¹EMBL, Meyerhofstr. 1, 69117 Heidelberg, Germany

²Center for Cancer Biology, University of Rochester Medical Center, 601 Elmwood Avenue, Box 633, Rochester, New York, NY 14642, USA

³Department of Genetics, Harvard Medical School, 200 Longwood Avenue, Boston, Massachusetts, MA 02115, USA

Correspondence to: D Bohmann, Center for Cancer Biology, University of Rochester Medical Center, 601 Elmwood Avenue, Box 633, Rochester, New York, NY 14642, USA

In recent years, studies in the model organism Drosophila melanogaster have contributed significant insights into the molecular and developmental biology of the AP-1 transcription factors Jun and Fos. Powerful genetic and biochemical approaches uncovered a baffling complexity and variability of the signaling connections to and from AP-1. The range of biological processes that Jun and Fos regulate in this organism is equally multi-faceted. Regulatory interactions between AP-1 and JNK, ERK, TGF, Notch or other signaling systems have been implicated in the control of a multitude of embryonic and adult events, including tissue closure processes, patterning of eye, gut and wing, as well as apoptosis. Here we review the information that has been gathered on Drosophila AP-1 in signal transduction and on the developmental and cellular functions controlled by AP-1-mediated signals in the fly. Lessons learned from the studies on AP-1 in Drosophila may contribute to our general understanding, beyond species boundaries, ofthis fundamental class of transcriptional regulators. Oncogene (2001) 20, 2347-2364.

Drosophila; AP-1; Fos; Jun; transcription factor; JNK

Introduction

The AP-1 proteins Jun and Fos represent prototypical examples of sequence-specific DNA-binding transcription factors in higher eukaryotes. Our current understanding of many of the paradigmatic functions of this class of cellular regulators, including dimerization, DNA-binding, transactivation, control by phosphorylation and ubiquitination derive in part from pioneering work on Jun and Fos. To date, AP-1 remains a fascinating topic of research as it provides inroads towards a molecular description of numerous signal transduction processes involving MAP kinase cascades and a number of other pathways of cellular information transfer. Another aspect that has increasingly gained attention concerns the biological processes orchestrated by signaling via AP-1. Genetic studies on AP-1 have provided insights into the molecular biology underlying diverse phenomena ranging from bone morphogenesis to neuronal cell death and differentiation, tissue remodeling and cancer.

The introduction of Drosophila as a model organism in which to study these model transcription factors is a powerful new addition to the arsenal of instruments at our disposal to understand AP-1. The power of Drosophila genetics and the fact that the fly is a comparatively simple organism with less genetic redundancy than encountered in vertebrates makes this system particularly attractive for studying principles of AP-1 function. Studies in Drosophila have already led to the identification of numerous new aspects of AP-1 biology. For example, mutant analysis has revealed a central function for AP-1 in tissue closure processes. In another example, studies in the fruitfly have shown that AP-1 and BMP signal transduction are intimately associated and together can control a whole panoply of complex developmental processes.

In the first part of this review we will summarize what is known about the molecular biology of Drosophila Jun and Fos and compare them to the mammalian homologues. Next, we will discuss a number of different developmental situations in which a function of Drosophila AP-1 transcription factors has been described. These include embryonic dorsal closure, pupal thorax closure, endoderm patterning, wing patterning and eye development. These different sections will be used to expand on certain molecular and cell biological aspects of AP-1 function in this organism, such as interaction with several signaling pathways, control of cytoskeletal events, apoptosis and tissue polarity.

Molecular biology

The Drosophila jun and fos genes, known as D-jun and D-fos respectively, have been cloned independently by low stringency screening of a Drosophila genomic library and by microsequencing of purified Drosophila AP-1 protein followed by screening of cDNA libraries (Perkins et al., 1990; Zhang et al., 1990). Only one member of the jun and fos families respectively appears to be present in Drosophila. The homology between the vertebrate Jun proteins and D-Jun is high in the bZIP DNA binding/dimerization region and still considerable in patches across the rest of the molecule (Figure 1). D-Jun shows a similar degree of sequence similarity to mammalian c-Jun as to JunD and is slightly less related to JunB. It is not possible to decide whether D-Jun is functionally or structurally more closely related to JunD or c-Jun.

In addition to the bZIP region, D-Jun shares several sequence features that have been functionally characterized in vertebrate Jun transcription factors. A comparison of the primary sequences of c-Jun and D-Jun suggests that the N-terminal phosphorylation sites, the targets for stimulatory phosphorylation by JNK, are conserved. As in the case of mammalian Jun, the activity of D-Jun is regulated at the protein level by MAP kinase phosphorylation. Three N-terminal phosphorylation sites (Ser 82, Thr 92 and Thr 107), homologous to the well-characterized regulatory sites at Ser 63 and 73 and Thr 91 (or 93) of c-Jun, are present. These residues are phosphorylated both in vivo and in vitro by two different MAPKs, JNK and ERK, encoded by the Drosophila genes basket and rolled respectively (Biggs et al., 1994; Brunner et al., 1994b; Riesgo-Escovar et al., 1996). The property of D-Jun to serve as an equally good substrate for both MAPKs may represent a more basic situation as compared to the mammalian situation where different Jun family members are more selective for their interacting kinases. For example, JNK and ERK preferentially phosphorylate c-Jun and JunD, respectively (Musti and Ventura, personal communication). Experiments with transgenic flies expressing mutant versions of D-Jun in which these phosphorylation sites are eliminated resemble D-jun loss of function mutants at least in certain genetic backgrounds, verifying the regulatory relevance of these sites in vivo (Peverali et al., 1996).

Interestingly, the only tyrosine that is present in D-Jun corresponds to Tyr 170 in human c-Jun. This residue has been shown to be phosphorylated by the nonreceptor tyrosine kinase Abl which subsequently allows the binding of c-Jun to Abl's SH2 domain (Barila et al., 2000). However, similar tyrosine phosphorylation of D-Jun or its interaction with Abl or any other nonreceptor tyrosine kinase has not yet been described.

Drosophila Fos (D-Fos) exhibits the clear signature of a Fos family member in the bZIP region. However, as opposed to the case of c-Jun and D-Jun discussed above, a comparison between D-Fos and the mammalian Fos family members outside this conserved domain reveals only moderate similarity at the primary sequence level (Figure 1). In fact, D-Fos is significantly larger than c-Fos and may contain additional functional features. For example, there is a cluster of possible JNK phosphorylation sites and a potential JNK-binding that show some resemblance to the delta domain in c-Jun (Ciapponi and Jackson, unpublished observations). In spite of this limited sequence conservation, D-Fos resembles mammalian Fos proteins in several functional aspects, indicating that it carries its name with justification.

Biochemical analyses showed that D-Jun and D-Fos, like their mammalian counterparts, can heterodimerize and bind to DNA containing the TRE motif. AD-Jun/D-Fos heterodimer can activate transcription from a template containing four AP-1 binding sites in vitro (Perkins et al., 1988, 1990). In contrast to c-Fos, D-Fos is able to homodimerize and constitute, at least in vitro, a functional transcription factor. However, the DNA-binding affinity of this homodimer is fourfold lower than that of a D-Jun/D-Fos heterodimer (Perkins et al., 1988, 1990).

Like c-Fos, D-Fos can be phosphorylated by ERK, and this event appears to be relevant in several RTK-controlled developmental processes, such as wing vein differentiation and photoreceptor recruitment. In addition, D-Fos serves as an effector of JNK signaling during dorsal closure (see below; Zeitlinger et al., 1997; L Ciapponi, unpublished). Several of the characterized phosphorylation sites of c-Fos are conserved in D-Fos (Figure 1).

Careful examination of the transactivation characteristics of D-Fos in the Drosophila embryo were carried out by Szüts and Bienz (2000a) and further highlighted the complexity of Fos' regulatory functions. It was shown that the transactivation domains of D-Fos works in a context-specific manner. In this regard, D-Fos needs other DNA-binding factors as well as corresponding extracellular signals to activate gene expression efficiently, at least on certain composite promoters. Proteins that can cooperate with D-Fos in this way were shown to include Jun, MAD, Dpp and JNK. This requirement for synergism was not observed when the 'classic' transactivation domain of the yeast transcription factor Gal4 was used. This type of cooperation between several transcription factors and signaling components might be a mechanism that regulates gene expression in a spatial and temporal manner. However, these results were not corroborated in conventional studies using tissue culture cells. Apparently, cell-based experimental systems are less suitable to study this type of regulatory interplay. Therefore, the study by Szüts and Bienz (2000a) illustrates the power of Drosophila genetics to address principal questions in transcription regulation.

Other AP-1-related bZIP transcription factors have also been found in Drosophila. These include two CREB proteins, CREB-A and CREB-B. Mutants deficient for CREB-A show several phenotypes notably in embryonic cuticle patterning (Andrew et al., 1997). Similarly, CREB-B has been implicated in a broad spectrum of regulatory processes, ranging from the patterning of the developing gut (Eresh et al., 1997) to brain functions such as learning and memory acquisition (Yin et al., 1995) and circadian rhythm (Belvin et al., 1999). During the development of the gut, D-Fos and CREB-like factors have been suggested to jointly regulate the expression of the homeotic gene labial (lab) in a process that appears to be independent of D-Jun (Eresh et al., 1997). Although biochemical data are not yet available, a straightforward interpretation of these findings would be that D-Fos and Drosophila CREB might form heterodimers which mediate the regulation of lab and ubx.

In mammals dimers between the ATF-2 bZIP factor and c-Jun can form and regulate gene expression. The only known Drosophila protein of the ATF family is dATF-4, which is related to the mammalian ATF-4 and is encoded by the cryptocephal (crc) gene. Analysis of crc mutants reveals pleiotropic functions of this transcription factor in later stages of development and metamorphosis (Hewes et al., 2000). The information available so far does not reveal any obvious connections between Crc and D-Jun or JNK regulated processes.

Dorsal closure

Drosophila mutants lacking either Jun, Fos or JNK function die as embryos because they cannot complete the process of dorsal closure (DC). Drosophila DC has thus become a prime system to study the genetics of AP-1 and JNK signal transduction. In the following section we will describe the mechanisms involved in the regulation and execution of DC and the role that AP-1 transcription factors play during these events.

At mid-embryogenesis, the extensive morphogenetic movements of germ band elongation and germ band retraction leave the embryo with only the ventral and lateral sides contained in tissue that will form the future epidermis (Figure 2). The dorsal side is covered with a different tissue, the amnioserosa. The DC process reorganizes the epidermal primordium to enclose the whole embryo (Knust, 1996).

At the point of development immediately preceding DC, all cell divisions of the epidermis are completed and DC is mainly mediated by the elongation of the epidermal cells along the dorsoventral axis and accompanying cell shape changes in the amnioserosa (Kiehart et al., 2000). DC is generally divided into three phases: initiation, spreading and suturing (Figure 2; Young et al., 1993). First, the dorsal-most row of cells (also called leading edge, LE) of the two epidermal sheets elongate along the dorso-ventral axis (initiation phase). LE cells fulfill specialized functions in regulating and coordinating the process and also in exerting mechanical forces that are partly responsible for the epidermal movement. Key regulators of DC are expressed in the LE (see below). Furthermore, the cells of the LE assemble a prominent actin cable that runs along the epidermal front and has been suggested to serve as either a 'purse string' or to provide mechanical support for the integrity of the closing LE.

In the intermediate phase of DC, epidermal cells located more laterally gradually lose their polygonal cell shape and elongate, thereby stretching the epidermal cell layer dorsally (spreading phase). In the final phase (suture), the cells of the two opposed leading edges meet each other and fuse at the dorsal midline of the embryo. Finally, all the epidermal cells retract to reassume their original polygonal cell shape.

Failure to complete the DC process is lethal and is characterized by a lack of the dorsal epidermis that is easily recognized in preparations of the embryonic cuticle ('dorsal open phenotype', see Figure 5). The analysis of one group of dorsal open mutants established the essential role for JNK signaling in DC. This field was pioneered by the work of Glise and Noselli (1995) with the analysis of mutants in hemipterous (hep; Glise et al., 1995). Molecular cloning and biochemical analysis showed that hep encodes the MKK7 homologue (a kinase of the JNKK type) of Drosophila (Holland et al., 1997). Cuticles of hep mutant embryos are dorsal open and epidermal cells are generally less elongated compared to wild-type epidermal cells.

The charting of this pathway (Figure 3) was extended by the characterization of the Drosophila JNK1 homologue Basket (Bsk; Riesgo-Escovar et al., 1996; Sluss et al., 1996). As in the case of hep, bsk mutants display a dorsal open phenotype due to defects in epidermal cell elongation. Experiments performed in vitro indicate that the JNKK Hep phosphorylates and thereby activates the JNK Bsk, which in turn becomes competent to phosphorylate D-Jun (Sluss et al., 1996).

Currently, the transcription factors D-Jun (Hou et al., 1997; Kockel et al., 1997; Riesgo-Escovar and Hafen, 1997b), D-Fos (Riesgo-Escovar and Hafen, 1997a; Zeitlinger et al., 1997) and Yan (also referred to as Anterior open, Aop; Riesgo-Escovar and Hafen, 1997b) have been shown to function downstream of Bsk in DC. Mutants of D-jun, and of D-fos, phenocopy hep and bsk with respect to cuticular phenotypes and epidermal cell elongation (see below). In contrast, the Ets factor Yan appears to act as a repressor for JNK signaling (Riesgo-Escovar and Hafen, 1997b). Removal of functional yan rescues the dorsal open phenotype of bsk embryos to a considerable extent.

The functional relationships between D-Jun, D-Fos, Bsk and Hep are also reflected by changes of target gene expression in embryos mutant for these genes. Both the BMP2/BMP4 (Bone Morphogenic Protein) homologue Dpp and the VH1 type phosphatase Puckered (Puc) are expressed in cells of the LE during DC (Figure 2; Ring and Martinez-Arias, 1993; St Johnston and Gelbart, 1987). This expression depends on an intact JNK signaling cascade since dpp and puc mRNA and enhancer trap activity is absent in hep, bsk, D-jun and D-fos mutant embryos (Martin-Blanco et al., 1998; Riesgo-Escovar and Hafen, 1997a; Sluss and Davis, 1997; Zeitlinger et al., 1997). Conversely, both transcriptional targets are ectopically induced in JNK gain of function contexts (Riesgo-Escovar and Hafen, 1997b). Consistent with Yan acting as an inhibitor of JNK-dependent transcription, dpp and puc mRNAs are also ectopically expressed in yan¹ mutants and reduced in embryos expressing a constitutively active, undegradable Yan protein. All available data establish puc and dpp as bona fide target genes of JNK signaling, although the direct action of AP-1 proteins on the regulatory regions of either gene remains to be proven (Riesgo-Escovar and Hafen, 1997b). In combination with the dorsal open cuticle, the absence of dpp and puc expression in the leading edge serves as a molecular marker for deficient JNK signaling.

By this stringent standard, the Protein Kinase C related kinase Pkn (Agnes et al., 1999; Martin-Blanco et al., 2000; Tateno et al., 2000; Zeitlinger and Bohmann, 1999) and the adapter protein Myoblast City (Mbc), although their mutant alleles give rise to defective DC phenotypes, are not components of the core JNK signal transduction pathway because the expression of dpp remains unaltered when their respective genes are mutated (Erickson et al., 1997; Lu and Settleman, 1999; Nolan et al., 1998).

Drosophila AP-1 couples Bsk activity with the expression of a Bsk inhibitor, Puc generating a negative transcriptional feedback loop. This negative feedback circuit might well be involved in fine-tuning and/or setting a precise time window of JNK activity during the DC process. puc mutant embryos consistently display all the features of a JNK gain of function phenotype such as ectopic expression of the target genes puc and dpp, as well as persistent elongation of epidermal cells at time points where wild-type cells would have already readopted their polygonal cell shape. Negative transcriptional circuits involved in regulation of signal transduction pathways represent a common theme during development (Freeman, 2000; Perrimon and McMahon, 1999). The posttranslational control of MAP kinases by inducible dual specificity phosphatases of the VH1 class might be a conserved motif in evolution, as similar mechanisms are described for yeast and mammals (Keyse, 1998).

No mutants for Drosophila homologues of a MAPKKK with a dorsal open phenotype, which would complete the three-tiered cascade found in 'classical' MAPK pathways, have been described so far. However, overexpression of the Drosophila homologue of TAK (TGF- activated kinase, a member of the MAPKKK group) in the embryonic epidermis results in ectopic activation of the JNK target genes dpp and puc (Mihaly et al., 2001; Takatsu et al., 2000). D-TAK is likely to act through JNKK/Hep and JNK/Bsk, as the rough eye phenotype of D-TAK expression is suppressed by mutations in the hep and bsk genes. Furthermore, Bsk is hyperphosphorylated in D-TAK overexpressing larvae, indicating that D-TAK, when provided exogenously, is sufficient to activate the JNK signal transduction pathway in flies. Nevertheless, many MAPKKK homologues are known to induce JNK signaling in mammalian tissue culture models, raising concerns about the specificity and physiological relevance (Davis, 2000; Fanger et al., 1997). Is D-TAK the physiological mediator of JNK signaling during DC? Expression of a dominant negative form of D-TAK in the dorsal region of the embryo gives rise to a dorsal open phenotype, albeit at very low penetrance (Takatsu et al., 2000). Further studies and especially the identification of a D-tak mutant are required to resolve the in vivo requirements for MAPKKKs during DC.

The situation is more clear in the case of the candidate MAPKKKK Misshapen (Msn), an Ste20 related enzyme. msn mutants manifest a DC phenotype. Genetic interactions with bsk and hep indicate that Msn acts in the same pathway, a conclusion supported by the absence of dpp expression in msn mutant embryos (Su et al., 1998). Ste20 kinases are considered to act upstream of kinases of the Ste11 type (Herskowitz, 1995) and are divided into two subgroups: PAKs (p21 activated kinases) and SPS1 proteins. PAKs are activated by binding directly to small GTPases such as Rac and Cdc42, while SPS1 proteins do not interact with small GTPases. By homology, Msn belongs to the SPS1 group, as it lacks a p21Rac and Cdc42 interaction domain and does not associate with D-Rac1 (Su et al., 1998).

Based on experiments with dominant forms of D-Rac and D-Cdc42, small GTPases have been suggested to play a role in DC (see below). Embryos expressing dominant negative D-Rac1 and D-Cdc42 display defective DC phenotypes (Harden et al., 1995; Riesgo-Escovar and Hafen, 1997b), whereas constitutively active forms result in ectopic activation of the JNK target genes dpp and puc (Glise and Noselli, 1997). The expression of target genes by constitutively activated D-Cdc42 requires intact hep, suggesting an upstream requirement of D-Cdc42 in JNK signaling (Glise and Noselli, 1997).

The proposed requirement for D-Rac1/2 in JNK activation raises the question of the epistatic relationship towards Msn, which does not interact with D-Rac1 but activates Bsk. Su et al. (1998) suggest a convergence of two signal transduction branches on an (as yet unidentified) MAPKKK: One coming in from D-Rac and D-Cdc42 and may be mediated by a Drosophila PAK homologue (Harden et al., 1996), the other branch involves Msn.

Nevertheless, two recent studies have brought the role of D-Cdc42 in JNK signal transduction into question (Genova et al., 2000; Ricos et al., 1999). In a careful analysis of dominant negative D-Cdc42 expressing embryos, Ricos et al. (1999) concluded that the cellular phenotypes of D-Cdc42DN resemble the loss of Dpp, characterized by epidermal 'bunching' rather than JNK signaling. Consistently, expression of constitutively activated D-Cdc42^V12 suppresses the dorsal open phenotype of embryos mutant for the Dpp receptor thick veins (tkv), supporting the view of D-Cdc42 acting downstream of Tkv mediated Dpp signaling. Additionally, dpp expression in the leading edge is clearly detectable in embryos derived from D-cdc42⁴/ D-cdc42⁶ mutant mothers (Genova et al., 2000). Bearing in mind that the allelic combination used might not lead to the complete elimination of D-Cdc42 function, this result supports the hypothesis that D-Cdc42 is not essential for JNK activation in Drosophila. These data leave the two known members of the Drosophila Rac subfamily and D-Ral as the only candidates for small GTPases mediating JNK activation during DC (Glise and Noselli, 1997; Sawamoto et al., 1999). Inasmuch as this conclusion is based entirely on the use of ectopic expression of dominant alleles and considering the surprising result obtained byD-Cdc42 mutants, loss of function studies are required for clarification.

In two recent studies, non-receptor tyrosine kinases entered the stage of DC. Mutations in the shark gene (SH2 domain akyrin repeat kinase) result in dorsal open embryos and loss of dpp expression in LE cells (Fernandez et al., 2000). As forced expression of activated c-Jun can rescue this phenotype at least to a certain extent, Shark was suggested to act upstream of Drosophila AP-1. However, the exact position of Shark, which is localized to the nucleus, within the JNK signaling pathway remains unknown.

Src42A and Tec29, two additional non-receptor tyrosine kinases were reported to regulate JNK activity (Tateno et al., 2000). Interestingly, the zygotic removal of each gene independently gives rise to a relatively mild head involution defect only (Roulier et al., 1998; Tateno et al., 2000). However, src42A, tec29 double mutant embryos show loss of dpp and puc expression in the leading edge and display a dorsal open phenotype. Conversely, expressing a dominantly activated Src42A in the wing pouch gives rise to ectopic puc enhancer trap activity, which is in part dependent on the presence of functional hep. The data indicate that signaling via non-receptor tyrosine kinases of the Tec and Src families is required and sufficient for the expression of JNK target genes and suggests their requirement upstream of Hep.

The upstream portion of the Drosophila JNK pathway is ill-characterized, and extra-cellular signals that induce it during DC remain hypothetical. However, the raw gene which, when mutated, causes defective DC may shed some light on this situation. Loss of function mutants of raw which encodes a pioneer protein, lead to an expanded dpp and puc expression domain in the lateral epidermis (Byars et al., 1999). This ectopic gene expression is D-jun-dependent, suggesting an upregulated JNK signaling activity in raw mutant backgrounds. Interestingly, on the basis of enhancer trap activities raw is expressed exclusively in the amnioserosa, raising the question of how an apparently non-secreted protein can control gene expression in several rows of cells of the adjacent epidermis. The authors propose two, equally plausible models: Either raw function restricts the range of a JNK activator, or it is involved in the production of a long range repressor of JNK signaling. Whatever the case, this report clearly points to the amnioserosa as a source of regulators of JNK activity during DC.

The only obvious candidate for a receptor of such a signal is the Drosophila insulin receptor (Inr). Inr mutant embryos display aberrant germ band retraction and a dorsal open phenotype (Fernandez et al., 1995). Nevertheless, neither biochemical nor genetic interaction experiments with other members of the dorsal closure pathway have been performed so far in order to address the function of Inr in DC. It is tempting to speculate about a pathway that might be triggered by insulin (or a related factor) as an autocrine signal that triggers JNK activation, as has been previously suggested by Glise et al. (1997).

Cytoskeletal regulation: cell matching and segmental patterning

It is becoming apparent that cytoskeletal control is essential for the accurate alignment, movement, and fusion of cells that occur during DC. The following section highlights the recent advances in understanding the forces, morphogenetic changes, and signal transduction mechanisms involved in DC. These studies provide new insight into the mechanics of cell sheet movement, imply a role for the JNK pathway in the alignment of cellular segments and characterize the Wnt/Wingless cascade (which is classically involved in embryonic segmental patterning) as a pathway that interacts with JNK in DC.

To better understand the molecular mechanisms driving cell sheet movement, one must first consider the nature of the forces involved. The forces responsible for cellular movement during DC were recently described in a study by Kiehart et al. (2000). In an elegant series of laser ablation experiments, they found that the cells of the LE as well as those of the amnioserosa contribute to tension that drives DC while the lateral epidermis retards the progression of the dorsal epidermis. These results rule out models where the lateral epidermis actively pushes the cells of the leading edge towards the dorsal midline. In addition, they showed that the actin cable in the apical end of the LE cells is not alone responsible for the driving force of DC, as DC continues despite cutting the cable with a laser. Therefore, it is clear that the morphogenetic movements in DC are not the result of a single tensile force, but rather result from the summation of the forces generated by the various players involved (amnioserosa, actin cable and lateral epidermis). This study does not exclude the possibility of other forces acting in DC, such as putative contractions of filopodia arising from LE cells (see below).

Recently, it has been shown that proper fusion of the two contralateral cell sheets in the final stages of DC is dependent on filopodial and lamellipodial extensions from cells of the LE (Jacinto et al., 2000). These protrusions were observed by expressing green fluorescent protein (GFP)-tagged actin in otherwise wild-type embryos and taking successive confocal images of the DC process. Filopodia from LE cells on one side were shown to reach out and contact their opposing partners and then possibly contract, as evidenced by a localized chinking in the opposed actin cables towards each other. In contrast to the coordinated and precise pairing of segments on each side of the epidermis observed when actin-GFP is expressed under engrailed-Gal4 (similar to LacZ expressed in a wingless pattern, Figure 4), coexpression of dominant negative Dcdc42 resulted in misalignment of the segments as a result of cell-cell mismatch. In this regard, filopodia are viewed as cell sensors that actively seek out the appropriate interaction partner and mediate the correct segmental alignment of the two epidermal sheets. Interestingly, the same analysis in embryos mutant for the JNKK hep demonstrated a complete lack of filopodia and lamellipodia. Also in these embryos, misalignment was observed where segments were actually able to meet at the posterior end. Since Dcdc42 acts downstream of Dpp (Ricos et al., 1999 and see above), whose transcript is upregulated by JNK's activation of AP-1 (Glise and Noselli, 1997; Riesgo-Escovar and Hafen, 1997b; Zeitlinger et al., 1997), AP-1 is likely to be involved in epidermal segment alignment through its activation of dpp.

The observation of initial LE cell movement in hep mutant embryos introduces an alternative view of the role for the JNK pathway in DC (Jacinto et al., 2000). It has been previously proposed (Hou et al., 1997; Riesgo-Escovar and Hafen, 1997a,b) that the JNK pathway is indispensable for the initiation phase of LE cell stretching and that embryos mutant for JNK pathway members do not exhibit stretching. However, data from Jacinto et al. (2000) show that there is an initial dorsal movement which results in mismatched segments, albeit in the posterior-most regions. A possible explanation for this discrepancy could be that one of the hep mutant alleles used by Jacinto et al. (2000) was a hypomorph and thus provided enough wild-type function to explain the observed cellular movement. However, careful confocal analyses of jun and fos mutant embryos (Zeitlinger et al., 1997; Kockel et al., 1997), as well as jun mutant germline clones (Figure 5) show that initial LE cell stretching does indeed occur in embryos devoid of any functional D-Jun or D-Fos protein. Therefore, initial LE cell stretching takes place in embryos deficient in AP-1 signaling, however this initial cell-elongation terminates prematurely and results in the cellular phenotype previously described for these mutants.

The signal that activates JNK in the LE is currently not known. However, a reversal of cause and consequence concerning JNK signaling and cell stretching offers an interesting possibility. If LE cell stretching occurs in the absence of JNK signaling, then it is conceivable that the stretching itself creates a stress on the cell that activates JNK. Cell stretch-induced JNK activation has been observed in vivo in endothelial cells following balloon angioplasty (Watson et al., 2000) and in mammalian tissue culture (Basdra et al., 1994; Li et al., 1996). Future research is likely to settle the debate on cell stretching and JNK signaling, and ultimately provide an explanation for the JNK activation stimulus, whether it is due to actin cable-mediated cell stretching or some as yet identified signal possibly acting through a cell surface receptor.

The need for proper segmental matching in DC suggests that signaling cascades classically involved in establishing spatial information along embryonic segments might also be involved in the process. Recent data implicate the Wnt/Wingless pathway, known for its role in embryonic segmental patterning, in DC (McEwen et al., 2000). Interestingly, embryos mutant for armadillo (arm), a component of the Wg signal transduction pathway, display a dorsal open phenotype in addition to the 'classical' defects in segment polarity. McEwen et al. (2000) observed that mutations in puc are able to suppress the DC defects of arm mutants. This initial observation led them to investigate the interactions of JNK and Wg pathways in DC and segment polarity. Subsequently, they found that artificial activation of the JNK pathway suppresses the armadillo phenotypes. wingless signaling mutants were shown to be defective in dpp expression in the LE, normally a hallmark of impaired JNK activity. Conversely, providing Dpp exogenously to armadillo mutants is sufficient to rescue its dorsal open phenotype. Taken together, these data show that the JNK and wingless pathways interact in DC, at least on the level of target gene expression. McEwen et al. (2000) note that activation of Wg signaling is not sufficient to rescue the DC defect of kay mutant embryos, suggesting its permissive role for dpp expression in the cells of the LE. Nevertheless, the molecular nature of the JNK-Wg interaction is not clear at present. The authors propose that the two pathways may operate in parallel or Disheveled may act as a branch point that activates either JNK or Armadillo. Indeed, Disheveled has been previously shown to mediate signaling from Frizzled, (a wingless receptor) to JNK in establishing tissue polarity in the Drosophila eye (Boutros et al., 1998 and see below).

Thorax closure: variation of the same theme?

The initial analysis of mutants defective in JNK signaling established the requirement for this pathway in epidermal morphogenesis in the Drosophila embryo. Interestingly, already in the first reports on the phenotypic abnormalities of hep or D-fos mutants, a malformation in thorax morphology had been described (Glise et al., 1995; Riesgo-Escovar and Hafen, 1997a; Zeitlinger et al., 1997). More specifically, the dorsal thorax, also known as the notum, of such mutant flies showed a cleft along the dorsal midline. In extreme cases, a deep furrow separates the two malformed heminota (Figure 6). These observations indicate a requirement for the JNK signal transduction pathway in thorax closure, in addition to its role in DC. However, thorax closure takes place later in Drosophila development, during metamorphosis, and is distinct and independent of the embryonic DC process.

Drosophila melanogaster is a holometabolous insect, i.e. the adult body plan is formed during metamorphosis. At this stage, a puparium is formed and the larval tissue becomes histolysed and replaced by the imaginal epithelia. The notum of the adult fly is derived from the two wing imaginal discs (Usui and Simpson, 2000). These discs are comprised of epithelial sacs that had been set aside early in embryonic development. Imaginal discs are divided into three structures: the imaginal disc proper (which consists of columnar cells), the peripodial membrane (a squamous epithelium) and the peripodial stalk (which attaches the disc to the larval epidermis), (Usui and Simpson, 2000). During larval stages, wing imaginal sacs proliferate and become patterned to form the anlagen of the wing and thorax (Blair, 1995; Brook et al., 1996). In the process of thorax closure, the proximal parts of the two contralateral wing imaginal discs approach each other and join to form the continuous notum of the adult fly. Thorax closure proceeds in four steps (Figure 7): Eversion of the imaginal disc epithelia through the peripodial stalk, dorsal migration of disc epithelia over larval tissue, cell shape changes towards the future dorsal midline and, finally, the fusion of the two heminota (Usui and Simpson, 2000).

Since thorax closure is frequently compared to DC, it is important to highlight the similarities and differences of these morphogenetic events. On the one hand, both processes share common regulatory circuits (discussed in more detail below). Both thorax and dorsal closure are defined by the approach of two contralateral cell layers that meet and fuse at a dorsal midline in order to form a continuous structure. Moreover, in both cases, the approaching cell layers differentiate a leading edge that is characterized by the expression of dpp and puc. In addition, both leading edges extend filopodia to the contralateral side (Jacinto et al., 2000; Martin-Blanco et al., 2000).

On the other hand, DC is driven largely by cell shape changes in the epidermis and amnioserosa, whereas thorax closure relies on cell migration over larval tissue and cell stretching of a few rows of cells at the final stages only (Martin-Blanco et al., 2000; Usui and Simpson, 2000; Zeitlinger and Bohmann, 1999). The leading edge in dorsal closure consists of a well-defined single row of epidermal cells, resembling a sharp boundary between the lateral epidermis and the squamous cells of the amnioserosa. In contrast, the leading edge of wing imaginal discs in thorax closure contains columnar epithelial cells of the disc proper as well as squamous cells of the former peripodial membrane, and is several rows of cells wide (Usui and Simpson, 2000).

Defective thorax closure manifests itself late during development. Thus, the genetic dissection of this process is hampered by the fact that mutations in putatively important genes result in lethality earlier in development. For example, embryos mutant for 'classical' JNK pathway components die in DC (see above). For this reason, thorax closure phenotypes cannot be observed in all mutants that have DC phenotypes. They are only apparent in some alleles of D-fos, hep and src42A (Figure 6) (Agnes et al., 1999; Martin-Blanco et al., 2000; Tateno et al., 2000; Zeitlinger and Bohmann, 1999).

Ectopic expression of the negative JNK regulator Puc phosphatase in the dorsal thorax anlagen results in a phenocopy of the thoracic cleft typical for D-fos and hep mutants. Furthermore, removal of one functional copy of puc rescues the thorax closure defects of flies partially defective in the kay or hep loci (Agnes et al., 1999; Martin-Blanco et al., 2000; Zeitlinger and Bohmann, 1999). Therefore, as in DC, Puc acts as a negative regulator of JNK signaling also in thorax closure.

Taken together, it appears that evolution has reused major parts of the signaling circuitry that regulates DC for thorax closure (or vice versa). This conservation extends to the non-receptor tyrosine kinases Shark and Src42A (Fernandez et al., 2000; Tateno et al., 2000). Partial loss of function shark and src42A flies show a thoracic cleft similar to hep and kay mutants. More strikingly, the strength of the thorax closure phenotype of src42A is enhanced by mutations in hep and suppressed by puc (Tateno et al., 2000).

All the mentioned similarities notwithstanding, one significant difference between dorsal and thorax closure is that transcription of dpp is not regulated by JNK in the latter (Agnes et al., 1999). However, Dpp signal transduction is required for thorax closure as it is in DC, as mutations in the Dpp receptors punt and tkv (Chen et al., 1998; Morimura et al., 1996; Simin et al., 1998) and the signaling transducer medea (Hudson et al., 1998) show thorax cleft phenotypes. Additionally, dpp is expressed in a similar domain as puc in the leading edge (Agnes et al., 1999). The analysis of flies expressing dominant negative thick veins in the region of the notum anlagen reveals a surprising similarity to the phenotypes elicited by the loss of Dpp signaling during DC (Martin-Blanco et al., 2000). In such animals, no filopodia are observed and the tissue of each heminotum contracts towards aberrant points at its own leading edge, giving rise to epidermal bunching. Consequently, although the two pathways are transcriptionally uncoupled, JNK and Dpp signaling act jointly to facilitate morphogenetic movements.

A common phenotype caused by loss of JNK signaling in DC and thorax closure is the detachment of an epithelium from the larval tissue (thorax closure) and from the amnioserosa (dorsal closure). This could mean that an important (and possibly the main) function of the JNK signal and AP-1 in these morphogenetic processes is to activate a gene expression program which generates a state of cell adhesion and cytoskeletal flexibility that is compatible with the observed cell shape changes and tissue movements. Whether JNK and AP-1 provide a permissive function as such a model suggests, or an instructive signal for thorax and DC remains to be addressed in the future.

JNK signaling and apoptosis

In mammals, JNK-mediated signaling has been reported to regulate apoptosis (Davis, 2000; Ip and Davis, 1998). In Drosophila, this issue has been addressed by analysing the role of JNK signaling in cell death during the development of the wing and the eye (Adachi-Yamada et al., 1999; Takatsu et al., 2000). Interestingly, JNK-induced cell death provides yet another example of the interplay between the JNK pathway and other signal transduction cascades, in this case, Decapentaplegic and Wingless.

The Dpp and Wg pathways are involved in the establishment of the anterior-posterior and dorsal-ventral axes, respectively. Both ligands are thought to act as morphogens as they specify cell fates in a concentration-dependent manner. The intersection of wg and dpp expressing cells defines the distal tip of the adult wing blade.

In addition to effects on wing patterning, viable combinations of hypomorphic (partial loss of function) dpp alleles and mutations with reduced Wg signaling activity show loss of the most distal wing structures, recognizable as visible notches in the adult wing margin (Adachi-Yamada et al., 1999). In the respective area of the wing imaginal disc, acridine-orange positive cells (an indicator of cell death) and an upregulation of puc are observed. The latter indicates the activation of JNK signaling. Augmentation of JNK signaling in these contexts by removing one functional copy of puc enhances the number of apoptotic cells and the loss of distal tissue. Interestingly, activation of JNK signaling alone is sufficient to trigger cell death, as the expression of an activated form of Hep leads to the appearance of acridine-orange positive cells in the wing disc and induces a drastic loss of adult wing tissue. Apoptosis caused by a reduction in Dpp signaling shows a strict requirement for JNK signaling, as removal of hep suppresses the cell death phenotype of certain viable mutants in the Dpp pathway.

These data indicate that reduction of Dpp (and perhaps Wg) signaling induces JNK-mediated apoptosis, suggesting a role for Dpp (and Wg) as a survival factor that suppresses a latent JNK activity. However, this is not universally true, as only cells located in the future distal area of the wing show this kind of response. Furthermore, ectopic activation of the Dpp signal transduction pathway induces Hep-dependent cell death in more proximal parts of the wing disc. Therefore, cells which are exposed to an improper dose of Dpp signaling within the activity gradient (too high or too low) undergo JNK-dependent apoptosis (Adachi-Yamada et al., 1999).

Also in the Drosophila eye imaginal disc, activation of JNK induces cell death, resulting in an eye ablation phenotype (Takatsu et al., 2000). Under these circumstances, the proapoptotic genes head involution defective (hid) and reaper (rpr) are transcriptionally upregulated. Consistent with a requirement for the endogenous gene activity of hid and rpr, simultaneously lowering the gene dosage by introducing a deficiency in hid, rpr and a third proapoptotic gene grim, restores the eye size significantly (Takatsu et al., 2000). This study suggests that JNK activation leads to the transcription of hid and rpr, both of which have been shown to suffice for the execution of an apoptotic program by inactivating the inhibitor of apoptosis proteins (DIAPs) (Goyal et al., 2000; Lisi et al., 2000; Wang et al., 1999). JNK-induced apoptosis in the Drosophila eye therefore requires transcription, even though it is not clear whether AP-1 is involved and if it induces hid or rpr directly. The potential involvement of D-Fos in the death of Drosophila photoreceptors is reminiscent to reports on c-Fos knock-out mice in which light-induced apoptosis of photoreceptors is defective (Hafezi et al., 1997). It is tempting to speculate that JNK signaling to AP-1 might be a conserved mechanism to control apoptosis in the eye.

The role of D-Fos in endoderm induction

The Drosophila embryonic midgut consists of two cell layers: the visceral mesoderm, forming the outer cell layer, and the subjacent endoderm, which initially does not carry any anteroposterior positional information (Lawrence, 1992). Therefore, the transcriptional initiation of the homeotic gene labial (lab) in the endoderm by the overlaying mesoderm has been a long-standing model system for an inductive process between two germ layers (Bienz, 1997).

The series of events leading to the induction of lab expression is initiated by the homeotic gene ultrabithorax (ubx) in the visceral mesoderm, which stimulates the expression of Dpp directly. Within the visceral mesoderm, Dpp induces the expression of three genes: first, it positively feeds back on ubx expression, second, it causes the expression of Wg in the neighboring parasegment, and third it stimulates the synthesis of the EGF-like ligand Vein. In turn, the three secreted molecules Dpp, Wg and Vein regulate the endodermal expression of lab (Immerglück et al., 1990; Panganiban et al., 1990; Szüts et al., 1998).

D-Fos is expressed broadly in the endoderm but not in the overlaying visceral mesoderm. This expression pattern is dependent on Dpp, as it is absent in dpp mutant embryos and expanded when Dpp is ubiquitously overexpressed (Brand and Perrimon, 1993; Riese et al., 1997). Furthermore, experiments using dominant-negative and constitutively activated alleles suggest that EGF receptor signaling and D-Fos itself might be required for the endodermal transcription of D-fos (Szüts and Bienz, 2000b).

D-Fos contributes to the Dpp-dependent activation of lab. The expression of D-fos itself remains unaltered in a lab mutant background, while dominant negative D-Fos reduces lab expression in the midgut (Riese et al., 1997). Additionally, the examination of embryos homozygous for a deficiency in D-fos (Zeitlinger et al., 1997) revealed that the expression of Labial is diminished in such a mutant background (Szüts and Bienz, 2000b). However, the loss of D-fos does not result in a complete absence of Labial positive cells in the embryonic endoderm, which might have been the reason why this effect passed unnoticed previously (Riesgo-Escovar and Hafen, 1997a). Furthermore, labial expression is embedded in a complex regulatory network, including a positive feedback loop with Labial activating the lab gene (Bienz, 1997; Tremml and Bienz, 1992). Therefore, the removal of a single component like D-Fos results only in a partial reduction in lab expression. In summary, the studies suggest that D-Fos cooperates with parallel signaling events triggered by Wg and Dpp in order to induce and maintain labial expression.

Eye development

The development of the Drosophila eye and especially the differentiation and survival of photoreceptors (R cells), has served as a paradigm for the molecular dissection of RTK-Ras signaling in an in vivo system (Bergmann et al., 1998; Raabe, 2000).

The adult eye consists of about 700-800 regularly arranged ommatidia or facets, each comprised of 11 accessory cells and eight neuronal photoreceptor cells. These ommatidial clusters differentiate and assemble during larval and pupal stages of development. Within a single facet, the photoreceptors containing the light-harvesting organelles called rhabdomeres are arranged in an asymmetrical, trapezoidal pattern (Wolff and Ready, 1993).

Ras-mediated receptor tyrosine kinase signaling is a prerequisite for the neuronal differentiation of photoreceptor cells. In combination with biochemical experiments, genetic analyses have led to a molecular blueprint of the Ras signal transduction pathway (Figure 8; Raabe, 2000). Briefly, the RTKs Sevenless (required for the differentiation of the R7 photoreceptor only) and the EGF receptor (EGFR) trigger the activation of Ras via the adaptors Drk and Dos and the GDP-GTP exchange factor Sos. In turn, Ras stimulates D-Raf, the first of three kinases comprising the 'classic' MAPK cascade which become sequentially activated by phosphorylation. The remaining two kinases are the MAPKK DSor which activates the ERK type MAPK encoded by the rolled gene.

Three nuclear substrates of Rolled have been implicated in the transduction of the Sevenless signal. The Ets protein Pointed P2 acts as a positive regulator of photoreceptor cell fate and has been shown to be a substrate of Rolled in vitro (Brunner et al., 1994a; O'Neill et al., 1994). The transcriptional inhibitor Yan (also named Aop), an Ets domain protein as well, acts as a negative regulator. Upon Rolled-mediated phosphorylation, Yan relocates to the cytoplasm and is rapidly degraded (Rebay and Rubin, 1995). The third transcription factor, D-Jun, has been reported to be a Rolled substrate in vitro and in vivo. (Peverali et al., 1996).

Several experiments suggest that D-Jun is a downstream effector of the Ras pathway. Drosophila Jun is expressed in the eye imaginal discs in a transient pattern which reflects the time and location where Ras signaling is activated to induce cell differentiation (Bohmann et al., 1994). Expression of a dominant-negative allele of D-Jun, comprised of the leucine zipper and basic domains only, results in a weak loss of photoreceptor phenotype and also suppresses gain of function phenotypes elicited by activated Ras pathway components (Bohmann et al., 1994). Additionally, activated versions of c-Jun (containing aspartic acid residues at MAP kinase phosphorylation sites, see Papavassiliou et al. (1995) for reference and D-Jun (N-terminal fusion to the VP-16 transactivation domain) induce R7 differentiation independent of the Sevenless RTK (Peverali et al., 1996; Treier et al., 1995). Since Rolled, downstream of Ras, phosphorylates and activates D-Jun (Peverali et al., 1996), it is clear that Ras may affect D-Jun activity.

Studies with D-jun mutant alleles confirmed that D-Jun may function as a mediator of Ras signaling during eye development (Kockel et al., 1997). However, complete removal of Jun in patches of eye disc tissue has only a very mild effect, indicating that D-Jun is not essential for photoreceptor recruitment and acts redundantly in this context (Kockel et al., 1997; Weber et al., 2000). It is presently unclear which other nuclear factors might be the mediators of the Ras signal that have a redundant relationship to D-Jun. Obvious candidates include Pnt and possibly D-Fos. Indeed, transfection experiments and genetic interaction studies indicate that D-Jun can cooperate with Pnt, possibly in a redundant fashion (Treier et al., 1995; Kockel, unpublished results).

Like D-Jun, D-Fos is expressed in a dynamic pattern behind the morphogenetic furrow of the eye imaginal disc (Zeitlinger, unpublished), suggesting that D-Fos might have a function during eye development as well. Indeed, a dominant negative form of Fos interferes with Sev-dependent photoreceptor differentiation (Bohmann et al., 1994). However, genetic analysis of D-fos mutants in eye development has been hampered by the unavailability of suitable mutant alleles. For example, kay¹ represents a deficiency and might be defective in other genes (Zeitlinger et al., 1997). All attempts to obtain homozygous kay¹ clones in eye or wing imaginal discs have failed and produced wild-type twin spots only. The second allele, kay², is a hypomorph and results in adult escapers without any eye phenotype.

The role of D-Jun and D-Fos in the developing eye is further complicated by the finding that these factors also mediate the response to a different signal transduction pathway that coordinates the orientation of ommatidia in the plane of the retina (see below).

Planar polarity

The JNK signal transduction pathway and AP-1 have been implicated in a second, independent patterning event during the course of eye development, namely the establishment of ommatidial planar polarity (Mlodzik, 1999; Strutt and Strutt, 1999). Already during late larval stages, the eye imaginal disc represents a highly organized epithelium displaying pronounced apical-basolateral asymmetry. Moreover, a second level of polarization within the plane of the epithelium can be observed. In the adult Drosophila eye, this planar polarity becomes evident by the mirror image orientation of the mature ommatidial clusters relative to the dorso-ventral midline, the so-called equator. This arrangement of facets is acquired during the course of photoreceptor differentiation, when R-cell precursors are recruited pairwise into the maturing ommatidia. Initially, the ommatidia exhibit an axial symmetry parallel to the dorso-ventral midline, but then rotate 90° in opposite directions on either side of the future equator and become asymmetric as differentiation proceeds.

The establishment of planar polarity and the precise determination of the equator is orchestrated by complex signaling involving counteracting gradients of Wg (Wehrli and Tomlinson, 1998), Jak/Stat (Zeidler et al., 1999b, 2000a) and Notch signaling activities (Blair, 1999; Cho and Choi, 1998; Dominguez and de Celis, 1998; Papayannopoulos et al., 1998). These combined signals ensure correct patterning across distances of several hundreds of cells in the eye field (Bray, 2000; Mlodzik, 1999; Strutt and Strutt, 1999). This series of signaling events culminates in the expression of one or several, long range acting 'factor(s) X' at the equator, which direct the ommatidial rotation (Bray, 2000; Zeidler et al., 1999a, 2000b).

The Frizzled (Fz) serpentine trans-membrane receptor has been suggested to act as a receptor of such a signal within the R3/R4 photoreceptor pair (Zheng et al., 1995). As the R3 cell precursor is recruited to a position closer to the equator, it receives a higher dose of the putative signal factor(s) X than R4, and this difference subsequently directs the rotation of the ommatidia (Cooper and Bray, 1999; Fanto and Mlodzik, 1999). Alternatively, each of these cells might sense the difference across their diameter individually (Tomlinson and Struhl, 1999). Whatever the case may be, in agreement with both hypotheses, compromising the differential Fz activity by loss of fz or forced expression of Fz protein in the R3/R4 pair results in randomly rotated ommatidia. This demonstrates the requirement of Fz for the establishment of the planar polarity in the Drosophila eye (Fanto and Mlodzik, 1999; Gubb, 1993; Theisen et al., 1994; Zheng et al., 1995). The family of Frizzled proteins are receptors for Wg/Wnt ligands (Bhanot et al., 1996; Bhat, 1998; Cadigan et al., 1998; Kennerdell and Carthew, 1998; Muller et al., 1999). Surprisingly however, none of the functionally characterized Drosophila Wnts are able to account for the Frizzled-mediated planar polarity. The only remaining candidate for factor(s) X so far, Four-jointed (Zeidler et al., 1999a, 2000b) does not reveal any homology to the Wnt family.

Nonetheless, an outline of a signaling pathway for planar polarity downstream of Fz arises from a series of genetic and biochemical studies (Figure 8) (Boutros and Mlodzik, 1999; Shulman et al., 1998). The multi-domain protein Disheveled (Dsh) appears to act as a downstream component of Fz since dsh loss of function alleles suppress the planar polarity defects caused by the forced expression of Fz (Boutros et al., 1998; Strutt et al., 1997). Strikingly, Dsh serves as a branchpoint between 'classical' Fz-mediated Wg signaling which mediates, for example the patterning of the embryonic cuticle or the wing blade (see above) and planar polarity signaling. The removal of components acting downstream of Dsh in the canonical Wg signaling cascade such as armadillo (Peifer et al., 1991) or pangolin/TCF (Brunner et al., 1997) do not elicit planar polarity defects (Axelrod et al., 1998; Boutros et al., 1998). Instead, the available data indicate that a planar polarity signal is mediated via the JNK cascade. Dsh expression activates JNK signaling in tissue culture, and mutations in hep, bsk or D-jun suppress the gain of function phenotype caused by Dsh overexpression. Conversely, the activation of JNK signaling rescues the polarity phenotype of eyes with impaired dsh function (Boutros et al., 1998; Strutt et al., 1997).

In accordance with the presented model of JNK-mediated planar polarity signaling, dominant interference with JNK signaling activity imposes a tissue polarity phenotype on the eye. Ectopic expression of dominant negative, as well as constitutively active Rac^V14 and Rho^V14 (Fanto et al., 2000), Msn^wt (Paricio et al., 1999), Hep^act, dominant negative and wt Bsk, and c-Jun^act (Weber et al., 2000) in the eye give rise to defective ommatidial rotation, which is reminiscent to gain- and loss of function fz and dsh flies. This view is complemented by the planar polarity phenotypes of rhoA (Strutt et al., 1997), msn (Paricio et al., 1999) and D-jun mutants (Weber et al., 2000).

However, the complete removal of hep function from developing eye tissue does not show patterning defects at all. In addition, the phenotypic lesions of bsk and D-jun are mild and are observed at low penetrance. These results suggest that JNK plays a redundant role in this process. As such, these findings suggest the existence of an alternative route acting in parallel to JNK in planar polarity signaling. Candidates for players in this pathway may involve MKK4 and the two p38 kinase homologues (Martin-Blanco, 2000; Paricio et al., 1999; Weber et al., 2000).

Taken together, it appears that a localized difference in JNK pathway activity (presumably between the R3 and the R4 cell of an ommatidium) determines the direction and the extent of rotation. If the signal is either too high or too low in both cells the difference cannot be sensed anymore and a misrotation phenotype ensues. A target gene of Fz and JNK-mediated planar polarity signaling that has been suggested to be involved in the sensing of this asymmetry is the Notch ligand Delta (D1) (Cooper and Bray, 1999; Fanto et al., 2000). As compared to the R4 precursor cell, dl expression is greater in R3 under wild-type conditions. The data imply that this difference in expression is because R3 is closer to the equator and thus receives a higher dose of factor(s) X. This initial difference in D1 expression (which might be small) is amplified by the Delta-dependent activation of Notch in the adjacent R4 photoreceptor, which negatively regulates D1 expression and determines the identity of the R4 cell. Such a mechanism translates a small initial difference in Fz activity into a binary cell fate decision: The Fz-mediated D1 expression directs the differentiation of photoreceptor precursors into an R3 cell, while elevated Notch activity induces R4 cell fate. The present evidence indicates that the activation of JNK signaling is sufficient to induce D1 expression and R3 cell fate, as targeted expression of Hep^act or c-Jun^act to the R3/R4 pair causes high D1 expression in both cells (Weber et al., 2000). Conversely, impaired JNK signaling in msn mutants results in R3 defects and disrupted polarity (Paricio et al., 1999).

This analysis reveals the coupling of the Notch and JNK signal transduction by D-Jun dependent expression of Delta. Similarly to c-Jun^act induced photoreceptor differentiation, Jun activation is sufficient to induce ectopic Delta expression. However, as mentioned above, the JNK kinase module and D-Jun do not account exclusively for the transduction of the polarity signal, leaving additional signaling events to be discovered.

Concluding remarks and perspectives

A wealth of knowledge on the function of Drosophila Jun and Fos in signal transduction and development has now been accumulated. The breadth of biological functions and multiplicity and variability of the regulatory connections of these transcription factors with other regulators of cell fate and function is remarkable. The emerging complex network illustrates that the AP-1 proteins can be regarded as almost generic switches that serve in a variety of settings to translate environmental information into a selective expression of genetic information.

What remains? The future of AP-1 research in Drosophila holds many promises. Three aspects deserve our special attention: genomics, proteomics and evolution.

For the ultimate understanding of the biology of a transcription factor it would be desirable to obtain a complete description of all genes that are regulated both positively and negatively by this factor in as many relevant situations and developmental stages as possible. While such an undertaking would have been deemed utopian 10 years ago, the deciphering of genomes and the advent of genomics technology has made this task a realistic, if ambitious, proposition. Drosophila offers a number of specific advantages for such an approach. Its genome is small, sequenced and comparatively well annotated. This greatly facilitates DNA array and SAGE based gene expression profiling experiments. A wealth of genetic information, the accessibility of large amounts of biological material and the possibility to obtain specific cell populations from wild-type and mutant flies by FACS offer exciting possibilities. Drosophila is also a good system for reverse genetics, i.e. permitting a top-down approach from gene to phenotype. This approach will be tremendously helpful in fitting the flood of genomic data into functional models.

Proteomics is promising in Drosophila for many of the same reasons: moderate complexity, the accessibility of material, the databases, and the conceivable interplay with forward and reverse genetics. With regard to D-Jun, D-Fos and the various signaling pathways these transcription factors are linked to, direct analysis of protein-protein interactions using sophisticated biochemistry and mass spectrometry may be expected to provide fresh insights into their function in transactivation and signal processing.

Finally and most importantly, the age of digital biology we are now entering will answer whether the notion that was introduced in the beginning of this review really holds water: is Drosophila really a model organism from which to derive significant and relevant conclusions that would also apply to vertebrates, i.e. us? Can we, for example, make predictions on the target genes and biological mechanisms regulated by AP-1 in certain pathologies of human patients based on what we learned in Drosophila? Is DC, for example, more than just a superficial model for wound healing? Initial reports on medically important results obtained in 'simple' systems such as the fly and the worm seem to indicate that this optimism is not entirely misplaced. The deciphering of genomes and the mind-boggling progress of bioinformatics will make it much easier to move between different model systems and to make valid comparisons of biological function across species boundaries.

Acknowledgements

The authors thank Lena M Staszewski for technical help, Julia Zeitlinger for contributing Figure 6, Beth Stronach for discussion and Erwin Wagner for initiating this review and for his patience with delayed contributions.

Adachi-Yamada T, Fujimura-Kamada K, Nishida Y, Matsumoto K. (1999). Nature 400, 166-169. Article MEDLINE

Agnes F, Suzanne M, Noselli S. (1999). Development 126, 5453-5462. MEDLINE

Andrew DJ, Baig A, Bhanot P, Smolik SM, Henderson KD. (1997). Development 124, 181-193. MEDLINE

Axelrod JD, Miller JR, Shulman JM, Moon RT, Perrimon N. (1998). Genes Dev. 12, 2610-2622. MEDLINE

Barila D, Mangano R, Gonfloni S, Kretzschmar J, Moro M, Bohmann D, Superti-Furga G. (2000). EMBO J 19, 273-281. MEDLINE

Basdra EK, Huber LA, Komposch G, Papavassiliou AG. (1994). Biochim. Biophys. Acta. 1222, 315-322. MEDLINE

Belvin MP, Zhou H, Yin JC. (1999). Neuron. 22, 777-787. MEDLINE

Bergmann A, Agapite J, McCall K, Steller H. (1998). Cell 95, 331-341. MEDLINE

Bhanot P, Brink M, Samos CH, Hsieh JC, Wang Y, Macke JP, Andrew D, Nathans J, Nusse R. (1996). Nature 382, 225-230. MEDLINE

Bhat KM. (1998). Cell 95, 1027-1036. MEDLINE

Bienz M. (1997). Curr. Opin. Genet. Dev. 7, 683-688. MEDLINE

Biggs WHI, Zavitz KH, Dickson B, van der Straten A, Brunner D, Hafen E, Zipursky SL. (1994). EMBO J. 13, 1628-1635. MEDLINE

Blair SS. (1995). Bioessays 17, 299-309. MEDLINE

Blair SS. (1999). Curr. Biol. 9, R356-R360. MEDLINE

Bohmann D, Ellis MC, Staszewski LM, Mlodzik M. (1994). Cell 78, 973-986. MEDLINE

Boutros M, Mlodzik M. (1999). Mech. Dev. 83, 27-37. Article MEDLINE

Boutros M, Paricio N, Strutt DI, Mlodzik M. (1998). Cell 94, 109-118. MEDLINE

Brand AH, Perrimon N. (1993). Development 118, 401-415. MEDLINE

Bray S. (2000). Curr Biol 10, R155-R158. Article MEDLINE

Brook WJ, Diaz-Benjumea FJ, Cohen SM. (1996). Ann. Rev. Cell. Dev. Biol. 12, 161-180.

Brunner D, Dücker K, Oellers N, Hafen E, Scholz H, Klämbt C. (1994a). Nature 370, 386-389. MEDLINE

Brunner D, Oellers N, Szabad J, Biggs WHI, Zipursky LH, Hafen E. (1994b). Cell 76, 875-888. MEDLINE

Brunner E, Peter O, Schweizer L, Basler K. (1997). Nature 385, 829-833. MEDLINE

Byars CL, Bates KL, Letsou A. (1999). Development 126, 4913-4923. MEDLINE

Cadigan KM, Fish MP, Rulifson EJ, Nusse R. (1998). Cell 93, 767-777. MEDLINE

Chen Y, Riese MJ, Killinger MA, Hoffmann FM. (1998). Development 125, 1759-1768. MEDLINE

Cho KO, Choi KW. (1998). Nature 396, 272-276. Article MEDLINE

Cooper MT, Bray SJ. (1999). Nature 397, 526-530. Article MEDLINE

Davis RJ. (2000). Cell 103, 239-252. MEDLINE

Dominguez M, de Celis JF. (1998). Nature 396, 276-278. Article MEDLINE

Eresh S, Riese J, Jackson DB, Bohmann D, Bienz M. (1997). EMBO J 16, 2014-2022. Article MEDLINE

Erickson MR, Galletta BJ, Abmayr SM. (1997). J. Cell. Biol. 138, 589-603. MEDLINE

Fanger GR, Gerwins P, Widmann C, Jarpe MB, Johnson GL. (1997). Curr. Opin. Genet. Dev. 7, 67-74. MEDLINE

Fanto M, Mlodzik M. (1999). Nature 397, 523-526. Article MEDLINE

Fanto M, Weber U, Strutt DI, Mlodzik M. (2000). Curr. Biol 10, 979-988. MEDLINE

Fernandez R, Tabarini D, Azpiazu N, Frasch M, Schlessinger J. (1995). EMBO J 14, 3373-3384. MEDLINE

Fernandez R, Takahashi F, Liu Z, Steward R, Stein D, Stanley ER. (2000). Genes Dev. 14, 604-614. MEDLINE

Freeman M. (2000). Nature 408, 313-319. Article MEDLINE

Genova JL, Jong S, Camp JT, Fehon RG. (2000). Dev. Biol. 221, 181-194. Article MEDLINE

Glise B, Bourbon H, Noselli S. (1995). Cell 83, 451-461. MEDLINE

Glise B, Noselli S. (1997). Genes Dev. 11, 1738-1747. MEDLINE

Goyal L, McCall K, Agapite J, Hartwieg E, Steller H. (2000). EMBO J. 19, 589-597. Article MEDLINE

Gubb D. (1993). Dev. Suppl 269-277.

Hafezi F, Steinbach JP, Marti A, Munz K, Wang ZQ, Wagner EF, Aguzzi A, Reme CE. (1997). Nat. Med. 3, 346-349. MEDLINE

Harden N, Lee J, Loh HY, Ong YM, Tan I, Leung T, Manser E, Lim L. (1996). Mol. Cell. Biol. 16, 1896-1908. MEDLINE

Harden N, Loh HY, Chia W, Lim L. (1995). Development 121, 903-914. MEDLINE

Herskowitz I. (1995). Cell 80, 187-197. MEDLINE

Hewes RS, Schaefer AM, Taghert PH. (2000). Genetics 155, 1711-1723. MEDLINE

Holland PM, Suzanne M, Campbell JS, Noselli S, Cooper JA. (1997). J. Biol. Chem. 272, 24994-24998. MEDLINE

Hou XS, Goldstein ES, Perrimon N. (1997). Genes Dev. 11, 1728-1737. MEDLINE

Hudson JB, Podos SD, Keith K, Simpson SL, Ferguson EL. (1998). Development 125, 1407-1420. MEDLINE

Immerglück K, Lawrence PA, Bienz M. (1990). Cell 62, 261-268. MEDLINE

Ip YT, Davis RJ. (1998). Curr. Opin. Cell. Biol. 10, 205-219. MEDLINE

Jacinto A, Wood W, Balayo T, Turmaine M, Martinez-Arias A, Martin P. (2000). Curr. Biol. 10, 1420-1426. Article MEDLINE

Kennerdell JR, Carthew RW. (1998). Cell 95, 1017-1026. MEDLINE

Keyse SM. (1998). Semin. Cell. Dev. Biol. 9, 143-152. Article MEDLINE

Kiehart DP, Galbraith CG, Edwards KA, Rickoll WL, Montague RA. (2000). J. Cell. Biol. 149, 471-490. MEDLINE

Knust E. (1996). Curr. Biol. 6, 379-381. MEDLINE

Kockel L, Zeitlinger J, Staszewski LM, Mlodzik M, Bohmann D. (1997). Genes Dev. 11, 1748-1758. MEDLINE

Lawrence PA. (1992). The Making of the Fly: The Genetics of Animal Design,. Oxford and Boston,

Li YS, Shyy JY, Li S, Lee J, Su B, Karin M, Chien S. (1996). Mol. Cell. Biol. 16, 5947-5954. MEDLINE

Lisi S, Mazzon I, White K. (2000). Genetics 154, 669-678. MEDLINE

Lu Y, Settleman J. (1999). Genes Dev. 13, 1168-1180. MEDLINE

Martin-Blanco E. (2000). Bioessays 22, 637-645. MEDLINE

Martin-Blanco E, Gampel A, Ring J, Virdee K, Kirov N, Tolkovsky AM, Martinez-Arias A. (1998). Genes Dev. 12, 557-570. MEDLINE

Martin-Blanco E, Pastor-Pareja JC, Garcia-Bellido A. (2000). Proc. Natl. Acad. Sci. USA 97, 7888-7893. MEDLINE

McEwen DG, Cox RT, Peifer M. (2000). Development 127, 3607-3617. MEDLINE

Mihaly J, Kockel L, Gaengel K, Weber U, Bohmann D, Mlodzik D. (2001). Mech. Development (in press.).

Mlodzik M. (1999). EMBO J. 18, 6873-6879. Article MEDLINE

Morimura S, Maves L, Chen Y, Hoffmann FM. (1996). Dev. Biol. 177, 136-151. MEDLINE

Muller H, Samanta R, Wieschaus E. (1999). Development 126, 577-586. MEDLINE

Nolan KM, Barrett K, Lu Y, Hu KQ, Vincent S, Settleman J. (1998). Genes Dev. 12, 3337-3342. MEDLINE

O'Neill EM, Rebay I, Tjian R, Rubin GM. (1994). Cell 78, 137-147. MEDLINE

Panganiban GE, Reuter R, Scott MP, Hoffmann FM. (1990). Development 110, 1041-1050. MEDLINE

Papavassiliou AG, Treier M, Bohmann D. (1995). EMBO J. 14, 2014-2019. MEDLINE

Papayannopoulos V, Tomlinson A, Panin VM, Rauskolb C, Irvine KD. (1998). Science 281, 2031-2034. Article MEDLINE

Paricio N, Feiguin F, Boutros M, Eaton S, Mlodzik M. (1999). EMBO J. 18, 4669-4678. MEDLINE

Peifer M, Rauskolb C, Williams M, Riggleman B, Wieschaus E. (1991). Development 111, 1029-1043. MEDLINE

Perkins KK, Admon A, Patel N, Tjian R. (1990). Genes Dev. 4, 822-834. MEDLINE

Perkins KK, Dailey GM, Tjian R. (1988). EMBO J. 7, 4265-4273. MEDLINE

Perrimon N, McMahon AP. (1999). Cell 97, 13-16. MEDLINE

Peverali FA, Isaksson A, Papavassiliou AG, Staszewski LM, Mlodzik M, Bohmann D. (1996). EMBO J. 15, 3943-3950. MEDLINE

Raabe T. (2000). Biochim. Biophys. Acta. 1496, 151-163. MEDLINE

Rebay I, Rubin GM. (1995). Cell 81, 857-866. MEDLINE

Ricos MG, Harden N, Sem KP, Lim L, Chia W. (1999). J. Cell. Sci. 112, 1225-1235. MEDLINE

Riese J, Tremml G, Bienz M. (1997). Development 124, 3353-3361. MEDLINE

Riesgo-Escovar JR, Hafen E. (1997a). Science 278, 669-672. Article MEDLINE

Riesgo-Escovar JR, Hafen E. (1997b). Genes Dev. 11, 1717-1727. MEDLINE

Riesgo-Escovar JR, Jenni M, Fritz A, Hafen E. (1996). Genes Dev. 10, 2759-2768. MEDLINE

Ring JM, Martinez-Arias A. (1993). Development 119, 251-259.

Roulier EM, Panzer S, Beckendorf SK. (1998). Mol. Cell 1, 819-829. MEDLINE

Sawamoto K, Winge P, Koyama S, Hirota Y, Yamada C, Miyao S, Yoshikawa S, Jin MH, Kikuchi A, Okano H. (1999). J. Cell. Biol. 146, 361-372. MEDLINE

Shulman JM, Perrimon N, Axelrod JD. (1998). Trends Genet. 14, 452-458. Article MEDLINE

Simin K, Bates EA, Horner MA, Letsou A. (1998). Genetics 148, 801-813. MEDLINE

Sluss HK, Davis RJ. (1997). J. Cell. Biochem. 67, 1-12. MEDLINE

Sluss HK, Han Z, Barrett T, Davis RJ, Ip YT. (1996). Genes Dev. 10, 2745-2758. MEDLINE

St Johnston RD, Gelbart WM. (1987). EMBO J. 6, 2785-2791. MEDLINE

Strutt DI, Weber U, Mlodzik M. (1997). Nature 387, 292-295. MEDLINE

Strutt H, Strutt D. (1999). Curr. Opin. Genet. Dev. 9, 442-446. MEDLINE

Su YC, Treisman JE, Skolnik EY. (1998). Genes Dev. 12, 2371-2380. MEDLINE

Szüts D, Bienz M. (2000a). Proc. Natl. Acad. Sci. USA 97, 5351-5356.

Szüts D, Bienz M. (2000b). Mech. Dev. 98, 71-76.

Szüts D, Eresh S, Bienz M. (1998). Genes Dev. 12, 2022-2035. MEDLINE

Takatsu Y, Nakamura M, Stapleton M, Danos MC, Matsumoto K, O'Connor MB, Shibuya H, Ueno N. (2000). Mol. Cell. Biol. 20, 3015-3026. MEDLINE

Tateno M, Nishida Y, Adachi-Yamada T. (2000). Science 287, 324-327. MEDLINE

Theisen H, Purcell J, Bennett M, Kansagara D, Syed A, Marsh JL. (1994). Development 120, 347-360. MEDLINE

Tomlinson A, Struhl G. (1999). Development 126, 5725-5738. MEDLINE

Treier M, Bohmann D, Mlodzik M. (1995). Cell 83, 753-760. MEDLINE

Tremml G, Bienz M. (1992). Development 116, 447-456. MEDLINE

Usui K, Simpson P. (2000). Dev. Biol. 225, 13-25. MEDLINE

Wang SL, Hawkins CJ, Yoo SJ, Muller HA, Hay BA. (1999). Cell 98, 453-463. MEDLINE

Watson L, Herdegen T, Buschmann T, Denger S, Wende P, Jahn L, Kreuzer J. (2000). Eur. J. Clin. Invest. 30, 11-17. MEDLINE