roughest

REGULATION

Sns, a member of the immunoglobulin superfamily that is essential for myoblast fusion and mediates heterotypic adhesion with Duf/Kirre and IrreC-rst-expressing cells

The body wall muscles in the Drosophila larva arise from interactions between Dumbfounded/Kirre and Irregular chiasm C-roughest (IrreC-rst)-expressing founder myoblasts and Sticks and stones (Sns)-expressing fusion competent myoblasts in the embryo. Sns, a member of the immunoglobulin superfamily that is essential for myoblast fusion (Bour, 2000), mediates heterotypic adhesion of S2 cells with Duf/Kirre and IrreC-rst-expressing S2 cells, and colocalizes with these proteins at points of cell contact. These properties are independent of their transmembrane and cytoplasmic domains, and are observed quite readily with GPI-anchored forms of the ectodomains. Heterotypic interactions between Duf/Kirre and Sns-expressing S2 cells occur more rapidly and to a greater extent than homotypic interactions with other Duf/Kirre-expressing cells. In addition, Duf/Kirre and Sns are present in an immunoprecipitable complex from S2 cells. In the embryo, Duf/Kirre and Sns are present at points of contact between founder and fusion competent cells. Moreover, Sns clustering on the cell surface is dependent on Duf/Kirre and/or IrreC-rst. Finally, although the cytoplasmic and transmembrane domains of Sns are expendable for interactions in culture, they are essential for fusion of embryonic myoblasts (Galletta, 2004).

The ability of Sns, Duf/Kirre and IrreC-rst to mediate cell–cell adhesion was examined using Drosophila S2 cells, which are predominantly non-adherent under normal conditions. As a prelude to examining the behavior of these molecules in combination, each was examined individually to evaluate their ability to direct homotypic aggregation. S2 cells were transiently transfected with Duf/Kirre, IrreC-rst, or Sns under the control of the copper inducible metallothionein promoter and allowed to aggregate. Following aggregation, the cells were fixed and examined by indirect immunofluorescence using anti-sera directed against specific domains or tags within each protein. As anticipated from previous studies (Dworak, 2001), Duf/Kirre-expressing S2 cells were frequently found in aggregates. Duf/Kirre protein accumulates at points of cell-cell contact in aggregates but is uniformly distributed on the surface of non-aggregated S2 cells. Similar to the behavior of Duf/Kirre, IrreC-rst mediates homotypic aggregation of S2 cells, and becomes enriched at points of cell-cell contact in the resulting cell clusters. Duf/Kirre and IrreC-rst enrichment is occasionally observed in regions where cell-cell contact is not apparent, possibly as a consequence of processes, visible by transmission electron microscopy, that extend around neighboring cells. In contrast to the behavior of Duf/Kirre or IrreC-rst, expression of Sns protein on the surface of S2 cells does not lead to homotypic cell adhesion (Dworak, 2001). A lower magnification view emphasizes the presence of many unassociated Sns-expressing cells. As anticipated, Sns is distributed uniformly on the surface in the absence of aggregation (Galletta, 2004).

To ensure that the Duf/Kirre and IrreC-rst clusters were the consequence of aggregation rather than cell division, the number of cells in aggregates of three cells or more were counted. Since cells should only divide at most once during the course of the experiment, clusters of three cells must represent those formed from adhesive events. In a survey of 4171 Duf/Kirre-expressing cells, 40% (1697) were found in aggregates of three or more. In a survey of 1002 IrreC-rst-expressing cells, 19% (192) were in aggregates of three or more cells. These results suggest that while Duf/Kirre may be more efficient in mediating homotypic aggregation than IrreC-rst, clearly both are capable of mediating such interactions. By contrast, a survey of 3275 Sns-expressing cells revealed only 26 cells in aggregates of three or more cells (Galletta, 2004).

While Sns-expressing cells do not aggregate homophilically, studies have indicated that these cells aggregate with cells expressing Duf/Kirre (Dworak, 2001). It was of interest to determine whether cells expressing Sns would interact with cells expressing IrreC-rst, and whether Sns and Duf/Kirre or IrreC-rst co-localize at points of cell-cell contact. To this end, S2 cells were independently, transiently transfected and the ability of Duf/Kirre and IrreC-rst-expressing cells to form aggregates and direct membrane co-localization of Sns in these aggregates was examined. All of these proteins were uniformly distributed on the cell surface in unaggregated cells. In contrast to their behavior in isolation, Sns-expressing cells readily associated in large clusters when combined with cells expressing Duf/Kirre. The Sns-expressing cells also associated with cells expressing IrreC-rst, with a similar efficiency. At least one of these IgSF members must be expressed on the cell surface for it to cluster, since no untransfected cells were observed in an analysis of 1109 small clusters of either Duf/Kirre:Sns or IrreC-rst:Sns-expressing cells. While the biological significance of such an interaction remains unclear, Duf/Kirre and IrreC-rst-expressing cells are capable of forming heterotypic aggregates with each other when expressed in S2 cells under similar conditions (Galletta, 2004).

Examination of individual proteins in small aggregates revealed clustering of Sns with either Duf/Kirre or IrreC-rst at points of cell contact. Thus, either Duf/Kirre or IrreC-rst can direct cells to associate with Sns-expressing cells, and co-localize with Sns at points of cell contact. Frequently much of the Sns protein in the cell accumulates at the points of cell-cell contact, leaving little if any protein on the rest of the cell surface. In rare cases, both proteins are observed in regions outside of obvious cell contacts. However, this pattern may reflect cell membranes that are extending around neighboring cells, mentioned earlier. Since Duf/Kirre and IrreC-rst serve redundant functions in the founder myoblasts, and behave similarly in the assays described above, subsequent experiments focused on Duf/Kirre (Galletta, 2004).

In some cases, the cytoplasmic domains of cell adhesion molecules play no role in their ability to direct cell interactions, while this domain can be critical in other cases. It was therefore of interest to determine whether these regions of Sns or Duf/Kirre were required for the S2 cell interactions. For these studies, the extracellular domains of Duf/Kirre and Sns were fused in frame to the GPI-anchor sequence of Fasciclin I. These constructs were separately, transiently transfected into S2 cells, and aggregation was examined. In the case of Duf/Kirre-GPI, the efficiency of homotypic aggregation was severely reduced compared to that of cells expressing full length Duf/Kirre. Since the relevance of Duf/Kirre homotypic aggregates in vivo is unclear, the role of the Duf/Kirre and Sns cytoplasmic and transmembrane domains in heterotypic aggregation was also examined. In an analysis similar to that done for Duf/Kirre homotypic aggregates, the ability of cells expressing the GPI-anchored or full length forms of Duf/Kirre to mediate heterotypic adhesion with cells expressing full-length or GPI-anchored forms of Sns was examined in pairwise comparisons. The influence of the Sns cytoplasmic and transmembrane domains was examined on adhesion with cells expressing full length Duf/Kirre. Within the limits of statistical significance, GPI-anchored Sns mediate aggregation at a level comparable to that of full length Sns. A similar analysis was carried out to examine the influence of the Sns cytoplasmic and transmembrane regions on adhesion with cells expressing Duf/Kirre-GPI. Again, Sns-GPI mediates adhesion with the Duf/Kirre-GPI-expressing cells at a level comparable to that of full length Sns. Thus, the Sns cytodomain and membrane spanning region appear to play no role in its ability to direct aggregation with Duf/Kirre-expressing cells (Galletta, 2004).

Since the Duf/Kirre-expressing cells were in excess in the above experiments, these data could not be used to determine the relative contribution of the Duf/Kirre cytodomain and transmembrane region. Therefore additional assays were carried out in which Sns or Sns-GPI-expressing cells were in a five-fold excess over either Duf/Kirre or Duf/Kirre-GPI-expressing cells to determine whether there was a requirement for the Duf/Kirre cytodomain or membrane spanning region in interactions with cells expressing Sns. These experiments were also set up as pairwise comparisons, and demonstrated that Duf/Kirre-GPI mediate aggregation at a level comparable to that of full length Duf/Kirre. These data suggest that there is no significant difference between the ability of Duf/Kirre or Duf/Kirre-GPI to aggregate with cells expressing full-length Sns. Lastly, the cytoplasmic and transmembrane domain of Duf/Kirre have a modest affect on its ability to direct aggregation with GPI-anchored Sns. However, the effect of the cytoplasmic or transmembrane domain on Duf/Kirre's ability to mediate heterotypic aggregation with Sns-GPI was not as great as its effect on the ability of Duf/Kirre to mediate homotypic cell adhesion. Of note, Duf/Kirre-GPI was enriched at points of cell-cell contact in both homotypic aggregates and in heterotypic aggregates with Sns and Sns-GPI. Thus, neither the cytoplasmic nor transmembrane domains of Sns or Duf/Kirre are essential for recruitment to cell-cell contacts (Galletta, 2004).

In summary, these results indicate that the cytoplasmic/transmembrane domains of Sns and Duf/Kirre do not influence the efficacy with which they direct heterotypic cell-cell adhesion. This observation is in contrast to that seen for Duf/Kirre, in which the cytodomain or membrane spanning region of Duf/Kirre plays a critical role in its ability to direct homotypic aggregation. One possible explanation for these results is that heterotypic association of Sns and Duf/Kirre is stronger, and does not require stabilization of the receptor through cytoplasmic or intramembrane interactions. Since the affinity of Duf/Kirre for homotypic versus heterotypic interactions could play a critical role in myoblast interactions in the embryo, the S2 cell aggregation assay was used to examine this preference (Galletta, 2004).

In the embryonic musculature, founder cells appear to fuse only with fusion competent myoblasts, and never fuse with each other. In principle, this directional fusion could be attributed to the differential expression of Duf/Kirre and Sns by these two cell types, and inability of these molecules to associate homotypically. However, results reported in this study and by Dworak (2001) demonstrate that Duf/Kirre-expressing cells do associate with each other in culture. It was therefore of interest to determine whether the affinity of Duf/Kirre-expressing cells for cells expressing Sns was greater than the affinity of Duf/Kirre-expressing cells for each other. To address this question, Duf/Kirre-expressing cells were aggregated in isolation or in the presence of an equal number of Sns-expressing cells. This analysis utilized stable cell lines in which approximately 30% of the corresponding population expressed Duf/Kirre and approximately 8% expressed Sns. Aggregation of Duf/Kirre-expressing cells was examined in three different conditions, all with the same total cell number. The goal was to ensure that any change in aggregation of Duf/Kirre cells was due to the specific addition of Sns-expressing cells rather than a consequence of doubling the number of adherent cells. For each time point, the number of Duf/Kirre-expressing cells free in solution was counted and the number that had been incorporated into aggregates. Duf/Kirre-expressing cells were incorporated into aggregates that included Sns-expressing cells at a faster rate and to a greater extent than those containing only Duf/Kirre-expressing cells. This behavior was not a simple consequence of the number of adherent cells present, since a two-fold increase in the number of Duf/Kirre-expressing cells did not have a dramatic effect on the rate or extent of aggregation. Thus, Duf/Kirre-expressing cells associate more readily into heterotypic aggregates with Sns-expressing cells than into homotypic aggregates with only Duf/Kirre-expressing cells (Galletta, 2004).

The striking colocalization of Duf/Kirre and Sns described earlier suggested the possibility that these proteins might physically associate in trans. To address this possibility, aggregates of stably transfected, Sns and Duf/Kirre-expressing cells were subjected to reversible protein cross-linking and lysed. HA-tagged Duf/Kirre was immunoprecipitated from the cell lysate using anti-HA resin, and the resulting immunoprecipitate examined by Western blot for the presence of Sns. HA-tagged Duf/Kirre was efficiently precipitated from both Duf/Kirre-only and Duf/Kirre-Sns mixed cell populations. As expected, Sns was not present in the anti-HA immunoprecipitate from cells expressing only Duf/Kirre or only Sns. However, it was clearly detected in immunoprecipitates from the mixed population of cells expressing Duf/Kirre-HA and Sns. Thus, Duf/Kirre and Sns are closely associated in an immunoprecipitable protein complex, possibly through a direct protein interaction (Galletta, 2004).

In the embryonic musculature, Duf/Kirre, IrreC-rst and Sns are necessary, either directly or indirectly, for the association of founder and fusion competent myoblasts. The striking co-localization of Sns with either Duf/Kirre or IrreC-rst in S2 cells prompted an examination of whether similar co-localization could be observed between embryonic myoblasts. First it was examined whether punctate clustering of Sns on the surface of embryonic myoblasts, previously described by Bour (2000), was dependent on the presence of Duf/Kirre or IrreC-rst. The distribution of Sns protein was examined in embryos deficient for both Duf/Kirre and IrreC-rst, and compared to that seen in wild-type embryos. As anticipated, Sns becomes localized to discrete sites in wild-type myoblasts (Bour, 2000). In contrast, Sns is distributed more uniformly on the myoblast surface in embryos lacking Duf/Kirre and IrreC-rst. Thus in embryos, as in S2 cells, the localization of Sns is dependent on the presence of Duf/Kirre or IrreC-rst (Galletta, 2004).

To determine whether Sns and Duf/Kirre co-localize in embryonic myoblasts in a manner similar to that observed in S2 cells, stage 13 embryos were examined by indirect immunofluorescence using polyclonal antisera directed against the Duf/Kirre and Sns proteins. As previously described for Sns (Bour, 2000), Duf/Kirre is expressed in a dynamic pattern that is restricted to discrete sites on the surface and in the cytoplasm of expressing cells. The pattern of Sns expression intersects that of Duf/Kirre, and is in close proximity to rP298-lacZ positive founder cell nuclei in the somatic mesoderm. Of note, punctate Sns expression is apparent at some sites in which Duf/Kirre expression is not detected. To address whether these sites might intersect points of IrreC-rst protein, which can interact with Sns-expressing cells and can substitute for Duf/Kirre in vivo, embryos were triple labeled with Duf/Kirre, IrreC-rst and Sns. IrreC-rst is readily detected at many sites of Sns enrichment that do not appear to colocalize with Duf/Kirre. In fact, examination of 204 discrete sites of Sns protein, derived from eight stage 13 embryos, revealed that 97% were colocalized with either Duf/Kirre and/or IrreC-rst (Galletta, 2004).

To determine whether sites of Duf/Kirre and Sns colocalization occur, as expected, on the cell surface, mesodermally expressed CD2 was used to visualize the cell membrane. CD2 staining revealed the surface of a growing myofiber and associated myoblasts. Duf/Kirre and Sns co-localize to points of contact between the fiber and a myoblast. Since the expression of both Duf/Kirre and Sns is dynamic and rapidly decreases upon fusion (Bour, 2000), co-localization of Duf/Kirre and Sns was examined in myoblast city (mbc) mutant embryos in which the myoblasts associate but remain unfused. By stage 14, the founder cells of these mutant embryos become morphologically distinct from the fusion competent cells, elongating and extending processes. As an apparent consequence of this fusion block, Duf/Kirre and Sns are stabilized at points of contact between the extended founder cell and several fusion competent cells. These data clearly show that Sns and Duf/Kirre co-localize in the embryo at critical contact points between founder cells and fusion competent myoblasts (Galletta, 2004).

IrreC/rst-mediated cell sorting during Drosophila pupal eye development depends on proper localisation of DE-cadherin

Remodelling of tissues depends on the coordinated regulation of multiple cellular processes, such as cell-cell communication, differential cell adhesion and programmed cell death. During pupal development, interommatidial cells (IOCs) of the Drosophila eye initially form two or three cell rows between individual ommatidia, but then rearrange into a single row of cells. The surplus cells are eliminated by programmed cell death, and the definitive hexagonal array of cells is formed, which is the basis for the regular pattern of ommatidia visible in the adult eye. This cell-sorting process depends on the presence of a continuous belt of the homophilic cell adhesion protein DE-cadherin at the apical end of the IOCs. Elimination of this adhesion belt by mutations in shotgun, which encodes DE-cadherin, or its disruption by overexpression of DE-cadherin, the intracellular domain of Crumbs, or by a dominant version of the monomeric GTPase Rho1, prevents localisation of the transmembrane protein IrreC-rst to the border between primary pigment cells and IOCs. As a consequence, the IOCs are not properly sorted and supernumerary cells survive. During the sorting process, Notch-mediated signalling in IOCs acts downstream of DE-cadherin to restrict IrreC-rst to this border. The data are discussed in relation to the roles of selective cell adhesion and cell signalling during tissue reorganisation (Grzeschik, 2005).

To summarise, IrreC-rst is colocalised with DE-cadherin in epithelial cells of pupal eye discs, and misdistribution of adherens junction components induces the mislocalisation of IrreC-rst, which then affects sorting of IOCs. However, although DE-cadherin forms a continuous belt in the apical regions of all cells (including all IOCs) in wild-type discs, IrreC-rst colocalises with DE-cadherin only at the border between 1° pigment cells and IOCs. What factor(s) might be responsible for the spatial restriction of IrreC-rst to this border? It has recently been shown that the removal of Notch or Delta function during cell-sorting results in the ubiquitous distribution of IrreC-rst to all plasma membranes and the prevention of programmed cell death. This study analysed whether this might be the result of defective DE-cadherin localisation. Antibody staining reveals no influence of Notch on the continuous apical localisation of DE-cadherin, but shows that IrreC-rst now colocalises with the latter on all plasma membranes of the IOCs. This suggests that Notch acts downstream of DE-cadherin in the control of IrreC-rst localisation. It is therefore tempting to speculate that it is the Notch pathway, which provides local signalling between the lattice cells to direct cell death, that prevents the accumulation of IrreC-rst at the borders between IOCs and thus restricts its localisation to the 1°/IOC cell boundary (Grzeschik, 2005).

Pattern formation in the Drosophila eye disc depends on a well-balanced system of signals that promote either the survival or the death of cells, mediated by the EGF and Notch receptor pathways, respectively. In addition, the morphogenetic events, which take place in a single-layered epithelium, crucially depend on factors that regulate the maintenance of cell polarity and cell shape, and modulate cell adhesion. Sorting of interommatidial cells (IOCs) during pupal development, which results in the conversion of several parallel rows of cells into a single ring, requires the weakening of pre-existing adhesive cell contacts and the establishment of new ones without interrupting the epithelial integrity of the tissue. During tissue morphogenesis, epithelial cells use different strategies to modify their adhesive contacts. One of these consists of regulating the amount and/or distribution of the homophilic cell-adhesion molecule E-cadherin, one of the central components of the adherens junctions. The first in vivo evidence for this kind of regulation came from the analysis of the Drosophila egg chamber. There, the localisation of the oocyte at the posterior pole depends on a higher level of expression of DE-cadherin in the oocyte and the posterior follicle cells, when compared with the nurse cells and other follicle cells. Differential adhesion can also be regulated by alterations in the composition or activity of intracellular binding partners, or by the integration of various other molecules into the adhesive complexes. No change in the distribution of the adherens junction components DE-cadherin and alpha-catenin could be detected in wild-type discs undergoing rearrangements of the IOCs. This behaviour contrasts with epithelial rearrangements during morphogenesis of the Drosophila tracheal system, which are associated with alterations in the amount of DE-cadherin, controlled by Drosophila Rac, another member of the Rho GTPase family. This in turn suggests that other cytoplasmic or transmembrane proteins are involved in the modulation of adhesion in IOCs. The adhesion protein IrreC-rst is involved in the control of the cell sorting process. Its predominant localisation at the border between primary pigment cells and IOCs (at the 1°/IOC border) has been suggested to provide an attractive interface that controls sorting. According to this proposal, IOCs tend to maximise their contacts with primary pigment cells. Failure to restrict IrreC-rst to this border results in the inability to sort the IOCs properly. Although IrreC-rst behaves as a homophilic adhesion molecule when expressed in cell culture, data from expression analysis argue for the presence of a different, as yet unknown, partner in the primary pigment cell (Grzeschik, 2005).

Of particular interest is the relationship between the localisation of DE-cadherin, a component of the zonula adherens (ZA) and IrreC-rst. In wild-type discs IrreC-rst colocalises with DE-cadherin at the 1°/IOC border in the apical ZA of the cell and removal of DE-cadherin completely abolishes IrreC-rst accumulation. Nothing is yet known about how IrreC-rst may integrate into the ZA at this border. In vertebrates, the Ca⁺⁺-independent cell adhesion molecule nectin, a transmembrane protein of the immunoglobulin superfamily, has been implicated in the organisation of cadherin-based adherens junctions, tight junctions and synapses. It is recruited into cadherin-based adherens junctions through interactions with the PDZ domain of l-afadin, an F-actin-binding protein. Intriguingly, the C-terminal sequence of IrreC-rst (T-A-V) matches the consensus binding site for class I PDZ domains (S/T-X-V). Interestingly, the protein encoded by the mutant allele irreC-rst^CT, which lacks the C-terminal 175 amino acids of the wild-type form, is no longer recruited into the ZA. It is, however, unlikely that IrreC-rst acts as a general adhesion molecule in IOCs of pupal eye discs, because the epithelial tissue structure is stable in the absence of irreC-rst function, as deduced from the formation of the continuous apical belt of DE-cadherin in irreC-rst mutants (Grzeschik, 2005).

The continuous belt of DE-cadherin can be disrupted by a number of different genetic conditions, such as overexpression of the membrane-bound intracellular domain of Crumbs, of DE-cadherin itself, or of a dominant-negative version of the monomeric GTPase Rho1. Overexpression of the membrane-bound intracellular domain of Crumbs in embryonic epithelia has been shown to lead to a redistribution of DE-cadherin throughout the plasma membrane and the formation of multilayered tissues. By contrast, IOCs overexpressing Crb_intra exhibit a fragmented DE-cadherin belt, which remains localized in the apical zone of the cells, and apicobasal organisation and tissue integrity are not affected. This suggests that IOCs may contain additional adhesion components which are independent of, or less affected by, Crb. Support for this view comes from the phenotype of discs lacking crb function, in which the apical belt of DE-cadherin expression is fragmented, yet there is no major effect on polarity or adhesion of the epithelium: the cells undergo nearly normal sorting and IrreC-rst is still restricted to the membrane at the 1°/IOC border. Overexpression of Crb_{intraDeltaERLI} (lacking four C-terminal amino acids [ERLI] of its short cytoplasmic domain, which serve to recruit a multiprotein complex that forms apical to the zonula adherens) does not interfere with sorting, suggesting that a protein complex similar to the one that controls apicobasal polarity in embryonic epithelia (which includes Stardust, DPATJ and D-Lin7) contributes to the development of the dominant phenotype (Grzeschik, 2005).

Overexpression of DE-cadherin similarly results in the fragmentation of the adhesion belt and defects in cell sorting. In various tissues, overexpression of full-length DE-cadherin can also reduce Wingless signalling by sequestering Armadillo from the cytoplasmic pool, thus making it unavailable to transduce the Wingless signal. However, the possibility that the defects in sorting are the result of a suppression of Wingless signalling can be excluded. Inactivation of components of the Wingless pathway in eye imaginal discs induces the initiation of ectopic morphogenetic furrows, and this phenotype was not observed upon overexpression of DE-cadherin. Overexpression of DE-cadherin in eye discs therefore seems to interfere with adhesion, rather than Wingless signalling (Grzeschik, 2005).

Rho GTPases play central roles in the organisation of the actin cytoskeleton and in cell adhesion. In mammals, inhibition of Rho activity results in the removal of cadherins from epithelial cell junctions, while increased Rho activity induces an invasive and metastatic phenotype. Members of the Rho GTPase family are recruited into the adherens junctions by direct interactions with junctional components. Thus, in Drosophila, Rho1 localises to the adherens junctions and interacts directly with alpha-catenin and p120^ctn, a homologue of ß-catenin. As in pupal epithelia expressing a dominant-negative form of Rho1, Rho1 mutant embryos exhibit a diffuse distribution of components of the ZA, such as DE-cadherin and alpha- and ß-catenin. Rho1 may either act directly on the accumulation of cadherins at the junctions, or indirectly by recruiting accessory proteins, which then modulate the amounts or activity of junctional and/or cytoskeletal proteins. Rho1 plays a different role in tracheal epithelia insofar as its inactivation does not disrupt DE-cadherin localisation, but rather interferes with the formation of the apical surface and the tracheal lumen (Grzeschik, 2005).

Although it is evident that DE-cadherin plays a crucial role in the accumulation of IrreC-rst at the adherens junctions, other mechanisms are required to explain the asymmetric localisation and restriction of the latter to the 1°/IOC boundary. It has been speculated that an as yet unknown ligand expressed in the primary pigment cell may account for this restricted accumulation. As an alternative, but not mutually exclusive model, it is suggested that signalling between the IOCs, mediated by Notch, which is expressed in IOCs during pupal development, prevents the accumulation of IrreC-rst at their borders. Interplay between adhesion and signalling molecules also directs other processes in which cellular polarisation is involved in tissue remodelling. The growth of the wing imaginal disc along the proximodistal axis, for example, is the result of cell shape changes and cell rearrangements during pupal development, which are controlled by the atypical cadherins Fat and Dachsous, as well as Four-Jointed, which is assumed to be a secreted molecule. This process, in turn, is responsible for the asymmetric localisation of components that control planar polarity, such as Frizzled, Dishevelled or Strabismus, that serves to ensure that bristles and hairs adopt a common orientation. During germ band elongation in the Drosophila embryo, adherens junction remodelling in intercalating ectodermal cells is facilitated by the polarised expression of non-muscle myosin II at the anteroposterior and of Bazooka at the dorsoventral cell boundaries. Future experiments will demonstrate whether cell sorting in pupal eye discs makes use of any of the components known to be involved in these processes (Grzeschik, 2005).

Preferential adhesion mediated by Hibris and Roughest regulates morphogenesis and patterning in the Drosophila eye

Cell adhesion is essential for morphogenesis; however, the mechanisms by which cell adhesion coordinates precisely regulate morphogenesis are poorly understood. This study analyzes the morphogenetic processes that organize the interommatidial precursor cells (IPCs) of the Drosophila pupal eye. The Drosophila immunoglobulin superfamily members Hibris and Roughest are essential for IPC morphogenesis in the eye. The two loci are expressed in complementary cell types, and Hibris and Roughest proteins bind directly in vivo. Primary pigment cells employ Hibris to function as organizers in this process; IPCs minimize contacts with neighboring IPCs and utilize Roughest to maximize contacts with primaries. In addition, evidence is provided that interactions between Hibris and Roughest promote junction formation and that levels of Roughest in individual cells determine their capacity for competition. These results demonstrate that preferential adhesion mediated by heterophilic interacting cell-adhesion molecules can create a precise pattern by minimizing surface free energy (Bao, 2005).

To properly organize the ommatidia into a precise pattern, the interommatidial precursor cells (IPCs) undergo dynamic cell rearrangements between 18 and 42 hr after puparium formation (APF). These cells will eventually differentiate as secondary and tertiary pigment cells (2ºs, 3ºs) and mechanosensory bristles. Emergence of the interommatidial lattice was further analyzed with an antibody to the β-catenin ortholog Armadillo (Arm), a core component of the adherens junction. Based on this work, IPC and ommatidial patterning was classified into four stages (hours are based on the approximate center of the eye field), which are briefly described (Bao, 2005).

(1) Initial cell sorting (18-24 hr APF). Initially, IPCs are scattered between ommatidia with a relaxed apical profile. As development progresses, two cells emerge from the IPC pool to enwrap the cone cells and become 1ºs; the remaining IPCs simultaneously line up in single file to contact 1ºs from adjacent ommatidia. Concurrently, some cells are removed by apoptosis (Bao, 2005).

(2) Emergence of 3ºs (24-27 hr APF). Typically, three cells are initially positioned equally at a vertex. One cell reaches past the other two to contact a third 1º ; this cell will then physically “invade” the vertex and mature as a 3º (Bao, 2005).

(3) Selection of 2º s (27-36 hr APF). Cells that fail to become 3ºs either become 2ºs or are removed by programmed cell death. During this final cell-fate decision, cell-cell adhesion becomes visibly polarized as IPCs form detectable junctional contacts with 1º s but not with other IPCs. In addition, a 'scalloping' of membrane profiles is observed as 1ºs push between IPCs, further confirming that the adhesion between 1ºs and IPCs is greater than between neighboring IPCs. By 36 hr APF, the hexagonal pattern is essentially complete: it is composed of a single 2º at each side and a 3º or bristle organule at each vertex (Bao, 2005).

(4) Maturation (36-42 hr APF). Visible adherens junctions return to the interfaces between IPCs (now 2ºs and 3ºs). Contacts are now smoothed as the scalloping caused by invasive 1º contacts is now relaxed (Bao, 2005).

One particularly striking feature of this morphogenetic process is the dynamic nature of the cell junctions, which were visualized with the junctional protein Arm. For example, the level of Arm in the cone cells was constant but the levels of Arm in the IPCs decreased: this was seen by comparing the levels of Arm in the two cell groups. This drop in Arm levels is followed by its complete loss between IPCs after 3ºs emerge and eventual reemergence at the final maturation stage to levels similar to cone cells. Thus, junctions appear to be diminished during the period of maximal cell rearrangement, suggesting that IPCs are free to move during these stages (Bao, 2005).

Using laser ablation studies, it has been demonstrated that 1ºs are centrally important for the process of organizing IPCs into a correctly patterned interommatidial lattice. However, the mechanism by which one cell can provide such remarkably precise patterning information to a larger collection of uncommitted cells has not been not clear. The dynamic interactions between Hibris and Roughest provide such a mechanism (Bao, 2005).

The 'differential adhesion hypothesis' (DAH) proposes that sorting-out and segregation of cell populations are driven by differences in the intensities of cell adhesions. Given motile and cohesive cell populations, DAH predicts that weakly cohesive cells will tend to be displaced by more strongly cohesive ones; this process can direct cells to segregate away from unlike cell populations, and it can control tissue spreading during, for example, germ layer maturation in the embryo. DAH has been supported by several observations. For example, quantitative differences in the level of cadherin expression can lead two cell populations to be mutually immiscible: less cohesive cells will envelope more cohesive ones, creating a 'sphere within a sphere' configuration. Recently, the importance of differential adhesion for patterning developing tissue has been demonstrated in the pupal retina. Cone cells segregate from other cells and assemble into a simple pattern by minimizing surface area, as do soap bubbles. This assembly is mediated at least in part by E- and N-cadherins, and manipulating cadherin levels within the cone cells or their neighbors can alter the final cone cell pattern. These experiments illustrate that differential adhesion caused by differences in cadherin expression can mediate morphogenesis and pattern formation (Bao, 2005).

The current data indicate that IPC patterning follows a mechanism that shows unique aspects when compared with these classical DAH experiments. First, manipulating E-cadherin levels does not alter the morphogenesis or arrangement of IPCs. Even when two neighboring IPCs have higher levels of E-cadherin, adhesion between these two IPCs or their final patterning is not affected. More critically, IPCs do not aggregate together or segregate away from their neighbors. Rather, they separate away from each other to minimize IPC:IPC contacts, and aggregate with ommatidial cores to maximize 1º :IPC contacts. That is, the data indicate that IPCs have a preference for adherence to 1ºs. This preference can be seen most clearly at 27 hr APF: the junctions between IPCs and 1ºs are strong and elaborate; the junctions between IPCs are indistinct, and 1ºs are seen to push between IPCs to maximize contact and create a scalloping effect. The result is the precise aggregation of two different cell populations (Bao, 2005).

Why do IPCs sort away from other IPCs and preferentially adhere to 1ºs? The data indicate that interactions between Hibris and Roughest provide the mechanism. The immunoglobulin-class proteins Roughest and Hibris are utilized by IPCs and 1ºs, respectively, to form heterophilic interactions. Several lines of evidence support this view: (1) both Hibris and Roughest are required for proper interommatidial lattice assembly; (2) hibris is expressed in 1ºs as well as in cone cells and roughest is expressed in IPCs at the time of IPC rearrangement in the eye; (3) expression of ectopic Hibris in either the 1º or IPC is sufficient to relocalize Roughest protein -- conversely, downregulation of Hibris in 1ºs leads to decreased levels of Roughest protein at the 1º :IPC interface; (4) Hibris and Roughest are capable of directly binding each other when isolated in tissue culture experiments (Bao, 2005).

After 1ºs are specified and start to express Hibris, levels of Roughest protein decrease between IPCs and increase in the borders between IPCs and 1ºs; for example, at 30 hr APF, Roughest protein is undetectable between IPCs. Furthermore, ectopic Hibris in 1ºs is sufficient to attract still more Roughest protein toward the 1º :IPC border; by contrast, ectopic Roughest in 1ºs does not attract additional Roughest. It is concluded that although Roughest can show homophilic interactions in S2 cells, it strongly prefers heterophilic interactions with Hibris in situ (Bao, 2005).

Ubiquitous Hibris expression greatly increases the levels of cell-junction proteins between IPCs. Similarly, individual IPCs that received ectopic Hibris form E-cadherin-rich borders with neighboring IPCs that are sharp, straight, and significantly enlarged (Bao, 2005).

The evidence indicates that 1ºs and IPCs prefer to adhere to each other based on their expression of Hibris and Roughest, respectively. One principle of thermodynamics states that the binding of two adherent molecules will lead to a reduction of free energy within the system, provided the equilibrium constant of association (k_a) is greater than the equilibrium constant of dissociation (k_d). The essential role of Hibris and Roughest in IPC morphogenesis prompts making an assumption: among the various molecules being displayed in the surfaces of 1ºs and IPCs, Hibris and Roughest play a major role in determining the flow of free energy. Roughest has a higher affinity for Hibris than itself, and therefore heterophilic binding between Roughest and Hibris leads to a greater reduction in free energy. As a result, contacts between IPCs and 1ºs contribute to a reduction of free energy and are favored, while contacts between IPCs and IPCs do not contribute to reduction of free energy and are disfavored (Bao, 2005).

Other features of the developing pupal eye provide important components to this patterning process. After 1ºs are specified, they establish cell junctions with each other and with cone cells. These cone cell/1º units are not free to move within the epithelial plane and form a functional patterning unit. Therefore, 1ºs function as the organizers in this context. In contrast, IPCs have reduced levels of junctional proteins and are free to move within the epithelium. Numerous filopodia from IPCs observed by SEM studies also point to their potential for high motility. Taken together, these data suggest that IPC morphogenesis follows a preferential adhesion model: IPCs exhibit preferential adhesion to 1ºs; 1ºs function as organizers for IPC morphogenesis, and IPC:1º contacts are free energy favored while IPC:IPC contacts are disfavored (Bao, 2005).

The ommatidial clusters are poorly organized until 18 hr APF, when the morphogenetic movements of the IPCs begin to organize clusters into a hexagonal array. Preferential adhesion of IPCs to 1ºs yields two major outcomes. (1) IPCs compete to adhere directly to the limited, Hibris-rich surface presented by the 1ºs. High motility of IPCs permits this competition to proceed and achieve a favored configuration. (2) Preferential adhesion can also lead to the removal of cells that fail to contact a 1º . Specifically, IPCs that adhere to 1ºs have an increased chance to survive since the Hibris:Roughest interactions provide a greater opportunity to establish a stable junction. By the same token, those cells that do not have access to 1ºs are disadvantaged and are commonly dropped from the apical surface; these cells are likely to be eventually removed by programmed cell death. As a result, each stage proceeds with a progressive reduction of the IPC:IPC contacting surfaces and an increase in IPC:1º contacting surfaces (Bao, 2005).

At the onset of IPC morphogenesis (18 hr APF), the average size of IPC:IPC contacts is not significantly different from the size of IPC:1º contacts. During the time cells in multiple layers are sorted into single file after the initial cell-sorting stage (24 hr APF), IPC:IPC contacts are significantly reduced. After emergence of 3ºs, this reduction in IPC:IPC contacts is particularly dramatic. The IPC:1º contacts are increased by a scalloped profile, a further demonstration that IPC:IPC contacts are disfavored. To complete this pattern, therefore, IPC:IPC contacts are further minimized by reducing the number of candidate 2ºs to one cell between each 3º and bristle. Thus, IPC morphogenesis reveals a mechanism by which pattern is determined through minimizing disfavored cell-cell contacts and maximizing preferred cell-cell contacts (Bao, 2005).

Finally, it is interesting to note how 2ºs are selected. After emergence of 3ºs, two IPCs are commonly found between a 3º and bristle. In many ways, these two IPCs are equal: each contacts two 1ºs and each establishes equally strong cell junctions; each forms a scalloped contour with two neighboring 1ºs, and each is exposed to the same molecular cues. However, evidence is provided that these two cells have a low affinity for each other, a situation that is not favored by minimum free energy principles. One cell will be removed. How is this cell chosen? Clues came from manipulating levels of Roughest, which altered each cell's capacity for competition. Artificially high levels of Roughest rendered a cell a supercompetitor: the targeted cell even replaced two cells to become both a 2º and a 3º. Presumably, high levels of Roughest promote a higher level of cell junctions, which makes a cell more competitive and determines the survivor. Conversely, low levels of Roughest put the targeted cell at a disadvantage during this competition. Therefore, during the selection of a 2º , differing levels of Roughest expressed by each cell may determine its fate: survival or death (Bao, 2005).

Neph1/Nephrin family members are required for the development of a wide array of tissues including axonal pathfinding and myoblast fusion in Drosophila and formation of the slit diaphragm in the developing mammalian kidney. The role observed for preferential adhesion in IPC morphogenesis and patterning in the Drosophila eye leads to the interesting possibility that similar mechanisms are utilized broadly in pattern formation (Bao, 2005).

Mutations in sec15 cause defects in synaptic specificity, axon targeting and localization of axon guidance components

The exocyst is a complex of proteins originally identified in yeast that has been implicated in polarized exocytosis/secretion. Components of the exocyst have been implicated in neurite outgrowth, cell polarity, and cell viability. An exocyst component, sec15, has been isolated in a screen for genes required for synaptic specificity. Loss of sec15 causes a targeting defect of photoreceptors that coincides with mislocalization of specific cell adhesion and signaling molecules. Additionally, sec15 mutant neurons fail to localize other exocyst members like Sec5 and Sec8, but not Sec6, to neuronal terminals. However, loss of sec15 does not cause cell lethality in contrast to loss of sec5 or sec6. The data suggest a role for Sec15 in an exocyst-like subcomplex for the targeting and subcellular distribution of specific proteins. The data also show that functions of other exocyst components persist in the absence of sec15, suggesting that different exocyst components have separable functions (Mehta, 2005).

Elevated levels of chaoptin in photoreceptor terminals have been described for another vesicle-trafficking mutant, the vesicle-SNARE neuronal-synaptobrevin (n-syb). This mutant also exhibits neuronal targeting defects. This observation raises the possibility that vesicle-dependent trafficking of transmembrane or other signaling molecules might be responsible for the neuronal targeting defects of sec15 mutant photoreceptors. Recently, Zhang (2004) identified Rab11 as an interacting partner of Sec15 in mammalian cell culture and proposed that Sec15 is an effector for some but not all Rabs. Indeed, an accumulation or upregulation of Rab11 immunoreactivity was seen in sec15 mutant photoreceptors, consistent with Rab11-positive vesicles failing to fuse with their target sites. To further test this hypothesis, the localization of cell adhesion and signaling molecules was examined in mutant photoreceptor cell bodies as well as terminals during photoreceptor development, precisely when target selection and cartridge formation occur (between P + 5% to P + 40% referring to time after pupation). Proteins were examined that have either been shown to be required for photoreceptor target selection, such as Dlar, N-cadherin, flamingo, and IrreC-rst, or that are likely to be required, based on work in other systems, such as Armadillo, Chaoptin, and Fasciclin II (Mehta, 2005).

Fasciclin II (Fas2) localization was examined in sec15 mutant photoreceptors, since chaoptin upregulation coincides with elevated levels of Fas2 in n-syb mutant photoreceptors. Fas2 appears to be present in aggregates in sec15 mutant photoreceptor cell bodies at P + 20%, in contrast to wild-type photoreceptors. In addition, the neuronal connections of the cell bodies exhibit Fas2 aggregated along the length of the mutant axons. Similarly, overexpression of Fas2 in photoreceptors causes neuronal targeting defects between P + 20% and P + 40%. In contrast to n-syb, however, no elevated levels of Fas2 are observed later in development. Hence, the data suggest that an aberrant localization of Fas2 in a specific developmental time window may at least partially underlie the observed phenotypes (Mehta, 2005).

Similar mislocalization phenotypes in photoreceptor cell bodies were also observed for other cell adhesion molecules such as Dlar and IrreC-rst during the developmental time window of photoreceptor target selection. Dlar is normally restricted apically in developing wild-type photoreceptors, at the center of the ommatidial array. In sec15 mutant photoreceptors it appears much more randomly distributed, such that a basal optical section through the eye shows Dlar at higher levels in mutant ommatidia. Although these results show mislocalization of cell adhesion molecules in the correct cell at the time when they are known to be required for proper target selection, no obvious defects were detected in the localization of Dlar or IrreC-rst in the developing lamina. This leaves open the question of whether mislocalization of Dlar and IrreC-rst beyond the resolution limit of confocal microscopy additionally contributes to the observed targeting defects (Mehta, 2005).

In vertebrates, Lar is known to localize to adherens junctions. Hence, a possible explanation for the mislocalization of Fas2, IrreC-rst, and Dlar in mutant photoreceptor cell bodies is a defect of adherens junctions. The subcellular localization of the adherens junction markers N-cadherin and armadillo was examined in the cell bodies as well as the terminals of mutant photoreceptors, but no mislocalization of N-cadherin was detected in either compartment. However, armadillo displayed localization defects selectively in the developing lamina, but not the photoreceptor cell bodies. Several other cell adhesion and signaling molecules, including flamingo, Crumbs, and Bazooka, were examined, all of which did not display aberrant localization at the level of light microscopy. It is concluded conclude that a specific subset of proteins is mislocalized in sec15 mutants (Mehta, 2005).

Single-minded, Dmef2, Pointed, and Su(H) act on identified regulatory sequences of the roughest

Roughest (Rst) is a cell adhesion molecule of the immunoglobulin superfamily that has multiple and diverse functions during the development of Drosophila melanogaster. The pleiotropic action of Rst is reflected by its complex and dynamic expression during the development of Drosophila. By an enhancer detection screen, several cis-regulatory modules have been identified that mediate specific expression of the roughest gene in Drosophila developmental processes. To identify trans-regulators of rst expression, the Gal4/UAS system was used to screen for factors that were sufficient to activate Rst expression when ectopically expressed. By this method the transcription factors Single-minded, Pointed.P1, and Su(H)-VP16 were identified. Furthermore, these factors and, in addition, Dmef2 are able to ectopically activate rst expression via the previously described rst cis-regulatory modules. This fact and the use of mutant analysis allocates the action of the transcription factors to specific developmental contexts. In the case of Sim, it could be shown to regulate rst expression in the embryonic midline, but not in the optic lobes. Mutagenesis of Sim consensus binding sites in the regulatory module required for rst expression in the embryonic midline, abolishes rst expression; indicating that the regulation of rst by Sim is direct (Apitz, 2005).

Rst has complex and multifaceted functions throughout the development of the fly, which include myogenesis, eye development, as well as axonal pathfinding in the optic lobes. To gain a better understanding of these functions at the levels of gene regulation and signal transduction, a number of tests were designed to identify both the transcriptional activators and their respective targets surrounding the rst locus. In a preceding study (Apitz, 2004), a number of DNA segments upstream of rst were characterized and regulatory regions were discovered that mediate gene expression in myoblasts, midline, and eyes, respectively. In the present study these results were supplemented with an in vivo screen to identify regulators of rst expression using the Gal4/UAS system. Several factors were discovered that are able to induce ectopic Rst expression and to activate reporter gene expression via rst cis-regulatory sequences (Apitz, 2005).

The experimental route taken to identify protein factors involved in the regulation of the rst gene is based on the detection of their potential to induce ectopic Rst expression in vivo. The use of sca-Gal4 as a driver line in this experimental approach is based on the following criteria. sca-Gal4 mediates expression in neuroectodermal cells of the embryo. At embryonic stage 10, these cells can be examined for ectopic Rst expression because no endogenous Rst expression is found at this time in these cells; this allows operators to obtain clear-cut and unequivocal results. The use of alternative Gal4 driver lines did not prove suitable because of the dynamic expression of Rst during all developmental stages, and due to its subcellular localization. For example, when dll-Gal4 is used as a driver, it is difficult to distinguish between ectopic Rst expression induced in the apical tips of cells of the leg discs, and endogenous Rst expression in the overlaying ectodermal cells. Furthermore, the use of sca-Gal4 has the advantage that neuroectodermal cells are not fully differentiated cells. This may more closely resemble the developmental state of the cells in which Rst expression is normally induced endogenously, e.g., in undifferentiated cells of the developing eye disc. However, this approach generally fails to reveal transcription factors that need a coactivator for induction of rst expression, which is not present in the cells of the neuroectoderm at embryonic stage 10. This may explain the failure of Dmef2 to induce rst expression at this stage. It was shown, however, that Dmef2 is able to induce rst expression at later stages by the use of rst-lacZ constructs. The function of the different rst-lacZ constructs has been linked to specific developmental circumstances (Apitz, 2004) and their activation by corresponding factors is consistent with the known roles of these proteins in development (Apitz, 2005).

Single-minded (Sim) is a basic helix-loop-helix-PAS (bHLH-PAS) transcription factor that serves as a master regulatory gene of CNS midline development. The CNS midline is derived from two mesectodermal cell rows that intermingle at the ventral midline following gastrulation of the embryo. The ventral midline cells move into the interior of the embryo where they function (among other roles) in the patterning of the CNS by secreting axonal guidance cues. It has been shown that expression of more than 20 genes depends on sim. Because Rst expression in the embryonic midline is lost in sim mutant embryos, rst is added to this list (Apitz, 2005).

Tgo, the coactivator of Sim, is ubiquitously expressed in the embryo. Thus, ectopic expression of target genes can be induced by Sim in transactivation studies. The early onset of strong ectopic expression of Rst in neuroectodermal cells under the transcriptional control of ectopic Sim suggests that rst may be its direct target. Several lines of evidence support this hypothesis. The expression of rst in the embryonic midline sets in shortly after gastrulation in the ventral midline, following accumulation of Sim protein in these cells. In an expression profiling study, rst was assigned to a group of 37 genes that mimic the expression pattern of sim in the embryonic midline. These genes are therefore likely direct targets of Sim/Tgo heterodimers. Since Sim does not act on the F5 fragment in the transactivation assay, the Sim-responsive rst cis-regulatory sequence was mapped to a 2.5-kb region contained within the F6 fragment, i.e., F6d. F6d was shown to contain a regulatory module for expression in the embryonic midline (Apitz, 2004). Two putative central midline elements (CME) are located in this region and both are conserved between D. melanogaster and D. pseudoobscura. It has been shown that heterodimers of Drosophila Sim and human Tgo bind to the CME and that Sim activates midline gene transcription through CME. Site-directed mutagenesis studies described in this paper show that both CME in F6d are required for reporter gene expression in the embryonic midline. Significantly, analysis of polytene chromosomal binding sites of ectopically expressed Sim in salivary gland nuclei revealed that Sim binds to chromatin within chromosomal subdivision 3C, coinciding with the rst locus. These data indicate that rst expression in the embryonic midline is directly regulated by Sim (Apitz, 2005).

Unexpectedly, ectopic reporter gene expression was observed in cells outside the midline using the mutated rst^F6d sequence. This finding indicates that by mutating the CME the affinity of this DNA sequence for other transcription factors may have been altered. It is conceivable that either a suppressor normally preventing expression outside the midline lost its affinity for the site, or an activator, which is active in the corresponding cells, gained affinity. Similar CME mutations were used in other studies, and no ectopic reporter gene expression was reported in these cases. The different outcome of these studies may be based on the different chromatin regions, in which the CME are embedded (Apitz, 2005).

The function of Rst in the embryonic midline is as yet unknown. The apical localization of Rst in cytoplasmic projections of midline cells that migrate in the interior of the embryo (Apitz, 2004), points to an adhesive function which may support the correct localization of the midline cells during migration. Since the cell adhesion molecule Hibris (Hbs) is the only known interaction partner of Rst that is also expressed in the midline, it will be interesting to examine the interactions of Rst and Hbs in CNS midline development (Apitz, 2005).

Sim is also involved in postembryonic optic lobe development. Although the expression pattern and axonal pathfinding defects of sim and rst point to a possible regulation of rst by Sim in the optic lobes, no evidence was found supporting this hypothesis using sim^ts mutants. This result is corroborated by analysis of rst cis-regulatory elements (Apitz, 2004). F6d, the rst regulatory sequence that mediates reporter gene expression in the embryonic midline and which is activated by ectopic Sim, does not display a corresponding reporter gene expression in the optic lobes. These results are unexpected since they point to a differential set of Sim target genes in postembryonic stages as compared to the embryo. It will be interesting to examine whether other known target genes of Sim are expressed independently of sim in the optic lobes as well, and whether this differential regulation depends on the differential use of alternative cofactors in the embryo and in the optic lobes, respectively. Since Tgo colocalizes with Sim in the optic lobes, one may suspect the involvement of additional uncharacterized cofactors. The existence of tissue-specific cofactors that direct Sim/Tgo activity to the embryonic midline have already been postulated by other authors (Apitz, 2005).

Ectopic expression of a constitutively active Pnt variant (Pnt.P1) mediates strong activation of Rst expression in neuroectodermal cells. Since Pnt.P1 recognizes the same target sequences as its splice variant Pnt.P2, the nuclear effector of the Ras-MAPK pathway, ectopic activation of Rst expression by Pnt.P1 is consistent with a regulation by the Ras-MAPK pathway. Similarly, the ectoptic activation of Rst expression in neuroectodermal cells by Su(H)-VP16 points to a regulation of rst by the Notch pathway (Apitz, 2005).

The Ras-MAPK and the Notch pathways display significant crosstalk during developmental processes in Drosophila, e.g., in cell fate specification of the eye disc. It is difficult to elucidate a possible regulation of a candidate gene by mutant analysis if it is activated by both pathways. In mutants, for one of the pathways, the activity of the other pathway will ensure residual expression of the candidate gene under scrutiny. rst expression is activated by both pathways and single mutant analysis did not reveal a significant loss of expression. Both pathways converge on regulatory elements contained within F6 and not in F5. Furthermore, a regulatory module is present in the nonoverlapping part of F6 that is activated in IOC before apoptotic decisions are made in these cells (Apitz, 2004). This module is located within an approximately 600-bp sequence and is active during several apoptotic decisions (Apitz, 2004). Consensus binding sites for Su(H) and Pnt in this module are conserved between D. melanogaster and D. pseudoobscura. Both the Ras-MAPK and Notch pathways are involved in apoptotic processes of IOC cells. Together, these data suggest that rst transcription is regulated by these pathways in the context of apoptotic decisions (Apitz, 2005).

The transactivation screen shows that Dmef2 acts on rst cis-regulatory sequences: the overlapping reporter gene constructs rst^F5 -lacZ and rst^F6 -lacZ are both ectopically activated by Dmef2 in the scabrous expression domain. Regions F5 and F6 contain partially overlapping sequences and it is likely that Dmef2 acts on a sequence interval bracketed by this overlap. This region contains a regulatory module for expression in the mesoderm (Apitz, 2004). This sequence contains a putative Dmef2 binding site that is also conserved between D. melanogaster and D. pseudoobscura. The sequence of this putative element exactly matches a Dmef2 binding site found in the enhancer of β3 tubulin. Therefore, it is argued that rst may be a direct target of Dmef2. This is consistent with similar mesodermal expression patterns and mutant phenotypes of Dmef2 and rst. In contrast, Rst is expressed in the mesoderm of Dmef2 loss-of-function mutants. An explanation for this result is provided by analysis of rst cis-regulatory sequences. rst is regulated in the mesoderm by at least two independent regulatory modules that mediate a differential expression pattern (Apitz, 2004). One is contained in F6p, the other in F5p. This points to a mesodermal regulation of rst by at least two factors, one of which is still active in Dmef2 mutants (Apitz, 2005).

Protein Interactions

The adaptor protein X11Lalpha/Dmint1 interacts with the PDZ-binding domain of the cell recognition protein Rst in Drosophila

The Drosophila cell adhesion molecule Roughest (Rst) plays key roles during the development of the embryonic musculature, spacing of ommatidia in the compound eye and of sensory organs on the antenna, as well as in the neuronal wiring of the optic lobe. In rst^CT mutants lacking the cytoplasmic domain of the Rst protein, cell sorting and apoptosis in the eye are affected, suggesting a requirement of this domain for Rst function. To identify potential interacting proteins, yeast two-hybrid screens were performed using as baits the cytoplasmic domains of Rst and its paralogue Kirre. Among several putative interactors, two paralogous Drosophila PDZ motif proteins related to X11/Mint were identified. X11/Mint family members in C. elegans (LIN-10) and vertebrates are believed to function as adaptor proteins and to regulate the assembly of multi-subunit complexes at the synapse, thereby linking the vesicle cycle to cell adhesion. Using genetic, cell biological, and biochemical approaches, the interaction of Rst with X11La has been shown to be of biological significance. The proteins interact, for example, in the context of cell sorting in the pupal retina (Vishnu, 2006).

The X11/Mint family comprises proteins with several distinct protein-protein interaction domains, including a phosphotyrosine-binding (PTB) domain, commonly found among signalling molecules, and two PDZ domains. These domains are specialized for binding to the C-terminus of transmembrane proteins and play a role in polarized protein targeting. PDZ proteins are most abundant in the cytoplasm and are known to link transmembrane proteins to the underlying cytoskeleton and cytosolic signalling proteins (Vishnu, 2006).

This analysis focused on the Rst-X11Lα interaction since several additional lines of evidence consistently pointed to a biologically meaningful interaction. Both proteins can be co-immunoprecipitated in vivo, and the association is mediated by the carboxyterminal three amino acids of Rst comprising the PDZ binding motif, and the PDZ domain of X11Lα. Conversely, this interaction is abolished in rst^CT flies lacking the C-terminal intracellular domain, confirming the importance of the PDZ binding domain for a functional interaction for these proteins in vivo. Flies heterozygous for rst^CT are sensitized with regard to loss of X11Lα. Reduction of X11Lα protein levels by X11Lα-targeted RNAi strongly interfered with cell sorting on this genetic background. Both proteins are not only co-expressed during eye development, but also during maturation of a specific layer of the distal medulla neuropile, which seems to indicate the importance of this interaction during the process of synaptogenesis (Vishnu, 2006).

In mammals, three distinct X11/Mint paralogue groups exist with partially overlapping functions. It has been argued that the roles of the carboxyterminal PTB and PDZ domains are fully redundant, whereas the highly divergent N-terminal domains confer isoform-specific protein interactions. For example, the aminoterminal domain of mammalian Mint1 can bind to CASK, whereas those of Mint2 and Mint3 cannot. Conversely, both mammalian Mint1 and Mint2 bind to Munc-18, a protein essential for synaptic vesicle exocytosis (Vishnu, 2006).

Phylogenetic reconstructions based on the divergence of the PTB and PDZ domains revealed that the two Drosophila X11/Mint homologs cannot be assigned to any of the three paralogue groups defined by the vertebrate X11/Mint families but rather form a separate and distinct clade along with other known invertebrate homologs. This may reflect the extended separate evolutionary history but it may also imply specialized functions of the proteins in both clades. It appears, however, that X11/Mint-coding genes have been repeatedly and independently duplicated after the split of the deuterostomes from the protostomes, and some of the paralogues have subsequently been lost. Significantly, Drosophila melanogaster and the mosquito Anopheles appear to be the only invertebrates with two X11L copies, whereas the sister species D. pseudoobscura only seems to have a single copy that represents an orthologue of X11Lα. Available sequence information does not permit an unequivocal assignment of the two near-identical Anopheles X11L homologues. However, these analyses indicate that both correspond to the X11β paralogue group, and that they may be the result of a duplication that followed the split of the two dipteran families. Although the Drosophila members of the family display distinguishing signature sequences that allow their categorization, both paralogues are sufficiently similar in sequence pattern and genomic organization to suspect that they were also the result of a more recent duplication event (Vishnu, 2006).

In the pupal eye disc X11Lα is expressed in the cytoplasm of interommatidial precursor cells and of primary pigment cells, and it accumulates at contact sites of these cells where it colocalizes with Rst. In addition, X11Lα is strongly expressed in primary pigment cells in a belt near the outer apical membrane which contacts interommatidial cells (Vishnu, 2006).

When the level of X11Lα protein is reduced in IOPs by driving one copy of X11Lα-targeted RNAi in IOPs mild cell sorting defects can still be observed at stage P40 and the numbers of interommatidial cells is moderately increased relative to controls since some supernumerary cells are not eliminated by apoptosis. Two copies of X11Lα-targeted RNAi have an even stronger effect on cell sorting and cell survival. The single copy X11Lα-targeted RNAi genetic background was therefore considered a sensitized system to detect putative synergistic effects of removal of one copy of Rst. The rst^CT/+; GMR–Gal4; X11Lα-targeted RNAi/+ pupae displayed strong sorting defects, thus supporting the proposed interaction of Rst and X11Lα at the genetic level. It has been postulated that cell sorting in the retina depends on an interaction of Rst in IOPs with a heterophilic ligand in primary pigment cells (identified as Hibris). It has been furthermore hypothesized that the strength of this selective interaction should be the driving force of cell sorting. As a result of the curent findings a modified model is now proposed. X11Lα does not seem to be required for the correct positioning of Rst in the membrane of IOPs. Its effect on cell sorting may therefore be better explained by Rst-mediated signalling following recognition of Hibris on primary pigment cells. The result of Rst-mediated signalling via X11Lα in IOPs could be to maximize their contacts to primary pigment cells. This could be achieved by stabilizing the DE-cadherin mediated cell adhesion (Vishnu, 2006).

Misexpression of full-length Rst in the eye imaginal disc by the Mz1369 driver leads to severe sorting defects as well as to occasional reduction of cone cells and primary pigment cells. The cell sorting defects are due to an interaction of the extracellular domain of ectopic Rst in cone cells with Hibris expressed by primary pigment cells. Misexpression of Rst-DD1 does not show this extracellular interaction, but the truncated molecule still seems to be active in signalling since it mediates differentiation defects, which are similar to those seen after misexpression of X11Lα. It was already reported by Hase (2002) that misexpression of X11Lα interferes with eye development. Increased levels of apoptosis were found in the eye disc of the third Instar larva. The data confirm that X11Lα misexpression severely disturbs the general regularity of the ommatidial lattice. The reduction in cone cell number is a highly penetrant feature of these eye discs. Most ommatidia contain only two or even fewer cone cells. This could explain the partial loss of primary pigment cells, since primary pigment cell formation is triggered by induction from the cone cells. The decrease in number of cone cells per ommatidium may also contribute to the deep rhabdomere phenotype. The expanded tips of cone cell processes participate in the formation of the fenestrated membrane, which contains the exit portals for the axons of retinula cells in an ommatidium. The disturbance of the portals could well play a role in the deep rhabdomere phenotype of X11Lα misexpressing flies. The phenocopy obtained with Rst-DD1 misexpression suggests that Rst is signalling via the X11Lα pathway (Vishnu, 2006).

Loss of X11Lα by RNAi knockdown methodology causes the reduction of rhabdomeres and retinula cells in adult flies. This is not due to a failure to form retinula cells, since retinula cells are specified and start differentiating during larval development. Their axons project into the brain. In addition, the initiation of rhabdomere formation can be observed as well in pupal stages. Mutants causing retinula cell degeneration at the adult stage, have compound eyes of normal depth, whereas loss of X11Lα leads to compound eyes with severely reduced thickness. It is therefore concluded that this is the consequence of cell degeneration in the late eye imaginal disc (Vishnu, 2006).

Loss of X11Lα by RNAi knockdown has much more severe consequences than any of the known Rst mutations. This could well be due to the redundancy between Rst and Kirre. However, it also cannot be ruled out that X11Lα has Rst and Kirre independent functions (Vishnu, 2006).

Is there a role for Rst-X11Lα interaction in synapse formation? Rst and X11Lα display a distinct overlapping expression pattern in a layer of the pupal distal medulla neuropil. This finding deserves special consideration, since it may indicate that both proteins and their interaction may play a specific role during synapse formation. In support of this hypothesis, it was shown at the neuromuscular junction of Drosophila, that downstream of the IgCAM Fasciclin II, X11Lα functions in bouton formation by interacting with the amyloid precursor protein, indicating a more general role for X11Lα in synaptogenesis (Vishnu, 2006).

IgCAM proteins of the Rst-family also function in the process of synapse formation. The Rst-like SYG-1 protein in C. elegans has been reported to play a role in initiating the assembly of presynaptic sites. Clustering of SYG-1 in HSNL neurons opposite the epithelial SNS-like SYG-2 signal determines the localization of the presynaptic transmitter release machinery in these neurons. So far, the chain of events leading from SYG-1 to the recruitment of presynaptic vesicles is not yet known. The Rst-X11Lα interaction in Drosophila suggests a putative mechanism. In vertebrates, X11/Mints functions in transmitter release and its role in the coupling of the vesicle cycle to cell adhesion has been convincingly shown. Mint1 knockout mice exhibit impaired GABAergic synaptic transmission. The overlap of X11Lα and Rst in the distal medulla of Drosophila in just one unique neuropil layer could well indicate a similar function here and will prompt further work (Vishnu, 2006).

In conclusion, the physical interaction between the X11Lα and Rst proteins has been demonstrated and it has been shown that they are involved in the same pathway that regulates cell sorting in the pupal retina. The involvement of this interaction in other processes requiring cell recognition is likely (Vishnu, 2006).