Myosin II regulation during C. elegans embryonic elongation:LET-502/ROCK, MRCK-1 and PAK-1, three kinases with different roles

Morphogenesis is an essential feature of embryonic development, during which cells change their overall shape to build tubular organs, extend cell sheets or organize into specific patterns. These changes depend on the organization of the cytoskeleton as well as on its dynamic behavior. Characterization of different morphogenetic processes has revealed the central role played by myosin II and its regulators, the Rho-binding kinase (ROCK) and myosin phosphatase (Conti and Adelstein,2008; Quintin et al.,2008).

Myosin II is an hexamer composed of two myosin heavy chains, two essential light chains and a pair of regulatory light chains (RMLCs), which converts chemical energy into mechanical force(Conti and Adelstein, 2008). Its activity is regulated in part through phosphorylation of the RMLC by serine/threonine kinases, in particular by ROCK(Somlyo and Somlyo, 2003). Although the importance of myosin II in morphogenesis is undisputed, its spatial organization and specific mechanism of activation have seldom been examined in vivo. Furthermore, its role has so far been documented mainly in morphogenetic processes involving a homogenous cell population in which all cells behave similarly (Quintin et al.,2008).

As an example of morphogenesis, we study C. elegans embryonic elongation, during which a bean-shaped embryo is changed into a worm-shaped larva. This process is driven by coordinated shape changes of epidermal cells that lead to a decrease in circumference and a 4-fold increase in length(referred to here as the 4-fold stage)(Priess and Hirsh, 1986). The epidermis is composed of three major cell groups, namely dorsal, ventral and lateral seam cells, with distinct anatomical properties: (1) dorsal and ventral cells, but not lateral seam cells, are in contact with muscles; and(2) most dorsal cells form a syncytium at the 1.5-fold stage(Chisholm and Hardin,2005).

Treatment of embryos with cytochalasin D established that actin microfilaments (MFs) present in epidermal cells are crucial for elongation(Priess and Hirsh, 1986). Genetic analysis revealed that mutations affecting MLC-4 (RMLC), NMY-1 (one of the two myosin II heavy chains) or LET-502 (ROCK) induce severe elongation defects (Piekny et al., 2003; Shelton et al., 1999; Wissmann et al., 1997). Conversely, mutations affecting MEL-11, the myosin-binding subunit of myosin phosphatase (MYPT), induce embryonic rupture probably due to hypercontraction of the actomyosin cytoskeleton, as an nmy-1 mutation prevents the rupture of mel-11 mutants (Piekny et al., 2003). Thus, the pharmacological and genetic analysis performed so far has not examined whether the different epidermal cells play distinct roles in elongation. Furthermore, although genetic analysis would predict that LET-502/ROCK activates MLC-4/RMLC, this has never been formally established during elongation.

C. elegans represents a powerful system to dissect myosin II function and regulation, in part because the organization of actin microfilaments in the epidermis provides exceptional spatial resolution. Here,we use new fluorescent GFP reporters to provide a detailed analysis of how myosin II dynamically functions during embryonic morphogenesis. We identify the myosin essential light chain and characterize its role in elongation. Using tissue-specific promoters and mutant transgenes, we determine in which cells myosin II acts to drive elongation, and dissect the molecular regulation of MLC-4 by phosphorylation. In particular, we examine the potential role of several kinases whose vertebrate orthologs can phosphorylate RMLC. We find that ROCK, PAK and MRCK homologs play distinct functions during embryonic elongation, and discuss what their potential targets might be.

Control N2 and other animals were propagated at 20°C (unless noted otherwise) (Brenner, 1974). The following strains were used: ML1395, rrf-3(pk1426) II;mcEx227[lin-26p::ABD_VAB-10::GFP, rol-6(su1006)]; EU618, mlc-4(or253)/qC1[dpy-19(e1259), glp-1(q339)] III; ML1148, mlc-4(or253)/qC1;mcEx399[mlc-4p::GFP::mlc-4WT, rol-6(su1006)];ML1151, mlc-4(or253)/qC1;mcEx402[mlc-4p::GFP::mlc-4DD,rol-6(su1006)]; ML1204, mlc-4(or253)/qC1;mcEx317[mlc-4p::GFP::mlc-4AA, rol-6(su1006)];ML1111, mlc-4(or253)/qC1;mcEx323[mlc-4p::GFP::mlc-4T17A,rol-6(su1006)]; ML1114, mlc-4(or253)/qC1;mcEx326[mlc-4p::GFP::mlc-4S18A, rol-6(su1006)];ML1567, mlc-4(or253)/qC1;mcEx540[ceh-16p::GFP::mlc-4WT, myo-2p::GFP];ML1178, mlc-4(or253)/qC1;mcEx407[elt-3p::GFP::mlc-4WT, myo-2p::GFP];ML1568, mlc-4(or253)/qC1;mcEx541[ceh-16p::GFP::mlc-4WT,elt-3p::GFP::mlc-4WT, myo-2p::GFP]; HR1157, let-502(sb118ts) I;KK332, mel-11(it26) unc-4(e120) sqt-1(sc13)/mnC1[dpy-10(e128)unc-52(e444)] II; RB689, pak-1(ok448) X; ML1594, mrck-1(ok586)/nT1[qIs51] V; ML1312, mcIs49[mlc-4p::GFP::mlc-4WT,rol-6(su1006)]; ML1153, mcIs46[dlg-1::rfp;unc-119(+)]; ML1310, mcEx467[lin-26p::ABD_VAB-10::GFP, myo-2p::GFP]; ML1152, mcEx403[mlc-5p::GFP::mlc-5, rol-6(su1006)]. pak-1 and mrck-1mutations were outcrossed four times. For genetic epistasis experiments,mothers were transferred to fresh plates for 2 hours. Half of this synchronized progeny was observed 5-7 hours after egg laying to determine the terminal phenotype by DIC microscopy (at least 50 embryos were observed). The second half was observed 24 hours after egg laying to quantify embryonic lethality.

RNAi was performed either using the Ahringer-MRC feeding RNA interference(RNAi) library (Kamath et al.,2003), or by injection of double-stranded RNA synthesized from PCR-amplified genomic fragments using a T3 or T7 mMESSAGE mMACHINE Kit(Ambion, Austin, TX, USA).

A feeding RNAi screen was performed on ML1395. L3-L4 stage larvae were placed for 36-40 hours on bacteria producing dsRNA. From these adults, a synchronized egg progeny was observed for elongation and actin organization defects 5 hours after egg laying. Primers used to inactivate mlc-5and mlc-4 were:5′-AATTAACCCTCACTAAAGGGAGCACATGGTATTGGGGGAAGTGG-3′ and 5′-AATTAACCCTCACTAAAGGGAGCACATGGTATTGGGGGAAGTGG-3′;5′-AATTAACCCTCACTAAAGGGCTGGGAGAGAGCGAATAAGAAAT-3′ and 5′-AATTAACCCTCACTAAAGGGCTCATTCTCTCTTTTATCGCCAA-3′,respectively.

Double-stranded RNA was injected into wild-type and let-502(sb118ts) young adults. Injected mothers were stored at the restrictive temperature of 25.5°C overnight and synchronized for egg laying the next morning. The progeny was scored 6-8 hours after egg laying,∼24 hours after dsRNA injection in the mothers.

An RT-PCR fragment corresponding to codons 1-290 of vab-10a, which encode two predicted calponin-homology domains(Bosher et al., 2003), was cloned at the NheI site of a 4 kb lin-26 promoter fragment[eFGHi construct, see (Landmann et al.,2004)] upstream of green fluorescent protein(GFP).

pML1517 construction was similar to pML1572, except that the GFPwas replaced by the mcherry coding sequence (gift from K. Oegema,Ludwig Institute for Cancer Research - UC San Diego, La Jolla, CA, USA), and a 5 kb lin-26 promoter was used.

The final construct contained 2 kb of the mlc-4 promoter, the GFP coding sequence and the mlc-4 coding DNA downstream of GFP, all assembled in vector pPD97.82, from which the NLS had been removed (gift from A. Fire, Stanford University, Stanford, CA, USA). The different mlc-4 variants were obtained by PCR-driven mutagenesis onto pML1522 using mutated primers. To generate ceh-16p- and elt-3p-driven GFP::mlc-4 constructs, the respective promoter regions of ceh-16 (2.9 kb) and elt-3 (1.9 kb) genes were introduced into pML1522 to replace the mlc-4 promoter.

The final construct contained 2.2 kb of the mlc-5 promoter, the GFP coding sequence and a 2.7 kb region containing mlc-5 and taf-9, which were PCR-amplified and cloned between the NotI and BamHI sites of pBSKII⁺.

The final construct was assembled from a PCR-amplified region including 2.9 kb of the mrck-1 promoter and most of the coding region, except that the final coding part was substituted with a 1.1 kb PCR cDNA fragment. These fragments were cloned between the PstI and KpnI sites of pPD95.75.

Constructs were injected at 10 ng/μl with either pPD118.33(myo-2p::GFP at 5 ng/μl), pCFJ90 (myo-2p::mcherry at 5 ng/μl) or pRF4 (rol-6(su1006) at 100 ng/μl) as a selection marker and pBSKII⁺ (up to 200 ng/μl final concentration).

Images were captured with a Leica TCS SP5 confocal microscope (using a 63× NA 1.4 Plan Apochromat HCX oil objective) controlled by the Leica LAS AF imaging software. Illumination was via 458 nm and 561 nm lasers. Images were computationally projected using the ImageJ software (ImageJ, NIH,Bethesda, Maryland, USA; http://rsb.info.nih.gov/ij/). Short time-lapse experiments were performed on live embryos using a Leica DMI6000 spinning-disk confocal microscope (100× NA 1.43 Plan Apochromat objective), coupled with an Andor Revolution Spinning Disk System (iXon EMCCD 512 × 512 camera and Yokogawa spinning disk head). For spinning-disk imaging of live embryos, single focal planes are shown.

Embryos were mounted on a 2% agarose pad in M9 buffer and the coverslip was sealed with paraffin oil. For SP5 confocal recordings, oxygen was depleted from the chamber by adding OP50 aerobic bacteria to the medium. This procedure should halt ATP production by the respiratory chain, and might thus indirectly affect protein phosphorylation. For each panel, at least 15 embryos were examined.

All mlc-4 constructs were tested in the mlc-4(or253)/qC1[dpy-19(e1259)glp-1(q339)] background. Freshly hatched L1 (less than 1.5 hours) were mounted in M9 buffer and photographed using a Zeiss Axioplan microscope under a 10× objective; their length was measured using the ImageJ software. Larvae were recovered from the slide and allowed to grow for 2.5 days at 25°C. The resulting adult homozygous mutants and heterozygotes were discriminated by PCR genotyping. Transgenes injected at 0.5 ng/μl failed to rescue, and transgenes injected at 2.5 ng/μl gave a rescue comparable to 10 ng/μl (not shown).

Until now the actin cytoskeleton has been visualized by phalloidin staining on fixed embryos, with a high background due to muscle actin staining(Costa et al., 1997; Priess and Hirsh, 1986). To overcome this limitation, we developed a fluorescent probe to visualize actin specifically in the epidermis, which confirmed and extended previous descriptions of MFs. We fused the predicted actin-binding domain (ABD) of the spectraplakin VAB-10 (Bosher et al.,2003) to green fluorescent protein (GFP), and put the resulting construct under the control of the epidermal lin-26 promoter. The corresponding ABD from the VAB-10 homologs Drosophila Shot and mouse MACF1 can bind actin in vitro (Lee and Kolodziej, 2002; Sun et al.,2001). We confirmed that the distribution of this probe (referred to below as ABD::GFP) coincides with phalloidin staining (see Fig. S1A-C in the supplementary material). The ABD::GFP fluorescence was initially detected along the lateral membranes of epidermal cells before dorsal intercalation,and then was restricted apically (see Fig. S1F,G in the supplementary material).

Starting at the 1.2-fold stage, MFs formed a complex meshwork in all epidermal cells (Fig. 1A-A″) that was highly dynamic. Indeed, time-lapse experiments showed rapid polymerization and shortening events (between 3 μm and 10 μm per second, n=5 cables); although we are observing an actin-binding reporter rather than actin itself, these measures are quite comparable to observations performed in cell culture using mDIA1 (DIAP1 -Mouse Genome Informatics) (Higashida et al., 2004). Subsequently, in seam cells, MFs remained rather short and fuzzy throughout elongation (Fig. 1A-D,A′-D′). By contrast, in dorsal and ventral epidermal cells (DV cells), MFs progressively formed evenly spaced parallel circumferential bundles (CFBs) (Fig. 1B-B″), particularly after the 2-fold stage when muscle contractions begin (Fig. 1C-C″). Some apparent discontinuities could be reproducibly observed in DV cells at the 2-fold stage(Fig. 1C″, arrowhead; see Fig. S1H-I′ in the supplementary material), but were disappearing after the 3- to 3.5-fold stage (Fig. 1D-D″). Their position was consistent with that of muscle anchoring structures (fibrous organelles), indicating that CFBs might run from adherent junctions to fibrous organelles rather than from one seam border to the other. The number of CFBs was constant between the 2- and 3.5-fold stages in DV cells, with ∼351±18 CFBs per embryo (n=15 embryos),in agreement with previous estimates(Costa et al., 1997). At the pretzel stage, actin bundles disassembled (data not shown).

Remodeling of actin MFs is often driven by non-muscle myosin II(Conti and Adelstein, 2008). To determine whether this is the case during C. elegans embryonic elongation, we examined the precise organization of myosin II. We generated a GFP N-terminal translational fusion for MLC-4/RMLC (referred to here as GFP::MLC-4WT), which could rescue the embryonic lethality and elongation defects of a strong mlc-4 mutant and displayed a filamentous pattern(see Fig. S2A,C in the supplementary material; Fig. 2B,C). As embryos move vigorously beyond the 1.8-fold stage, we anaesthetized embryos by oxygen deprivation to precisely describe the pattern (see Materials and methods). However, this treatment induced the rapid disappearance of filaments (see Fig. S2B in the supplementary material).

As part of our efforts to dissect the molecular regulation of MLC-4 (see below), we noticed that a phosphomimetic form of MLC-4, in which residues T17 and S18 were changed into aspartate (GFP::MLC-4DD), formed a stable filamentous-like pattern even in the absence of oxygen(Fig. 1E-H). We therefore used this modified form to examine MLC-4 distribution, as the GFP::MLC-4DD reporter showed a pattern similar to the wild-type reporter (compare Fig. 1H with Fig. S2C in the supplementary material). We observed that MLC-4 is zygotically first detected during dorsal intercalation (data not shown). The expression level of GFP::MLC-4DD (Fig. 1E-H) was higher in seam cells than in DV cells at all stages. In seam cells, starting at the 1.2-fold stage, GFP::MLC-4DD formed aggregates(Fig. 1E-E″) that were not systematically associated with cell junctions(Fig. 1J). These aggregates evolved into a meshwork (Fig. 1F′-H′), which was reminiscent of the actomyosin meshwork that establishes embryonic polarity in C. elegans zygotes(Munro et al., 2004). By contrast, in DV cells, the GFP::MLC-4DD fusion protein formed puncta(Fig. 1E″) that progressively became aligned and denser, until they formed filaments(Fig. 1F″-H″)overlapping actin (Fig. 1I). GFP::MLC-4 was also detected in adults (see Fig. S3 in the supplementary material). Together, the distinct patterns of MFs and myosin II in seam versus DV cells argue that both cell types should play different roles during elongation.

To determine whether the high MLC-4 expression in seam cells is physiologically relevant, we performed tissue-specific rescue experiments by expressing the GFP::mlc-4WT construct either in DV cells using the elt-3 promoter (Gilleard et al.,1999) or in seam cells using the ceh-16 promoter(Cassata et al., 2005) (see Fig. S4 in the supplementary material). GFP::MLC-4WT expression in DV cells could not rescue the elongation defects of mlc-4(or253) mutants,despite having a normal localization, and was even reducing the size of mlc-4 larvae for two out of three lines(Fig. 2B,C). By contrast,GFP::MLC-4WT expression in seam cells could significantly rescue mlc-4 mutants, suggesting that seam cells are the main epidermal cells driving elongation. This rescue was not improved by the addition of both ceh-16- and elt-3-driven transgenes(Fig. 2C), indicating that the most likely explanation for this partial rescue resides in the strength of the ceh-16 promoter relative to that of mlc-4. Together, these results provide the first direct demonstration of the major role played by seam cells in driving elongation.

Phosphorylation of two highly conserved RMLC residues controls stress fiber formation in cell culture and cytokinesis in Drosophila(Jordan and Karess, 1997; Watanabe et al., 2007). To test whether phosphorylation of the corresponding residues in MLC-4 (threonine 17 and serine 18, T17 and S18, respectively)(Fig. 2A) regulates myosin II activity in C. elegans, we introduced point mutations into the GFP::mlc-4WT transgene. We generated four variants that were introduced into the mlc-4(or253) mutant: a constitutive phosphomimetic form, GFP::mlc-4(T17DS18D) (referred to as mlc-4DD); a non-phosphorylatable form, GFP::mlc-4(T17AS18A)(referred to as mlc-4AA); and two mono-phosphorylatable forms, GFP::mlc-4(S18A) and GFP::mlc-4(T17A) (referred to as mlc-4S18A and mlc-4T17A, respectively)(Fig. 2B). Both GFP::MLC-4WT and GFP::MLC-4DD proteins could fully rescue the embryonic elongation defect of mlc-4(or253) mutants (Fig. 2B,C). Rescued L1 larvae further developed into sterile adults of normal length, presumably due to poor germline expression, as described previously (Shelton et al.,1999). Conversely, GFP::MLC-4AA could not rescue the mlc-4(or253) elongation defect(Fig. 2B,C), whereas constructs in which either S18 or T17 are replaced by alanine (mlc-4S18A or mlc-4T17A) partially rescued elongation of mlc-4 mutants(Fig. 2B,C). Taken together,these experiments show that both T17 and S18 are required for MLC-4 function.

Our results highlight the importance of MLC-4 phosphorylation for its activity and actin association. Mains and co-workers predicted the existence of another kinase and possibly other myosin II regulators to account for the normal elongation of animals defective for both LET-502/ROCK and MEL-11/MYPT(Piekny et al., 2000; Wissmann et al., 1997). To identify such components, we designed two RNAi screens based on the following observations. Partial RNAi knockdown of MLC-4(Fig. 3F) or of the myosin heavy chain NMY-2 in the nmy-1 mutant background bypasses their early requirement for cytokinesis and leads to elongation arrest at the 1.2-fold stage (Piekny et al., 2003). We assumed that this phenotype reflects what should be observed when non-muscle myosin II is inactive, with essentially no maternal contribution specifically during elongation. Therefore we performed two RNAi screens in parallel, focusing in particular on genes whose knockdown causes arrest at the 1.2-fold stage.

In the first screen, we looked for severe elongation defects using the feeding RNAi library (Kamath and Ahringer,2003). In the second screen, we targeted kinases known to regulate myosin II in other systems. As described below, the first screen led to the identification of the myosin II essential light chain, whereas the second identified kinases acting in parallel to LET-502/ROCK.

Using a strain hypersensitive to RNAi for the first screen, we focused on genes previously shown to be essential for embryogenesis(Kamath and Ahringer, 2003),but eliminated genes predicted to be involved in general metabolism or transcription. Among the seven candidates retained (see Table S1 in the supplementary material), we characterized T12D8.6, the knockdown of which induced a strong elongation defect very similar to that observed in embryos lacking the two myosin heavy chains NMY-1 and NMY-2(Piekny et al., 2003). We named T12D8.6 mlc-5, as its predicted translation product displays strong homology (45% and 55% identity over its EF-hand domains)(Fig. 3B) to vertebrate and Drosophila myosin essential light chains(Edwards et al., 1995; Robert et al., 1984) (see Fig. S5 in the supplementary material).

mlc-5 RNAi by injection in adults mainly induced embryonic elongation defects ranging from 1.2- to 2-fold arrest(Fig. 3C,D), as observed for mlc-4 RNAi (Fig. 3F,G). Embryos failed to shorten circumferentially (compare dashed lines in Fig. 1D and Fig. 3E). Although the overall pattern of actin bundles was not dramatically affected after knocking down mlc-5 (Fig. 3E) or mlc-4 (Fig. 3H), it was partially distorted in ventral cells, making sometimes thicker bundles(Fig. 3E,H, arrow), indicating that myosin II only plays a minor role in organizing CFBs. This conclusion can only be tentative because such embryos must still contain some MLC-4 or MLC-5,as their complete absence arrests embryogenesis at the 1-cell stage (see Fig. S6E,G in the supplementary material)(Shelton et al., 1999).

The phenotype observed in MLC-5-depleted embryos suggests that MLC-5 could be part of the myosin II complex. To test this hypothesis, we generated an MLC-5 N-terminal GFP translational fusion(Fig. 4A). GFP::MLC-5 was first detected during dorsal intercalation (data not shown). It was then highly expressed in seam cells and to a lesser extent in DV cells(Fig. 4B), where it formed oxygen-sensitive parallel bundles starting at the 2-fold stage(Fig. 4C,D) overlapping with actin (Fig. 4E-G), like GFP::MLC-4. In mlc-4 knockdown embryos, GFP::MLC-5 failed to form bundles and instead formed aggregates (Fig. 4H-J), as previously observed for NMY-1 in mlc-4 mutant embryos (Piekny et al., 2003). Conversely, in mlc-5 knockdown embryos, GFP::MLC-4 formed small aggregates or was rather diffuse (data not shown), as observed for the GFP::MLC-4AA mutant form described above (see Fig. S2E in the supplementary material). To confirm that MLC-5 is part of the myosin II complex, we also observed the distribution of the myosin heavy chain NMY-2 after strong mlc-5 RNAi induced by injection in L4 larvae. Because NMY-2 is highly expressed in early embryos, we looked at the first cell division, a process that requires myosin II activity (see Fig. S6 in the supplementary material). We observed that mlc-5(RNAi) induced a strong cytokinesis defect (see Fig. S6E,G in the supplementary material) and that NMY-2::GFP failed to form a contractile ring (see Fig. S6F,H in the supplementary material). Thus, MLC-5 plays a major role during early embryonic development, like NMY-2(Guo and Kemphues, 1996) and MLC-4 (Shelton et al., 1999). Altogether, our results suggest that myosin II assembly and/or interaction with actin filaments require(s) the presence of both MLC-4 and MLC-5.

As mentioned above, at least one additional kinase has been predicted to act in parallel to LET-502. We assumed that its inactivation in the let-502 mutant background would lead to an early elongation arrest such as that observed when mlc-5 or mlc-4 are depleted. To identify this unknown kinase, we performed an enhancer RNAi screen in the thermosensitive let-502(sb118ts) mutant, targeting kinases known to activate the RMLC in other systems (see Table S2 in the supplementary material) (Batchelder et al.,2007; Matsumura,2005). The let-502(sb118ts) single mutant shows a characteristic 2-fold stage arrest at 25.5°C(Fig. 5C)(Diogon et al., 2007). Among the ten kinases tested (see Table S2 in the supplementary material), two of them aggravated the elongation phenotype of let-502(sb118ts) mutants,causing a 1.2-fold stage arrest (Fig. 5E,G; see Fig. S7 in the supplementary material) but did not display any elongation phenotype in control animals (not shown). These two kinases are the p21-activated kinase homolog PAK-1 and the myotonic dystrophy kinase-related cdc42-binding kinase homolog, MRCK-1(Fig. 7B).

To define the cause of this severe elongation defect, we first characterized the actin cytoskeleton in double mutants. As in mlc-4or mlc-5 knockdown embryos, there were only minor MF organization defects in single let-502(sb118ts) mutants and let-502;pak-1 and let-502;mrck-1 double mutants at the restrictive temperature (see Fig. S8 in the supplementary material), indicating that their elongation defect was not due to actin disorganization.

Using a functional assay, we then tested whether either kinase might be involved in MLC-4 phosphorylation. We examined whether the phosphomimetic MLC-4DD form described above could rescue the elongation defects of let-502, let-502;pak-1(RNAi) and let-502;mrck-1(RNAi)embryos. Indeed, in Drosophila, a phosphomimetic form of Sqh/RMLC allows 4% hemizygous rok (Drosophila ROCK) mutant survival until adulthood (Winter et al.,2001). We found that the mlc-4DD transgene could not rescue the elongation defects of let-502(sb118ts)(Fig. 5D,I) and poorly rescued the let-502(sb118ts);pak-1(RNAi) mutant elongation defects(Fig. 5F,I). These observations do not exclude the possibility that LET-502 and PAK-1 could phosphorylate MLC-4 (see Discussion). By contrast, the mlc-4DD transgene enabled let-502(sb118ts);mrck-1(RNAi) embryos to elongate until the 2-fold stage (Fig. 5H,I), indicating that a constitutively active form of myosin II can bypass the requirement for MRCK-1 kinase during elongation.

We observed a normal GFP::MLC-4WT pattern in let-502(sb118ts);mrck-1(RNAi) and let-502(sb118ts);pak-1(ok448) embryos (data not shown; see below for a description of the pak-1(ok448) allele), rather than the fuzzy pattern observed for MLC-4AA. This result might stem from the fact that neither let-502(sb118ts) nor RNAi against mrck-1 are complete knockouts.

To further define the respective contribution of LET-502/ROCK, PAK-1 and MRCK-1 to embryonic elongation, we performed a genetic analysis. Specifically,we created all double and triple mutants for let-502, pak-1, mrck-1and the myosin phosphatase regulatory chain mel-11, to order them relative to each other in a genetic pathway.

To perform this analysis, we used a presumptive mel-11 null allele, mel-11(it26), which causes mutant embryos to rupture when they reach the 1.7-fold stage (Wissmann et al., 1999) (Fig. 6B). First, we used a pak-1 deletion allele, pak-1(ok448), which removes the kinase domain (a second allele, tm403, with a deletion in the CRIB domain confirmed most results). The pak-1(ok448) mutant induced L1 larval lethality at 25.5°C(Fig. 6A, line 3; Fig. 6B), and blocked elongation at the 1.2-fold stage when combined with let-502(sb118ts),as observed after pak-1(RNAi) in the let-502(sb118ts)background (Fig. 6A, line 6; Fig. 6B). Second, we used a presumptive mrck-1 null allele, mrck-1(ok586), which removes the central part of the gene and causes a frameshift(Fig. 7A,B). Homozygous mrck-1(ok586) mutants can grow to adulthood, but subsequently give rise to progeny that arrest at the L1 stage(Fig. 7C,D) or during embryogenesis before elongation (Fig. 6B; Fig. 7E),indicating a maternal effect lethality. The let-502(sb118ts);mrck-1(ok586) double mutant displayed penetrant maternal-effect pre-morphogenesis defects, which we did not characterize further (Fig. 6A, line 7). This indicates that both genes are also redundant during early embryogenesis. Therefore, in further tests involving mrck-1, to circumvent the early requirement for MRCK-1, we used RNAi in adults, which avoids depletion of the maternal pool of mrck-1 mRNA and preferentially affects the elongation stage.

When combined with pak-1(ok448) or mrck-1(RNAi), the mel-11(it26) allele induced embryos to rupture, as observed in mel-11 mutants alone (Fig. 6A, lines 8 and 12), indicating that pak-1 and mrck-1 act upstream of or in parallel to mel-11. Triple let-502;mel-11;pak-1 mutants arrested at the 1.2-fold stage, like let-502;pak-1 double mutants (Fig. 6A, lines 6 and 9). Hence, introducing pak-1 cancels cosuppression by let-502 and mel-11. By contrast, ∼20%of triple let-502;mel-11;mrck-1(RNAi) mutants could reach adulthood,which is comparable to the proportion of let-502;mel-11 double mutants surviving (Fig. 6A,lines 5 and 14). It confirms that MRCK-1 acts upstream of the let-502/mel-11 pathway, and hence does not directly phosphorylate MLC-4. In the quadruple let-502;mel-11;mrck-1(RNAi);pak-1combination, all embryos failed to elongate(Fig. 6A, line 17). Thus, the let-502;mel-11;mrck-1(RNAi) embryos that elongated probably did so because PAK-1 could substitute for LET-502 and MRCK-1 in activating MLC-4.

In order to describe further the role of MRCK-1 during elongation, we developed a C-terminal GFP translational fusion, MRCK-1::GFP(Fig. 7A). The construct was able to fully rescue the mrck-1(ok586) homozygous mutant. We observed that MRCK-1 is enriched in the seam cells, and forms parallel punctate bundles in the DV cells (Fig. 7F) very similar to those of GFP::MLC-4, but seemed to be excluded from junctions. The expression in seam cells persisted until adulthood(Fig. 7G). Furthermore MRCK-1::GFP was strongly expressed in the pharynx throughout development (not shown) and in the sublateral nerve cords(Fig. 7G), presumably at the level of synapses (Fig. 7G,inset). The presence of MRCK-1 in the nervous system might explain the L1 arrest. Altogether, MRCK-1 localization is consistent with the idea that it is involved in myosin II activation.

Non-muscle myosin II is thought to mediate most morphogenetic events involving epithelial cells. However, the mechanisms controlling its assembly and activity in vivo remain unclear. Here we provide an in-depth analysis of both of these aspects during C. elegans embryonic elongation, a process that involves junction remodeling and partial apical constriction (for a review, see Quintin et al.,2008).

Part of our work has relied on the introduction of novel GFP probes for live imaging of the actomyosin cytoskeleton. Instead of GFP::actin reporters,which we found to induce embryonic lethality and abnormally thick actin bundles (F.L., unpublished), we relied on the acting-binding domain of the spectraplakin VAB-10 (Bosher et al.,2003), which was expressed specifically in the epidermis. This probe faithfully recapitulates the distribution of actin as visualized by phalloidin staining, and has already been successfully used by other labs(Lockwood et al., 2008; Patel et al., 2008). Other researchers have used the Moesin actin-binding domain for the same purpose(Edwards et al., 1997; Motegi and Sugimoto, 2006), a tool that is becoming very popular in the fly community.

Similarly, we generated a N-terminal GFP::MLC-4/RMLC fusion, which revealed a complex meshwork in the lateral seam cells and thin filaments in the dorsal and ventral epidermal cells. Although these structures had not been observed with a C-terminal GFP fusion (Shelton et al., 1999), we believe they reflect the bona fide position and organization of myosin II, because our novel GFP::MLC-5 construct revealed a similar pattern. Moreover, antibodies against the myosin heavy chain NMY-1 suggest a similar distribution (Piekny et al., 2003). Last, consistent with actin and myosin forming a common higher order complex, we found that the localization of both reporters mainly overlaps when they are combined.

The myosin II complex normally includes a heavy chain and two light chains. Our work strongly suggests that MLC-5 corresponds to the essential myosin light chain. First, sequence analysis shows that MLC-5 is more related to the human and Drosophila essential light chain than to any other EF-hand-containing proteins. Second, mlc-5 knockdown induces cytokinesis defects and strong elongation defects, as reported when MLC-4/RMLC or both heavy chains are depleted (Piekny et al., 2003; Shelton et al.,1999). Third, the colocalization of GFP::MLC-5 with actin microfilaments depends on the presence of MLC-4 and, conversely, formation of the NMY-2-enriched contractile ring during cytokinesis depends on MLC-5.

Two lines of evidence indicate that MLC-4/RMLC phosphorylation favors its assembly into the myosin II complex. First, its distribution became rapidly diffuse in oxygen-deprived embryos, in which the ATP pool necessary for phosphorylation should rapidly dissipate. Second, the distribution of a non-phosphorylatable form of MLC-4 (MLC-4AA) was similarly diffuse in a homozygous mlc-4 mutant. Conversely, a phosphomimetic form (MLC-4DD)retained its subcellular distribution even in the absence of oxygen. These results indicate that MLC-4 phosphorylation influences actin-myosin interactions. In vertebrate smooth muscles, RMLC phosphorylation appears to induce a conformational change in the heavy chain that destabilizes the interaction between its tail and the head-neck region, where the RMLC resides,to facilitate the formation of myosin filaments (for a review, see Ikebe, 2008). Likewise, in Drosophila, RMLC phosphorylation favors its recruitment to the cortex(Royou et al., 2002).

In contrast to most other well-characterized morphogenetic processes(Quintin et al., 2008), such as gastrulation and germband extension in Drosophila or C. elegans ventral enclosure, C. elegans embryonic elongation involves several types of epithelial cells (seam, ventral and dorsal cells). It had been predicted, based on their microtubule distribution and their enrichment for LET-502/ROCK and MLC-4, that seam cells should play a central role in elongation (Piekny et al.,2003; Priess and Hirsh,1986; Shelton et al.,1999; Wissmann et al.,1999). However, this had not been formally established. Using seam-specific or DV-specific promoters to express MLC-4 in an mlc-4mutant, we indeed found that myosin II acts predominantly in the lateral seam cells. Our results are consistent with our previous finding showing that the RhoGAP RGA-2, which behaves as a negative regulator of myosin II, acts predominantly in DV cells (Diogon et al.,2007). Altogether, our data suggest that tension should be released in the DV cells, whereas it must be increased in seam cells, to achieve embryonic elongation. Along these lines, a physical modeling of the elongation process, taking into account the incompressibility of the embryo and the mechanical properties of the cytoskeleton, suggests that lateral cells are sufficient to induce the initial lengthening of the embryo(Ciarletta et al., 2009).

Biochemical analysis in vertebrates suggests that RMLC phosphorylation on serine 19 (S19) triggers myosin II motor activity, and that phosphorylation on threonine 18 (T18) together with S19 further increases its activity(Ikebe et al., 1986). Likewise, phosphorylation of Drosophila RMLC at the position equivalent to S19 is sufficient to induce myosin II activity(Jordan and Karess, 1997; Vereshchagina et al., 2004). Consistent with these data, we show that a phosphomimetic variant of MLC-4/RMLC (MLC-4DD) fully rescues the elongation defect of mlc-4mutants, whereas an unphosphorylatable form (MLC-4AA) fails to rescue the defect. Since the MLC-4T17A variant, which can still be phosphorylated on the adjacent serine (S18), rescued more significantly than the MLC-4S18A variant,we conclude that in C. elegans too, the serine plays a more crucial role in MLC-4/RMLC activation.

The kinase(s) that mediate RMLC phosphorylation during epithelial morphogenesis has not been systematically investigated before, although it has been assumed that ROCK should play a key role. We show that depletion of let-502 with either pak-1 or mrck-1 leads to arrest at the 1.2-fold stage, which we consider as an absence of elongation. This phenotype is not due to dramatic actin disorganization, suggesting that these kinases play a role either in myosin II activation or in another process. We could not directly test whether the loss of these kinases affects MLC-4/RMLC phosphorylation using vertebrate phospho-RMLC antibodies due to technical problems. Although these antibodies have been successfully used in early C. elegans embryos (Jenkins et al., 2006; Lee et al.,2006; Piekny and Mains,2002), we suspect that fixation protocols for immunohistochemistry modify the physiological properties of the embryo and mimic the oxygen-free conditions that affect MLC-4 distribution during elongation. Moreover, in early embryos, MLC-4 forms a local patch (cytokinetic ring, invaginating endoderm precursor, cortex of zygote), whereas we mainly need to see individual filaments during elongation.

To circumvent this technical limitation, we instead used a functional test. We demonstrated that a constitutively active form of myosin II can bypass the requirement for the MRCK-1 kinase in let-502(sb118ts);mrck-1(RNAi)embryos, suggesting that MRCK is a kinase responsible for myosin II activation, probably via an indirect process involving MEL-11 inactivation. Consistent with this idea, MRCK-1::GFP distribution is very similar to that of MLC-4. The same phosphomimetic form of MLC-4/RMLC could not rescue the elongation defects of let-502(sb118ts) and very poorly, yet significantly, rescued those of let-502(sb118ts);pak-1(RNAi) mutants. These observations do not exclude the possibility that LET-502/ROCK and PAK-1 could phosphorylate MLC-4 (see below), as occurs in vertebrates.

Our genetic epistasis data suggest that MRCK-1, but not PAK-1, acts upstream of MEL-11/MYPT in vivo. Indeed, we found that a significant proportion of let-502;mel-11;mrck-1 triply defective embryos could normally elongate, but that removing pak-1 in this background abrogates their elongation. In vertebrates, ROCK, PAK and MRCK can directly phosphorylate RMLC (Amano et al.,1996; Chew et al.,1998; Leung et al.,1998), as well as MYPT, to inhibit myosin phosphatase activity(Kimura et al., 1996; Takizawa et al., 2002; Tan et al., 2001), both of which induce myosin II activation. Integrating these biochemical data, we suggest that two parallel pathways can activate MLC-4/RMLC(Fig. 8): first, the redundant activities of LET-502/ROCK and MRCK phosphorylate, and thereby inactivate,MEL-11/MYPT; second, PAK-1, in parallel to LET-502/ROCK, can directly phosphorylate MLC-4/RMLC. Interestingly, our in vivo analysis is consistent with in vitro kinase assays with carcinoma cell extracts showing that the simultaneous inhibition of ROCK and MRCK dramatically decreases MYPT phosphorylation without affecting RMLC phosphorylation(Wilkinson et al., 2005). Wilkinson and co-workers proposed that ROCK and MRCK redundantly activate myosin II through the inhibition of myosin phosphatase, and that additional kinases phosphorylate RMLC, which the authors suggest are MLCK and ZIPK(Wilkinson et al., 2005). The closest C. elegans MLCK and ZIPK homologs (ZC373.4 and DAPK-1,respectively) do not act in parallel to let-502 (see Table S2 in the supplementary material). Therefore, in addition to LET-502/ROCK, PAK-1 appears to be the best candidate kinase responsible for MLC-4/RMLC phosphorylation in the process of embryonic elongation. Although this is the simplest model, we cannot exclude the possibility that a fourth kinase distinct from PAK-1 phosphorylates MLC-4/RMLC to substitute for LET-502/ROCK and MRCK-1 when they are both absent.

Although we propose that LET-502/ROCK, PAK-1 and MRCK-1 directly and indirectly activate MLC-4/RMLC, their respective contributions are probably different. Indeed, the elongation phenotype of let-502 single and let-502;mrck-1 double mutants suggests that the resulting increase in MEL-11/MYPT activity is not compensated by the kinase activity of PAK-1. It argues that the ability of PAK-1 to phosphorylate MLC-4/RMLC is low compared with the ability of MEL-11/MYPT to dephosphorylate it.

Two genetic observations argue that PAK-1 and LET-502/ROCK activate a second process through phosphorylation, which occurs before or independently of myosin-II-based microfilament shortening(Fig. 8). First, as already recalled, the phosphomimetic form of MLC-4 could not rescue the let-502;pak-1 double mutant. Second, a pak-1 null allele prevented the mutual suppression observed between let-502 and mel-11. What could this other process be that is regulated by LET-502/ROCK, PAK-1 and their targets? The elongation phenotype of let-502;pak-1 is more similar to that observed when actin dynamics is pharmacologically inhibited than when microtubule dynamics is inhibited(Priess and Hirsh, 1986). The process could thus be actin based, even though we did not detect major actin defects, which is probably due to a lack of resolution and/or proper reagents.

The simplest hypothesis is that LET-502 and PAK-1 share a common target,besides MLC-4, that would be important for elongation. In vertebrates, ROCK and PAK phosphorylate common targets, such as intermediate filaments (IFs) or ERM (ezrin, radixin and moesin) proteins, promoting their reorganization(Goto et al., 1998; Goto et al., 2002) or ability to bind actin (Kissil et al.,2002; Matsui et al.,1998), respectively. However, C. elegans IF and ERM homologs are unlikely to be the relevant LET-502 and PAK-1 targets, as mutations affecting epidermal IFs or the unique ERM homolog, ERM-1, have either a less severe elongation phenotype or no elongation phenotype at all(Van Furden et al., 2004; Woo et al., 2004). Alternatively, one can imagine that LET-502 and PAK-1 phosphorylate two distinct targets that have redundant roles during elongation. For instance,human PAK1 can also phosphorylate filamin and a subunit of the Arp2/3 complex(for a review, see Arias-Romero and Chernoff, 2008). Therefore, changes in the actin network at the level of the cortex might facilitate its visco-elastic properties and allow junction remodeling. Alternatively, cortical myosin II might regulate actin network visco-elasticity, as observed for instance during cytokinesis(Girard et al., 2004). Future investigations should attempt to clarify these issues and allow the identification of target X.

To summarize, our study addresses for the first time in vivo the role of the three kinases LET-502/ROCK, PAK-1 and MRCK-1 in myosin II activation in a complex system involving different cell types. We show that these kinases have redundant roles during elongation of the C. elegans embryo, opening perspectives in the understanding of how myosin II regulation can promote complex morphogenetic processes.