regulator of EGFR activity, when overexpressed this molecule completely inhibits the activity of the EGFR. The extracellular domain of Kek1 binds the extracellular domain of the EGFR and prevents its autophosphorylation. Collaborative work done with Drs. K. Carraway and P. Leder (Harvard Medical School) has revealed that the Drosophila Kek1 protein is a potent inhibitor of the EGFR and ErbBs mammalian receptors (Amundadottir et al., 2000). Since vertebrate homologs of Kek1 are present in the database, they are excellent candidates for novel vertebrate extracellular inhibitors of the mammalian EGFR family of RTKs.

4. Studies on the JAK/STAT pathway

Previously, from our analysis of hopscotch (hop, JAK) and marelle (mrl, STAT), we showed that the Drosophila JAK/STAT pathway is involved in embryonic patterning as well as blood cell proliferation (Binari et al., 1994; Hou et al., 1996; see review in Zeidler et al., 2000). In addition, we have documented a novel developmental function for the pathway in establishment of ommatidial polarity in the eye (Zeidler et al., 1999). A major advance in the characterization of this pathway has been our finding that Unpaired (Upd) encodes a secreted protein that can activate the JAK/STAT pathway (Harrison et al., 1998). With the completion of this work, the Upd receptor is the major missing component of this pathway. Interestingly, we have recently discovered a novel interaction between the Tor and JAK/STAT pathway (Li and Perrimon, in preparation). We have shown that removal of the Mrl/STAT protein from early embryos partially suppresses the ectopic tll expression and phenotypic defects caused by a gain-of-function tor (TorGOF) mutation. Further, expression of a TorGOF in Drosophila S2 cells causes activation of STAT, indicating that activation of Mrl transduces part of the TorGOF activity. Interestingly, lack of mrl or hop gene activities has little or no effects on tll expression in the wild type, indicating that under normal conditions Tor signals mainly through the Ras/Raf/MAPK module, and that the JAK/STAT pathway plays a minimal role in the regulation of tll expression. Such a situation is analogous to certain types of human diseases and cancers, where a particular STAT molecule is activated and mediates the effects of an hyperactivated RTK. However, in most cases this particular STAT is not essential for the normal development of the tissues affected (Su et al., 1997). Therefore, STAT may be an essential mediator of a pathogenic or oncogenic effect, which under normal conditions does not function as the result of RTK activation.

5. Studies on Wingless signaling pathway

The current model of Wg signaling is that secretion of Wg in the extracellular space requires a putative acyltransferase encoded by porcupine (Kadowaki et al., 1996). Extracellular Wg interacts with both Heparan Sulfate Proteoglycans (HSPGs) of the Glypican family (Dally and Dally-Like Protein (DLP); Lin and Perrimon, 1999; Tsuda et al., 1999; Baeg et al., 2000) and Frizzled receptors (Fz and DFz2; Bhanot et al., 1999). Activation of Fz receptors modulates the activity of Dishevelled (Dsh) which in turn regulates the activity of a complex of proteins that include GSK3, APC, Axin and Armadillo (Arm). Subsequent accumulation of Arm in turn regulates the activity of the transcription factor Tcf. In addition to this signaling pathway, Fz regulates the planar cell polarity pathway (PCP), which requires Dsh and the JNK pathway, but not Arm and GSK3 (Axelrod et al., 1998; Boutros et al., 1998). Our major contributions to the Wg/Wnt field in the past 4 years has been to demonstrate the role of Dsh in the PCP pathway and to establish the role of HSPGs in Wg reception (Haecker et al., 1997; Lin and Perrimon, 1999; Baeg et al., 2000). The involvement of HSPGs in Wg signaling originated from the identification of a number of mutations in enzymes involved in HSPG biosynthesis. From our screens for zygotic lethal mutations associated with specific maternal effect phenotypes, we have identified mutations in two genes required for HS GAG biosynthesis: sugarless (sgl, UDP-glucose dehydrogenase; Haecker et al., 1997; also identified by Binari et al., 1997 and Haerry et al., 1997) and sulfateless (sfl, N-deacetylase/N-sulphotransferase; Lin and Perrimon, 1999). These enzymes are involved in the biosynthesis of the Heparan Sulfate Glycosaminoglycan (HS GAG) chains, which are long unbranched chains of sulfated repeating di-saccharides attached to a protein core (such as Syndecan and Glypicans; see review by Perrimon and Bernfield, 2000). Numerous biochemical and cell-culture assays have suggested that HSPGs are critical for a variety of biological phenomena such as organogenesis, embryonic development, angiogenesis, regulation of blood coagulation, cell adhesion and lipid metabolism (Lindahl et al., 1998; Perrimon and Bernfield, 2000). In the context of signal transduction, HSPGs have been implicated in a number of signaling pathways, in particular the Fibroblast growth factor (FGF), Wnt, TGF-§, and Hh pathways. In the context of Wg signaling, analyses of both the GAG enzymes sfl and sgl, as well as the protein cores (Dally and DLP), have revealed that the HSPGs are required positively in Wg signaling. Interestingly, over-expression of DLP leads to an accumulation of extracellular Wg, suggesting that DLP plays a role in Wg extracellular distribution. Finally, we found that sfl mutant cells, which do not properly synthesize HSPGs, do not trap extracellular Wg proteins. Taken together, our results indicate that HSPGs are critical for the extracellular distribution of Wg. From our analysis of sgl and sfl mutants, we have also shown that in addition to Wg signaling, HSPGs are absolutely required for FGF and Hh signaling (Lin et al., 1999; Bellaiche et al., 1998; The et al., 1999). In the case of FGF, we have shown that embryos that develop in the absence of either sgl or sfl gene activity are defective in FGF signaling. Further, over-expression of an FGF ligand can bypass the requirement for the HSPGs indicating that at non-physiological levels the ligand can activate the receptor (Lin et al., 1999). These results are consistent with the model that HSPGs, in the context of FGF signaling, act as a co-receptor that stabilizes the FGF-FGFR complex. Indeed, recent structural studies have provided evidence that the HSPG is required to stabilize ligand and receptor clustering (Plotnikov et al., 1999). Unexpectedly, we found that HSPGs are involved in the movement of the heparin-binding Hh proteins through tissues. Most of this evidence originated from our studies on the tout-velu (ttv) gene, identified in our maternal/zygotic screens (Perrimon et al., 1996), which encodes a type II transmembrane HS polymerase enzyme (Bellaiche et al., 1998). In the wing imaginal disc, Hh travels and acts at a distance of 8-10 cell diameters from the site of its production to induce the expression of its target gene patched (ptc) and decapentaplegic (dpp) along the anterior-posterior (A/P) boundary. Results from mosaic analyses of ttv mutations (Bellaiche et al., 1998), have revealed that the activity of the ttv gene is required in Hh-receiving cells for the movement of Hh. The current model is that a Ttv-modified HSPG is required for the proper distribution through tissue of the membrane-targeted cholesterol-modified Hh (HhNp) molecule (The et al., 1999). Recently, a novel Patched-like transmembrane protein, Dispatched (Disp), has been identified and shown to act exclusively in Hh-secreting cells to liberate HhNp from the cellular membrane (Burke et al., 1999). One current model is that the Ttv-modified cell surface HSPG interacts with HhNp to release it from Disp, thus allowing HhNp to move through tissues. Interestingly, ttv is a member of the Ext gene family, which has been implicated in the human syndrome multiple exostoses (Ext, Stickens et al., 1996). We have proposed that the basis of the Ext syndrome corresponds to a defect in Indian Hedgehog signaling in the bone. Finally, we have recently completed the characterization of a new segment polarity gene, fringe connection (frc), that was originally isolated from our maternal/zygotic screens. Embryos that lack both maternal and zygotic frc activities exhibit defects in their Wg, Hh and FGF signaling pathways (Selva et al., 2000). In addition, large frc mutant clones, induced in heterozygous animals, exhibit a fringe-like phenotype. Previously, work by others has shown that Fringe modulates the activation of Notch by its ligands, Serrate and Delta. Further, it has recently been shown that Fringe encodes a glycosyltransferase that adds N-acetylglusosamine onto Notch (Fortini, 2000). Molecular characterization of frc has revealed that it encodes a putative sugar transporter. In collaboration with Dr. S. Turco (University of Kentucky), we have shown, following the expression of frc in Leishmania, that Frc conferred UDP-Glucuronic acid and N actetylglucosamine (but not UDP-Gal or GDP-Man) uptake in isolated microsomes. This result explains the diversity of frc mutant phenotypes since Frc transports, from the cytoplasm into the Golgi, the sugar residues necessary for both HSPGs biosynthesis and Fringe activity.

6. Signal transduction pathways that regulate cytoskeleton reorganization during embryonic development

To characterize which signaling pathways regulate cytoskeleton reorganization during embryonic development, we have began studies on two developmental processes that rely on precise cell shape changes, gastrulation and dorsal closure. Gastrulation begins immediately after cellularization with the invagination of mesodermal and endodermal primordia and is associated with a major morphological rearrangement of the embryo. The mechanism underlying these invaginations is based on a series of coordinated cell shape changes. The force driving these cell shape changes is thought to be generated by the actin/myosin network underlying the cell surface. We have a novel guanine nucleotide exchange factor for Rho family GTPases, encoded by the gene, akkordeon (akk) or D-rhoGEF2, that is essential for the generation of cell shape changes and maintenance of an intact actin cytoskeleton during gastrulation. Previous analyses by others have identified two genes involved in gastrulation, folded gastrulation (fog) which encodes a secreted molecule and the G-protein concertina (cta). The current model is that the activity of Akk is regulated by Fog/Cta. However, because the phenotype of either fog and cta is much weaker than akk, it indicates that another activity is involved in regulating akk activity. Currently, in collaboration with Dr. A. MichelsonŐs laboratory (HHMI, Brigham) we have initiated the molecular characterization of another gene, named Smog, that has a gastrulation phenotype very similar to akk mutants. Possibly, Smog encodes a component of the Fog-independent pathway that regulates Akk activity. During germband retraction, cell shape elongation along the dorsal-ventral axis starts from the dorsal-most cells of the epithelium and is driven by a structure at the dorsal side of the cells referred to as "the leading edge". Dorsal closure is completed with the fusion at the dorsal midline of both edges and the internalization of the amnioserosa. We have characterized mutations in the Drosophila homologue of the mammalian proto-oncogene c-Jun gene (Djun) and shown that Djun regulates the process of dorsal closure. We, as well as a number of other groups studying the dorsal closure process, have proposed that the Jun kinase signaling pathway phosphorylates and activates Djun. Subsequently, Djun regulates the expression of dpp in the dorsal-most cells. Finally, the DPP protein is secreted, binds to its receptors on the cell surface and ultimately regulates the reorganization of the cytoskeleton. We have identified a number of additional genes involved in dorsal closure. In particular, one of them encodes a nuclear protein with WD40 motifs suggesting that it effects the specificity of Jun binding to AP-1. Further characterization of this novel Jun partner is in progress as well as characterization of other members of this pathway.

7. Studies on the establishment of apical-basal polarity
in epithelial cells

We have conducted a screen for mutations that disrupt the morphology of embryonic and follicle cell epithelia. Drosophila epithelial tissues are an excellent model system to understand how cells organize their internal cytoskeleton and organelles to adopt particular shapes, polarities, and connections to neighboring cells. Our long-term goal, which is described in more detail in the Proposed Research section below, is to analyze the molecular links between cell polarity and signaling. Using the directed mosaic method (Duffy et al., 1998), we have searched for mutations that affect the organization of the follicle cell columnar epithelium. Among the most interesting genes that we have identified, thus far, are scribble (scrib) and capulet (cap). Scrib activity is required for proper localization of apical proteins in polarized cells, and maintenance of the embryonic epithelium in a monolayer (Bilder and Perrimon, 2000). We have found that Scrib encodes a protein with multiple PDZ domains and localizes to septate junctions. Interestingly, we have characterized two additional genes, the tumor suppressors lethal(2) giant larvae (lgl; Jacob et al., 1987) and discs-large (dlg, Woods and Bryant, 1991), that have identical effects on epithelial morphogenesis. Scrib and Dlg colocalize tightly throughout development, and overlap significantly with Lgl. The activity of all three genes is required for the cortical localization of Lgl and the junctional localization of both Scrib and Dlg. dlg and lgl show genetic interactions with scrib, where reduction of dlg or lgl function enhances scrib mutant phenotypes. The similar phenotypes of dlg, lgl and scrib mutants, along with both colocalization of their gene products and their genetic interactions, suggests that the three proteins act together in a common pathway to maintain cell polarity. Finally, zygotic loss of scrib, like those of dlg or lgl, causes dramatic overproliferation of imaginal disc cells, indicating that scrib also acts as a tumor suppressor. The observation that Dlg and Lgl act together with Scrib to dictate cell polarity suggests possible mechanisms of action for this group of proteins. While the PDZ domains of Dlg and Scrib are likely to bind to transmembrane proteins that organize the epithelial cell surface, the role of Lgl in the determination of polarity may derive from its function in targeted secretion.Yeast Lgl homologs bind to plasma membrane SNARE proteins and act along with the 'exocyst' complex to fuse vesicles to the growing bud tip (Lehman et al., 1999). In vertebrate epithelia, exocyst components are found at the tight junction, a structure analogous to the septate junction where Dlg and Scrib are found. We have proposed that, in Drosophila epithelia, Dlg and Scrib act at cell-cell junctions to provide a spatial cue that control the activity of Lgl and the exocyst in vesicle trafficking. These observations suggest a link between the transmembrane proteins that establish polarity and the protein targeting system that maintains it. In summary, our results suggest that a group of membrane-associated proteins act in concert to regulate both cell structure, secretion and cell proliferation. The second gene that we have identified from our screen for mutations affecting epithelial morphology is cap, a Drosophila homolog of Cyclase Associated Proteins (Baum et al., 2000). CAP homologs have been shown to inhibit actin polymerization in vitro, by sequestering monomeric actin, and have been found to physically associate with Adenylyl Cyclase and Abl tyrosine kinase. Thus CAP proteins may play a conserved role linking actin organization to signal transduction. In Drosophila, cap mutant clones exhibit a dramatic re-organization of F-actin, with ectopic filaments forming at the apical surface of cap mutant epithelial cells. Thus, CAP limits actin filament formation in vivo and in vitro, and regulates the global distribution of actin filaments within cells. Abl and Ena control actin organization within the growth cones of neurons to mediate axonal pathfinding. Since mammalian CAP has been shown to bind Abl, we investigated, in collaboration with the laboratory of Dr. VanVactor (Harvard Medical School), whether CAP, Abl and Ena function in a common pathway (Wills et al., 2000). We find that CAP is required for proper pathfinding in the Drosophila embryo, since ectopic midline crossing is observed in cap mutant embryos. Moreover, this axonal migration defect is dramatically enhanced by changes in the dosage of Abl, implying that CAP is part of the Abl signal transduction cascade within neurons. Further, in follicle cells, we find that CAP acts together with Ena and Abl to control the simple epithelial actin organization, with CAP acting to inhibit actin polymerization catalyzed by Ena (Baum and Perrimon, 2000). Finally in Drosophila tissue culture cells, CAP, Ena and Abl associate in a complex. Thus CAP, Ena, Abl appear to act together to transduce signals into changes in the distribution and level of F-actin, in migrating neurons, epithelial cells and other polarized cells.

8. Functional genomic analysis of cellular morphology
using high-throughput RNAi screens

The sequencing of the Drosophila genome provides us with an unprecedented resource. New technologies, however, are required to systematically analyze the functions of the ~14,000 predicted genes. The simple addition of double-stranded RNA (dsRNA) to Drosophila cells in culture reduces or eliminates the expression of target genes by RNA-interference (RNAi), efficiently phenocopying loss-of-function mutations. Thus, genome-wide screens using RNAi methodology offers an approach to systematically identify genes with cell-based functions. Our laboratory is conducting high-throughput RNAi screens using Drosophila cells cultured in 384-well plates and automated microscopic imaging. For pilot screens, we generated a set of 1000 dsRNAs from genes predicted to encode regulators central to many fundamental cellular processes, including all kinases, phosphatases, small GTPases, and GTPase regulators. We screened this collection to identify genes that control cell morphology and cytoskeletal organization in different cell assays. By visualizing actin filaments, microtubules, and DNA, we identified many distinct cellular defects that allowed us to group the genes into phenotypic classes. In this way, we identified components of linear signaling pathways that differentially affect cell form and function. To identify additional components in these pathways, we then carried-out RNAi modifier screens. This work shows that RNAi screens in cell culture assays are ideal for the identification of genes and additional pathway components that affect specific cell functions, including aspects of cell morphology.