Polarity establishment, asymmetric division and segregation of fate determinants in early C. elegans embryos^*

The AP embryonic axis is put in place during the first cell cycle and this process can be decomposed into three principal steps: symmetry breaking, establishment of AP polarity, and maintenance of this polarity (Figure 2). These three steps are discussed in turn below and can be viewed with high temporal and spatial resolution using simple time-lapse Differential Interference Contrast (DIC) microscopy (Movie 1).

Fertilization occurs in the spermatheca, with the union of a mature oocyte and a sperm cell. The newly fertilized zygote is then displaced into the uterus where meiosis is completed and early embryogenesis takes place. Starting from the time of anaphase of meiosis II, the entire actomyosin network located in the cell cortex that underlies the plasma membrane undergoes contractions that depend on the non-muscle myosin II heavy chain NMY-2 and the associated non-muscle myosin light chain 4 MLC-4 (Munro et al., 2004; Shelton et al., 1999) (Figure 2A). Moreover, such surface contractions require the Rho family GTPase RHO-1 and the Rho-binding kinase LET-502, which together promote NMY-2 activity and contraction of the cortical actomyosin network (Kumfer et al., 2010; Motegi and Sugimoto, 2006; Schonegg and Hyman, 2006).

After symmetry breaking, the non-contractile cortical area expands towards the anterior of the embryo until reaching ∼50% egg-length (Hird and White, 1993) (Figure 2B). This anterior movement entails a flow of cortical material away from centrosomes, which is compensated by posterior-directed streaming of cytoplasmic material (Hird and White, 1993). Quantitative analysis and computer simulations confirm that the hydrodynamic properties of the cytoplasm can explain such compensatory streaming (Niwayama et al., 2011).

What are the properties of the actomyosin network during the anteriorly-directed flow of cortical material? An answer comes from experiments in which the actomyosin network was severed locally with a laser microbeam and the resulting surface movements perpendicular to the cut monitored as a measure of the underlying cortical tension (Mayer et al., 2010). Performing such severing along the longitudinal axis demonstrated that cortical tension orthogonal to the AP axis is twice as high in the anterior, contractile, region than in the posterior, non-contractile, region. By contrast, performing such severing along the shorter axis established that cortical tension along the AP axis is uniform throughout the embryo. Thus, there is an anisotropy in cortical tension in the anterior, being higher orthogonal to the AP axis. Such cortical anisotropy is lacking in embryos that do not have anteriorly-directed cortical flows (for instance, following defective centrosome assembly), compatible with cortical anisotropy being a consequence of such flows (Mayer et al., 2010). A physical model of these phenomena that treats the cortex as a thin film of active viscous fluid demonstrates that local contractions of the actomyosin network can result in long-range movements of the cell cortex towards the embryo anterior (Mayer et al., 2010). Thus, anterior-restricted anisotropy in cortical contractility can be instrumental in polarity establishment.

The partitioning of the cortex into a retracting anterior contractile domain and an expanding posterior non-contractile domain is accompanied by the asymmetric distribution of PAR proteins at the cell cortex. The six par genes (for partitioning defective) were identified amongst maternal-effect mutations as having defective AP polarity, resulting in the generation of daughter cells with altered fate, size, spindle positioning and cell cycle progression (reviewed in Goldstein and Macara, 2007; Asymmetric cell division and axis formation in the embryo; Rose and Kemphues, 1998a; Kemphues and Strome, 1997). PAR-3 and PAR-6 are PDZ-containing proteins that, together with the atypical protein kinase C PKC-3, form the so-called “anterior PAR complex”. Both PAR-1 and PAR-4 are Ser/Thr kinases, whereas PAR-2 is a RING-finger protein and PAR-5 a 14-3-3 protein. The anterior PAR complex marked by GFP::PAR-6 is present throughout the cortex initially, but recedes from the posterior following symmetry breaking, concomitant with the anteriorly-directed movement of the contractile cortex (Munro et al., 2004; Cuenca et al., 2003). Conversely, GFP::PAR-2 is first detected on the posterior cortex shortly after the local cessation of cortical contractions and then expands anteriorly to approximately 50% egg-length (Movie 2) (Bienkowska and Cowan, 2012; Cuenca et al., 2003). PAR-1 is also enriched on the posterior cortex (Figure 2), while PAR-4 and PAR-5 are present uniformly in both cytoplasm and cortex.

Although anteriorly-directed movement of the contractile actomyosin network marked by NMY-2::GFP foci occurs concomitantly with expansion of the posterior domain marked by GFP::PAR-2, the dynamics of these two cortical components differs (Petrasek et al., 2008). Indeed, beam-scanning fluorescence correlation spectroscopy (FCS) experiments revealed that the motility of GFP::PAR-2 is greater than that of NMY-2::GFP (Petrasek et al., 2008). This suggests that PAR-2 distribution is somehow uncoupled from actomyosin dynamics, which is also in line with the finding that PAR-2 can bind phospholipids (Motegi et al., 2011; see below in this section).

The dynamics of anterior and posterior PAR proteins have been investigated using fluorescence recovery after photobleaching (FRAP) (Goehring et al., 2011a; Cheeks et al., 2004). These experiments indicate that exchange between the cytoplasmic and the cortical pool, while occuring (Cheeks et al., 2004), is slow compared to lateral diffusion (Goehring et al., 2011a). In addition, analysis by Total Internal Reflection Fluorescence Microscopy (TIRF) reveals that GFP::PAR-6 resides in puncta that have a limited residency time at the cell cortex (Hyenne et al., 2012). Together, these studies illustrate that PAR proteins constitute a dynamic ensemble, whose kinetics must be taken into consideration when reflecting upon the mechanisms of polarity establishment.

Anteriorly directed movement of the actomyosin network is critical for polarity establishment, as evidenced, for instance, by the fact that movements of cortical GFP::PAR-6 are attenuated in embryos partially depleted of NMY-2 (Munro et al., 2004). Polarity establishment is also altered in other cases in which cortical contractility is compromised, such as in embryos with impaired RHO-1 activity, including those depleted for NOP-1 as mentioned earlier, or of the TAO kinase KIN-18 (Spiga et al., 2013; Tse et al., 2012). Reciprocally, par genes are required for efficient anteriorly-directed cortical flows. Thus, anteriorly-directed movements of NMY-2::GFP are nearly abolished in embryos lacking par-3, par-4, par-5 or par-6 function, as well as in embryos depleted of MEX-5/6, two partially redundant CCCH finger proteins that contribute to polarity establishment (Cheeks et al., 2004; Munro et al., 2004; Cuenca et al., 2003; Schubert et al., 2000). In par-2(RNAi) embryos, by contrast, cortical flows are normal initially (Cheeks et al., 2004; Munro et al., 2004), in line with the fact that anterior movements of GFP::PAR-6 are not altered during the establishment phase (Cuenca et al., 2003).

In summary, polarity establishment entails contraction of the active actomyosin network towards the embryo anterior, accompanied by the segregation of anterior and posterior PAR proteins into two largely mutually exclusive domains.

Recent work has unveiled a role for PAR-2 during the early steps of polarization that is masked by the principal mechanism discussed above relying on local inactivation of RHO-1 (Motegi et al., 2011; Zonies et al., 2010). Embryos carrying a partial loss-of-function allele of ect-2, ect-2(ax751), exhibit inefficient anteriorly-directed cortical flows and delayed PAR-3 clearance from the posterior cortex, but eventually achieve a polarized distribution of PAR-3 (Zonies et al., 2010). Because par-2 is essential for such delayed polarization in ect-2(ax751) embryos, it was proposed that there is a partially redundant pathway dependent on PAR-2 for polarity initiation that is revealed when the actomyosin network is compromised (Zonies et al., 2010).

What are the mechanisms that enable PAR-2 to exert this function? Live imaging revealed that PAR-2 associates with centrosomes, whereas biochemical experiments established that PAR-2 is a microtubule binding protein (Motegi et al., 2011). Interestingly, microtubule association protects PAR-2 from phosphorylation by PKC-3 in vitro. Thus, in vivo, association of PAR-2 with microtubules emanating from the centrosomes at the posterior would prevent the action of PKC-3, which would otherwise inhibit PAR-2 cortical localization (Motegi et al., 2011). In search of a mechanism underlying this protective effect, it was discovered that PAR-2 binds phospholipids in vitro. This binding is curbed upon phosphorylation of PAR-2 by PKC-3, but this negative regulation does not operate when microtubules are present in addition. Furthermore, membrane-bound PAR-2 recruits more PAR-2, creating a positive feedback loop that promotes expansion of the posterior domain. Whilst important when the principal pathway of polarity establishment is lacking, this PAR-2-dependent pathway is not essential under normal circumstances. Thus, although embryos expressing only a mutant version of PAR-2 that cannot bind microtubules initiate polarity with slower kinetics, they manage to establish it correctly thereafter (Motegi et al., 2011), consistent with the notion that microtubules contribute to but are not essential for polarity establishment (Tsai and Ahringer, 2007; Cowan and Hyman, 2004b; Sonneville and Gönczy, 2004;).

In summary, initiation of polarity establishment in the vicinity of centrosomes is aided by a PAR-2-dependent pathway that relies on the ability of this protein to bind microtubules and thus be shielded from PKC-3-mediated phosphorylation.

The phase of polarity establishment is followed by a maintenance phase during which centrosome ablation no longer interferes with AP polarity (Cowan and Hyman, 2004b), although microtubules have been suggested to contribute to refining polarity during this phase (Ai et al., 2011). During polarity maintenance, mutual inhibition between the anterior and posterior cortical domains is critical and par-2 is essential to maintain the anterior PAR complex restricted to the embryo anterior. Thus, in par-2 mutants, GFP::PAR-6, which had been restricted to the anterior during the establishment phase, returns to the posterior to occupy the entire cortical circumference (Cuenca et al., 2003), concomitantly with abnormal posterior-directed cortical movements of NMY-2::GFP (Munro et al., 2004). Although it has been proposed that these NMY-2 movements contribute to preventing the maintenance of PAR asymmetries (Munro et al., 2004), more recent evidence indicates that cortical NMY-2 plays a minor role in this phase (Beatty et al., 2013).

FRAP experiments indicate that GFP::PAR-6 and GFP::PAR-2 can diffuse slightly across the boundary during this maintenance phase, highlighting the absence of a strict physical separation between the two domains (Goehring et al., 2011a). What, then, are the mechanisms preventing the spread of anterior PAR proteins to the posterior and reciprocally? Phosphorylation of PAR-2 by PKC-3 is again part of the answer (Hao et al., 2006). Structure/function analysis led to the identification of a PAR-2 fragment that is sufficient for localization to the posterior cortex and which contains 7 predicted PKC-3 phosphorylation sites. Atypical protein kinase C can phosphorylate this PAR-2 fragment in vitro, and mutating these 7 serine/threonine residues to non-phosphorylatable alanine leads to GFP::PAR-2 being present throughout the cortex. Conversely, altering these residues to phosphomimetic glutamic acid prevents GFP::PAR-2 cortical localization (Hao et al., 2006). The RING finger of PAR-2 is critical to counteract the impact of PKC-3, perhaps by making PAR-2 a poorer substrate for the kinase (Hao et al., 2006). In turn, the presence of PAR-2 at the cell cortex promotes the recruitment of PAR-1 through a direct interaction between the two proteins (Motegi et al., 2011), although PAR-1 can localize to the cortex in a PAR-2-independent manner when PAR-3 is lacking (reviewed in Kemphues and Strome, 1997). In addition, PAR-1 can phosphorylate PAR-3 in vitro, and PAR-1 kinase activity is required for the exclusion of the anterior PAR complex from the posterior cortex in embryos that rely solely on the PAR-2-dependent pathway described above (Motegi et al., 2011).

Apart from the actomyosin network and the PAR proteins, other components participate in generating AP polarity in the one-cell embryo. This includes LGL-1, the worm homologue of the Drosophila tumor-suppressor protein Lethal Giant Larvae, which functions redundantly with PAR-2 during polarity maintenance (Beatty et al., 2010; Hoege et al., 2010). LGL-1 is enriched on the posterior cortex, and its depletion, while not resulting in a phenotype on its own, affects the distribution of cortical NMY-2 and enhances the lethality incurred from partial PAR-2 inactivation. Conversely, increasing LGL-1 levels alleviates aspects of the par-2(RNAi) phenotype. Therefore, LGL-1 can be thought of as defining a second pathway that contributes to polarity establishment in the absence of PAR-2. What restricts LGL-1 cortical distribution to the posterior? Phosphorylation by PKC-3 is key here as well. Mutating three PKC-3 predicted phosphorylation sites to non-phosphorylatable residues leads to uniform cortical distribution of GFP::LGL-1, whereas the converse phosphomimetic variant does not localize (Hoege et al., 2010) or localizes less well (Beatty et al., 2010) to the cell cortex. These and other observations suggest that a PAR-6/PAR-3/LGL-1 complex forms initially, but that PKC-3-mediated phosphorylation of LGL-1 prevents its localization on the anterior side during the course of polarity establishment (Beatty et al., 2010; Hoege et al., 2010).

Two other components important for the maintenance of AP polarity are CDC-37, an Hsp90 co-chaperone that in other systems stabilizes kinases, and the small G protein CDC-42. In C. elegans, CDC-37 appears to function primarily by stabilizing PKC-3, such that cdc-37(RNAi) embryos exhibit defective mutual exclusion of the anterior and posterior cortical domains (Beers and Kemphues, 2006). However, slight differences between the cdc-37(RNAi) and the pkc-3(RNAi) phenotypes are suggestive of additional roles for CDC-37 (Beers and Kemphues, 2006). C. elegans CDC-42, like its homologue in other organisms, physically interacts with PAR-6 (Gotta et al., 2001), and the two components are mutually dependent for their cortical localization (Aceto et al., 2006). Furthermore, embryos expressing a mutant version that cannot interact with CDC-42 as sole source of PAR-6 exhibit polarity defects mirroring those of partial CDC-42 depletion (Aceto et al., 2006). Where does active CDC-42 GTP localize in one-cell embryos? A transgenic line expressing GFP fused to a binding domain of the Wiskott-Aldrich syndrome protein WSP-1 that specifically interacts with the GTP-bound form of CDC-42 (GFP::GBDWSP-1) has been generated to address this question (Kumfer et al., 2010). This reporter revealed that cortical CDC-42 GTP is enriched initially on the posterior, then on the anterior, before becoming undetectable. A targeted RNAi screen amongst potential GEFs and GAPs using this reporter led to the identification of the GEF CGEF-1 and the GAP CHIN-1 as regulators of CDC-42 activity (Kumfer et al., 2010). CGEF-1 depletion results in lower cortical CDC-42 GTP throughout the cell cycle, whereas CHIN-1 depletion affects CDC-42 activity specifically when GFP::GBDWSP-1 is normally present on the anterior cortex. Compatible with antagonistic influences on cortical CDC-42, CGEF-1 depletion leads to reduction of the domain occupied by cortical PAR-6, whereas CHIN-1 depletion leads to an increase in this domain (Beatty et al., 2013). Embryos lacking both PAR-2 and LGL-1 function still exhibit polarized CDC-42 GTP distribution in a CGEF-1- and CHIN-1-dependent manner. Together with other supporting evidence, these results suggests that CHIN-1 and CGEF-1 are part of a third pathway that acts in parallel to the principal PAR-2-dependent pathway and the redundant LGL-1-dependent pathway for polarity maintenance (Beatty et al., 2013). Moreover, the MRCK-1 kinase is needed for CDC-42 GTP cortical enrichment when it is normally present on the anterior (Kumfer et al., 2010). Although cdc-42(RNAi) embryos have been reported to have defects primarily in the maintenance phase of polarization (Gotta et al., 2001; Kay and Hunter, 2001), upon more severe RNAi-mediated depletion, CDC-42 becomes critical for cortical PAR-6 distribution also during polarity establishment (Schonegg and Hyman, 2006). This earlier role appears to be mediated through removal of cortical PAR-2, since PAR-6 remains present at the cell cortex in embryos doubly depleted of CDC-42 and PAR-2 (Schonegg and Hyman, 2006). Overall, these observations indicate that binding of CDC-42 to PAR-6 plays a critical role in setting up and maintaining AP polarity.

Additional components, including ones that are important for basic cellular processes, are harnessed to ensure polarity maintenance. For instance, the dynamin GTPase, a universal modulator of endocytosis, membrane trafficking, and actin dynamics, participates in polarity maintenance in one-cell C. elegans embryos (Nakayama et al., 2009). Embryos partially depleted of DYN-1 exhibit less cortical GFP::PAR-6 on the anterior side during the maintenance phase and, conversely, an expansion of the GFP::PAR-2 domain. Moreover, FRAP analysis reveals that dynamin promotes association of GFP::PAR-6 with the cell cortex during this phase. Compatible with this view, embryos expressing DYN-1::GFP and mCherry::PAR-6 exhibit green and red puncta that are often juxtaposed, raising the possibility that PAR-6 associates in a dynamin-dependent manner with endocytic vesicles (Hyenne et al., 2012; Nakayama et al., 2009). Interestingly, DYN-1::GFP is present initially throughout the cortex as small punctae, but becomes restricted to the anterior cortex during polarity establishment, in a manner analogous to GFP::PAR-6, and this in a PAR-6- and PAR-3-dependent manner (Nakayama et al., 2009). Moreover, the distribution of DYN-1::GFP is altered in embryos depleted of RACK-1, which was identified for its role in cytokinesis, but whose depletion results in polarity phenotypes reminiscent of those observed in dyn-1(RNAi) embryos (Ai et al., 2011). Whereas dynamin contributes to polarity maintenance, conversely, PAR-6, PAR-3, PKC-3, and CDC-42 regulate endocytosis (Balklava et al., 2007), indicative of a mutual interplay between the anterior PAR complex and the trafficking machinery.

Just like dynamin, the early endosomal protein RAB-5 is enriched on the anterior side during the maintenance phase, in a manner that depends on AP polarity cues (Andrews and Ahringer, 2007). RAB-5 depletion results in a slight alteration of the GFP::PAR-6 domain, albeit without an accompanying alteration in the GFP::PAR-2 domain (Hyenne et al., 2012). However, pleiotropic phenotypes often complicate analyzing the consequences of depleting components required for basic cellular processes. For instance, RAB-5 depletion also alters the actin cytoskeleton (Hyenne et al., 2012; Audhya et al., 2007), and it may be that the impact on GFP::PAR-6 is secondary to alterations in actin distribution or dynamics. Overall, whether trafficking events are instructive for directing AP polarity or merely elements that contribute in a more passive manner through their general requirement for cell physiology remains to be clarified.

In summary, the maintenance of AP polarity relies principally on the mutual inhibition of anterior and posterior cortical domains, which, together with associated components, ensures the robust maintenance of polarity along the AP embryonic axis.

The wealth of information regarding the relationships between anterior and posterior polarity components, as well as knowledge about their dynamics and that of the underlying actomyosin network, have fueled mathematical modeling of how AP polarity is achieved in one-cell C. elegans embryos. An initial model captured the essence of the behavior of the actomyosin network, as well as of the PAR proteins and of their relationships (Tostevin and Howard, 2008). This model consists of reaction-diffusion equations that rest on the assumption that PAR protein dynamics at the cell cortex result from their diffusion within the cortical plane and from exchange with the cytoplasmic pool. A further assumption is that association of anterior PAR proteins with the contractile actomyosin cortex increases over time. In the output from this model, anterior PAR proteins exclude posterior PAR proteins (and reciprocally), generating effective positive feedback loops that lead to mutually exclusive cortical distributions. The molecular tenets of such mutual exclusion remain open in the model, but could correspond for instance to phosphorylation of PAR-2 by PKC-3 in the case of inhibition of the posterior complex by the anterior one. This initial model left open several questions, including how a postulated asymmetry in the distribution of microtubules that is essential in the model to maintain polarity relates to the situation in vivo. Further modeling indicates that mutual inhibition can maintain distinct domains even without a contribution of the actomyosin network, provided oligomerization of the anterior PAR proteins is included (Dawes and Munro, 2011). Interestingly, PAR-3 proteins are known to dimerize in other systems and have thus been postulated to correspond to the molecular basis for such oligomerization. This revised model features a non-linear relationship in mutual cortical exclusion, which predicts that polarity should be lost abruptly below a threshold level of anterior PAR proteins, which is indeed observed experimentally (Dawes and Munro, 2011). A related but more elaborate model incorporating quantitative analysis of parameters that characterize the behavior of GFP::PAR-6 and GFP::PAR-2 has been developed (Goehring et al., 2011b). In this model, the cell cortex is considered as a thin film that transports embedded molecules towards the anterior by advection, much like a river would carry leaves floating on the water surface. Compatible with this model, simulations indicate that advection by such fluid flow can indeed displace anterior PAR proteins transiently and thus be critical for polarization of one-cell C. elegans embryos (Movie 3).

In summary, mutual interactions between anterior and posterior PAR proteins have lent themselves to mathematical modeling, and future work in this direction is anticipated to unveil further tenets of the underlying mechanisms of AP polarity.

Once properly established and maintained, AP polarity is translated into several processes that contribute to correct asymmetric division of the one-cell embryo, including centrosome and spindle positioning (Figure 3). These processes are crucial for correct asymmetric division because, in animal cells, the cleavage furrow is specified so as to bisect the position of the mitotic spindle at the end of anaphase, thus dictating the relative size and the spatial relationship of daughter cells. Anaphase spindle position is determined by the location of the two centrosomes prior to mitosis, as well as that of the two spindle poles derived from the centrosomes during mitosis. In the one-cell C. elegans embryo, both centrosome and spindle positioning are regulated by AP polarity cues, ensuring coordination with the segregation of cell fate determinants (Figure 3) (reviewed in McNally, 2013; Morin and Bellaiche, 2011; Cowan and Hyman, 2007; Galli and van den Heuvel, 2008; Asymmetric cell division and axis formation in the embryo; Betschinger and Knoblich, 2004; Cowan and Hyman, 2004a).

Centrosome and spindle positioning in one-cell C. elegans embryos can be readily observed by simple differential interference contrast (DIC) time-lapse microscopy (Movie 1). Centrosomes and the associated male pronucleus are located initially close to the posterior cortex, and then migrate jointly towards the anterior. Concomitantly, the female pronucleus migrates towards the posterior, resulting in the meeting of the two pronuclei in the posterior half of the embryo. Migration of both male and female pronuclei depends on microtubules and microtubule associated proteins, including the minus-end directed motor dynein and members of the dynein complex (reviewed in McNally, 2013; Morin and Bellaiche, 2011; Galli and van den Heuvel, 2008). Dynein present on the surface of the female pronucleus is thought to capture the plus-ends of astral microtubules nucleated by the centrosomes, and thus power migration of the female pronucleus towards the male pronucleus associated with the centrosomes (Gönczy et al., 1999). After pronuclear meeting, the centrosomes and the joined pronuclei move towards the embryo center, while undergoing a 90° rotation. This process is referred to collectively as centration/rotation and results in the centrosomes becoming aligned along the AP axis in preparation for cell division. Following nuclear envelope breakdown (NEBD), the spindle assembles in the cell center and along the AP axis, but is displaced posteriorly during metaphase and anaphase (reviewed in Asymmetric cell division and axis formation in the embryo). Moreover, the posterior spindle pole exhibits vigorous transverse oscillations during anaphase. As a result of the asymmetric spindle position achieved at the end of mitosis, the first cleavage is unequal and generates the larger anterior blastomere AB and the smaller posterior blastomere P₁.

What is the molecular nature of the force generators that exert pulling forces on centrosomes and spindle poles? The available evidence indicates that the same components are utilized during male pronuclear migration, centration/rotation and spindle positioning, either at the cell cortex and/or in the cytoplasm. Because most experiments aimed at elucidating the nature and the mechanism of action of these components were focused initially on spindle positioning, this will be discussed first.

A ternary complex is critical for generating pulling forces on astral microtubules. This ternary complex is comprised of two partially redundant Gα subunits, GOA-1 and GPA-16, two essentially identical GoLoco containing proteins, GPR-1 and GPR-2, as well as the large coiled-coil protein LIN-5 (Colombo et al., 2003; Gotta et al., 2003; Srinivasan et al., 2003; Tsou et al., 2003a; Gotta and Ahringer, 2001). The data indicates that LIN-5 binds GPR-1/2, which associate through their GoLoco motif with GOA-1/GPA-16 bound to GDP (Kotak et al., 2012; Park and Rose, 2008; Couwenbergs et al., 2007; Nguyen-Ngoc et al., 2007; Srinivasan et al., 2003). As Gα proteins are myristoylated, the entire ternary complex is anchored at the plasma membrane, and perhaps at endomembrane compartments. Spindle severing and spindle pole fragmentation experiments established that the ternary complex, presumably acting at the cell cortex, is required for generating pulling forces during mitosis (Nguyen-Ngoc et al., 2007; Colombo et al., 2003; Grill et al., 2003).

How does the ternary complex result in asymmetric net pulling forces on the two spindle poles? Although the answer to this important question is not entirely clear, the asymmetry in net pulling forces correlates with an asymmetric distribution of some ternary complex components, and this in a PAR-dependent manner. Thus, whereas the cortical distribution of GOA-1 and GPA-16 appears uniform (Park and Rose, 2008; Afshar et al., 2005; Afshar et al., 2004; Gotta and Ahringer, 2001; Miller and Rand, 2000), that of GPR-1/2 is slightly asymmetric during metaphase and thereafter, with more protein present on the posterior side (Colombo et al., 2003; Gotta et al., 2003; Tsou et al., 2003a) (Figure 3). It should be noted also that some mitotic embryos exhibit bipolar asymmetry, in which both anterior and posterior cortices have higher GPR-1/2 levels than the lateral cortex (Park and Rose, 2008). LIN-5 distribution is bipolar in most mitotic embryos, yet all embryos undergo posterior spindle displacement (Park and Rose, 2008), indicating that LIN-5 does not need to be asymmetrically distributed for such displacement to occur (see Section 3.3 below). Importantly, GPR-1/2 asymmetry is under the control of AP polarity cues, indicating that it could be critical for the observed asymmetry in net pulling forces (Colombo et al., 2003; Gotta et al., 2003; Srinivasan et al., 2003; Tsou et al., 2003a; Gotta and Ahringer, 2001). Further support for the levels of GPR-1/2 being key comes from transgenic animals that express YFP-GPR-1 in addition to endogenous GPR-1/2, and which exhibit exaggerated spindle oscillations (Kotak et al., 2012; Redemann et al., 2011). Moreover, during centration/rotation, levels of cortical GPR-1/2 and LIN-5 also correlate with the direction of pulling forces, with more protein present on the anterior cortex (Park and Rose, 2008). This observation supports the notion that cortical force generators play a role also during centration/rotation. A further correlation is observed in embryos depleted of the casein kinase I CSNK-1, which exhibit an increase of cortical GPR-1/2 levels and accompanying exaggerated centration/rotation movements (see also Section 3.4).

In summary, these findings indicate that GPR-1/2 and probably LIN-5 are critical for imparting the proper extent of pulling forces on centrosomes and spindle poles in one-cell C. elegans embryos. The mechanisms that regulate the distribution or function of ternary complex components are described in Section 3.4, after we discuss the mechanical basis of cortical force generators below.

How does the ternary complex generate pulling forces on astral microtubules? Dynamic microtubules are an important component. Indeed, spindle severing experiments revealed that pulling forces are lacking when microtubules are stabilized by taxol in wild-type embryos or compromised in tbb-2(qt1) embryos, which carry a temperature-sensitive mutation in a β-tubulin gene confering resistance to microtubule depolymerization (Nguyen-Ngoc et al., 2007). In addition, the minus-end directed microtubule dependent motor protein dynein is required. Spindle severing experiments established that pulling forces are compromised in embryos carrying a temperature-sensitive reduction-of-function allele of the dynein heavy chain gene DHC-1 (Nguyen-Ngoc et al., 2007; Schmidt et al., 2005). Similarly, pulling forces are compromised upon depletion of the dynein light chain DYRB-1 or the dynein-interacting protein LIS-1, both of which co-immunoprecipitate with GPR-1/2 and LIN-5 (Couwenbergs et al., 2007; Nguyen-Ngoc et al., 2007).

Importantly, ternary complex components are required for the presence of dynein and of associated proteins at the cell cortex (Nguyen-Ngoc et al., 2007). Given that the LIN-5 homologue NuMA can associate with dynein in vertebrate systems (Kotak et al., 2012; Merdes et al., 1996), a plausible working model is that LIN-5 likewise mediates association with the dynein complex, thus anchoring the motor at the cell cortex (Figure 3B). Dynein could function as a bona fide motor protein, generating force by attempting to move towards the minus end of astral microtubules. With dynein anchored at the cell cortex, this would result in a pull on the astral microtubule. Alternatively, dynein could merely serve as a coupling device to retain interaction with the plus-end of depolymerizing microtubules, which would be generating force in this scenario. Intriguingly, levels of cortical YFP-GPR-1 increase locally after mechanical stimulation of the embryo with a glass needle, indicating that GPR-1/2 distribution is modulated by mechanical properties of the cell (Bringmann, 2012). Furthermore, this observation raises the possibility that local increase of GPR-1/2 serves as a positive feedback mechanism to promote dynein anchorage at sites experiencing a cortical pull.

An important step forward came from elegant imaging experiments in which cortical events were visualized with high spatial and temporal resolution (Kozlowski et al., 2007). Microtubule plus-ends in the immediate vicinity of the cortex were imaged using GFP::βtubulin, which marks both polymerizing and depolymerizing microtubules, as well as with GFP::EB2, which marks strictly the plus-end of growing microtubules (Movie 4). These experiments established that the cortical residency time of microtubule plus-ends is ∼1 second (Kozlowski et al., 2007), significantly lower than previously estimated (Labbé et al., 2003). Interestingly in addition, imaging during spindle oscillations revealed that contacts between microtubules and the cortex are more frequent on the lateral side towards which the spindle pole is moving (Kozlowski et al., 2007). Moreover, the posterior spindle pole moves with a velocity that correlates with the maximal polymerization speed of microtubules over a range of temperatures. These observations led to a 3D model of spindle displacement (Kozlowski et al., 2007). In this model, differences in net pulling forces between the anterior and posterior could be due either to a higher attachment rate of microtubules to cortical force generators on the posterior, to a larger fraction of cortical force generators being active on that side or to differential mechanical properties of the actin cytoskeleton, for instance higher rigidity in the anterior. Each force generator in this model is estimated to exert a pull of ∼50 pN. Since this value is similar to the force that can be exerted by a single depolymerizing microtubule, but about 7 times higher than that exerted by a single dynein molecule, it was suggested that microtubule depolymerization is more likely to power force generation, with dynein merely serving as a coupling device (Kozlowski et al., 2007).

Another study postulated that modulation of dynein processivity is central for spindle displacement (Pecreaux et al., 2006). It was observed that partial depletion of a dynein light intermediate chain or of GPR-1/2 abolishes oscillations, whereas posterior spindle displacement is nevertheless achieved. These findings indicate that a threshold of force generation must be reached for oscillations to occur (Pecreaux et al., 2006). Modeling suggests that all the features of oscillations in the wild-type, including their onset and their termination, can be accounted for by an increase in the processivity of a motor, for instance dynein. Further analysis of oscillation frequencies lead to the proposal that such an increase could, in principle, result from a decrease in the detachment rate of microtubules (Pecreaux et al., 2006).

A major change in cortical residency times of microtubules occurs in embryos depleted of the small G protein of the ARF family EFA-6 (O'Rourke et al., 2010), which was identified as a suppressor of the lethality provoked by conditional dynein inactivation (O'Rourke et al., 2007). efa-6(RNAi) embryos fail to undergo spindle oscillations and exhibit drastic prolongations of microtubule cortical residency times, although the mechanisms by which this occurs remain to be determined (O'Rourke et al., 2010). GFP::EFA-6 is enriched on the anterior cortex, and rescue experiments with truncated fusion proteins indicate that the N-terminal domain and the PH domain that anchors EFA-6 to the cortex, but not its GTPase activity, are required for spindle oscillations. Other phenotypes observed upon EFA-6 depletion include transient separation of centrosomes from the nuclear envelope and increased pole-to-pole distance during mitosis, suggestive of excess pulling forces (O'Rourke et al., 2010). Whereas it remains to be determined whether this is the case, for instance through spindle severing or spindle pole fragmentation experiments, this raises the possibility that excess and perhaps uniform pulling forces also prevents oscillations.

Interactions between the plus-end of astral microtubules and the cell cortex are numerous, and it is likely that most are not productive for force generation, given that merely ∼50 force generators have been estimated to be active at any given time (Grill et al., 2003). Therefore, changes that are relevant for force generators may be difficult to identify among a large number of unproductive events at the cell cortex. An attempt has been made to monitor the activity of cortical dynein using a so-called EB1 velocity assay (EVA) (Gusnowski and Srayko, 2011). This assay is conducted in mutant embryos in which microtubules are released from the centrosomes due to a dominant mutation in the katanin MEI-1, thus allowing observation of microtubule fragments gliding along the cortex. The fastest amongst such movements are dependent on dynein activity and have been dubbed “dynein dependent microtubule velocities” (ddMTVs). Such ddMTVs are reduced in embryos depleted of LIN-5, but not of GPR-1/2 or of GOA-1/GPA-16. This suggests, first, that LIN-5 can exert some function independently of other ternary complex components, and second, that ddMTVs do not reflect cortical pulling forces responsible for centrosome and spindle displacement. ddMTVs are more numerous during pronuclear migration and centration/rotation (Gusnowski and Srayko, 2011), potentially reflecting the fact that the former type of interaction is more important during these earlier stages.

In summary, the ternary complex serves to anchor dynein to the plasma membrane, and perhaps to endomembranes. Together with microtubule depolymerization, this generates pulling forces on astral microtubules, thus enabling proper centrosome and spindle positioning during the first asymmetric division of C. elegans embryos.

Because Gα proteins are an integral part of the ternary complex, modulating their availability has an impact on pulling forces. Depleting the Gβ protein GPB-1 or the Gγ protein GPC-2, which could normally associate with the Gα proteins in a heterotrimeric configuration, results in excess movements during centration/rotation as well as increased net pulling forces on the anterior spindle pole during mitosis (Afshar et al., 2005; Afshar et al., 2004; Tsou et al., 2003a). These results are compatible with the Gβγ dimer acting as a negative regulator of force generation, presumably by competing with GPR-1/2 for binding to Gα-GDP. Consistent with this interpretation, GPR-1/2 distribution in gpb-1(RNAi) embryos increases during centration/rotation and becomes uniform during mitosis (Thyagarajan et al., 2011; Tsou et al., 2003a). Furthermore, the phenotypes generated by depletion of GPB-1 or GPC-2 are suppressed by the concomitant depletion of GOA-1/GPA-16 (Tsou et al., 2003a). The potential for negative regulation offered by the Gβγ dimer appears to be harnessed to ensure appropriate cortical GPR-1/2 distribution in the wild-type (Thyagarajan et al., 2011). Indeed, GPB-1 levels at the cell membrane are lower during mitosis, which correlates with increased cortical GPR-1/2. As in other systems, GPB-1 traffics through the endosomal network in C. elegans (Thyagarajan et al., 2011), and because RAB-5-positive early endosomes are enriched on the anterior in a PAR-dependent manner (Andrews and Ahringer, 2007), GPB-1 trafficking may be more pronounced on the anterior side. Compatible with this view, asymmetric accumulation of GPB-1 is observed inside cells, presumably in the endosomal network, following partial inactivation of intracellular trafficking (Thyagarajan et al., 2011). Overall, these results suggest that modulation of Gβγ trafficking plays a critical role during the first asymmetric division of C. elegans embryos.

After the first asymmetric division, the embryo has an AP axis, with AB and P₁ having inherited different PAR domains (Figure 4A), as well as different cytoplasmic factors that influence cell fate (see Section 5, Figure 5). At second cleavage, the AB cell undergoes a symmetric division in which the spindle is oriented perpendicular to the previous division axis; however, due to the constraints of the eggshell, one AB daughter shifts more posteriorly and is thus designated ABp (Figure 4C). In contrast, the P₁ cell exhibits polarized domains and divides asymmetrically again on the AP axis to produce EMS and P₂; the position of the EMS cell defines the ventral side of the embryo (Figure 1, Figure 4) (reviewed in Maduro, 2010; Schnabel and Priess, 1997). The division of the ABa and ABp cells at third cleavage results in the left-hand AB daughter cells being more anterior in position, which ultimately defines the left/right axis of the worm (see Section 5), while the EMS and P₂ cells both undergo asymmetric, oriented divisions (Figure 1, Figure 4D). Not only are division orientations during these cleavages highly defined, but also division timing is specific to each blastomere and its progeny. Thus, AB and P₁ exhibit cell cycle asynchrony, with P₁ dividing later than AB; similarly, after the asymmetric division of P₁, the P₂ daughter divides later than EMS (Figure 4). In this section, we first outline what is known about the molecular basis of cell cycle asynchrony at the two-cell stage and then cover polarity and spindle positioning at the second and third cleavages.

In support of the first mechanism, a candidate approach revealed that PLK-1 and CDC-25, two positive regulators of mitotic entry, are enriched in AB compared to P₁ (Budirahardja and Gönczy, 2008; Nishi et al., 2008; Rivers et al., 2008). Across eukaryotic evolution, polo-like kinases and Cdc25 phosphatases promote entry into mitosis through activation of the cyclin dependent kinase Cdk1. In C. elegans, the polo-like kinase PLK-1 also promotes entry into mitosis and is thought to act by phosphorylating CDC-25.1. PLK-1 is present at higher levels in the cytoplasm of the AB cell than in the P₁ cell (Budirahardja and Gönczy, 2008; Nishi et al., 2008; Rivers et al., 2008; Chase et al., 2000). CDC-25.1 is also asymmetrically distributed; the nucleus of AB contains more CDC-25.1, apparently because PLK-1-dependent nuclear import occurs more readily in AB than in P₁ (Rivers et al., 2008). Determining the significance of these asymmetries for cell cycle duration in AB versus P₁ has proven difficult because complete depletion by RNAi results in earlier cell cycle arrest, and partial depletion slows the cell cycles of both AB and P₁. Nonetheless, several lines of evidence from partial RNAi experiments support a role for PLK-1 and CDC-25.1 in regulating cell cycle asynchrony (Budirahardja and Gönczy, 2008; Rivers et al., 2008). First, the observed changes in cell cycle rates in various par mutants correlate with PLK-1 and CDC25.1 localization. For example in par-3 mutants, the cell cycle rates of AB and P₁ are like that of a wild-type P₁ cell, and the levels of PLK-1 and CDC25.1 in both cells are similar to those in a wild-type P₁ cell. Second, although partial depletion of CDC-25.1 and PLK-1 does not abolish asynchrony, it does produce a more profound effect of slowing the P₁ cell cycle. This result indicates that P₁ is more sensitive than AB to reduction of PLK-1 and CDC-25.1, consistent with these proteins being key regulators of cell cycle asynchrony between the two blastomeres in the wild-type.

The second mechanism for cell cycle asynchrony is coupled to the unequal first cleavage. The observation that interference with DNA replication has a stronger effect on delaying entry into mitosis in P₁ compared to AB pointed to a potential role for checkpoint control in influencing cell cycle asynchrony (Brauchle et al., 2003). Such delayed entry into mitosis results from the activation of a DNA replication checkpoint utilizing the conserved ATL and CHK-1 kinases. Importantly, depletion of ATL-1 and/or CHK-1 in wild-type embryos, without interfering with DNA replication, results in the AB and P₁ cell cycles being closer together (Brauchle et al., 2003). Further, examination of RNAi-treated embryos that have normal polarity but an equal first cleavage revealed a similar decrease in the time difference. These and other results are consistent with the model that DNA replication factors are uniformly distributed in the one-cell and become rate limiting in the P₁ cell because of its smaller size after unequal division (Brauchle et al., 2003). This activates the checkpoint and preferentially slows the P₁ cell cycle. PLK-1 asymmetry is normal in embryos lacking ATL-1 and/or CHK-1 function, and analysis of embryos simultaneously compromised for the two pathways suggests that they act in parallel (Budirahardja and Gönczy, 2008; Rivers et al., 2008). Thus, the preferential slowing of the P₁ cell cycle by checkpoint activity, combined with the preferential acceleration of the AB cell cycle by PLK-1 together contribute to the final cell cycle asynchrony observed in wild-type embryos. Similar mechanisms may contribute to setting the specific cell cycle rates exhibited after subsequent asymmetric divisions.

Asymmetric division of P₁ likely depends on the PAR proteins. Although PAR-2 and PAR-1 are initially present around the entire cortex in P₁ (Figure 4A), an anterior PAR-3 domain forms during prophase, and PAR-2 and PAR-1 recede to the posterior cortex (Figure 4B) (reviewed in Schubert et al., 2000; Kemphues and Strome, 1997). These observations suggest that polarity is actively reestablished in P₁ in a manner that mirrors the situation in P₀, a view supported by the observation that GFP::PAR-6 localization is coincident with an anteriorly-directed flow of cortical NMY-2::GFP also in P₁ (Munro et al., 2004).

Although PAR dynamics during P₁ polarization are similar to those in P₀, there also appear to be unique requirements for P₁ polarity based on the phenotypes of ooc-3 and ooc-5 mutants (Basham and Rose, 2001; Pichler et al., 2000; Basham and Rose, 1999). In embryos derived from ooc mutant mothers, PAR localization is largely normal in the one-cell, but not in P₁. The transmembrane protein OOC-3 and the torsin family member OOC-5 both localize to the endoplasmic reticulum and contiguous nuclear envelope. In addition, depletion of a subset of nuclear pore complex proteins results in analagous P₁ polarity defects (Schetter et al., 2006). These observations suggest a link between nuclear envelope biology and polarity that remains to be elucidated.

The movements that determine spindle position in response to PAR polarity domains in P₁ are also similar to those described for P₀ (reviewed in Asymmetric cell division and axis formation in the embryo; Kemphues and Strome, 1997). The P₁ spindle orients on the AP axis via a 90° rotation of centrosomes and the associated nucleus prior to mitosis (Figure 4B). Mutation or depletion of Gα, GPR-1/2, LIN-5, and LET-99 show that all these components are required for nuclear rotation in P₁ (Colombo et al., 2003; Gotta et al., 2003; Srinivasan et al., 2003; Tsou et al., 2003a). Analysis of temperature sensitive mutations likewise reveals a role for dynein and some of its regulators (Zhang et al., 2008). Embryos with impaired Gβγ function have abnormal centrosome positioning and retarded rotation in P₁ (Gotta and Ahringer, 2001), but it remains to be determined if these defects are due to excess Gα, as for P₀, or whether Gβγ plays a separate role in P₁ spindle positioning. Anaphase spindle positioning has not been analyzed extensively in P₁, but the posterior spindle pole oscillates as in P₀, and LET-99 and GPR-1/2 are similarly localized (Figure 4).

There are several other genes required for P₁ nuclear rotation, but their precise roles remain to be determined. For example, spn-4 mutant embryos exhibit a P₁ spindle orientation perpendicular to the AP axis (Gomes et al., 2001). However, as SPN-4 is known to be involved in translational regulation (see Section 5), this phenotype is likely due to defects in properly translating a yet-to-be identified protein. Regardless, an interesting aspect of the spn-4 phenotype is that PAR domains in P₁ are initially normal, but then realign with the transverse spindle later in the cell cycle (Gomes et al., 2001). This suggests the existence of a mechanism that reinforces alignment of the spindle with cell polarity in P₁, which could be analogous to the refinement of polarity domains during cytokinesis in the one-cell embryo mentioned earlier (Schenk et al., 2010).

PAR domains are also established prior to the asymmetric divisions of P₂ and P₃, but their positions are reversed with respect to the earlier divisions, with spindle positioning being similarly reversed (reviewed in Kemphues and Strome, 1997). That is, the PAR-2 domain is located along the ventral side of P₂ and the P₂/EMS contact site, rather than at the posterior of P₂ (Figure 4C, D). The P₂ nucleus rotates so that one centrosome is associated with the P₂/EMS cell contact (or the P₃/E contact) and the spindle sets up with a dorsal/ventral tilt; as a consequence, the smaller P cell is born “anteriorly” and ventrally, adjacent to the E cell (Figure 4, Figure 5). Cell recombination experiments in which wild-type blastomeres were separated and then put back together with different partners revealed that the cue for polarity reversal is contact with the EMS cell (Arata et al., 2010). Polarity reversal requires the activity of the MES-1/SRC-1 pathway. MES-1 has similarity to receptor tyrosine kinases and is enriched at the cell contact region of EMS and P₂. MES-1 is required in both EMS and P₂ to induce polarity reversal, while the SRC-1 tyrosine kinase is only required in P₂ (Arata et al., 2010; Bei et al., 2002; Berkowitz and Strome, 2000; Kemphues and Strome, 1997).

Spindle positioning in P₂ likely utilizes some of the same components as in P₀. GPR-1/2 and LIN-5 are enriched at the EMS/P₂ cell contact site prior to division of EMS and P₂ (Figure 4C) (Srinivasan et al., 2003; Tsou et al., 2003a). Elegant cell recombination experiments using cells differentially labeled with YFP-GPR-1 and mCherry::GPR-1 revealed that the enrichment at the contact site occurs in P₂ rather than in EMS (Werts et al., 2011). Therefore, the asymmetry of GPR-1/2 and LIN-5 at the cell contact likely reflects PAR dependent pathways for localization of these proteins operating in the P₂ cell and needed for P₂ spindle orientation. Indeed, MES-1 and SRC-1 are required for the localization of GPR-1/2 at the P₂ cell contact and for P₂ spindle orientation, consistent with their roles in polarity domain reversal.

The EMS cell also undergoes an asymmetric division. A nuclear rotation event positions the spindle on the AP axis of the embryo, and one spindle pole becomes positioned at the EMS/P₂ contact site (Figure 4C, D). These movements coupled with cell polarization result in one daughter, the E cell, being born next to P₂ and assuming endodermal fate (see Section 5). PAR domains are not polarized with respect to AP polarity in the EMS cell. Once again, cell isolation and reconstitution experiments illustrated that signals from P₂ both polarize EMS for cell fate and induce spindle positioning. Both processes are regulated redundantly by the MES-1/SRC-1 pathway and a Wnt signaling pathway (Sawa, 2012; Werts and Goldstein, 2011; Werts et al., 2011; Bei et al., 2002; Schlesinger et al., 1999; Goldstein, 1995b). Only some of the Wnt pathway components used to specify the E cell (see Section 5) are also used for EMS spindle positioning. For example, upstream components, including the Wnt signal (MOM-2), the Frizzled receptor (MOM-5), and Disheveled family proteins are required. Recent studies determined that the divergent β-catenin homolog WRM-1 also participates in regulating EMS spindle orientation, although not through transcription as predicted for canonical Wnt signaling. Instead, a WRM-1::GFP reporter becomes asymmetrically depleted from the EMS/P₂ side of the EMS cell in a Wnt pathway-dependent manner (Kim et al., 2013; Takeshita and Sawa, 2005). Genetic analysis indicates that it is the removal of WRM-1 from the posterior (P₂ interacting cortex) that is key, rather than its presence in the cell (Kim et al., 2013). This explains why previous studies failed to uncover a role for WRM-1 in spindle orientation. Surprisingly however, cell isolation and reconstitution experiments have shown that a Wnt expressing P₂ cell can actively orient the spindle (Goldstein et al., 2006). Thus, the Wnt pathway may influence spindle positioning in more than one way, or other factors influence spindle orientation in intact embryos.

Although there is no obvious asymmetry of G protein pathway components in EMS, the analysis of temperature sensitive alleles of LET-99 and GPA-16 indicates that these proteins are nonetheless important for EMS spindle orientation (Zhang et al., 2008; Tsou et al., 2003a). Further, the defects observed in these mutants appear stronger than those observed following single loss of the Wnt or MES-1 pathways. This result could indicate that LET-99 and GPA-16 function in both pathways, or alternatively act in a separate parallel pathway promoting spindle orientation. In addition, analysis of the dynactin components GFP::DNC-1 and GFP::DNC-2 showed that these proteins are enriched at the EMS/P₂ contact site (Zhang et al., 2008). Although cell reconstitution experiments like those utilizing tagged-GPR fusions should be conducted to definitively determine if dynactin is asymmetrically localized in the EMS cell, dynactin localization is dependent on both MES and Wnt signaling. Further, dnc-1(RNAi) results in partial failures of EMS rotation. Thus, asymmetric dynactin accumulation may provide a useful tool to further dissect how Wnt signaling regulates spindle movements. For example, it will be of interest to detemine if the removal of WRM-1 is necessary for dynactin accumulation in EMS.

Although the AB cell and its daughters divide symmetrically, their division pattern is nonetheless tightly regulated. The centrosome migration pattern around the AB nucleus results in the spindle poles being perpendicular to the AP axis at NEBD, and no rotation events are seen in wild-type embryos. However, in par-3, par-6, and pkc-3 mutant embryos, the AB nuclear centrosome complex rotates on to the AP axis, as in the P₁ cell (reviewed in Kemphues and Strome, 1997). Cell shape manipulations revealed that the nuclear centrosome complex rotates towards flat cell contact areas in par-3 mutants (Tsou et al., 2003b). Thus the anterior PARs act not only to orient spindles onto a specific axis, but to prevent cell shape from affecting division pattern, through an unknown mechanism.

After division of the AB cell, the centrosomes migrate in ABa and ABp so that they are on the future left/right embryonic axis, and the spindles set up in this orientation. However, at telophase, the future left spindle pole shifts anteriorly so that the spindles are skewed 30° relative to the precise left/right axis (Figure 4) (Pohl, 2011; Bergmann et al., 1998). The only well-characterized component to play a role in this spindle skew is the Gα protein GPA-16 (Bergmann et al., 2003). Temperature sensitive gpa-16 mutants raised at semi-permissive temperature can have normal first and second divisions, but variable ABa and ABp spindle positions. In some cases, the telophase shift in ABa and ABp is reversed, which correlates with the production of adult worms with reversed handedness. It remains to be determined if GPA-16 is part of the cue for left/right handedness, or whether the G protein pathway is acting with GPR-1/2/LIN-5 to orient spindles responding to an asymmetry established earlier during development.

After AP polarity domains have been established in the one-cell embryo, several components act downstream of the PAR proteins to generate cytoplasmic asymmetries (Figure 5), which contribute to specifying the fates of blastomeres or their daughter cells during early embryogenesis (Figure 1). In the following we refer to such components as polarity mediators because they are involved in the asymmetric localization of multiple downstream cell fate regulators and act in many cells. Polarity mediators include several proteins that contain tandem CCCH zinc-finger domains: MEX-5, MEX-6, MEX-1, and POS-1. In addition, the KH domain protein MEX-3 and the RRM domain protein SPN-4 act as polarity mediators. The CCCH zinc-finger, KH, and RRM protein domains interact with mRNAs, and most of these polarity mediators have been implicated in translational control (reviewed in RNA-binding proteins; Asymmetric cell division and axis formation in the embryo; Translational control of maternal RNAs). We use the term cell fate regulators for those molecules that directly regulate the fate of one or a few founder cells. These regulators include PIE-1 and NOS-2 (P cell fate), SKN-1 and POP-1 (EMS fates), PAL-1 (C and D fate), as well as GLP-1 and APX-1 (AB fates) (reviewed in Wang and Seydoux, 2013; Sawa, 2012; Maduro, 2010; RNA-binding proteins; Asymmetric cell division and axis formation in the embryo; Notch signaling in the C. elegans embryo; Translational control of maternal RNAs; Wnt signaling; Schnabel and Priess, 1997; Bowerman, 1995). The localization patterns of these regulators are described below in Sections 5.2-5.4, and their developmental activities will be explained in Section 5.5.

MEX-5 and MEX-6 are almost identical proteins that act partially redundantly and thus will be referred to as MEX-5/6, unless studies were carried out on one protein only. MEX-5/6 are required to generate asymmetric localization of the other polarity mediators in addition to downstream cell fate determinants, and thus can be considered primary polarity mediators.

MEX-5 is localized at higher levels in the anterior cytoplasm by the end of the one-cell stage, and thus is inherited primarily by the AB daughter cell upon division (Figure 5A, B) (Cuenca et al., 2003; Schubert et al., 2000). Analysis of the dynamics of GFP::MEX-5 fusions gave no evidence for directed transport to the anterior, but revealed that the cytoplasmic mobility of MEX-5 is faster in the posterior compared to the anterior (Daniels et al., 2010; Tenlen et al., 2008). Examination of a photoconvertible Dendra::MEX-5 fusion, which allows proteins to be marked and then monitored over time, yielded similar results and further showed that neither degradation nor synthesis promotes MEX-5 asymmetry (Griffin et al., 2011). Several sites in MEX-5 are phosphorylated in vivo in a PAR-4- and PAR-1-dependent manner and are needed for normal MEX-5 asymmetry and function (Griffin et al., 2011; Tenlen et al., 2008). PAR-1 is likely to be the direct regulator of MEX-5 asymmetry. PAR-1 is present in a cortical and cytoplasmic gradient that correlates with the gradient of increased diffusivity of MEX-5 in wild type. Although it remains to be determined how the cytoplasmic PAR-1 gradient is produced, this kinase can phosphorylate MEX-5 at serine residue 404 (S404) in vitro. Further, mutation of S404 in Dendra::MEX-5 results in slower mobility and loss of asymmetry, similar to what is observed for unaltered Dendra::MEX-5 in par-1 mutant embryos. Conversely, knockdown of the catalytic subunit of PP2A phosphatase, which is uniformly distributed in the cytoplasm, results in increased MEX-5 mobility (Griffin et al., 2011). Mutations in the tandem zinc-finger domains of MEX-5, including ones that alter the affinity of MEX-5 for RNA, also cause an increase in MEX-5 mobility and loss of asymmetry (Griffin et al., 2011; Tenlen et al., 2008; Pagano et al., 2007). Finally, FCS experiments and biochemical analyses indicate that multiple MEX-5 complexes exist in vivo (Griffin et al., 2011; Daniels et al., 2010). These results, along with mathematical modeling, lead to the proposal that phosphorylation by a gradient of cytoplasmic PAR-1 promotes the release of MEX-5 from RNA-containing complexes in the posterior (Griffin et al., 2011). The resulting increased mobility of MEX-5 allows it to diffuse to the anterior where the lower levels of PAR-1 are insufficient to counteract dephosphorylation of MEX-5 by PP2A. Dephosphorylated MEX-5 associates with RNP complexes and cannot redistribute into the posterior as readily, generating a stable anterior to posterior gradient of MEX-5.

In summary, by the end of the one-cell stage, anterior-posterior polarity of cortical PAR proteins has been translated into opposing cytoplasmic gradients of PAR-1 and MEX-5. MEX-5/6 then regulate further downstream asymmetries as outlined below.

The mechanism by which MEX-5/6 restrict the localization of the polarity mediators POS-1 and MEX-1, as well as the cell fate regulator PIE-1, to the germ cell precursors is well understood (Figure 5). Live imaging of PIE-1::GFP showed that the fusion protein is localized in two steps (Movie 5) (Reese et al., 2000). First, PIE-1::GFP becomes enriched in the posterior cytoplasm of the one-cell embryo before division. Second, any protein remaining in the AB cell is degraded, and a similar degradation occurs in the somatic daughter cell of subsequent P cell divisions. The degradation events are dependent on MEX-5/6. PIE-1 contains tandem CCCH finger domains, as do POS-1 and MEX-1, even though the overall sequence of these proteins is distinct. A structure/function analysis of PIE-1::GFP revealed that the first zinc finger ZF1 is necessary and sufficient for degradation in the anterior cell. ZF1 binds to ZIF-1, a SOCS box protein that is a substrate recruitment factor for the CUL-2 E3 Ubiquitin ligase complex (DeRenzo et al., 2003). The ZF1 of POS-1 and MEX-1 are likewise necessary for their degradation in somatic cells (Reese et al., 2000).

How do MEX-5/6 activate protein degradation in somatic cells? Regulation of ZIF-1 by MEX-5 at the translational level is key. The zif-1 mRNA appears uniformly distributed, but the expression of a reporter transgene bearing only the zif-1 3’UTR is repressed during oogenesis and then in the P lineage of embryos (Oldenbroek et al., 2012; Guven-Ozkan et al., 2010). Analysis of reporter gene expression in different mutant backgrounds indicates that MEX-3, SPN-4, and POS-1 all act negatively on zif-1 mRNA translation in the embryo, while MEX-5/6 act positively. All of these factors bind to the zif-1 3’UTR in vitro. MEX-5/6 and POS-1 bind to the same region of the 3’UTR, with MEX-5/6 binding being preferred. These and other data suggest that in vivo, high levels of MEX-5/6 antagonize translational repression of the zif-1 mRNA, allowing its expression in AB and other somatic cells (Oldenbroek et al., 2012). Because ZIF-1 also promotes POS-1 degradation in AB, this further relieves zif-1 mRNA translational repression in somatic cells, whereas the high levels of POS-1 in the germ line precursors keep zif-1 expression off (Reese et al., 2000). Moreover, POS-1 has been suggested to repress MEX-6 translation based on in vitro binding to the mex-6 3’UTR and genetic analysis (Tenlen et al., 2006). Such mutual antagonism of MEX-5/6 and POS-1 would reinforce their asymmetric localization patterns and thus those of their downstream targets.

While the studies above suggest a role in translational regulation, other work indicates that MEX-5/6 can also mediate asymmetry through interactions at the protein level. MEX-5/6 are required for the anterior cytoplasmic enrichment in the one-cell embryo of the polo-like kinases PLK-1 and PLK-2, which are important for cell cycle asynchrony at the two-cell stage as described earlier (Budirahardja and Gönczy, 2008; Nishi et al., 2008; Rivers et al., 2008). Both PLK-1 and PLK-2 bind to MEX-5 and MEX-6 in yeast two hybrid assays via their polo box domains (PBD), and a PBD::GFP fusion protein is anteriorly enriched in embryos. These and other results suggest that physical association with MEX-5/6 generates the asymmetric localization of PLK-1/2. Conversely, mutations that affect the ability of MEX-5 to bind to PLK-1/2 do not affect MEX-5 asymmetry, but do affect MEX-5 activity, including its ability to promote PIE-1 degradation (Nishi et al., 2008).

MEX-5/6 also appear to regulate the anterior distribution of the polarity mediator MEX-3 at the protein level. Based on quantitative analysis of MEX-3 levels, the activity of MEX-5/6 stabilizes MEX-3 in the AB lineage, potentially protecting it from degradation. MEX-6 can physically associate with MEX-3 in yeast two-hybrid assays and this interaction could promote MEX-3 stability (Huang et al., 2002). For example, the association of MEX-3 with MEX-6, and thus PLK-1/2, could modify MEX-3 stability through phosphorylation. Alternatively, because MEX-3 and MEX-5/6 both bind to zif-1 mRNA (Oldenbroek et al., 2012), the MEX-3/MEX-6 protein interaction could instead reflect a transient association that is part of the mechanism for zif-1 regulation.

By the end of the two-cell stage, the polarity mediators POS-1, MEX-1, and MEX-3 are asymmetrically localized (Figure 5), while SPN-4 is still uniform at this stage. In addition to their role in repressing zif-1 translation discussed above, these polarity mediators determine the protein localization patterns of several cell fate regulators (Figure 5) (Ogura et al., 2003; Gomes et al., 2001; Tabara et al., 1999; Guedes and Priess, 1997; Draper et al., 1996). The best characterized examples involve translational repression by MEX-3 and POS-1.

Although the mechanisms described above can account for many of the asymmetric protein localizations in the early embryo, interesting variations on these themes contribute to the segregation of other components.

For instance, the pre-divisional posterior enrichment of PIE-1 has been linked to its diffusional mobility, reminiscent of the studies of the MEX-5 gradient, but in this case with diffusion of PIE-1::GFP being slower in the posterior. Pre-division enrichment is separable from somatic degradation and involves a region of the PIE-1 protein that includes the second CCCH-finger (ZF2) (Daniels et al., 2009; Reese et al., 2000). It has been postulated that the formation of complexes of PIE-1 with P granules and/or RNA could slow PIE-1 diffusion in the posterior and prevent its redistribution (Daniels et al., 2009); however, an asymmetric distribution of P granules is not needed for proper PIE-1 localization (Gallo et al., 2010; Wallenfang and Seydoux, 2000). Although in theory PAR-1 could phosphorylate PIE-1 to regulate diffusion, this seems unlikely given that MEX-5/6 act downstream of PAR-1 and are required for the pre-division enrichment of PIE-1, in addition to regulating PIE-1 after division (Cuenca et al., 2003; Schubert et al., 2000).

Mobility regulated by the state of assembly may also play a role in the localization of P granules to the germ cells (Gallo et al., 2010; Brangwynne et al., 2009; Cheeks et al., 2004). P granules were initially thought to move towards the posterior with the internal cytoplasmic flow in the early one-cell embryo, with P granules left in the anterior postulated to be degraded (reviewed in Asymmetric cell division and axis formation in the embryo). However, more recent studies show that P granules also travel anteriorly with subcortical flow, such that flow per se would not provide an efficient localization mechanism (Gallo et al., 2010; Brangwynne et al., 2009; Cheeks et al., 2004). Further, detailed quantitative analysis of fluorescent P granule reporters showed that the major mechanism for posterior enrichment involves regulated assembly and disassembly (Gallo et al., 2010; Cheeks et al., 2004), which can be modeled as a system of condensing versus dissolving liquid droplets (Brangwynne et al., 2009). In particular, P granulexs grow and shrink throughout the cell cycle, with the portions of the cell destined to the P lineage showing a higher degree of assembly. Mutations in the pptr-1 gene (protein phosphatase 2a regulatory subunit) prevent P granules assembly during mitosis and lead to a lack of their enrichment in the P cells (Gallo et al., 2010). However, pptr-1 mutants exhibit normal posterior enrichment of several other germ line proteins. In addition, although MEX-5/6 act to promote P granule disassembly in the anterior cytoplasm in wild-type embryos, double mutant analysis indicates that this effect is separable from that of PPTR-1 in the posterior (Gallo et al., 2010). These and other results suggest that this assembly mechanism is distinct from the other MEX-5/6 dependent posterior localization mechanisms described earlier.

Another mechanism ensures the segregation of SKN-1, a bZip-like transcription factor needed for the fates of EMS and its daughter blastomeres (Bowerman et al., 1993; reviewed in Maduro, 2010; Maduro, 2006, Schnabel and Priess, 1997). SKN-1 protein is enriched in the nucleus of the P₁ cell and its daughters, due to enrichment in the P₁ cell combined with degradation in the AB cell. Although its posterior localization depends on MEX-5, SKN-1 does not have a CCCH-type finger that could mediate its degradation via the ZIF-1 pathway. Instead, reduction of SKN-1 in the AB cells is regulated in part by a different E3 ligase, EEL-1 (Page et al., 2007). The polarity mediator MEX-1 is also needed for proper SKN-1 distribution. However whether MEX-5 and MEX-1 are involved in the degradation or posterior enrichment of SKN-1 remains to be determined.

In contrast to some organisms, the localization of mRNAs to specific cells does not appear to play a major role in the restricted localization of most proteins in C. elegans (Tranlational control of maternal RNAs; Schnabel and Priess, 1997). Although the pos-1 and mex-1 mRNAs are enriched in the P₁ cell, the corresponding proteins are also subjected to regulation at the level of degradation as described earlier. In addition, a number of mRNAs are localized specifically to the germ cell precursor P₄ by the 28 cell stage, but this is due to degradation in the somatic blastomeres beginning at the 4-cell stage. MEX-5/6 are required for this degradation via a mechanism that appears distinct from their role in localizing PIE-1 and POS-1 to the P₄ cell (D'Agostino et al., 2006; Tenenhaus et al., 2001). In addition, germ-line restricted mRNAs are likely regulated at the level of translation as well, as illustrated above for nos-2.

In summary, although the details regarding the mechanisms underlying some of these interactions remain to be elucidated, their existence adds to the diversity of means to localize proteins in the early embryo. Further, for many of the proteins described here and earlier, localization patterns are generated through more than one of these mechanisms, presumably conferring robustness to the system.

The division pattern of the embryo leads to a specific arrangement of cells, which, combined with cytoplasmic asymmetries, results in spatially restricted cell signaling events that further refine the fates of the early blastomeres. As mentioned earlier, contact with the P₂ cell induces EMS to divide asymmetrically to produce E, the precursor of all intestinal tissue, and MS, a major muscle precursor (Figure 1, Figure 5) (reviewed in Sawa, 2012; Munro and Bowerman, 2009; Notch signaling in the C. elegans embryo; Wnt signaling; Schnabel and Priess, 1997). Additional asymmetric divisions in the P lineage produce C, a muscle and hypodermis precursor, D, a muscle precursor and P₄, the germ-line precursor (Figure 1D). Concurrently, signals from P₂ influence ABp so that it generates a different set of cell types than those produced by its more ventral sister ABa. In addition, the specific cell contacts generated by the skew in the division of the ABa and ABp cells at third cleavage leads to cell-signaling events that determine the left/right axis of the embryo (reviewed in Pohl, 2011; Schnabel and Priess, 1997).

How does contact with the P₂ cell result in such different outputs in the AB descendants versus the EMS descendants? The answer is that the signaling pathways involved are distinct. The specification of ABa and ABp will be discussed below, together with the establishment of left/right differences, as both involve the Notch signaling pathway. In contrast, the specification of E versus MS depends on an asymmetric division induced by Wnt and MES-1/SRC-1 signaling pathways introduced earlier. Cell isolation and reconstitution experiments using wild-type and mutant cells indicated that the Wnt signal (MOM-2) is required in P₂, while the Frizzled receptor (MOM-5) and other downstream components are needed in EMS for endoderm specification (Goldstein and Hird, 1996; Goldstein, 1995a; Goldstein, 1995b; reviewed in Werts and Goldstein, 2011; Schnabel and Priess, 1997). MES-1, on the other hand, is required in both P₂ and EMS for proper specification of endoderm (Bei et al., 2002; Schlesinger et al., 1999). It remains to be determined how early polarity cues localize Wnt and MES-1 activity.

The outcome of both Wnt and MES-1/SRC-1 signaling is that the EMS daughter cell closest to P₂, the E cell, exhibits lower nuclear levels of the TCF family transcription factor POP-1 (Sugioka et al., 2011; Sumiyoshi et al., 2011; Lin et al., 1998; Thorpe et al., 1997; Lin et al., 1995; reviewed in Sawa, 2012; Munro and Bowerman, 2009). The most progress has been made in determining how Wnt signals polarize POP-1 levels. In particular, the E cell has higher nuclear levels of WRM-1, which correlates with the loss of cortical WRM-1 prior to the EMS division (Kim et al., 2013). WRM-1 is important for the nuclear export of POP-1 in the E nucleus. Another β-catenin family protein, SYS-1, is also enriched in the E cell nucleus. SYS-1 acts as a transcriptional co-activator with POP-1, similar to β-catenin in other systems. The POP-1 remaining in the E cell nucleus appears to complex with SYS-1 and thus activate endodermal (E cells) gene targets. In contrast, the high levels of nuclear POP-1 in MS, where only low levels of the SYS-1 co-activator are present, results in repression of endoderm genes; body wall and pharyngeal muscle gene targets are expressed in MS instead (Sugioka et al., 2011; Huang et al., 2007; Phillips et al., 2007; Rocheleau et al., 1999; Shin et al., 1999). SKN-1 is required for activation of gene targets in both E and MS, and thus it is the specific combination of activators and repressors that leads to differential gene expression and determines their respective fates.

The fates of the somatic daughters of P₂ and P₃ appear to result mainly from the presence of PAL-1, a caudal domain transcription factor that activates genes required for the production of muscle cell types (reviewed in Maduro, 2010; Schnabel and Priess, 1997). As described in Section 5.3, PAL-1 protein is restricted to the nuclei in the posterior half of the embryo, but it is not specific for C and D (Hunter and Kenyon, 1996). Given the overlap in posterior localization of SKN-1 and PAL-1, why do the C and D cells activate different targets than the EMS daughters? The timing of SKN-1 protein availability is thought to be the reason, since SKN-1 is present from the 2-cell stage but remains at only low levels by the 12-cell stage (Bowerman et al., 1993). By contrast, PAL-1 is present at similar levels from the 4- to the 12-cell stage, and thus remains at high levels when the C and D founders are generated.

Although present, neither SKN-1 nor PAL-1 activates gene expression in the P cells. This is because the germ line precursors are protected from activating somatic programs of gene expression by PIE-1, which is restricted to the P cell at each asymmetric division (Figure 5) (Mello et al., 1996). Germ-line precursors in C. elegans, as in many organisms, are transcriptionally quiescent. Inhibition of transcription correlates with the failure to phosphorylate the C-terminal domain of RNA polymerase on Serine-5 (Ser-5) and Ser-2, which are needed for transcription initiation and elongation, respectively. These phophorylation events are inhibited by PIE-1 in the P cells of early C. elegans embryos (Tenenhaus et al., 1998). In vitro studies indicate that PIE-1 directly binds to, and inhibits, the kinase needed for Ser-2 phosphorylation (Batchelder et al., 1999). Altered forms of PIE-1 that do not bind the kinase fail to inhibit Ser-2 phosphorylation in vivo, but can still suppress transcription and promote germ-line fate through effects on Ser-5 phosphorylation (Ghosh and Seydoux, 2008). Inhibition of transcription is thought to prevent SKN-1, PAL-1 and other transcription factors from activating gene expression in the P cells. However, after asymmetric division of P₂, the degradation of PIE-1 from the somatic daughter via the MEX-5/ZIF-1 pathway described above relieves transcriptional repression, allowing SKN-1 activation of target genes in EMS. Similarly, after the P₂ and P₃ divisions, disappearance of PIE-1 allows PAL-1 activation of target genes to commence. PIE-1 is also important for activating the translation of NOS-2 in the P₄ cell; although the mechanism is unknown, this function of PIE-1 is separable from its role in preventing transcription (D'Agostino et al., 2006; Tenenhaus et al., 2001).

At the same time that signals from P₂ induce the asymmetry of EMS, P₂ contact influences the fate of ABp via a separate signaling pathway involving GLP-1, a Notch family receptor (reviewed in Notch signaling in the C. elegans embryo; Translational control of maternal RNAs; Schnabel and Priess, 1997). Only the AB daughters express the GLP-1 receptor (Figure 5; see Section 5.3) due to the repressive actions of polarity mediators in the posterior (Crittenden et al., 1997; Evans et al., 1994). Conversely, only P₂ has the ligand for the GLP-1 receptor, APX-1 (Figure 5C) (Mickey et al., 1996). Thus, GLP-1 signaling can only be stimulated in the AB cells, and it is asymmetrically activated because only ABp makes contact with P₂. This signaling interaction induces ABp's descendants to produce dorsal cell types (Mickey et al., 1996; Shelton and Bowerman, 1996; Mango et al., 1994).

The formation of the left/right (LR) body axis also requires Notch signaling (reviewed in Pohl, 2011; Notch signaling in the C. elegans embryo; Schnabel and Priess, 1997). Although C. elegans is overtly bilaterally symmetric, there are specific left/right differences in cell types and organ positioning, which are virtually invariant in wild-type worms. The mechanism that establishes left/right handedness begins at the third cleavage (see Section 4); the initial anterior shift in position of the left-hand AB daughter cells is amplified by coordinated cell movements that ultimately result in specific cell contacts between certain AB descendants and the MS cell (Pohl and Bao, 2010; Bergmann et al., 2003; Bergmann et al., 1998). Physical manipulation during the third cleavage such that the right AB daughters are more anteriorly positioned results in a reversal of cell arrangements and produces viable worms with reversed left/right handedness. These observations and the lack of any asymmetric localization of cell fate regulators within the AB lineage suggest that most aspects of the left-right axis are specified by cell-cell signaling events. Indeed, the use of a temperature sensitive glp-1 allele and cell ablations revealed that the GLP-1 receptor and the MS cell are required for the distinct lineages on the left and right sides; however the ligand in this case is not APX-1 and remains to be identified (reviewed in Pohl, 2011; Notch signaling in the C. elegans embryo; Schnabel and Priess, 1997).

The early C. elegans embryo has emerged as a major model for the study of asymmetric division. In addition to the usual benefits of C. elegans, such as excellent genetic and genomic tools, the transparency of the worm embryo has allowed for major advances in imaging subcellular processes with utmost spatial and temporal resolution, through the use of countless fusion proteins. Combined with laser manipulations to test biophysical aspects of these process, as well as mathematical modeling and computational simulations, this has translated into a comprehensive approach to tackle fundamental biological questions during polarization and asymmetric division of the one-cell C. elegans embryo. In addition to elucidating the mechanisms that transform an initially isotropic system into a polarized one in the first cell cycle, studies of the subsequent early divisions have yielded insights into how cell-signaling events build upon the initial polarity to further refine cell fates. Comparative studies will be of particular interest to investigate how subtle variations on these themes may explain the fascinating diversity among early nematode embryos (Schulze and Schierenberg, 2011; Brauchle et al., 2009).

Many findings made in the C. elegans embryo system have had an important impact on understanding related processes in other organisms. Since the discovery of the PAR proteins in C. elegans over 20 years ago, homologs have been identified in numerous vertebrates, in other invertebrates such as Drosophila and even sponges (reviewed in Morin and Bellaiche, 2011; Fahey and Degnan, 2010; St Johnston and Ahringer, 2010; Galli and van den Heuvel, 2008; Betschinger and Knoblich, 2004). PAR proteins are localized in reciprocal cortical domains not only in asymmetrically dividing cells during development, but also after differentiation in polarized epithelial cells and neurons. Where examined, the PAR proteins have been shown to play important roles in establishing or maintaining polarity in these diverse cell types. Similarly, the use of heterotrimeric G proteins to link PAR polarity cues to spindle positioning is conserved in many organisms and cell types. Finally, the principles underlying the formation of cytoplasmic asymmetries and refinement of cell fates by signaling events in the early embryo have parallels in other systems. At the same time, the work in C. elegans is yielding new paradigms such as asymmetric cytoplasmic localization generated by regulated diffusion. Further analysis in C. elegans as well as studies in various other organisms are likely to continue to synergize in the future to further our knowledge of universal polarity mechanisms.

We are grateful to Eugenel Espiritu, Sachin Kotak, Zoltán Spiró, Michael VanGompel, and the reviewers for critical reading of the manuscript. We thank Eugenel Espiritu for help with the tables, Simon Blanchoud and Zoltán Spiró for contributing Movies 1-4, as well as Ed Munro for providing the images of NMY-2::GFP in Figure 1. Research on asymmetric division in the Gönczy laboratory is supported by the Swiss National Science Foundation (grants 3100A0-102087 and 3100A0-122500/1), that in the Rose laboratory by the NIH (R01GM068744).