Cell polarity in early C. elegans development

Most animals have a fairly asymmetric body form. The anteroposterior axis is generally the most obviously polar, exhibiting the highest degree of asymmetry, the dorsoven- tral axis is generally intermediate in this respect and the left- right axis is often superficially symmetrical, although it is rarely so in detail. On the other hand the fertilized egg has little obvious symmetry.

Many differentiated cells exhibit an obvious polarity; epithelial cells have clearly defined apical and basal domains and many neurons have spatially separated axonal and dendritic regions. The cells of an embryo in the early stages of development are not so obviously polar, yet many exhibit polarity in that they divide asymmetrically, giving rise to daughters that differ in their states of determination and often also in size. The question can then be posed: is the polarity of the organism a consequence of the combined polarities of all the asymmetric divisions during embryogenesis, or is a co-ordinate system set up in the early embryo which acts to specify the polarities of embryonic cell divisions? Polarity in the developing embryo of the nematode Caenorhabdilis elegans has been studied both al the organismal and individual blastomere level by a variety of means including embryonic manipu- lations. cell biological characterization, and genetic analysis.

In this review we discuss experiments that address polarity in C. elegans blastomeres. The first section discusses how the embryonic axes are established and the state of axial polarity in early embryonic cells. In the second section we discuss a cell interaction that specifies polarity in the responding cell. The third section reviews the case in C. elegans for which most is known about the acquisition of cell polarity: the establishment of polarity in the zygote. Lastly, we discuss how cell division axes are set up in polarized cells.

Early development in C. elegans is shown in Figs 1 and 2 (see Wood. 1988 for review). The sperm enters at the future posterior end of the embryo, opposite the end where the oocyte pronucleus resides. The oocyte pronucleus, which was arrested in meiolic prophase, resumes mciosis and both polar bodies are extruded. After mciosis. contractions of the anterior membrane occur and a transient pseudocleavage furrow forms near the center of the egg. During this lime the oocyte pronucleus migrates posteriorly toward the sperm pronucleus. The sperm pronucleus also migrates somewhat toward the oocyte pronucleus. and the two meet in the posterior half of the egg. The pronuclci break down and first cleavage dix ides the egg into a larger anterior cell (AB) and a smaller posterior cell (Pi). This cleavage is unequal because of the posterior placement of the mitotic spindle. AB then divides with the spindle initially set up orthogonal to the anteroposterior axis. During anaphase B. the elongating spindle skews towards the anteroposterior axis so that one daughter. ABa. is positioned anterior to the other. ABp. A few minutes later. l’i divides along the anteroposterior axis to form a larger anterior cell (EMS) and a smaller posterior cell (P?). Divisions in the AB lineage continue along axes orthogonal to the preceding division axes, while divisions of the Pi progeny lend to place them in an anterior-posterior line, though constraints of the eggshell skew them somewhat off this axis. Gastrulation begins al the 28-cell stage; Ea and Ep are the first cells to migrate into the center of the embryo. By the lime gastru- lation ends approximately 550 cells have been generated and most embryonic cell divisions are complete. The embryo elongates inside the eggshell until it is folded in three. During this period of elongation, muscle cells begin to function and the embryo begins to twitch. Al approxi- mately 12 hours after fertilization, the LI larva hatches from the eggshell.

Unlike many other animals, in C. elegant early fate segre- gation does not establish the three germ layers (ectoderm, mesoderm and endoderm). Most tissues are assembled from bits of several distinct lineages. Up to the final cell division there are still mother cells that give rise to two daughters whose lates are normally associated with differing germ layers. The lineage is not completely chaotic though: the ectoderm forms largely from progeny of AB. C’pl and Cpr. the mesoderm forms largely from progeny of MS. D. Cal. and Car. the endoderm forms completely from the progeny E, and the germline from Pi (Fig. 3).

All three embryonic axes in C. elegant are established by the six-cell stage. For each axis, evidence is reviewed on how the axis is established, and whether blastomere polarity might play a role in determining the axis.

Anteroposterior asymmetries are evident soon after fertil- ization (Fig. 4): they include the ruffling of the anterior membrane, the polarized Hows of cortical and cytoplasmic material (Nigon. 1960; Hird and White. 1993)and the local- ization of P granules al pseudocleavage (Strome and Wood. 1982. 1983). At first cleavage, the asymmetries include the posterior positioning of the mitotic apparatus, and the dif- ference in the size and shape of the two asters (Albertson. 1984). which has been shown to be position-dependent and not due to intrinsic differences between the asters (Hyman and White. 1987). The cell biology of the events that establish polarity in the zygote w ill be discussed in a later section. Here we discuss the anteroposterior asymmetries apparent in the egg at the lime of fertilization, as candidates for the initial cue for establishing anterior-posterior polarity.

Some minutes before each oocyte passes into the sper- matheca. the maternal nucleus moves to the end of the oocyte opposite the end that will first enter the spermatheca. The end near the oocyte nucleus will become the anterior end of the embryo. The polar bodies, which are a useful mark of the place where the oocyte nucleus underwent mciosis. are normally found at this end of the embryo following fertilization, but occasionally can be found at the posterior end. The finding that the location of the oocyte nucleus at the time of mciosis can vary with respect to the future anterior-posterior axis suggests it may not act as a cue for determining the anterior-posterior axis.

As the oocyte enters the spermatheca, a sperm enters at the end opposite the maternal pronucleus. Experimentally altering the position of the mitotic apparatus in early blas- tomeres can generate cortical and cytoplasmic Hows similar to those of pscudoclcavage, whose directions are predicted by the position of the mitotic apparatus (Hird and White. 1993). This finding raises the possibility that the sperm asters (analogous to the mitotic apparatus) could direct the Hows seen during pseudocleavage, and by doing so may establish the anteroposterior axis (Hird and White. 1993). However, the sperm entry point becomes anterior rather than posterior in some other nematodes, such as Turbal rix aceti (Pai, 1928) and Panagrellus redivivas (Goldstein, unpub- lished data).

Another possible cue for establishing the anteroposterior axis could be a pre-existing asymmetry in the cytoskeleton or of determinants in the unfertilized oocyte. However no asymmetric distributions of materials have been found in the oocyte by ultrastructural or molecular data.

Early cleavages in nematodes follow a stem cell-like pattern. The zygote (P0) cleaves unequally, forming a larger anterior daughter (AB) which contributes to the somatic tissues, and a smaller posterior daughter (P₁) which continues to cleave in a stem cell-like pattern. P₁ reiterates this pattern, dividing to form a larger anterior daughter (EMS) which contributes to somatic tissues, and a smaller posterior daughter (P₂) which continues to divide as a stem cell. Stem cell-like divisions continue for two more cycles, however their polarity varies between different nematode species: in some nematodes the smaller stem cell-like daughter continues to form posteriorly, making the germ line founder cell (P₄) the posterior-most daughter. In C. elegans and some other nematodes, a reversal of polarity takes place such that P₃ and P₄ come to lie anterior to their somatic sister cells. Schierenberg (1987) has shown that a posterior fragment of P₀ or P₁ retains the ability to divide in a stem-cell like pattern. However this ability moves out of the posterior region of P₂. If a posterior fragment of P₂ containing the nucleus and centrosomes is extruded early in the cell cycle, the fragment continues to divide as a stem cell, whereas if it is taken late in the P₂ cell cycle, it divides equally. The result suggests that there is a site in P₂ responsible for germline polarity, which changes position midway through P₂’s cell cycle, placing the stem cell-like daughter to the anterior in the cleavages that follow.

The daughters of AB. ABa and ABp. give rise to distinct lineages. In an experiment designed to lest whether the ABa and ABp blastomeres are initially equivalent. Priess and Thomson (1987) switched the positions of ABa and ABp by exerting pressure on one side of AB as it divided. If ABa and ABp fates were determined prior to the division of AB. then this operation would have caused inappropriate lineages to be executed in that part of the embryo. However, a normal worm developed, demonstrating that ABa and ABp are equivalent cells, in that each has the potential to differ- entiate as the other normally does. ‘Hie subsequent differ- ences in their fates must come from cell interactions.

The experiment also altered the dorsoventral axis of the four cell stage (Fig. 5). After the manipulation the EMS blas- tomere was on the side that would have become dorsal if unmanipulated, and a daughter of AB was on the side that would have become ventral if unmanipulated. This reversed the dorsoventral axis of the developing worm but gave an otherwise normal worm. Hence the differences between blastomeres along the dorsoventral axis are established by the four-cell stage.

This experiment provides evidence on the determination of dorsoventral polarity within each blastomere: each of the four blastomeres probably was not rotated 180’ by the manipulation, yet the embryo’s axis reversed and a normal worm developed, suggesting that dorsoventral polarity at this stage is a function only of the blastomeres’ positions and not of a polarity within each of the cells. This point may be illustrated by contrasting this case with an animal in which a functional polarity has been shown to exist in early blastomeres. In the starfish Aslerinapectinifera, rotating one of the blastomeres of the two-cell stage 90° or 180° causes axis rotation in half the resulting embryo, resulting in twinned or double partial embryos (Kuraishi anil Osanai. 1989).

Left-right asymmetries are lirst apparent in C. dedans at the six-cell stage with the synchronous division of ABa and ABp. As these cells divide, an invariant skewing of their spindles causes the AB progeny on the left side of the embryo (ABal and ABpI) to shift slightly anteriorly relative to the two AB progeny on the right (ABar and ABpr). Wood (1991) has shown that the left-right handedness of the worm is established via this shift. When the left side progeny of AB were pushed posteriorly al the six-cell stage these cells executed lineages appropriate for the right hand side. Similarly, the cells on the right hand side (now positioned anteriorly as a result of the manipulation) executed left hand lineages. A worm developed but structures normally found on the worm’s right side had developed on the left, and vice versa (Fig. 6).

That handedness could be reversed without reversing the left-right positions of blastomeres demonstrated that AB’s left side progeny are initially equivalent to AB’s right side progeny. The eventual differences between the fates of the left and right AB progeny must be established later via interactions with other cells. The skewing of the AB progeny at the six-cell stage causes AB cells on the right and left sides to have different neighbors later in embryo- genesis: these differences in contacts, and hence differences in interactions, are probably responsible for left right dif- ferences. It is not known why the spindles in ABa and ABp always skew in the same direction in the unmanipulated embryo.

This experiment also demonstrates that each of the early blastomeres have not yet determined their left-right axes: reversal of the left-right axis could be accomplished without reversing the left-right polarity of each blastomere by rotating it 180°.

The blastomere repositioning experiments of Priess and Thompson (1987) and Wood (1991) discussed above suggested that AB cell fates are established largely via inductions. Induction has also been found to be required for establishment of gut fate in the E lineage, via an induction which functionally polarizes EMS.

The EMS blastomere divides longitudinally to form two cells with distinct fates. E and MS. The progeny of E form the entire gut of the worm. The progeny of MS contribute to the posterior half of the pharynx, some anterior body wall muscle, and some neurons. Blastomere isolation and recom- bination experiments have shown that gut is specified in E via an induction that occurs during the four-cell stage between P₂ and EMS (Goldstein. 1992). Isolating the EMS cell early in the four-cell stage prevented gut differentiation. Placing an early-isolated EMS back in contact with the P₂ cell rescued the gut defect, whereas contact with ABa or ABp did not.

Blastomere recombination experiments have shown that gut induction establishes a functional polarity in EMS (Goldstein. 1993). In a normal embryo p₂ contacts the posterior side of EMS, from which the E blastomere will form. When P₂ is removed, gut docs not form and both of EMS’s daughters take on MS-like lineage timings. This is not an artefact of manipulation, as replacing P₂ after removing it rescues gut specification and the liming defect in E. In an experiment designed to test whether the other side of EMS (the presumptive MS side) can respond to gut induction. P₂ was removed from a four-cell stage embryo and was placed onto the opposite side of EMS. In this case the blastomere that normally would have been E took on an MS-like lineage timing, and gut formed from the descen- dants of what normally would have been the MS blastomere (Eig. 7). The experiment shows that P₂ establishes gut fate in the side of EMS it contacts, and suggests that the two sides of EMS are otherwise equivalent in developmental potential.

Mutants in which the gut does not form (Morton et al., 1992; T. Stiernagle. D. C. Sigurdson. and .1. Shaw, personal communication) may be useful in determining how gut induction establishes a functional polarity in EMS. In these mutants both daughters of EMS take on MS-like lineage timings, as occurs when P₂ is removed.

Gul induction occurs within a single cell cycle, specify- ing gut late in one side of a single cell. EMS. Previous work has suggested that inductions generally take hours or days, and establish the fates of large groups of cells, however most of the work on liming of inductions has been done in verte- brates (Jacobson. 1966; Jones and Woodland, 1987). In invertebrate embryos there are several inductions which may occur quickly and affect the fate of only part of a single cell (Clement. 1962: Nishida and Satoh, 1989), however little is known yet about the timing of inductions in invertebrates.

First cleavage in C. elegans generates two cells, AB and P₁. which differ in size, cell cycle period, cell contents, and cell fates. The events that occur before first cleavage are being studied to elucidate how an asymmetric division is accom- plished and how cellular components are segregated.

Germ-line specific granules, termed P granules, have been detected in C. elegans by immunofluorescence (Strome and Wood, 1982) and by electron microscopy (Wolf et al., 1983). P granules are distributed uniformly throughout the egg’s cytoplasm at the lime of fertilization. During pseudo- cleavage they coalesce and become redistributed posteriorly, where they will be inherited by P₁ at first cleavage. The P granules are inherited similarly by the successive germline precursors P₂. P₃ and P₄. The redistribution of the P granules is achieved as early as prophase of each division. Although the function of P granules is unknown, they serve as a useful model for investigating how cellular contents are asymmetrically partitioned into particular daughter cells.

Microfilament-based and microtubule-based motors are known to move cellular components within cells (Zalokar,1974; Beckerle and Porter. 1983). Strome and Wood (1983) have tested whether microfilaments or microtubules are required for any of the motile events that occur in the first cell cycle. Treating fertilized eggs with inhibitors of micro- tubule polymerization prevented only the migration of the pronuclei toward each other, whereas P granule segregation to the posterior, pseudocleavage, and contraction of the anterior membrane occurred normal. Treatment with microfilament polymerization inhibitors did not affect pronuclear migration, but did prevent signs of asymmetry in the zygote. P granules coalesced but did not segregate pos- teriorly, and pseudocleavage and contraction of the anterior membrane did not occur.

The distribution of microfilaments in oocytes and early embryos has been investigated by Strome (1986). In oocytes and also just after fertilization actin microfilaments are found predominantly in the cortex, as a mesh of libers and foci. As pseudocleavage occurs and the P granules are redis- tributed to the posterior pole, foci of actin concentrate pro- gressively in the anterior side.

In order to determine when microfilaments are required for generating asymmetry. Hill and Strome (1988) pulsed embryos with a microfilament inhibitor (cytochalasin I)) al three distinct times in the first cell cycle. Embryos pulsed for 2 to 10 minutes before psuedocleavage and pronuclear migration went on to segregate P granules and cleave normally. Embryos pulsed during pseudocleavage and pronuclear migration did not show normal signs of asymmetry in the zygote: P granules coalesced but were not segregated posteriorly, and the mitotic spindle formed in the centre of the zygote and either stayed central or drifted slightly in cither direction, leading to a two-cell stage with equal or nearly equal-sized blastomeres. Embryos pulsed after pscudoclcavage (and hence alter P granule segregation) but before cytokinesis maintained the asymmetric distribution of P granules and cleaved normally. Hence microfila- ments are required only briefly late in the first cell cycle to generate asymmetry in the zygote. This is the time that pscudoclcavage occurs, the anterior membrane contracts. P granules are redistributed posteriorly, and pronuclear migration occurs. Microfilaments are not required before this time, nor are they required after this time to maintain asymmetry.

The localization of developmental potential has been investigated by disrupting microfilaments as described above (Hill and Strome. 1988) and then studying the developmental consequences (Hill and Strome. 1990). Pulsing embryos before or after the critical interval did not affect subsequent development. Pulsing embryos during the critical interval led to cither normal development or developmental defects consistent with abnormal establishment of embryonic asymmetry. In those embryos where early cleavages appeared normal, tissue differentiation also occurred normally, whereas in those where early cleavages were abnormal, differentiation rarely occurred. In the abnormally cleaving embryos early cell divisions occurred in a way that suggested cither a duplication of the anterior hall of the embryo (no small P₂-like cell at the four-cell stage, all four cells equal-sized), a duplication of the posterior half (P₂-like cells at both ends), or a complete reversal of polarity (P₂- like cell at the anterior end). The position of the P granules al this stage was consistent with this interpretation, as were the cell cycle times. The finding that cell size, cell cycle limes, and positioning of the P granules were correlated in these embryos suggests that a single microfilament- mediated event establishes embryonic asymmetry, late in the first cell cycle.

Hird and White (1993) have examined the cortex in live embryos during pseudocleavage and found that cortical material flow s anteriorly. This is the time when foci of actin microfilaments found in the cortex become concentrated anteriorly (Strome. 1986) and is also the critical period when microfilaments are required to generate asymmetry in the zygote (Hill and Strome. 1988. 1990). Both the cortical and cytoplasmic flows of pseudocleavage have been found to be microfilament-mediated: flows occur with a velocity con- sistent with actin-mediated panicle movements, and can be prevented by depolymerizing microfilaments but not by depolymerizing microtubules. One reasonable model that all of these observations suggest is the following: during pseudocleavage there is a flow of cortical actin towards the anterior of the egg where it assembles into a contractile network. In order to balance this cortical How. cytoplasm flows in the opposite direction, i.e. towards the posterior. This flow is driven by the continuous contraction of the anterior cortex. This bulk cytoplasmic How could sweep determinants posteriorly where they could be anchored and inherited by the posterior daughter.

Cytoplasmic and cortical flows also occur at the two-cell stage (Hird and White. 1993). In P₁. cortical material flows anteriorly and cytoplasmic material (including the nucleus and centrosomes) flows posteriorly, around the time when P granules are known to become redistributed posteriorly. These flows do not depend on the presence of intact micro- tubules. but may require microfilaments. P₁ then goes on to divide unequally. Such flows do not occur in AB. which undergoes equal cleavage and gives rise to daughter cells with equivalent developmental potential (Priess and Thomson. 1987). In this case, the flows in P₁ may arise from the close proximity of the posterior aster at first cleavage to the cortex.

Hird and While (1993) have shown that many aspects of pseduocleavage can be replicated in later blastomeres. Treating blastomeres with the microtubule polymerization inhibitor nocodazole before cytokinesis causes the mitotic apparatus to be formed close to the cortex: when treated cells attempt to divide, there is a flow of cortical and cytoplasmic material in opposite directions, as in pseudocleavage. The direction of subsequent flows could be predicted by the position of the attenuated mitotic apparatus: cortical material always flowed away from it and cytoplasmic material always flowed towards it. In addition, a transient furrow is formed, the cortex distant from the mitotic apparatus undergoes a series of contractions, and actin becomes heavily concentrated in these contracting regions, as in pseudocleavage (Fig. 8). The results show that an eccentrically placed mitotic apparatus will direct flows in early blastomeres, suggesting that pseudocleavage may be brought about by the asters of the sperm pronucleus (analogous to the mitotic apparatus) inducing a cortical How away from them and a cytoplasmic flow toward them.

Four genes required for cytoplasmic localization have been identified by Kemphues et al. (1988), in screens for non- conditional maternal effect lethal mutations. Each of the/w mutants (for cytoplasmic partitioning) go through many cell divisions and often produce some differentiated cell types, but show defects in the redistribution of P granules and cortical microfilaments (Kemphues et al., 1988; Kirby et al., 1990; Motion et al., 1992). Cleavage orientations and blas- tomere sizes are often abnormal.

Specific defects vary between the four par mutants. In par 1 mutants first cleavage generates two daughters of nearly equal size, and the P granules are distributed equally among early blastomeres. Pharyngeal muscle and body wall muscle both differentiate, but the gut does not. In par-2 mutants P granules are cither incompletely localized or not localized al all. and first cleavage generates two equal-sized daughters. Gul differentiation occurs rarely, while other tissues differentiate normally. In par-3 mutants the P granules are incom-pletely localized or not localized, and microfilaments are not redistributed anteriorly during pseudocleavage. however differentiation of gut and other tissues occurs. In par-4 mutants the cells of the two and four-cell stages are normal in size, however. P granules are distributed equally or nearly equally, and tissue differentiation is widely affected. Gut and body wall muscle cells differentiate only rarely, and pharyn- geal muscle cells differentiate in fewer than half of the embryos. The defects suggest that par gene products are part of a system for intracellular localization.

The sku-1 gene is required to specify the fate of the EMS blastomere (Bowerman et al., 1992). Bowerman et al. (1993) have shown recently that the SKN-I protein is distributed unequally al the two-cell stage. SKN-I is present from just before first cleavage through the eight-cell stage, accumu- lating to much higher levels in the nuclei of Pi and its progeny than in AB and its progeny.

Mothers homozygous for maternal-effect mutations in the mex-l gene produce embryos in which SKN-I accumulates in both AB and P₁ (Bowerman et al., 1993), and the AB lineage produces descendants with MS-like fates (Mello et al., 1992). suggesting that the mex-l gene product is involved in unequally distributing SKN-I to P₁. par-1 (Kemphues et al., 1988) is also required for the unequal dis- tribution of SKN-1 al the two-cell stage (Bowerman et al., 1993).

At the four cell stage both EMS and P2 have high levels of SKN-I, yet mutations in ska-1 affect only EMS and not P₂. SKN-I activity appears to be repressed in P₂ by pie-1. mothers homozygous for maternal-effect mutations in pie-1 produce embryos in which the distribution of SKN-I protein is normal (Bowerman et al., 1993). bul P2 lakes on an EMS- like fate (Mello et al., 1992).

If an early blastomere partitions cytoplasmic components, such as P granules, along a particular axis, then it is crucial that the subsequent division is along the same axis so that only one daughter cell inherits the segregated material. Cells normally divide at right angles to the preceding division as a consequence of the migration pattern of the centrosomes (which form the poles of the spindle) during interphase (Hyman and White. 1987). so an active mechanism is needed to impose alternative division axes. In a number of C. elegans blastomeres the diametrically opposed centrosomes rotate onto the anteroposterior axis prior to the assembly of the spindle. These cells therefore dix ide along the same axis as their mothers.

Hyman and While (1987) studied the rotational choreog- raphy of the centrosomes and the attached nucleus in P₀. P₁. P₂ and EMS (Eig. 9). In P₀. P₁ and EMS the centrosomes rotate through 90 degrees; in P₂ they rotate 45 degrees onto the same axis as the preceding segregation of the P granules. The centrosomes and nucleus in P₀ and P₁ move some distance anteriorly prior to rotation, then one centrosome, apparently at random, leads the rotation with the complex pivoting around the other centrosome. The leading centro- some often ends up very close to a discrete region of the anterior cortex; a membrane invagination is occasionally seen at this site indicating that tension is being exerted on the cortex. Rotation requires the presence of microtubules, and antitubulin immunofluorescence show’s that micro- tubules extend from the centrosomes to the anterior cortex during rotation.

Hyman (1989) studied the dynamics of nuclear/centro- some rotation in P₁ using a laser microbeam. It was shown that disruption of either the rotating centrosome or of microtubules between the anterior cortex and the rotating centrosome stops rotation. Rotation resumes shortly after, but the other centrosome leads. Laser ablation at the membrane invagination also blocks rotation. These results have generated the following model for P₁ centrosome rotation (Hyman.1989): the centrosomes (initially at 90 degrees to the anteroposterior axis) extend microtubules to a capture site on the anterior cortex where their ends are stabilized. The microtubules exert tension on this site, and pull the centrosomal-nuclear complex anteriorly, initially without rotation. A “tug of war” then ensues where one cen- trosome finally establishes a stronger connection to this site than the other. This generates torque, causing one centro- some to swing onto the axis. As centrosomes are diametri- cally opposed to each other on the nucleus, the rotation places the other centrosome along this axis. Microfilaments also play a role in this process, possibly by keeping the cortical site localized to a discrete spot (S. N. H.. unpub- lished observations). Candidates for providing the rota- tional force include kinesin-like microtubule motors located cither al the centrosomes themselves or al the anterior cortical site. There is immunocytological evidence that kinesin-like proteins are present at the centrosomes and at the rotational site in some blastomeres (S. N. IL. unpub- lished observations). Definitive identification of the rota- tional machinery may come from analysis of mutants defective in centrosome rotation (Lesilee Simpson Rose, personal communication).

Work by B. G. was supported by an NSF grant to Gary Freeman and an Nil! training grant at the University of Texas, and an American Cancer Society fellowship al the Medical Research Council. Work by S. N. II. and J. G. W. was supported by the Medical Research Council.