|Home | About | Journals | Submit | Contact Us | Français|
A hallmark of polarized cells is the segregation of the PAR polarity regulators into asymmetric domains at the cell cortex1, 2. Antagonistic interactions involving two conserved kinases, atypical protein kinase C (aPKC) and PAR-1, have been implicated in polarity maintenance1, 2, but the mechanisms that initiate the formation of asymmetric PAR domains are not understood. Here, we describe one pathway used by the sperm-donated centrosome to polarize the PAR proteins in Caenorhabditis elegans zygotes. Before polarization, cortical aPKC excludes PAR-1 kinase and its binding partner PAR-2 by phosphorylation. During symmetry breaking, microtubules nucleated by the centrosome locally protect PAR-2 from phosphorylation by aPKC, allowing PAR-2 and PAR-1 to access the cortex nearest the centrosome. Cortical PAR-1 phosphorylates PAR-3, causing the PAR-3/aPKC complex to leave the cortex. Our findings illustrate how microtubules, independent of actin dynamics, stimulate the self-organization of PAR proteins by providing local protection against a global barrier imposed by aPKC.
Newly fertilized C. elegans zygotes have no predetermined anterior/posterior polarity. Before symmetry breaking, the PDZ domain proteins PAR-3 and PAR-6 and the kinase aPKC/PKC-3 (“anterior PARs”) are uniformly at the cell cortex, and maintain the kinase PAR-1 and the RING protein PAR-2 (“posterior PARs”) in the cytoplasm. During symmetry breaking, the sperm centrosome (or microtubule-organizing center, MTOC) contacts the cortex1, 2 eliciting two changes: actomyosin flows directed away from the MTOC3, and recruitment of PAR-2 to the cortex nearest the MTOC (Fig. 1a)4. Actomyosin flows and PAR-2 function in parallel to displace anterior PARs from the cortex, allowing PAR-1 to also load on the posterior cortex. After cortical flows cease, PAR-2 becomes essential to prevent anterior PARs from returning to the posterior cortex (polarity maintenance)3, 5. In this study, we investigate how the MTOC recruits PAR-2 to the cortex, and how PAR-2 in turn displaces anterior PARs.
We first examined PAR-2 dynamics in fixed zygotes depleted of the myosin regulatory light chain MLC-46 (Fig. 1 and Supplementary Fig. S1a). mlc-4(RNAi) zygotes do not develop cortical flows and depend solely on PAR-2 for symmetry breaking4. Before symmetry breaking, PAR-2 was in the cytoplasm and weakly enriched at the MTOC core (Fig. 1b and Supplementary Fig. S1a). During symmetry breaking, PAR-2 appeared on the cortex. In 21 of 28 (endogenous PAR-2) and 25 of 30 (GFP::PAR-2) zygotes fixed at this stage, PAR-2 was unevenly distributed on the cortex, with the highest levels at the microtubule-dense core of the MTOC (Fig. 1c and Supplementary Fig. S1a). The plasma membrane marker mCherry::PHPLC was uniformly distributed at this stage (Supplementary Fig. S1a). After symmetry breaking, PAR-2 distribution on the cortex became more uniform (Fig. 1d) and the PAR-2 domain expanded to reach 32 +/− 5.2% of the embryo’s circumference (Fig. 1f). Live imaging confirmed that the PAR-2 domain correlates with the site of MTOC/cortex contact (Supplementary Fig. S1c). Treatments that interfere with microtubule nucleation, yielded zygotes that formed no cortical GFP::PAR-2 domains, or domains that were significantly smaller (12.1 +/− 5.1%) and appeared later than controls (Fig. 1e,f and Supplementary Fig. S1d). We conclude that, in the absence of cortical flows, PAR-2 loading depends on microtubules and correlates spatially and temporally with MTOC/cortex contact.
Enrichment of PAR-2 on the MTOC core during symmetry breaking suggested that PAR-2 has microtubule-binding activity. We found that recombinant PAR-2 could be pelleted with, but not without, microtubules by high-speed centrifugation (Kd apparent: 1.19 μM, Fig. 2b and Supplementary Fig. S2). Visualization of recombinant GFP::PAR-2 mixed with rhodamine-labeled microtubules confirmed that PAR-2 binds microtubules in vitro (Fig. 2c). Deletion analysis identified three microtubule-binding regions in PAR-2 (Fig. 2a and Supplementary Fig. S3a). A fusion [GFP::PAR-2(1–221)] containing the first microtubule-binding region but lacking the cortical localization domain localized to spindles in vivo (Supplementary Fig. S3a). Full-length GFP::PAR-2 also localized to spindles, but only in zygotes treated with the microtubule-stabilizing drug taxol (Supplementary Fig. S3b). Mutagenesis of basic residues conserved in C. briggsae and C. remanei PAR-2 yielded two mutations (R163A and R183-5A) that significantly reduced microtubule binding in vitro (Fig. 2a–c and Supplementary Figs. S2 and S4). R183-5A also interfered with the localization of GFP::PAR-2(1–221) to spindles (Supplementary Fig. S3a), and with the localization of full-length PAR-2 to taxol-stabilized spindles (Supplementary Fig. S3b), and to the MTOC at symmetry breaking (Supplementary Fig. S1b). We conclude that the first microtubule-binding domain of PAR-2 is necessary and sufficient for interactions with microtubules in vitro and in vivo, and that microtubule binding is required to enrich PAR-2 on the MTOC core during symmetry breaking.
To determine the function of microtubule binding, we expressed GFP::PAR-2(R163A) and GFP::PAR-2(R183-5A) from RNAi-resistant transgenes in mlc-4(RNAi) zygotes depleted of endogenous PAR-2 (see Methods). For positive controls, we used wild-type GFP::PAR-2 and a mutation (K162A) in the first microtubule-binding domain that does not affect microtubule-binding affinity in vitro (Fig. 2b and Supplementary Fig. S2). All fusions were expressed at comparable levels (Supplementary Fig. S5a). Whereas GFP::PAR-2 and GFP::PAR-2(K162A) localized to the posterior cortex, GFP::PAR-2(R163A) and GFP::PAR-2(R183-5A) remained in the cytoplasm in the majority of zygotes (Fig. 3a and Supplementary Table 1). We obtained identical results under three other conditions that eliminate MTOC-induced cortical flows; ect-2(ax751)4, mat-1(ax227)7, and spd-5(RNAi)8 (Fig. 3a and Supplementary Table 1). Localization to the cortex was restored by reducing PKC-3 by RNAi or by eliminating the PKC-3 phosphorylation sites in PAR-2 (Fig. 3a and Supplementary Table 1). GFP::PAR-2 and GFP::PAR-2(R183-5A) exhibited identical cortical dynamics in pkc-3(RNAi) zygotes as revealed by fluorescence-recovery after photobleaching (FRAP)(Fig. 3b and Supplementary Fig. S5b). GFP::PAR-2(R163A) and GFP::PAR-2(R183-5A) localized to the posterior cortex in mlc-4(+) embryos (where PKC-3 is mobilized by flows) (Fig. 3a) and could rescue the embryonic lethality of par-2(RNAi) and/or par-2(lw32) zygotes to the same extent as wild-type GFP::PAR-2 (Supplementary Table 2). We conclude that microtubule binding is essential for symmetry breaking but not for polarity maintenance, or for PAR-2 to associate with the cortex in the absence of PKC-3.
How does microtubule binding facilitate PAR-2 cortical loading at symmetry breaking? The microtubule-binding regions of PAR-2 contain several PKC-3 phosphorylation sites (Fig. 2a), raising the possibility that microtubule binding protects PAR-2 from phosphorylation by PKC-3. Consistent with this possibility, addition of microtubules inhibited the phosphorylation of PAR-2 by human aPKC in vitro (Fig. 2d and Supplementary Fig. S6a). Inhibition was not observed in the presence of nocodazole (Supplementary Fig. S6b), or when the PAR-2 microtubule-binding mutants (R163A and R183-5A) were used as substrates (Fig. 2d and Supplementary Fig. S6a). Consistent with microtubules acting as competitive inhibitors, 0.8 μM polymerized tubulin was sufficient to increase the Km by 65% without affecting the Vmax of the aPKC kinase reaction (Supplementary Fig. S6c). The average intracellular tubulin concentration has been estimated at ~20 μM9 and should be even higher at the MTOC core, consistent with the possibility that microtubules protect PAR-2 from PKC-3 at symmetry breaking.
To test this hypothesis further, we developed an in vitro microtubule/PKC-3 competition assay in the presence of a “cortex mimic”. Interactions with plasma membrane phospholipids have been implicated in the localization of PAR-1 and PAR-3 homologs to the cortex10, 11. Using a protein-lipid binding assay, we found that PAR-2 interacts with phospholipids including phosphoinositides (Fig. 2e,f and Supplementary Fig. S7). Phosphorylation by aPKC, or phosphomimetic mutations in the PKC-3 sites, interfered with PAR-2 binding to lipids (Fig. 2e,f and Supplementary Fig. S7), as they interfere with PAR-2 cortical localization in vivo12. Remarkably, preincubation with 1.5 μM polymerized tubulin rescued PAR-2’s ability to bind to lipids in the presence of aPKC (Fig. 2f and Supplementary Fig. S7c). Microtubules did not restore lipid binding to PAR-2(R183-5A) (Fig. 2f and Supplementary Fig. S7c), even though this mutant could bind lipids as efficiently as wild-type PAR-2 in the absence of aPKC (Fig. 2e and Supplementary Fig. S7a,b). We conclude that binding to microtubules is sufficient to protect PAR-2 from aPKC/PKC-3 and retain binding to plasma membrane lipids.
After reaching the cortex, PAR-2 becomes partially resistant to exclusion by PKC-3, and this resistance depends on the PAR-2 RING domain12. FRAP analyses revealed faster cortical dynamics for the RING mutant GFP::PAR-2(C56S) compared to GFP::PAR-2 and GFP::PAR-2(R183-5A) (Fig. 3b and Supplementary Fig. S5b). GFP::PAR-2(C56S) was enriched on the MTOC at the time of MTOC/cortex contact in most mlc-4(RNAi) zygotes (14 of 25), but did not form a posterior cortical domain (Fig. 3c and Supplementary Fig. S1b). Endogenous PAR-2 could rescue the cortical localization of both GFP::PAR-2(C56S) and GFP::PAR-2(R183-5A) in mlc-4(RNAi) zygotes (Fig. 3c). These results indicate that cortical PAR-2 is stabilized at the cortex by its RING domain, and recruits additional PAR-2 molecules from the cytoplasm independent of microtubule binding.
By pronuclear meeting, PAR-3 and PKC-3 were excluded from the PAR-2 domain (Fig. 4a and Supplementary Table 3). This exclusion was dependent on PAR-1 (Fig. 4a and Supplementary Table 3). PAR-1 co-localized with PAR-2 on the posterior cortex in mlc-4(RNAi) zygotes expressing wild-type GFP::PAR-2, but not in zygotes expressing GFP::PAR-2(R183-5A), where PAR-2 does not load and PAR-3 and PKC-3 are not excluded (Fig. 4a and Supplementary Table 3). In Drosophila oocytes, PAR-1 phosphorylates PAR-3, causing PAR-3 to lose its cortical association13. In vitro kinase assays confirmed that C. elegans PAR-1 can phosphorylate PAR-3 (Supplementary Fig. S8a,b). Furthermore, we found that PAR-3 and PKC-3 were not excluded in zygotes where PAR-1 lacked kinase activity14 or its cortical localization domain15, or in zygotes expressing a PAR-3 fusion missing the PAR-1 phosphorylation sites16 (Fig. 4a and Supplementary Table 3). We conclude that recruitment of PAR-1 to the PAR-2 domain leads to exclusion of the PAR-3/PKC-3 complex, likely by direct phosphorylation of PAR-3 by PAR-1.
How does PAR-2 recruit PAR-1? In mammalian cells, PAR-1 cortical localization depends on a C-terminal domain that contains a conserved aPKC phosphorylation site required for cortical exclusion by aPKC17. We confirmed that the corresponding domain of C. elegans PAR-1 (965–1192 aa) is necessary and sufficient to target PAR-1 to the cortex (Supplementary Fig. S8c), and that the conserved aPKC site T983 can be phosphorylated by aPKC in vitro (Supplementary Fig. S8d) and is required to exclude PAR-1 from PKC-3(+) cortices in vivo (Supplementary Fig. S8c). Remarkably, we found that GFP::PAR-1(965–1193 aa) and PAR-1(it51), which cannot exclude PAR-3 and PKC-3, were still able to localize with PAR-2 on the posterior cortex in mlc-4(RNAi) zygotes (Fig. 4a and Supplementary Table 3), indicating that PAR-2 can recruit PAR-1 to cortices also occupied by PKC-3. To determine if PAR-1 and PAR-2 interact, we first immunoprecipitated GFP::PAR-1 and GFP::PAR-2 from worm extracts. We detected endogenous PAR-2 in GFP::PAR-1 immunoprecipitates and endogenous PAR-1 in GFP::PAR-2 immunoprecipitates, indicating that at least a subset of PAR-2 and PAR-1 molecules are in a complex (Fig. 4b). Using purified recombinant proteins, we found that PAR-1 and PAR-2 interact directly, and that the PAR-1 C-terminus is sufficient for the interaction (Supplementary Fig. S8e). We conclude that PAR-2 recruits PAR-1 to the cortex, via a direct interaction involving the PAR-1 C-terminal domain.
In wild-type embryos, depletion of tubulin delays symmetry breaking18, raising the possibility that microtubule-dependent loading of PAR-2 contributes to symmetry breaking even in the presence of flows. Consistent with this possibility, at symmetry breaking, GFP::PAR-2(R183-5A) zygotes showed either uniform PAR-3 and no PAR-2 at the cortex (2 of 12), or asymmetric PAR-3 and no (5 of 12) or low PAR-2 (5 of 12). In contrast, 9 of 10 zygotes expressing wild-type GFP::PAR-2 already had complementary PAR-2/PAR-3 domains at this stage (Fig. 5a). Live imaging experiments revealed that GFP::PAR-2(R183-5A) loads on the posterior cortex 29.0 +/−11.2 seconds later than GFP::PAR-2 (Fig. 5b). After this initial delay, GFP::PAR-2(R183-5A) cortical levels increased rapidly and were indistinguishable from GFP::PAR-2 levels by mitosis (Fig. 5a,b), and all zygotes fixed at this stage excluded PAR-3 from the GFP::PAR-2 domain (Fig. 5a). We conclude that, in wild-type embryos, microtubule binding by PAR-2 contributes to the fast kinetics of PAR-2 loading/PAR-3 clearing, but is not essential after cortical flows displace anterior PARs.
Our observations suggest a simple model for polarization of the C. elegans zygote (Fig. 5c). When the MTOC contacts the cortex, the high density of microtubules transiently protects PAR-2 from phosphorylation by PKC-3, allowing a few molecules of unphosphorylated PAR-2 to interact productively with the cortex. Cortical PAR-2, stabilized by its RING domain, recruits PAR-1 as well as additional PAR-2 molecules (“PAR-2 feedback loop”), allowing the PAR-2/PAR-1 domain to expand beyond the site of MTOC/cortex contact. PAR-1 phosphorylates PAR-3, causing the PAR-3/PKC-3 complex to leave the cortex. The anterior PARs are also displaced by cortical flows triggered by the MTOC. Both symmetry-breaking functions of the MTOC (induction of cortical flows and protection of PAR-2 from PKC-3) are transient and depend on the PAR-2 feedback loop and PAR-1 for PAR domain maintenance (also see12).
This model clarifies several observations in the literature. First, although some studies support a role for microtubules in symmetry breaking7, 18, others have suggested that microtubules are not required19, 20. Our findings demonstrate a role for microtubules to load PAR-2 at the earliest stage of symmetry breaking, but leave open the possibility that the MTOC uses a second, microtubule-independent cue to initiate cortical flows21. Our model also explains why PAR-2 is not essential to exclude anterior PARs in par-6/+ zygotes22, 23 or zygotes that overexpress LGL-1, which, like PAR-1, antagonizes the cortical localization of anterior PARs24, 25. We suggest that the primary function of the PAR-2 feedback loop is to maintain sufficient PAR-1 on the posterior cortex to ensure permanent exclusion of anterior PARs. This function may not be needed in embryos where cortical levels of anterior PARs are already biased by flows and further reduced by mutation or LGL-1 overexpression. A remaining question is what prevents the PAR-2 domain from spreading to the entire cortex. Anterior and posterior PARs exchange with the cytoplasm and diffuse freely across the PAR boundary26. One possibility, therefore, is that as the anterior PARs become restricted to a smaller region of the cortex, the concentration of PKC-3 at the boundary reaches a threshold sufficient to block further PAR-2 spreading.
The PAR system has been implicated in the polarization of several cell types, including some that do not undergo cortical flows27. Our findings illustrate how the self-organizing properties of the PAR network are sufficient to polarize a cell in the absence of long-range actin dynamics. In principle, any localized cue that favors the binding of one class of PARs with the cortex will be sufficient to initiate a cascade of self-organizing interactions within the network. We suggest that cortical flows, while non-essential, contribute to the polarization process by increasing the robustness of the response. Cortical flows may also serve to align PAR asymmetry with the cell’s intrinsic geometry, as PAR domains are often miss-aligned with respect to the long axis of mlc-4(RNAi) zygotes (Fig. 4a and reference4).
Microtubules have been proposed to polarize cells by transporting polarity regulators to specific regions of the cell (reviewed in reference28). Our findings identify another way microtubules break symmetry: by protecting polarity regulators from cortical exclusion by aPKC.
This study was supported by the Japan Society for the Promotion of Science ([F.M.]), the American Cancer Society (PF-08-158-01 [E.G.]), and the National Institute of Health (R01HD37047 [G.S.]). G.S. is an investigator of the Howard Hughes Medical Institute. We thank J. Ahringer, A. Audhya, L. Boyd, A. Desai, P. Gonczy, M. Gotta, R. Green, C. Hoege, K. Kemphues, Y. Nishimura, K. F. O’Connell, K. Oegema, L. S. Rose, A. Sugimoto, C. M. Waterman, H. Zaher, and the Caenorhabditis Genetic Center for reagents and expertise.