Roles of myosin phosphatase during Drosophila development

Myosins are a superfamily of actin-dependent molecular motor proteins, among which the bipolar filament forming myosins II have been the most studied. The activity of smooth muscle/non-muscle myosin II is regulated by phosphorylation of the regulatory light chains, that in turn is modulated by the antagonistic activity of myosin light chain kinase and myosin light chain phosphatase. The phosphatase activity is mainly regulated through phosphorylation of its myosin binding subunit MYPT. To identify the function of these phosphorylation events, we have molecularly characterized the Drosophila homologue of MYPT, and analyzed its mutant phenotypes. We find that Drosophila MYPT is required for cell sheet movement during dorsal closure, morphogenesis of the eye, and ring canal growth during oogenesis. Our results indicate that the regulation of the phosphorylation of myosin regulatory light chains, or dynamic activation and inactivation of myosin II, is essential for its various functions during many developmental processes.


INTRODUCTION
Myosins are a superfamily of actin-dependent molecular motor proteins involved in a variety of essential processes that include muscular contraction, cytokinesis, vesicle transport, cell movement and cell shape change (reviewed by Mermall et al., 1998;Oliver et al., 1999;Sellers, 2000;Sokac and Bement, 2000;Wu et al., 2000). Among the 17 subclasses of myosins, conventional myosins, known as myosin IIs, have been the most studied. Myosin IIs form bipolar filaments that drive contractile events by bringing together actin filaments of opposite polarity. Myosin II molecules are hexameric enzymes consisting of two heavy chains, two regulatory light chains (MRLC), and two essential light chains. They can be subclassified into four groups based on their motor domain (or tail) sequences: sarcomeric myosins, vertebrate smooth muscle/non-muscle myosins, Dictyostelium/Acanthamoeba type myosins and yeast type myosins.
The activity of smooth muscle/non-muscle myosin II is regulated by the phosphorylation of MRLC that is modulated by the antagonistic activity of myosin light chain kinase (MLCK) and myosin light chain phosphatase (MLCP). MLCP is composed of three subunits: a catalytic subunit made up of protein phosphatase 1c β (also called δ), a myosin binding or targeting subunit (MYPT), and a small subunit of unknown function. MYPT binds and confers the selectivity of PP1c for myosin (Hartshorne et al., 1998).
The phosphatase activity of MLCP can be regulated in several ways (reviewed by Hartshorne et al., 1998;Somlyo and Somlyo, 2000). Rho-kinase (ROCK) phosphorylates an inhibitory phosphorylation site on MYPT and inhibits the phosphatase activity in smooth muscle. This phosphorylation may occur through ZIPK (leucine zipper interacting protein kinase)-like kinase (MacDonald et al., 2001) or integrin-linked kinase (Kiss et al., 2002). Myotonic dystrophy protein kinase phosphorylates the same inhibitory phosphorylation site , although it is not clear whether this phosphorylation event also goes through ZIPK. In addition, protein kinase C (PKC) can phosphorylate the ankyrin repeat region of MYPT, and thus attenuate the interaction of MYPT with PP1c and MRLC (Toth et al., 2000). Furthermore, CPI-17, a smooth muscle-specific inhibitor of MLCP, can also regulate the phosphatase activity of MLCP. Phosphorylation of CPI-17 by PKC, or ROCK, or protein kinase N, or p21activated kinase (PAK) dramatically enhances the inhibition ability of CPI-17 (Eto et al., 1997;Koyama et al., 2000;Senba et al., 1999;Takizawa et al., 2002a;Takizawa et al., 2002b). Finally, MRLC can also be phosphorylated by ROCK and PAK, which itself is a substrate of Rac and Cdc42. Thus ROCK can regulate MRLC phosphorylation both through direct phosphorylation of MRLC and through inactivation of MLCP. Importantly, although the biochemistry of these phosphorylation events is well characterized, the physiological significance of these regulatory steps in vivo remains to be explored.
The in vivo function of non-muscle myosin II has been extensively analyzed in Drosophila melanogaster, Dictyostelium discoideum and Saccharomyces cerevisiae. Drosophila has a single non-muscle myosin II heavy chain encoded by zipper (zip), as well as a single non-muscle myosin II regulatory light chain encoded by spaghetti squash (sqh). Analysis of the phenotypes associated with mutations in zip and sqh have revealed that non-muscle myosin II regulates cell shape changes and cell movements in multiple processes such as cytokinesis, dorsal closure and oogenesis (Edwards and Kiehart, 1996;Jordan and Karess, 1997;Wheatley et al., 1995;Young et al., 1993). In addition, mutations in both zip and sqh affect planar cell polarity during development (Winter et al., 2001).
The temporal requirement of zip has been studied in sqh 2 mutant animals that carry a sqh transgene driven by a heat shock promoter (Edwards and Kiehart, 1996). This analysis showed that sqh activity is needed for eye and leg imaginal discs morphogenesis. Also, during oogenesis, sqh is required for morphogenesis of interfollicular stalks, border cell migration, centripetal cell ingression, dorsal appendage cell migration, and rapid transport of the nurse cell cytoplasm into the oocyte. Inhibition of this transport was also observed in animals that carry homozygous sqh 1 germline clones (GLCs) (Wheatley et al., 1995).
The in vivo function of MRLC phosphorylation was determined by expression of sqh transgenes that contain mutated phosphorylation sites in a sqh null mutant background (Jordan and Karess, 1997). Embryos carrying the null mutation sqh AX3 die, mostly during the first larval instar, and sqh AX3 GLCs develop extensive defects, including failure in cytokinesis, during oogenesis. SqhA20A21, which has both the primary and secondary phosphorylation sites changed to alanine, failed to rescue sqh AX3 , indicating that phosphorylation of Sqh is important for myosin II function. In support of this, a change of serine 21 to glutamic acid (SqhE21), that presumably mimics constitutive phosphorylation of Sqh, substantially rescues the sqh AX3 oogenesis phenotype.
To gain further insight into the regulation of Zip and to define precisely the in vivo function of MLCP, we have cloned the Drosophila homologue of the MYPT gene (DMYPT). We find that DMYPT is essential for cell sheet movement during dorsal closure, morphogenesis during eye development, and ring canal growth during oogenesis. Our results indicate that regulation of the phosphorylation state of MRLC, and dynamic activation and inactivation of myosin II, are essential for its various functions during many developmental processes.

Cloning of DMYPT
The DMYPT cDNAs, AT12677, RE34228, RE63915 and AT31926 were obtained from ResGen, Invitrogen Corporation. AT12677 was sequenced completely in both directions, while the other clones were only sequenced from their most 5′ and 3′ ends. To determine the organization of the DMYPT locus, sequences from all clones were assembled into one contig and then compared with the Drosophila genome sequence. AT12677, RE34228, RE63915 contain the entire DMYPT open reading frame (ORF). RE34228, RE63915 include all of exon 1, AT12677 starts from the middle of exon 2, 43 bp 5′ of the start codon. AT31926 starts within exon 4. AT12677, RE34228 and AT31926 share the same 3′ end.
Generation and rescue of DMYPT mutations P-element insertions were excised using the ∆2-3 transposase following conventional methods. These excision lines were then analyzed for lethality and fertility. For rescue experiments, the DMYPT ORF from AT12677 was cloned by PCR into CaSpeR-hs between NotI and XbaI, and injected with helper DNA (a source of ∆2-3 transposase) into w 1118 flies to generate transgenic lines. One of the transgenic lines, hs2-4, is a viable insertion on the X chromosome and was used to generate hs2-4/+; DMYPT 03802 /TM3,Sb animals. The numbers of rescued DMYPT 03802 mutant progeny were scored following a 1-hour daily heat-shock treatment at 37°C.

Cuticle and eggshell preparation
For cuticle preps, overnight egg collections were aged for 30 hours at 25°C, dechorionated with a 50% solution of commercial bleach, washed with PBST (PBS and 0.1% Triton X-100), mounted in Hoyers mounting medium with lactic acid, and heat-treated at 60°C overnight. Eggshells were prepared similarly without the bleaching step. Images were taken with a SPOT™ digital camera (Diagnostic Instruments) using phase-contrast or dark-field optics on a Zeiss Axiophot microscope and processed with Adobe Photoshop.

Embryo staining
In situ hybridization to embryos was performed as described previously (Patel et al., 1987). For immunofluorescence staining, embryos from mutant/TM3,actinGFP were fixed and stained with rabbit anti-GFP serum (1:2000, Molecular Probes) and mouse antifasciclin III (7G10, 1:20, DSHB) or anti-phosphotyrosine (4G10, 1:1500, UBI). Homozygous mutant embryos were identified by the absence of GFP. Role of MYPT in Drosophila  [A20A21], and RhoA 720 /CyO. The progeny were raised at 29°C and the resulting adults were dehydrated in an ethanol series, dried in SAMDRI PVT-3B and coated with Hummer V Sputter Coater. Scanning electron micrographs were generated using a LEO 1450 VP electron microscope.

Structure of DMYPT
A BLAST search of the Drosophila database (http://flybase.bio.indiana.edu) with mammalian MYPT sequences reveals that the Drosophila genome has a single related gene, CG5891. CG5891 is predicted to encode a protein with limited homology to mammalian MYPT at the N terminus (~300 aa). However, sequence analysis of several cDNAs derived from CG5891 (see Materials and Methods) uncovered additional regions of homology between the mammalian and fly homologues suggesting that the predicted CG5891 gene was incorrectly annotated. A representative cDNA, AT12677, encodes an ORF of 1101 amino acids (aa) that we named Drosophila MYPT (DMYPT) to follow the nomenclature of the mammalian protein. A comparison of the compiled DMYPT cDNA and genome sequences shows that the DMYPT locus contains 18 exons and 17 introns (Fig. 1A). The start codon lies in the second exon and the stop codon in the last. Sequence alignment shows that DMYPT shares significant homology with human MYPTs in three regions (Fig. 1B), the N terminus containing several ankyrin repeats, the C terminus, and a short peptide in the middle that contains the highly conserved inhibitory phosphorylation site ( Fig. 1C) (Kawano et al., 1999;Kimura et al., 1996).

Genetics of the DMYPT locus
To characterize the consequences of loss of DMYPT function during development, we searched for mutations in the DMYPT gene. Two P-element transposon insertions in the DMYPT locus have been defined molecularly by recovery of flanking genomic sequence (Fig. 1A). EP(3)3727, in the first intron, is homozygous viable and l(3)03802, in the tenth intron, is associated with zygotic lethality. We also identified several deficiencies that remove DMYPT sequences based on genetically defined breakpoints as well as their failure to complement l(3)03802 (Fig. 1D). Df(3L)th102 deletes DMYPT entirely and thus serves as a complete loss-of-function allele for use in this study.
To determine whether the l(3)03802 P-element insertion within the DMYPT locus is responsible for the lethality, and to generate new deletion alleles, we excised both DMYPT Pelement insertions using the ∆2-3 transposase. Mobilization of each element resulted in the recovery of both viable precise excisions and lethal imprecise excisions. Among the >200 excisions derived from l(3)03802, over half were viable, indicating that the lethality associated with the l(3)03802 chromosome is due to disruption of DMYPT and not another lethal hit. Thus l(3)03802 is renamed as DMYPT 03802 and EP(3)3727as DMYPT 3727 . Two of the strongest embryonic lethal excision lines, DMYPT 2-188 and DMYPT  , like the original insert, DMYPT 03802 , fail to complement Df(3L)th102 and are described in detail below. Eleven of the 39 lethal excisions derived from DMYPT 3727 failed to complement with DMYPT 03802 and Df(3L)th102, which is consistent with the notion that they disrupt DMYPT activity.
To confirm that the DMYPT 03802 insertion disrupts DMYPT function and that the cDNA derived from the DMYPT locus encodes all the functions associated with DMYPT activity, we rescued the original lethal P insertion with a transgene containing a heat shock promoter driving a DMYPT cDNA. Following 1-hour heat treatments daily from embryogenesis to eclosion, hs-DMYPT fully rescues DMYPT 03802 homozygous animals to adulthood. Stopping heat treatment 1 to 2 days before eclosion lead to incomplete rescue of DMYPT 03802 , with adults developing wing and leg defects similar to those noted for zip or sqh mutants partially rescued by a transgene (Edwards and Kiehart, 1996;Halsell et al., 2000) (data not shown). Stopping heat treatment 3 days prior to eclosion resulted in no rescue to adulthood. The complete rescue of the lethality associated with DMYPT 03802 by the hs-DMYPT transgene demonstrates that loss of DMYPT activity is responsible for the lethal phenotype.

Loss of DMYPT activity during embryogenesis is associated with a dorsal closure phenotype
To assess the timing and cause of lethality associated with the DMYPT 03802 insertion, embryos were collected and analyzed. Lethal phase analysis showed that 44% of homozygous DMYPT 03802 animals die during embryogenesis, while the remaining 56% die during early first larval instar (485 total embryos counted). More than 80% of the dead mutant embryos displayed a failure of dorsal closure with a characteristic dorsal hole in their cuticles (Fig. 2B,C). The size of the hole in such flies is variable and is also influenced by the genetic background (data not shown). Homozygous Df(3L)th102 embryos (Fig. 2D), as well as DMYPT 03802 /Df(3L)th102 embryos (Fig. 2E) also showed dorsal closure defects. The embryonic cuticle phenotype of DMYPT 03802 /Df(3L)th102 is more severe (more embryos displayed large dorsal holes) than homozygous DMYPT 03802 , suggesting that DMYPT 03802 is a hypomorphic allele. In addition, all of the embryonic lethal excision lines analyzed that were derived from DMYPT 03802 (data not shown), and ten of the lethal excision lines from DMYPT 3727 (Fig. 2F), produced embryos with dorsal closure defects. Altogether, these results indicate that DMYPT is required for dorsal closure.
Dorsal closure involves a cell sheet movement where the dorsal-lateral ectoderm on both sides of the developing embryo moves toward the dorsal midline to cover a degenerative squamous epithelium, the amnioserosa (reviewed by Knust, 1997;Noselli and Agnes, 1999;Stronach and Perrimon, 1999). This epithelial cell sheet movement encloses the embryo in a continuous protective epidermis. Genetic loss-of-function studies have identified the Jun N-terminal kinase (JNK) signal transduction cascade as one of the key modulators of dorsal closure morphogenesis (Noselli and Agnes, 1999). Transcriptional targets of JNK signaling include decapentaplegic (dpp), a secreted morphogen related to the bone morphogenetic proteins (BMPs), and puckered (puc), a dual-specificity phosphatase that mediates a negative feedback loop of the JNK signal transduction pathway via dephosphorylation of JNK.
To determine whether the failure of dorsal closure in DMYPT mutants is due to an influence on JNK signaling, we assayed for dpp expression in the leading cells of the ectoderm during closure. In situ hybridization revealed that the spatial and temporal expression pattern of dpp is normal in DMYPT mutant embryos (data not shown), suggesting that DMYPT does not function through the JNK pathway during dorsal closure.
To further examine the cause of dorsal closure defects in the mutants, we stained DMYPT mutant embryos for markers that allowed us to analyze the cell polarity and shape in the dorsal ectoderm. We observed apically localized phosphotyrosine immunoreactivity similar to wild-type flies (Fig. 3A).
Moreover, there was normal basolateral fasciclin III immunostaining (Fig. 3B). Altogether, these results suggest that there are no gross defects in cell orientation or polarity. However, we did notice that older mutant embryos began to show abnormal cell shapes at the leading edge of the epidermis (Fig. 3B), which could account for the defects in dorsal closure observed in the DMYPT mutants.
Consistent with the late embryonic defects observed in DMYPT zygotic mutants, we find that DMYPT is maternally contributed and ubiquitously expressed during embryogenesis (data not shown). This maternal supply of DMYPT is likely the reason that the dorsal closure phenotype is variable among embryos and is influenced by genetic background. However, we cannot address this question directly since DMYPT is required during oogenesis (see below).
C. Tan, B. Stronach and N. Perrimon DMYPT is required for ring canal growth during oogenesis During oogenesis, each cystoblast divides four times with incomplete cytokinesis and produces one oocyte and fifteen support nurse cells that are all connected through cleavage furrows. These cleavage furrows subsequently develop into ring canals. These open rings allow the nurse cells to transport cytoplasm into the oocyte, slowly from stage 6 to stage 10, then rapidly at stage 11. This fast phase of transport is referred to as 'dumping', and has been shown previously to require the activity of Sqh (MRLC). In sqh mutant germline egg chambers, dumping is blocked (Wheatley et al., 1995).
To analyze the role of DMYPT during oogenesis, we generated homozygous mutant germline clones (GLCs) of DMYPT 03802 using the FLP-FRT/dominant female sterile technique (Chou and Perrimon, 1996). Females carrying DMYPT 03802 homozygous GLCs lay few tiny eggs, about a quarter of the size of wild type eggs (Fig. 4, compare A and C), which do not develop. Examination of the mutant egg chambers revealed that the dumping of nurse cell cytoplasm to the oocyte was blocked (Fig. 4, compare B and D). This is similar to the dumpless phenotype observed with sqh homozygous mutant GLCs as well as for mutants in other actin binding proteins (reviewed by . To investigate the basis of the dumpless phenotype associated with DMYPT 03802 GLCs, we stained actin filaments using Texas Red phalloidin. The most obvious defect involves the ring canals. At stage 8, wild-type egg chambers had large bagel-shaped ring canals (Fig. 5A). In contrast, the ring canals of DMYPT 03802 GLC egg chambers were very small (Fig. 5B).
To determine whether the ring canals of DMYPT 03802 GLCs never enlarged, or whether they grew and then collapsed, we examined the ring canals in different stage egg chambers. In wild-type egg chambers, ring canals grow from 1 µm at stage 2 to 10 µm at stage 11 ( Fig. 5C) (see also, Tilney et al., 1996). In contrast, the ring canals of DMYPT 03802 GLCs barely grew (Fig. 5D). Mutation of DMYPT in follicular cells have no effects on the ring canal growth (data not shown), suggesting that DMYPT is required in the germline for ring canal growth. Presumably, these small ring canals cannot support the fast phase cytoplasmic transport and thus cause the dumpless phenotype resulting in tiny eggs.
In addition to actin, several other proteins, including Hu-li tai shao (Hts), Kelch, and phosphotyrosine (pY)-containing proteins (Robinson et al., 1994;Xue and Cooley, 1993), are recruited to ring canals as they form. Immunolocalization experiments revealed that both Hts and Kelch were localized  to the small DMYPT mutant ring canals (Fig. 6A,B). Interestingly, although pY staining was present in the mutant ring canals, we also observed an ectopic accumulation of pY staining in the nurse cells (Fig. 6D arrows). The basis of this ectopic accumulation remains to be determined.
Next, we analyzed the subcellular distribution of Zip. It has been reported that mutation of Sqh caused Zip to form aggregates (Edwards and Kiehart, 1996;Jordan and Karess, 1997;Wheatley et al., 1995), thus we expected to detect an effect on Zip distribution in the absence of DMYPT. Surprisingly, no major changes in Zip distribution were detectable between wild-type egg chambers and DMYPT GLCs. In both cases, Zip was uniformly distributed at low level with enhanced cell cortex localization (Fig. 6C). Our observations are consistent with the result that DMYPT mutations have no effect on Zip localization during dorsal closure (Mizuno et al., 2002).

Interaction between DMYPT and the small GTPases during eye development
Previous studies have shown that the Rho family GTPases, Rac1, RhoA, and Cdc42, each play a role in dorsal closure (Glise and Noselli, 1997;Harden et al., 1995;Harden et al., 1999;Hou et al., 1997;Magie et al., 1999;Strutt et al., 1997), and may influence myosin activity through a RhoA mediated signal. Programmed overexpression of these genes by the eyespecific GMR promoter causes distinct rough eye phenotypes (Hariharan et al., 1995;Nolan et al., 1998). To pinpoint the C. Tan, B. Stronach and N. Perrimon   relationship of DMYPT with these GTPases, we examined the effects of reducing DMYPT activity on the rough eye phenotypes. Interestingly, reduction of DMYPT strongly enhanced the eye phenotype caused by GMR-Rac 7A (Fig. 7 compare B and C). The eyes of GMR-Rac 7A /DMYPT 03802 flies were much smaller, with fewer bristles and hexagonal-shaped ommatidia, than those of GMR-Rac 7A /OreR flies. Consistent with the idea that the P-insertion and the excisions are hypomorphic alleles, Df(3L)th102 enhanced the GMR-Rac 7A eye phenotype to an even greater extent than either DMYPT 03802 , DMYPT 2-188 or DMYPT 2-199 (data not shown). However, reduction of DMYPT had no effect on the size of the rough eye caused by either GMR-RhoA or GMR-Cdc42, although it did enhance the rough eye phenotype caused by GMR-RhoA as fewer bristles formed (Fig. 8 compare B and C). Together, these data suggest that DMYPT plays a role in eye development and functions downstream of, or in parallel with Rac and Rho.

DMYPT is a negative regulator of the Rho/myosin signaling pathway in vivo
RhoA functions downstream of Rac in determining ommatidia polarity in the eyes (Fanto et al., 2000). Reducing the dosage of RhoA enhances the effect of sev-Rac N17 , a dominant negative form of Rac driven by the sevenless (sev) enhancerpromoter in the eye, and suppresses the activity of sev-Rac V12 , which encodes a constitutively active form of Rac. Consistently, overexpression of RhoA (sev-RhoA) rescues sev-Rac N17 , while reduction the amount of Rac using a deficiency that uncovers Rac has no effect on the gain-of-function RhoA phenotype. Thus, similar to the Rho dependence on Rac function observed in mammalian fibroblasts, some developmental events in Drosophila also rely on a hierarchy of GTPase function (Nobes and Hall, 1995).
Consistent with these observations, reducing the dosage of RhoA partially suppresses the rough eye phenotype caused by GMR-Rac (Fig. 7, compare B and D). In fact, mutations of all the putative positive regulators of myosin activity (RhoA-Zip signaling pathway), including RhoA, Drok and zip itself, moderately suppress the rough eye phenotype of GMR-Rac, opposing the effect of DMYPT mutants (Fig. 7 compare B with D, E and F). This suggests that the RhoA-Zip signaling pathway functions downstream of Rac, and that DMYPT is a negative regulator of the pathway.
Importantly, replacing the phosphorylation sites of Sqh with alanine remarkably suppressed the rough eye phenotype, while replacing them with glutamic acid to mimic phosphorylation slightly enhanced the phenotype (Fig. 7 compare B with H, Fig. 8 compare D with E and F). This suggests that dephosphorylation of Sqh is important in eye morphogenesis and that DMYPT may be involved in regulating the dephosphorylation of myosin light chain in eye development.
To examine whether other myosins are also involved in this process, we tested the effect of myosin VIIA, an unconventional myosin encoded by crinkled (ck), in the same assay. Myosin VIIA was chosen because ck and zip behave antagonistically in wing hair number determination in the Drosophila adult wing (Winter et al., 2001). Interestingly, ck behaves oppositely to myosin II (Zip) during eye morphogenesis since a reduction in ck activity enhances the GMR-Rac rough eye phenotype, nearly to the same extent as a reduction in DMYPT (Fig. 7 compare B and G).   03802 inhibits the formation of bristles in GMR-RhoA eyes, but has little effect on the overall eye size. In addition, the phospho-mimicking sqh mutation enhances, while the non-phosphorylatable sqh mutation suppresses, the rough eye phenotype associated with GMR-Rac.

DISCUSSION
We have generated and characterized the phenotypes associated with mutations in the Drosophila MYPT gene. Our analyses indicate a role for this protein in various developmental processes. These include a role in dorsal closure, oogenesis and eye development. Dominant genetic interactions with DMYPT alleles reveal that DMYPT is a negative regulator of the RhoA-myosin signaling pathway, which acts downstream of, or in parallel with Rac.
Cell movement during dorsal closure Several lines of evidence suggest that DMYPT is essential for dorsal closure. First, the DMYPT 03802 P-element, which disrupts DMYPT activity, leads to embryos with a dorsal open phenotype that can be reverted by precise excision of the insertion. Second, deficiencies of the DMYPT locus, as well as a number of imprecise excisions of the P-element insertions in DMYPT, are also associated with embryonic lethality and dorsal closure defects. Third, the lethality associated with DMYPT 03802 is rescued to adulthood using a DMYPT transgene.
Given that RhoA and zip are required for dorsal closure (Strutt et al., 1997;Young et al., 1993), it is not surprising that DMYPT, a regulator of myosin function presumed to act downstream of RhoA, is also implicated in dorsal closure. Like RhoA (Lu and Settleman, 1999;Magie et al., 1999), DMYPT mutations do not affect dpp expression suggesting that the failure of dorsal closure in DMYPT mutants is independent of JNK signaling. Nonetheless, it is somewhat unexpected that the loss-of-function mutants of both zip and its putative negative regulator, DMYPT, have similar rather than opposite phenotypes, each displaying late defects in cell shape and elongation (Young et al., 1993). One possibility is that the activity of myosin II has to be regulated spatially. Sqh is phosphorylated at the leading edge, indicating activation of Zip at that site (Mizuno et al., 2002). In the DMYPT mutant, in addition to the leading edge, phosphorylated Sqh is also localized to the dorsal boundaries of the leading edge cells. Thus this pool of mislocalized active Sqh may increase the activity of Zip where it is normally less active, ultimately interfering with dorsal cell movement. Another possibility is that dynamic regulation of Zip activity is important for cell sheet movement. Perhaps activation of Zip is required for cell shape changes in the ectoderm and for maintaining tension as the epithelial front moves forward, but concomitant inactivation of Zip is also necessary for the cells to modulate adhesion allowing forward motility. This paradoxical requirement of myosin II activity is similar to the function of cell adhesion in cell movement; some cell adhesion is necessary for cell movement, but strong adhesion inhibits cell movement.

Role of DMYPT in ring canal growth during oogenesis
Drosophila oogenesis starts with cystoblasts undergoing 4 rounds of cell division. Through an unknown mechanism, cytokinesis of the cyst cell is arrested and the cleavage furrow that separates the cells does not close completely. The cleavage furrow is then stabilized and transformed into an early ring canal, which contains only an outer rim including the actin binding protein anillin (Field and Alberts, 1995), glycoprotein mucin-D (Kramerova and Kramerov, 1999), and phosphotyrosine proteins (Robinson et al., 1994). Then, filamin (cheerio) (Li et al., 1999;Sokol and Cooley, 1999), aducin-like protein hts-RC and filamentous actin are recruited to the ring canal to form an inner-rim (Robinson et al., 1994;Yue and Spradling, 1992). At the same time, phosphotyrosine proteins are also detected in the inner rim. Src64 and Tec29 are responsible for most of the phosphotyrosine staining Guarnieri et al., 1998;Roulier et al., 1998). Later, the inner rim is further stabilized by the actin bundling protein kelch (Kelso et al., 2002;Xue and Cooley, 1993).
Ring canals grow in diameter, thickness and length in two phases (Tilney et al., 1996). First, the canal increases in thickness (~6 fold) from stage 2 to 5, while its diameter and length barely grow. At the same time the number of actin filaments increase from 80 at stage 2 to ~700 at stage 6. Second, the diameter and length grow enormously, while the thickness stays the same. During the second phase, the actin filaments are changed into discrete bundles. Astonishingly, the total number and density (number of filaments per cm 2 ) of actin filaments remain the same.
Fluorescence recovery after photobleaching experiments have shown that the ring canal actin is highly dynamic, constantly cycling between polymerization and depolymerization (Kelso et al., 2002). This, together with the involvement of the actin-nucleating protein complex Arp2/3 in ring canal growth , argues that ring canals grow by de novo actin polymerization and regulated cross-linking. This model requires that the newly assembled actin filaments must slide past other bundles since there is no seam in the ring canal.
DMYPT could function at several times during ring canal formation, including cytokinesis arrest, initiation of ring canal formation, or growth of the ring canal. Since the ring canal starts as a cleavage furrow of cytokinesis, myosin II is presumably there. DMYPT may be necessary to inhibit myosin-powered contraction because in the DMYPT mutant we observe that the ring canals are smaller, presumably as a result of overcontraction. Secondly, the sliding of the anti-parallel actin filaments is likely to be driven by myosin. In the absence of myosin activity, such as in the sqh mutant GLC, ring canals are deformed, often not smooth and round, but pointed, and loosely packed (Jordan and Karess, 1997). The deformed ring canals may also contain seams. However, as during dorsal closure, the activity of myosin must be precisely regulated. Unregulated myosin II activity, for example, in the DMYPT mutant, may cause over-sliding of the actin filaments, thus blocking ring growth, while maintaining overall ring canal morphology. Finally, myosin II may be involved in actin filament turnover or bundling. In this case, hyperactivated myosin in the DMYPT mutant may cause the actin filaments to be constitutively locked, unable to incorporate new actin to promote ring canal expansion.
Loss of sqh activity also blocks dumping of the nurse cell cytoplasm into the oocyte (Wheatley et al., 1995). This is related in part to the inactivity of myosin II in the cell cortex in sqh mutants, which under normal circumstances provides the force for rapid transport of nurse cell contents to the oocyte during dumping. Currently from our analysis of DMYPT Role of MYPT in Drosophila GLCs, it is not clear whether DMYPT also regulates the activity of zip in the cell cortex.
Mutations in sqh have also pinpointed roles of myosin II in morphogenesis of interfollicular stalks, border cell migration, centripetal cell ingression, and dorsal appendage cell migration, all processes that involve the somatic tissue surrounding the germline during oogenesis (Edwards and Kiehart, 1996). Centripetal cell ingression has been compared to dorsal closure because myosin II is highly localized and forms a ring at the leading edge of the migrating cells like those at the leading edge of the ectoderm during dorsal closure. It will be very intriguing to see if DMYPT has functions in these somatic cells of the egg chamber.

Regulation of myosin II
The regulation of MRLC phosphorylation is essential to modulate myosin II activity and can be controled by several distinct mechanisms. For instance, RhoA can activate its effector ROCK that in turn phosphorylates MYPT, either directly or indirectly. MYPT phosphorylation inhibits the phosphatase activity of MLCP and leads to elevation of MRLC phosphorylation. Phosphorylation of MRLC can also be increased by activation of MLCK, another downstream target of RhoA (reviewed by Hartshorne et al., 1998;Somlyo and Somlyo, 2000). Thus, the antagonistic activity of kinase and phosphatase is thought to engender a delicate balance of myosin II activity modulated through the phosphorylation state of its regulatory light chain.
To assess the relationship between DMYPT regulation of myosin II and signaling via the Rho GTPase family members, we turned to the Drosophila eye where sensitive genetic interactions can be observed. One study has implicated RhoA function downstream of, or in parallel with, Rac during orientation of ommatidia in the eye (Fanto et al., 2000). Consistent with this, we found that reducing the amount of RhoA, Drok and zip partially alleviates the eye defect associated with overexpression of Rac, while reducing the dosage of a putative negative regulator of myosin enhances the rough eye phenotype. Furthermore, expression of a nonphosphorylatable form of Sqh, which presumably reduces the activity of Zip, dramatically rescues the phenotype, while overexpression of a phospho-mimicking Sqh mutant, which should increase the activity of myosin, exacerbates the eye defects. Taken together, these data indicate that the regulation of myosin II activity via balancing the phosphorylation level of Sqh is critical for proper morphogenesis of the Drosophila eye. Based on our results, we propose that it is DMYPT that mediates myosin II downregulation in this system.
Recently, Winter and colleagues have identified similar genetic interactions between RhoA, Drok and zip in restricting the number of F-actin based prehairs in the development of wing cells (Winter et al., 2001). Not surprisingly, the same genetic relationship holds true during dorsal closure (Mizuno et al., 2002). Overexpression of Drok, a positive regulator of Zip, phenocopies a mutation in the negative regulator, DMYPT. Moreover, a loss-of-function mutation of zip potently suppresses the embryonic lethality caused by mutation of DMYPT or over expression of Drok. In other developmental contexts, myosin II functions downstream of Rho and/or MYPT (Halsell et al., 2000;Mizuno et al., 2002;Piekny et al., 2000) suggesting that similar mechanisms underlie all of these very diverse biological processes. Since they all require actin cytoskeletal reorganization, it suggests that RhoA regulates cytoskeletal remodeling in non-muscle cells in vivo through the RhoA kinase-MYPT-myosin II pathway.
Interestingly, crinkled (myosin VIIA), an unconventional myosin, behaves antagonistically to Zip/myosin II in both eye morphogenesis (this study) and wing hair number restriction (Winter et al., 2001). This suggests that various myosins interact in different cell types to regulate reorganization of the actin cytoskeleton. It will be interesting to determine the specificity of functions of different myosins and their modes of regulation. Since there are many different myosins, and yet a single MYPT in Drosophila, it remains to be determined whether, and how, DMYPT interacts with other myosins.
In conclusion, we have identified the Drosophila homologue of mammalian MYPT, named DMYPT accordingly. DMYPT plays multiple roles during Drosophila development. Loss of DMYPT function leads to blockage of rapid transport of nurse cell cytoplasm, inhibition of ring canal growth, failure of dorsal closure, defects of eye morphogenesis, and other unidentified processes during pupae development. Furthermore, our data indicate that dynamic regulation of myosin II activity via regulating phosphorylation level of myosin regulatory light chain by DMYPT is critical for the function of myosin II.