Two prominent characteristics of epithelial cells, apical-basal polarity and a highly ordered cytoskeleton, depend on the existence of precisely localized protein complexes associated with the apical plasma membrane and on a separate machinery that regulates the spatial order of actin assembly. ERM (ezrin, radixin, moesin) proteins have been proposed to link transmembrane proteins to the actin cytoskeleton in the apical domain, suggesting a structural role in epithelial cells, and they have been implicated in signalling pathways. The sole Drosophila ERM protein Moesin functions to promote cortical actin assembly and apical-basal polarity. As a result, cells lacking Moesin lose epithelial characteristics and adopt invasive migratory behaviour. These data demonstrate that Moesin facilitates epithelial morphology not by providing an essential structural function, but rather by antagonizing activity of the small GTPase Rho. Thus, Moesin functions in maintaining epithelial integrity by regulating cell-signalling events that affect actin organization and polarity. Furthermore, these results show that there is negative feedback between ERM activation and activity of the Rho pathway (Speck, 2003).

ERM proteins, highly related members of the larger protein 4.1 superfamily, can exist in an active or inactive conformation. The inactive conformation is brought about by head-to-tail interactions between the FERM (four-point-one, ERM) domain and the carboxy terminus. In this state, ERM actin-binding and membrane association sites are not accessible. A conformational change induced by Rho pathway activation produces the active ERM form. Mammalian ERM proteins bind actin and cell-type-dependent transmembrane proteins such as CD44, CD43 and several I-CAMs. Genetic tests of hypotheses regarding ERM protein involvement in signalling and cytoskeletal support have been impeded by functional redundancies between these three proteins in mammals. In contrast, Drosophila has only one ERM gene, Moesin, enabling such analyses to be performed (Speck, 2003).

To test Moesin function in vivo, a transposable element insertion in the Moesin gene, P{lacW}l(1)G0323, was identified. Males hemizygous for this insertion display late larval lethality. This lethality can be rescued by a duplication of region 8B, Dp(1;Y)619, as well as by a complementary DNA-based Moesin+ transgene. Taken together, these results indicate that l(1)G0323 disrupts Moesin function but does not affect any other essential genes, and is hereafter referred to as Moe^G0323 (Speck, 2003).

To determine the extent to which Moesin function is disrupted by this allele, Moesin protein level and distribution were examined in Moe^G0323 mutants. Moesin protein is not detected in imaginal discs from Moe^G0323 hemizygous larvae by immunofluorescence or immunoblotting. Furthermore, animals trans-heterozygous for the Moe^G0323 allele and a deficiency of the 8B region, Df(1)lz-90b24, display the same lethal period and have phenotypes identical to Moe^G0323 hemizygous and homozygous mutant larvae, establishing that Moe^G0323 is genetically null (Speck, 2003).

To investigate Moesin function in the morphogenesis of epithelial structures the imaginal discs were examined from Moe^G0323 hemizygous larvae. In wild-type epithelial cells, filamentous actin associates with the cell cortex and accumulates in the apical domain, where it largely coincides with Moesin. In Moe^G0323 cells, F-actin undergoes a marked redistribution: it accumulates in ectopic sites within cells and becomes depleted from the apical region. To demonstrate that actin is reduced in the apical domain, a Moe⁺ transgene was examined only in the posterior compartment of the mutant disc, generating a 'half-rescued' disc. The rescued cells in the posterior compartment have a wild-type phenotype, and have significantly more filamentous actin in the apical domain than adjacent Moe^G0323 cells (Speck, 2003).

The imaginal epithelium normally consists of a single layer of tall columnar cells with an obvious apical-basal polarity. In Moe^G0323 imaginal discs, many cells are situated basal to the epithelial monolayer. These cells do not express the junctional marker E-Cadherin or other epithelial polarity markers, including ß-Catenin, Coracle, Crumbs or Discs lost. These data suggest that in Moe^G0323 organisms, epithelial cells lose intercellular junctions and epithelial polarity, and are extruded basally from the epithelium (Speck, 2003).

To determine whether cells lacking Moesin not only lose epithelial morphology but also adopt migratory behaviour, a subset of cells within the imaginal epithelium was marked by expressing green fluorescent protein (GFP) under control of the region-specific patched-Gal4 driver. In a wild-type wing imaginal disc, marked cells stay within a continuous stripe along the anterior-posterior boundary. In a Moe^G0323 disc, cells expressing GFP under control of patched-Gal4 are found away from the stripe, suggesting that these cells have dispersed. Furthermore, these cells are able to invade adjacent regions of the disc comprising cells that still retain epithelial morphology. Thus, Moe^G0323 cells not only lose epithelial markers and intercellular adhesion, but also become motile and show invasive behavior, suggesting that cells lacking Moesin undergo an epithelial-to-mesenchymal transition (Speck, 2003).

Rho-induced phosphorylation of a C-terminal threonine has been shown to relieve the inhibitory intramolecular association and generate the active conformation of the ERM molecule (Matsui, 1998; Shaw, 1998; Simons, 1998). This residue is conserved in Drosophila Moesin at amino acid position 559. To test the role of Thr 559 phosphorylation in Moesin activation, two mutant cDNA constructs were generated: Moe(T559D), a phosphomimetic form, and Moe(T559A), a form that cannot be phosphorylated at this residue. Moe^G0323 hemizygous males are rescued to viability by transgenes expressing Moe(T559D), but not by Moe(T559A), implying that Moe(T559D) is an active form whereas Moe(T559A) is inactive (Speck, 2003).

The phosphomimetic form of ERM proteins has constitutively active properties because it may be locked in the open conformation (Gautreau, 2000). To examine the effects of ERM activation, Moe(T559D) was overexpressed under region-specific Gal4 drivers in wild-type imaginal discs. A marked upregulation of cortical F-actin was observed in cells that express Moe(T559D). By contrast, little or no effect was seen with Moe⁺ and Moe(T559A) overexpression. The accumulation of cortical actin induced by the gain-of-function mutant is complementary to the reduction of apical actin seen in the Moesin loss-of-function background, indicating that Moesin controls assembly of cortical actin (Speck, 2003).

The pleiotropic nature of the Moe^G0323 phenotype suggests an involvement in a signalling pathway that controls cell adhesion and motility. Given the crucial function carried out by the Rho family of small GTPases in regulating these processes and the evidence for Rho regulation of ERM function (Shaw; 1998; Hirao; 1996; Mackay, 1997) the relationship between Moesin and Rho signalling in vivo was examined by manipulating the genetic dose of Rho1, a Drosophila RhoA homolog, in a Moe^- background. The null Rho172R allele was used to reduce by half maternal and zygotic Rho function. Halving the dose of Rho1 not only ameliorates the Moe^G0323 disc epithelium organization and actin localization phenotypes, but also strongly suppresses Moe^G0323 lethality. A similar suppression of lethality was observed with other Rho1 alleles and when the dose of Drok, a Rho effector kinase, was reduced, and a weaker effect was observed with another downstream effector of Rho1 signalling, the non-muscle myosin-II heavy chain gene zipper). The observed suppression was not due to enhanced Moesin protein production or stability. Taken together, these results suggest that Moesin functions antagonistically to activity of the Rho pathway in regulating epithelial polarity and integrity (Speck, 2003).

The ultrastructure was examined of wild-type and Moe^G0323 epithelial cells as well as that of Moe^G0323 cells that are also heterozygous for Rho172R. In transmission electron micrographs, wild-type epithelial cells have apical finger-like microvilli that tend to congregate near the areas of contact between adjacent cells. In the Moesin mutant, the microvilli are replaced by large apical protrusions that are often of irregular shape. Reduction of Rho activity partially suppresses this Moe^- phenotype, and results in cells that lack the protrusions but still display abnormally large microvillus-like structures. These results suggest that Moesin may have a Rho-independent role in maintaining apical integrity that is not essential for viability (Speck, 2003).

If Moesin normally functions as an antagonist of Rho pathway activity, then the effects of Rho pathway hyperactivation should be similar to the Moe loss-of-function phenotype. To test this hypothesis, whether Rho1 overexpression phenotypes in wing imaginal discs mimic those seen in Moe loss-of-function mutants was investigated. Cells overexpressing Rho1 mislocalize F-actin and lose epithelial characteristics. As with Moe^G0323 cells, Rho1-expressing cells in the blade region of the wing imaginal disc lose junctional markers and drop basally. These cells also assume a mesenchymal character, become motile and invade between wild-type epithelial cells. Similar, but more disruptive phenotypes were observed when a constitutively active form of Rho1, Rho1V14, was overexpressed in wild-type imaginal discs. Therefore, Rho1 overexpression strongly resembles Moe loss of function, suggesting that Moesin and Rho1 exert opposite effects on the epithelium (Speck, 2003).

To test further the relationship of Moesin with the Rho pathway, whether a reduction in Moe function would suppress a phenotype associated with downregulation of Rho1 was investigated. Rho1 has been shown to function downstream of dishevelled (dsh) in the planar cell polarity (PCP) pathway that regulates the polarity and number of hairs generated by each wing blade cell. Each wild-type wing cell produces a single hair; however, multiple wing hairs result when Rho pathway function is impaired. In flies that are mutant for the dsh1 allele (which inactivates the PCP but not the wingless-signalling function of dishevelled), double-hair-producing cells occurred at a frequency of 6.3%. Removal of a single dose of Moe suppresses the number of cells with double hairs in dsh¹ mutants to 1.2%, suggesting that Rho pathway function is upregulated in response to reduction in Moesin activity (Speck, 2003).

These genetic results in Drosophila suggest that Moesin functions antagonistically to Rho pathway activity. To distinguish between an effect on Rho itself and effects on a downstream component of the pathway, and to extend these results from Drosophila to mammalian cells, the effect of reducing ERM function in mammalian LLC-PK1 epithelial cells was examined. Use was made of an ERM truncation, ezrin^Act, that has dominant-negative properties in flies and reduces function of all three ERM proteins in these cells (as assayed by phospho-ERM staining, a marker for ERM activation (Hayashi, 1999). Expression of ezrin^Act resulted in increased levels of Rho activity, whereas expression of wild-type ezrin caused either no effect, or slightly decreased the level of Rho activity. These results indicate that ERM proteins negatively regulate Rho pathway function by altering the activation state of Rho itself, rather than a downstream component of the pathway (Speck, 2003).

The phenotypic analysis presented in this study clearly demonstrates that Moesin function is required to maintain apical-basal polarity, epithelial integrity, and correct actin distribution. This is consistent with Moesin's association with the apical membrane in epithelial cells and the ability to bind filamentous actin (McCartney; 1996; Edwards, 1997). These results also show that Moesin promotes epithelial morphology by functioning antagonistically to Rho pathway activity, rather than by providing structural linkage between the plasma membrane and the cytoskeleton. Moe^- defects, including lethality, are strongly suppressed when Rho pathway activity is reduced, suggesting that Moesin does not have essential structural functions. Furthermore, the diverse Moe^- phenotypes -- decreased cortical actin accumulation, loss of epithelial polarity, and acquisition of migratory cell behavior -- can all be attributed to increased Rho pathway activity. Consistent with this model for Moesin function, increased Rho activity has been linked to invasive and metastatic phenotypes in mammalian epithelial cells. Because ERM proteins are known downstream effectors of Rho signalling, the results further suggest the existence of a negative feedback loop between ERM proteins and the Rho pathway. This negative feedback mechanism may be important in inhibiting migratory and invasive cellular behaviors that characterize metastatic cells (Speck, 2003).

The results with mammalian epithelial cells suggest that the effect of ERM function on Rho signalling occurs upstream to or at the level of Rho itself. Previous studies have detected direct binding between ERM proteins and two known regulators of Rho function, RhoGDI (Rho GDP dissociation inhibitor: Takahash, 1997) and Dbl GEF (guanine nucleotide exchange factor for Rho: Takahashi, 1998). Although one of these studies suggested that ERM proteins function to promote Rho activation in vitro (Takahashi, 1997), the in vivo significance of these interactions has not yet been examined. Elucidation of the functional significance of these interactions may be essential to the further understanding of the relationship between ERM proteins and Rho pathway activity. Of note, the proposed negative feedback loop between ERM and Rho parallels the antagonistic relationship between the ERM-related tumour suppressor Merlin and another Rho-family GTPase Rac (Shaw, 2001), and suggests that ERM and Merlin may have similar roles in regulating Rho-family signalling (Speck, 2003).

GENE STRUCTURE

cDNA clone length - 2418

Bases in 5' UTR - 247

Exons - 9

Bases in 3' UTR - 343

PROTEIN STRUCTURE

Amino Acids - 575

Structural Domains

Moesin-like shares 58% overall identity with both human moesin and human ezrin, 57% identity with mouse radixin, and only 41% identity with human merlin. Designation of this gene as Moesin-like, rather than as ezrin or radixin, is based primarily on its lack of the polyproline tract characteristic of the latter two proteins. furthermore, Moesin-like shares a greater degree of identity in the C-terminal divergent region with human moesin (26%) than with human ezrin (25%) or mouse radixin (22%) (McCartney, 1996).

Analysis of the Drosophila genome identified a single gene encoding an ERM protein, which had been previously cloned and named Moesin-like or Moesin. However, when compared with human ERMs, Drosophila Moesin has strongest conservation with Radixin in the FERM domain, with Ezrin in the C-terminal tail, and shares with mammalian Moesin the absence of a polyproline track that is found in both Ezrin and Radixin. To gain further insight into the evolution of ERMs, sequence conservation of ERM proteins was analyzed from various species. Each available ERM protein from vertebrates seems to be a true orthologue of Ezrin, Radixin or Moesin, respectively, suggesting that three ERM genes exist in these species. By contrast, in common with Drosophila, ERMs from Caenorhabditis elegans and the snail Biomphalaria glabrata are equally related to Ezrin, Radixin and Moesin, thus suggesting that the presence of three ERM genes is restricted to deuterostomes. However, Merlin/NF2 proteins from protostomes and deuterostomes are more related to each other than to ERM, indicating that divergence between NF2 and ERM preceded the separation of these two taxa. Within the C-terminal tail, Drosophila Moesin also contains residues characteristic of Ezrin, Radixin or Moesin, and the threonine involved in ERM activation is conserved in Drosophila Moesin (position 559). Thus, Drosophila Moesin seems to be a prototypical ERM protein, sharing both common and specific features with all three vertebrate ERMs (Polesello, 2002).

date revised: 25 October 2004

Home page: The Interactive Fly © 1995, 1996 Thomas B. Brody, Ph.D.

The Interactive Fly resides on the
Society for Developmental Biology's Web server.