Microtubules are involved in anterior-posterior axis formation in C. elegans embryos

In Caenorhabditis elegans embryos, the oocyte has no developmentally important polarity (Goldstein and Hird, 1996). After fertilization, the oocyte and sperm pronuclei are usually at opposite ends of the oblong embryo, and the oocyte pronucleus undergoes two meiotic divisions, extruding two polar bodies before polarity induction and mitosis (Oegema and Hyman, 2006). Posterior polarity is induced via an unknown centrosome- dependent signal brought in by the sperm (Goldstein and Hird, 1996; O'Connell et al., 2000; Sadler and Shakes, 2000; Wallenfang and Seydoux, 2000; Cowan and Hyman, 2004, 2006). This is thought to locally down-regulate actin contractility, leading to anterior movement of the anterior PAR (partitioning defective) polarity proteins PAR-3, PAR-6, and PKC-3 through the anterior contraction/movement of actin and nonmuscle myosin (Munro et al., 2004; Cowan and Hyman, 2007). Anterior PAR protein localization leads to the posterior cortical localization of PAR-2 (Cuenca et al., 2003). Establishment of the anterior and posterior cortical PAR domains is critical for all downstream polarized events (Gonczy and Rose, 2005).

Sperm in which centrosome maturation is defective, such as spd-5 and spd-2 mutants, often fail to induce posterior polarity (O'Connell et al., 2000; Hamill et al., 2002; Cowan and Hyman, 2004). The defect in centrosome maturation in spd-5 and spd-2 mutants causes a delay in the nucleation of microtubules, which led to the proposal that microtubules might be required for inducing posterior polarity (O'Connell et al., 2000; Hamill et al., 2002). Consistent with this idea, Wallenfang and Seydoux (2000) showed using a variety of mutant backgrounds that a persistent meiotic (acentrosomal) metaphase spindle can induce posterior polarity, with the posterior polarity protein PAR-2 being found near the arrested meiotic spindle at the presumptive anterior end rather than near the sperm pronucleus. This posterior polarity induction depends on microtubules, supporting the view that microtubules can provide polarity signaling and that centrosomes are not required (Wallenfang and Seydoux, 2000).

Recent studies have challenged a role for microtubules in polarity induction (Cowan and Hyman, 2004; Sonneville and Gonczy, 2004). After the knockdown of tubulin by RNAi or inhibition with the drug nocodazole, embryos can still establish embryonic polarity, with normal anterior and posterior PAR domains (Cowan and Hyman, 2004; Sonneville and Gonczy, 2004). These results, together with work showing that removal of the centrosome by laser ablation leads to a failure of polarity induction, led to the conclusion that centrosomes induce polarity independently of microtubules (Cowan and Hyman, 2004; Sonneville and Gonczy, 2004). However, this model fails to explain how a meiotic spindle, which lacks centrosomes, can provide a polarity signal. In addition, although greatly reduced, tubulin is still detectable at the centrosome after RNAi of tubulin or nocodazole treatment (Hyman and White, 1987; Cowan and Hyman, 2004; Sonneville and Gonczy, 2004), leaving open the possibility that microtubules could be required for the posterior signal.

In this study, we explore the relationship between polarity signaling and PAR protein levels and carry out a detailed analysis of embryos in which tubulin levels are reduced. Our results strongly support a role for microtubules in polarity establishment.

To identify embryonic polarity regulators, we tested a set of embryonic lethal genes by RNAi for inducing defects in the localization of GFP–PAR-2 in the one-celled embryo (unpublished data). One gene that showed a strong defect was spd-5, as previously reported (Hamill et al., 2002). SPD-5 is a centrosome component required for centrosome maturation and mitotic spindle assembly (Hamill et al., 2002).

We observed that most spd-5(RNAi) embryos showed reversed embryonic polarity, with GFP–PAR-2 near the meiotic polar bodies and opposite the sperm pronucleus (Fig. 1, E–H and Q). Both PAR-3 and the nonmuscle myosin NMY-2 also show a reversed distribution in spd-5(RNAi) embryos, suggesting that a normal process of polarity induction is occurring (Fig. 1 N and Fig. S1). Reversed embryonic polarity has previously been associated with mutants or RNAi knockdowns inducing a persistent or abnormal meiotic spindle (Wallenfang and Seydoux, 2000; Liu et al., 2004; Sonneville and Gonczy, 2004). However, no meiotic defects have been reported for spd-5 mutants (Hamill et al., 2002). We confirmed that meiotic timing (n = 5) and polar body and spindle formation during meiotic divisions (n = 6) are normal in spd-5(RNAi) embryos. To test whether microtubules are responsible for the reversed posterior polarity signal, we used a β-tubulin mutant (Ellis et al., 2004) combined with RNAi to inhibit tubulin function in spd-5(RNAi) embryos (see Materials and methods). In such embryos, anteriorly localized PAR-2 was never detected (Table I, seventh column). In contrast, the total percentage of embryos that showed posterior PAR-2 in spd-5(RNAi) embryos was similar irrespective of the RNAi knockdown of tubulin (Table I, last column). These results show that a normal meiosis produces a microtubule-dependent signal that can induce posterior polarity. Consistent with these data, wild-type embryos sometimes show a transient anterior cap of PAR-2 protein (Boyd et al., 1996). However, the timing of meiotic polarity induction differs from that of centrosome-dependent induction: reversed polarity in spd-5(RNAi) embryos occurs 10 min later than the initiation of normal posterior polarity in wild-type embryos (Fig. S2).

We observed a higher frequency of reversed polarity after RNAi of spd-5 than in spd-5(or213ts) mutants (Fig. 1 Q, second and third rows; Hamill et al., 2002). A possible explanation for this is that spd-5(or213ts) embryos might have more SPD-5 activity because the mutant may not be a null allele. To test this idea, we used RNAi of spd-5 to try to further reduce SPD-5 activity in spd-5(or213ts) embryos. Such embryos display a range of PAR-2 localization defects similar to those of spd-5(or213ts) mutants (Fig. 1 Q). This indicates that the difference in penetrance of the reversal of polarity is not caused by a difference in SPD-5 activity.

Our experiments assaying polarity defects of spd-5(RNAi) embryos were performed in a background harboring GFP–PAR-2 in addition to endogenous PAR-2 and, thus, have increased PAR-2 protein levels (Wallenfang and Seydoux, 2000; Hao et al., 2006). This raised the possibility that the difference in penetrance of reversed polarity could be caused by a difference in the relative levels of anterior and posterior PAR proteins. To investigate this hypothesis, we compared the localization of PAR-2 in spd-5(RNAi) embryos (1) carrying GFP–PAR-2, (2) carrying no PAR transgenes, and (3) in which the level of the anterior PAR protein PAR-6 was increased through a GFP–PAR-6 transgene (Cuenca et al., 2003). Strikingly, we found that the embryonic polarity phenotypes induced by spd-5(RNAi) in the three genotypes strongly differ: PAR-2 is exclusively at the anterior end of 80% of GFP–PAR-2 embryos, of 52% of embryos with wild-type PAR levels, and of 0% of GFP–PAR-6 embryos (Fig. 1 Q). Posterior polarity completely failed to be induced in 97% of these latter embryos, as they showed uniform GFP–PAR-6 and no PAR-2 on the cortex (Fig. 1 Q and not depicted). Because the polarity signaling events in the three genotypes are expected to be equivalent, these results suggest that the ability of a signal to induce polarity depends on the relative levels of the PAR proteins.

The similarity in the reversed polarity induced by a meiotic microtubule-dependent (acentrosomal) signal and a sperm centrosome-dependent signal prompted us to reexamine a role for microtubules in posterior polarity induction. We used RNAi to simultaneously deplete α and β tubulins (hereafter referred to as tubulin(RNAi); see Materials and methods). Extended RNAi of hermaphrodites led to maternal sterility, indicating that embryos completely devoid of tubulin cannot be produced (see Materials and methods). To achieve as strong a depletion as possible, we analyzed the embryos produced in a time window when 50–90% of double-stranded RNA (dsRNA)–injected hermaphrodites were sterile. To assess the effects of tubulin knockdown at different early embryonic stages, we carefully classified embryo age using the state of DNA condensation: stage 1, meiosis; stage 2, no DNA condensation; stage 3, initiation of DNA condensation; stage 4, intermediate DNA condensation; and stage 5, full DNA condensation (see Fig. 2 for staging criteria). We then compared the pattern of tubulin staining in wild-type and tubulin(RNAi) embryos.

Wild-type embryos display a strong cortical network of microtubules at all of these stages (Fig. 2, A–E). At stage 1 (meiosis), microtubules are additionally visible in the meiotic spindle (Fig. 2 A). In stage 2, only the cortical network is observed (Fig. 2 B). In stage 3, most (92%) embryos have apparent centrosomal asters, which are found in 100% of stage 4 and stage 5 embryos (Fig. 2, C–E and K). In tubulin(RNAi) embryos, tubulin immunoreactivity is strongly diminished at all stages (Fig. 2, F–J). During meiotic stages, embryos lack cortical microtubules but show a small concentration of tubulin around the maternal DNA (Fig. 2 F). At stage 2, a weak network of cortical tubulin fibers is visible (Fig. 2 G). In contrast to wild type, only 22% of stage 3 and 34% of stage 4 tubulin(RNAi) embryos show a concentration of tubulin staining at the centrosome, with the remainder having weak cortical microtubules (Fig. 2, H, I, and K). At stage 5, 89% have detectable microtubules at the centrosome. This analysis shows that after strong tubulin knockdown, a small centrosomal microtubule aster eventually forms, but at a later time than in wild-type embryos.

To confirm that tubulin knockdown does not impair the accumulation of other centrosomal proteins, we examined two centrosome markers in tubulin(RNAi) embryos. We found that the centrosomal accumulation of SPD-5 (Hamill et al., 2002) and TAC-1 (Bellanger and Gonczy, 2003; Le Bot et al., 2003; Srayko et al., 2003) are normal in tubulin(RNAi) embryos (n = 13 and n = 8, respectively; Fig. 3 and not depicted).

We then compared polarity induction in wild-type and tubulin(RNAi) embryos. In wild-type embryos, polarity is initiated at stage 3 (initiation of DNA condensation; Fig. 1 K). At this stage, we observed that 69% of such embryos had posterior PAR-2, all of which had visible asters; a further 23% had lower but detectable levels of tubulin at the centrosome but no obvious cortical PAR-2 (Fig. 2, C and K). At stage 4 (intermediate DNA condensation) and stage 5 (full DNA condensation), all wild-type embryos were polarized and had robust asters (Fig. 2, D, E, and K). These results show that similar to a previous study (Cuenca et al., 2003), the initiation of polarity is strongly correlated with growth of the sperm aster.

We found that polarity induction in tubulin(RNAi) embryos was delayed relative to wild type and was only apparent in embryos with centrosomal asters. Whereas the majority of wild-type embryos are polarized at stage 3, only 11% of tubulin(RNAi) embryos are polarized at this stage (Fig. 2, H and K). Similarly, at stage 4, when all wild-type embryos are polarized, only 17% of tubulin(RNAi) embryos are polarized (Fig. 2, I and K). Importantly, although only a minority of stage 3 and stage 4 tubulin(RNAi) embryos have asters (22% and 34%, respectively), all of the polarized tubulin(RNAi) embryos had visible asters (Fig. 2 K). At stage 5, 11% of tubulin(RNAi) embryos still lacked asters, and these were not polarized (Fig. 2 K). Therefore, there is a tight correlation between the time of polarity induction and microtubule aster growth in both wild-type embryos and tubulin(RNAi) embryos. Because centrosomal microtubule growth is delayed in the context of a normal centrosome in tubulin(RNAi) embryos, these results strongly argue that microtubules are required for polarity induction.

Previous work demonstrated that the polarity signal is dependent on the centrosome and that centrosome maturation is critical for the signal (O'Connell et al., 2000; Hamill et al., 2002; Cowan and Hyman, 2004). Mutants that impair centrosome maturation fail to polarize (O'Connell et al., 2000; Hamill et al., 2002). Regulation of centrosome maturation timing by cyclin E (CYE-1) is also important, as RNAi of cye-1 delays maturation and prevents polarity induction (Cowan and Hyman, 2006). Our work supports the view that the delay in microtubule growth caused by centrosome maturation defects in these mutants is responsible for the impairment in polarity induction.

Microtubule involvement in both the centrosome-dependent signal and a meiosis-dependent signal suggests that both processes may use a similar signaling mechanism. However, the timing of these signals appears to be different. We found that the meiotic signal induces PAR polarity ∼13 min after the completion of meiotic divisions and 10 min later than centrosome- dependent posterior polarity (Fig. S2). Therefore, it seems unlikely that it is the meiotic spindle itself that is delivering a signal. The meiotic divisions produce two polar bodies that have associated midbodies, or spindle remnants where they are attached to the embryo. Because these spindle remnants contain microtubules, one possibility is that this is the source of the meiotic polarity signal. The time of meiotic signaling is coincident with rapid mitotic centrosomal microtubule growth, which might induce growth/activity of the spindle remnant ends.

What could be the nature of the microtubule-dependent signal? It is surprising that components have not yet been found given the extensive genetic and RNAi screening that has been conducted for cell polarity genes. A possible reason for this is that polarity induction might involve partially redundant signaling pathways, which are commonly seen in other processes. Good candidates for involvement are microtubule plus end–binding proteins, many of which either regulate microtubule dynamics or influence cortical processes (Akhmanova et al., 2005). Identifying the signaling and receiving factors is the most crucial future task for understanding the mechanism of polarity induction.

The following strains were used, culturing by standard methods (Brenner, 1974): wild-type Bristol N2, JJ1473: zuIs45 [nmy-2–NMY-2–GFP; unc-119(+)] (Munro et al., 2004), zuEx69 [par-6–GFP–PAR-6] (Nance et al., 2003), JH1512: axIs1137 [rol-6(d); pie-1–GFP–PAR-6] (Cuenca et al., 2003), tbb-2(or362ts) (Ellis et al., 2004), KK866 itIs153 [pie-1–PAR-2–GFP] (Wallenfang and Seydoux, 2000), EU856: spd-5(or213) (Hamill et al., 2002), JA1390 itIs153 [pie-1–PAR-2–GFP, unc-119(+)], and ruIs57 [pie-1–β-tubulin–GFP), unc-119(+)] (Wallenfang and Seydoux, 2000; Praitis et al., 2001).

Synthesis of dsRNA and RNAi was performed by injection as described previously (Ahringer, 2006) using RNAi feeding clones from Fraser et al. (2000) and Kamath et al. (2003) as templates for RNAi synthesis (sjj_F56A3.4 for spd-5, sjj_C47B2.3 for tba-2, and sjj_C36E8.5 for tbb-2). RNAi of α- and β-tubulin genes was performed by combining dsRNA with both tba-2 and tbb-2. Because of the high level of sequence identity among tubulin genes, tba-2 dsRNA will additionally target tba-1, and tbb-2 dsRNA will target tba-1. In all tubulin(RNAi) experiments, tba-1 and tbb-2 dsRNAs were coinjected, and experiments were conducted at 25°C. 50–90% of tubulin(RNAi) hermaphrodites are sterile at 22–30 h after injection; at 40 h, all hermaphrodites are sterile even though sperm are still present. spd-5(RNAi) in Fig. 1 was performed at 25°C, and embryos were dissected 30 h after injection. tubulin(RNAi) experiments were performed for 22–30 h at 25°C.

Antibody staining was performed as described previously (Andrews and Ahringer, 2007). Antibodies used were rabbit anti–SPD-5 (Hamill et al., 2002), rat anti–PAR-3 (Dong et al., 2007), rabbit anti–PAR-2 (Dong et al., 2007), rabbit anti–TAC-1 (Le Bot et al., 2003), mouse antitubulin (clone DM1 A1; Sigma-Aldrich), and chicken anti-GFP (Chemicon). Secondary antibodies were purchased from Jackson ImmunoResearch Laboratories. Confocal images were taken on either an LSM 510 Meta microscope (Carl Zeiss MicroImaging, Inc.) or a Radiance instrument (Bio-Rad Laboratories). Wild-type and RNAi experiments were conducted in pairs, and images were taken using the same settings.

For live recordings in Fig. S2, itIs153 [pie-1–PAR-2–GFP] and ruIs57 [pie-1– β-tubulin–GFP] were mounted in egg buffer (118 mM NaCl, 40 mM KCl, 3 mM CaCl₂, 3 mM MgCl₂, and 5 mM Hepes, pH 7.2) on 18 × 18-mm coverslips coated with 0.3% poly-lysine (Sigma-Aldrich), and the coverslips were inverted onto 3% agar pads and sealed with petroleum jelly. Paired images were taken every 10 s using a 63× lens on a fluorescence microscope (Axioplan 2; Carl Zeiss MicroImaging, Inc.) and 3DM software (Improvision).

Fig. S1 shows that spd-5(RNAi) embryos show a reversal of anterior and posterior PAR domains. Fig. S2 shows that reversed PAR-2 polarity in spd-5(RNAi) embryos is delayed relative to wild-type PAR-2 polarity.

We thank B. Bowerman, E. Munro, J. Nance, and G. Seydoux for strains and reagents and D. Rivers, I. Latorre, and D. St. Johnston for comments on the manuscript. We also thank Y. Dong for excellent technical assistance. Some strains used in this work were provided by the Caenorhabditis Genetics Center, which is funded by the National Institutes of Health National Center for Research Resources.

J. Ahringer is supported by a Wellcome Trust Senior Research Fellowship (grant 054523).