The let-99 gene is required for proper spindle orientation during cleavage of the C. elegans embryo

In organisms ranging from yeast to mammals, certain cells have reproducible patterns of division during growth and development (Kropf, 1992; Jürgens, 1995; Rhyu and Knoblich, 1995; Chant, 1996; White and Strome, 1996). The timing and orientation of division contributes to the spatial organization of cells and thus can be important for morphogenesis, tissue organization and cellular interactions that influence developmental fate (Mello et al., 1994; Reinsch and Karsenti, 1994; Torres-Ruiz and Jürgens, 1994; Smith et al., 1996). Division orientation also plays a role in the segregation of cellular components during intrinsically asymmetric divisions (Strome and Wood, 1983; Chenn and McConnell, 1995; Kraut et al., 1996). In animal cells, division orientation is determined by the position of the mitotic spindle: the cleavage furrow always bisects the spindle (Salmon, 1989). Studies of oriented divisions in invertebrates suggest that astral microtubules interact with the cell cortex to position the spindle (Hyman and White, 1987; Schroeder, 1987; Hyman, 1989; Kawamura and Yamashiki, 1992; McGrail and Hays, 1997). The dependence of spindle position on an intact actin and/or microtubule cytoskeleton has also been demonstrated in algae and budding yeast (e.g. Snyder et al., 1991; Kropf, 1992; Palmer et al., 1992).

The nematode Caenorhabditis elegans is an excellent system for studying mechanisms underlying cell division patterning in a multicellular organism. The pattern of cell division is virtually invariant from animal to animal, and the transparency of the organism facilitates examination of division pattern in living specimens (Sulston et al., 1983). The embryonic division pattern is characterized by a series of unequal divisions. During the division of the one-cell, P₀, the spindle is aligned along the longitudinal (anterior-posterior) axis of the embryo and becomes positioned closer to the posterior pole. First cleavage is therefore unequal, giving rise to a larger anterior cell, AB and a smaller posterior cell, P₁. At second cleavage, the P₁ spindle is oriented on the same axis as the first spindle. At third cleavage, similar orientations are seen in the daughters of P₁ (EMS and P₂). In both cells, the spindle pole is oriented towards the membrane furrow that separates EMS and P₂; the EMS spindle is thus anterior-posterior, but the P₂ spindle is dorsal-ventral due to the cellular arrangements within the eggshell. The divisions of P₁, EMS and P₂ are asymmetric in terms of cell size, the inheritance of cytoplasmic factors and cell fate potential (Laufer et al., 1980; Strome and Wood, 1983). For example, only the P cells inherit P granules. In contrast, the spindle axes of AB and its daughters are perpendicular (transverse) to the previous axes and the divisions are symmetric.

The two patterns of spindle position at second cleavage (longitudinal and transverse) result from differences in centrosome movement (Hyman and White, 1987; Hyman, 1989). In the P₁ cell, there is a 90° rotation of the nuclear-centrosome complex during prophase so that the spindle aligns on the longitudinal axis once again. Rotation is dependent on an interaction between the astral microtubules and an anterior cortical site and also requires actin filaments. Waddle et al. (1994) have shown that actin and actin capping protein transiently localize to the anterior cortex, consistent with a role for these proteins in nuclear rotation. Similar rotations occur in the 1-cell (P₀) and in the daughters of the P₁ cell (EMS and P₂).

Studies of the par genes indicate that the difference in cleavage pattern between the AB and P₁ cells is dependent on the establishment of polarity in the embryo (Kemphues et al., 1988). Mutations in the par genes perturb many of the asymmetries normally present in the 2-cell embryo, such as the posterior localization of P granules and the different cell fate potentials of AB and P₁. Some par mutations also alter spindle orientations at second cleavage. In particular, genetic and molecular analyses indicate that the PAR-3 protein plays an important role in preventing nuclear rotation in the AB cell (Cheng et al., 1995; Etemad-Moghadam et al., 1995). It is unclear whether PAR-3 functions directly or indirectly in preventing nuclear rotation, or how many other factors are involved in controlling the difference in pattern between the AB and P lineages. One such factor may be the G protein β subunit encoded by the gpb-1 gene, which is required for normal spindle orientations in several cells during early cleavage (Zwaal et al., 1996). Thus, while several proteins required for normal spindle orientation have been identified, the precise mechanism of nuclear rotation and its regulation are unknown.

We have taken a genetic approach to identify additional molecules required for orienting spindles in C. elegans embryos. We have isolated several mutations that cause alterations in the early cleavage pattern and here present a characterization of mutations in the let-99 gene. Our observations suggest that the let-99 gene is required for normal spindle orientation in both the AB and P1 cells, and let-99 mutants can show a reversal of the normal pattern of spindle orientations. However, other aspects of asymmetry are normal in let-99 mutant embryos, suggesting that let-99 functions downstream of the par genes and the establishment of polarity in the control of spindle orientation.

C. elegans strains were cultured as described in Brenner (1974). Strains containing the following mutations or rearrangements were used: LGIV: him-03(e1147), egl-23(n601), dpy-4(e1166), dpy-20(e1282ts), unc-5(e152), unc-22(e66), unc-22(s7), unc-31(e169), eDf19, sDf2, sDf21, sDf22, sDf62; nT1[unc(n754) let] IV;V. LGIII: par-2(lw32), unc-45(e286ts), sC1 [dpy-1(e1) let ], sma-4(e729), par-3(it71), qC1. Bristol N2 or unc-22(e66) worms were used as wild-type (let-99+) controls. Strains were obtained from the Caenorhabditis Genetics Center or constructed in the Kemphues laboratory. For all experiments, worms were grown at 20°C, but during videomicroscopy, embryos were at 23–25°C. There were no differences in the phenotypes of let-99 mutants grown at 16°, 20° or 25°C (Table 1).

Maternal effect lethal (mel) mutations were isolated using the procedure outlined in Kemphues et al. (1988), using egl-23 him-3 hermaphrodites mutagenized with 50 mM EMS. From 3317 F₁ individuals, 508 lines that segregated F₂ worms bearing mel mutations were isolated. 221 lines, those in which embryos arrested with large numbers of cells but abnormal morphogenesis, were further analyzed by videomicroscopy for cleavage pattern defects in the embryos. 15 mutations that produced cleavage pattern phenotypes were isolated, including the it141 mutation. The it141 mutation was mapped to chromosome IV using Sequence Tag Site mapping (Williams et al., 1992) and was mapped by meiotic recombination to the unc-22 dpy-4 interval. The deletions eDf19 and sDf62 complemented it141, while sDf2, sDf21 and sDf22 failed to complement. The let-99(s1201) mutation, which maps to the sDf2 sDf22 overlap (Clark et al., 1988), failed to complement it141 for maternal effect lethality. it141 was outcrossed five times in the process of constructing unc-22(e66) let-99(it141)/nT1 [unc(n754) let] IV;V which was the source of let-99(it141) mutants for all experiments. unc-5 unc-22(s7) let-99(s1201) unc-31/nT1 IV;V was outcrossed five times to construct an unc-22(s7) let-99(s1201)/nT1 [unc(n754) let] IV;V strain which was used for experiments after the initial complementation test.

The let-99 alleles behave as maternal effect mutations. Virtually all embryos from let-99 mutant mothers fail to hatch (Table 1). A wild-type copy of let-99 contributed by male sperm does not rescue the embryonic lethality; 0/192 embryos hatched when unc-22 let-99(it141) hermaphrodites were sperm depleted and then mated to N2 males (3 crosses). In addition, virtually all of the embryos from let-99/+ hermaphrodites hatched (519/522 for unc-22 let-99 (it141), 499/509 for unc-22 let-99(s1201) ) suggesting that there is no zygotic effect of this mutation on embryonic viability. The recovery of unc-22 adults from these heterozygous mothers was slightly lower than expected (20.6% for unc-22 let-99(it141), n=519; 19.6% for unc-22 let-99(s1201), n=499; 26.4% for unc-22, n=284). Sperm-depleted N2 hermaphrodites mated to dpy-20 let-99(it141) males produced almost all viable progeny (589 males, 608 hermaphrodites, 7 unhatched), indicating that there is no paternal effect of let-99(it141) on embryonic viability.

Embryos were cut from hermaphrodites and filmed under Nomarski optics using a time-lapse videorecorder and CCD camera. For most observations, embryos were mounted on poly-lysine-coated coverslips in water, then inverted over a slide using additional coverslips as spacers to prevent flattening. The measurements in Table 2 were made from monitor tracings of videotaped embryos. Nuclear-centrosome and spindle orientations were scored from late prophase through metaphase; after this time, cell shape changes can obscure and distort the spindle orientation. For second cleavage, divisions were scored as longitudinal if the spindle was aligned within 0–45° of the long axis of the egg. Spindles at angles greater than 45° were scored as transverse. For third cleavage, orientation was scored with respect to the previous division’s spindle orientation and the resulting division furrow, since cell positions change relative to the longitudinal axis in both wild-type and let-99 embryos. For some observations, embryos were ‘flattened’ by mounting on 5% agar pads under a coverslip.

To determine the let-99 terminal phenotype, intact embryos were incubated on poly-lysine-coated slides at 20°C in a moist chamber until the time of wild-type hatching. For cell ablations, embryos were mounted on agar pads and viewed under Nomarski optics at 100× magnification. A VSL-337 nitrogen laser (Laser Science, Inc.) was used to irradiate nuclei with approximately 50-100 pulses of light until significant debris formed within the nucleus, as described in Mello et al. (1992). After ablation, partial embryos were allowed to develop at 20°C for 16-20 hours in a moist chamber. Embryos were examined for differentiated cell types or tissues using Nomarski or polarization optics as described in Bowerman et al. (1992) and Mello et al. (1992).

Embryos were fixed and stained for immunofluorescence using anti-P granule antibody OIC1D4 or K76, following the protocol of Strome and Wood (1983; provided by S. Strome or the Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA; contract N01-HD-2-3144, NICHD). Immunolocalization of tubulin, PAR-1, PAR-2 and PAR-3 were performed as described in Kemphues et al. (1986), Guo et al. (1995), Boyd et al. (1996) and Etemad-Moghadam et al. (1995), respectively. For all immunofluorescence experiments, embryos were also stained with 4′,6 diamidino-2-phenylindole dihydrochloride (DAPI) to visualize nuclei. Method II of Waddle et al. (1994) was used for anti-CP and anti-actin (monoclonal C4, ICN) staining, since it allows for fixation of small numbers of mutant embryos attached to poly-lysine slides. CP and actin accumulations appeared smaller using this method than with Method I and, in about 50% of embryos (both wild type and let-99), the accumulation was associated with a gap between the AB and P1 cells. In addition, the size of dots and plaques varied from one preparation to another, even in wild type, and so let-99 and wild-type embryos were always prepared in parallel. Division orientation in actin/CP-stained embryos was determined by double staining with anti-tubulin antibodies and/or DAPI. Specimens were viewed and photographed using a Zeiss or Olympus Photomicroscope.

In order to identify genes involved in controlling the orientation of cell division in C. elegans, we generated a collection of maternal effect lethal mutants. Embryos from homozygous mutant mothers were observed during the first few cleavage divisions using time-lapse videomicroscopy. We isolated 10 mutations that produced altered cleavage patterns (excluding alleles of previously known genes such as the pars). The characterization of one of these mutants, it141, is presented here. We mapped the it141 mutation to chromosome IV and found that it failed to complement let-99(s1201) (Clark et al., 1988). Embryos from homozygous let-99(s1201) mothers show the same phenotype as embryos from let-99(it141) mothers. Both mutations act as recessive maternal effect lethal mutations (see Methods) and show the same phenotype in trans to a deletion for the region (Tables 1,²–3). Thus, we believe that it141 and s1201 represent strong loss-of-function mutations in the let-99 gene. The two alleles are virtually indistinguishable in terms of maternal effect lethality and cleavage pattern phenotype (Tables 1,²–3). The observations presented in this report are based on observations of it141 unless otherwise noted. For simplicity, we will refer to embryos from homozygous mutant mothers as let-99 embryos.

To characterize the role of the let-99 gene in spindle orientation, we examined let-99 mutant embryos by time-lapse videomicroscopy and anti-tubulin immunofluorescence. Defects in nuclear positioning and spindle orientation during the first three cleavages were observed. The microtubule networks in let-99 embryos were otherwise indistinguishable from wild-type.

The first cell cycle events of let-99 and wild-type embryos are summarized schematically in Fig. 1 and quantified in Table 2. In wild-type embryos, the female pronucleus migrates to meet the male pronucleus in the posterior of the embryo. The pronuclear-centrosome complex then moves to the center of the embryo and rotates 90° (usually before nuclear envelope breakdown), resulting in mitotic spindle formation on the longitudinal axis of the embryo. The spindle moves and elongates asymmetrically in the posterior direction, and first cleavage is unequal. In let-99 mutant embryos, the pronuclei met in the posterior (Fig. 1F), but failed to move completely to the center of the embryo. In addition, in many embryos, the nucleus did not complete rotation before nuclear envelope breakdown, and the spindle formed at an angle (compare Fig. 1B and G; Table 2). In most embryos, as the misoriented spindle elongated it moved within the cytoplasm onto the correct axis (Fig. 1H). Thus, first cleavage in let-99 embryos produced a 2-cell stage resembling wild type (Fig. 2A,E) except that the cells were slightly more unequal in let-99 embryos (Table 2).

In wild-type embryos at second cleavage, the AB spindle is transverse while the P1 nucleus undergoes rotation and the spindle is longitudinal. At second cleavage in let-99 embryos, abnormal spindle orientations were observed in both the AB and P₁ cells; Table 3 tabulates the different patterns. In approximately half of the mutant embryos, the AB spindle was normal, but in half it was longitudinal (compare Fig. 1D,I). Intermediate positions of the nuclear-centrosome complex were seen in prophase in let-99 mutants, especially in AB, but these tended to resolve into either transverse or longitudinal orientations as the spindle formed. In contrast, in most let-99 embryos, the P₁ nucleus failed to undergo rotation during prophase and the P₁ spindle set up transversely (compare Figs 1D,I, and 2B,F). Thus, in about 40% of let-99 mutant embryos, the second cleavage spindle orientations were reversed with respect to wild type. The reversed pattern usually led to abnormal cellular arrangements (Figs 1J, 2G).

Alterations in spindle orientations were also seen at third cleavage in let-99 mutants. In wild-type embryos, the AB daughter cells have transverse spindles (Figs 1E, 2D). In the AB daughters of let-99 mutant embryos, the spindles were usually oriented with one pole towards the membrane that separates the AB daughters; that is, the spindles were oriented on the same axis as at the previous division (Figs 1J, 2H; 16/17 for ABa, 10/17 for ABp). The ABa and ABp orientations were independent of whether AB divided normally or abnormally at second cleavage. In wild-type embryos, the EMS and P₂ nuclei rotate 90° and 45°, and have longitudinal and dorsal-ventral spindle axes, respectively. In let-99 mutants, the P₁ daughter cell nuclei often failed to undergo rotation and spindles were positioned transversely (compare Fig. 1E,J); this was most easily scored in the anterior P1 daughter (10/15 transverse). The cell cycles of the P₁ daughters were also perturbed (described later). Analysis of later divisions was prevented by abnormalities in cell position caused by the early alterations.

In let-99 mutant embryos, many cells exhibit abnormal spindle orientations, but several observations suggest that spindle orientation is not simply random. First, in the AB lineage, one pole of the ‘longitudinal’ spindles appeared to be orienting to a site on the posterior cortex of AB in a manner similar to that seen in a wild-type P₁ cell during nuclear rotation. Spindle orientation was often late in AB, but we observed clear cases of nuclear rotation during prophase in ABa. Second, although spindle orientation in AB was quite variable, ABa spindles oriented on the same axis as the previous spindle in over 90% of embryos and P₁ failed to undergo nuclear rotation in 90% of embryos. In addition, the different spindle orientations seen in 2-cell let-99 embryos do not appear to be a secondary consequence of the difference in cell size between AB and P₁. We found no correlation between cell size and longitudinal orientation of the spindle in the AB cells or in the few cases where the P₁ spindle was longitudinal (not shown). As a further test for the effects of cell size, we examined let-99 par-1 double mutant embryos. In par-1 and let-99 par-1 double mutant embryos, first cleavage was more equal than in wild type; however, the second cleavage spindle orientations in the double mutant were like those of let-99 single mutants (Table 5). Thus, overall there appeared to be a trend towards a reversal of spindle orientations in the AB versus the P lineages in let-99 mutant embryos.

Time-lapse videomicroscopy also revealed that, in let-99 embryos, the nuclear-centrosome complexes rocked back and forth within the cytoplasm of the P₀, P₁ and AB cells. This rocking action was not restricted to the plane of normal rotation and was different from the movements of wild-type P1 nuclei during rotation. Rocking in let-99 embryos appeared most vigorous during prophase but continued as the metaphase spindle formed. The nuclei in cells at third cleavage appeared to undergo some rocking, but the motion was diminished relative to that in the first two cell cycles. Spindle position in let-99 embryos was also easily perturbed by mounting conditions. It has been observed that longitudinal spindle orientations are less frequent in par-3 mutant embryos that are mounted on agar pads and flattened slightly by the weight of a coverslip (Cheng et al., 1995); slight pressure does not perturb spindle orientations or the development of wild-type embryos. We examined let-99 embryos mounted on an agar pads and found that the AB spindle was longitudinally oriented much less often than in ‘unflattened’ let-99 embryos (Table 5).

The tendency towards reversal of spindle orientation pattern in let-99 mutants does not appear to reflect a more general defect in embryonic polarity. Second cleavage is asynchronous as in wild type, with AB dividing before P₁ (Fig. 1). We also examined the distribution of P granules as a marker for polarity. In wild-type embryos, germ-line-specific P granules become localized to the posterior pole of the embryo during the first cell cycle, and so are partitioned into the P₁ cell (Fig. 3A). The P granules are then segregated to the P daughter at each of the next three divisions (Fig. 3B; Strome and Wood, 1983). In 1-cell and 2-cell let-99 mutant embryos (Fig. 3D), the P granules were localized normally to the posterior (98%, n=81; 100% for wild type, n=37). Due to the abnormal cleavage plane in P₁, however, the subsequent distribution of the P granules in older embryos was abnormal (Fig. 3E). As an additional marker of asymmetry, we examined the distribution of the PAR-1 protein, which is required for normal polarity in the early embryo. PAR-1 protein is localized to the posterior periphery of the 1-cell and the P cells (Guo and Kemphues, 1995). In let-99 1-cell and 2-cell embryos, the PAR-1 protein was localized normally (97%, n=35; 100% for wild type, n=43).

To determine if the asymmetry in cell fate potential produced by first cleavage is normal in let-99 embryos, we carried out laser ablation experiments. In wild-type embryos, the AB and P₁ cells give rise to different cell types, a subset of which can be produced autonomously (Priess and Thomson, 1987; Mello et al., 1992). If the AB nucleus is ablated, the partial embryos produced by the ‘isolated’ P₁ has gut cells, body wall muscle, pharynx, hypodermis and neurons. AB will produce only hypodermis and neurons when P₁ is ablated; production of AB-derived muscle and pharynx requires cell interactions with P₁ descendants. We ablated either the AB or P₁ nucleus in wild-type and let-99 mutant embryos and allowed the partial embryos to develop to term. In most ablated let-99 embryos, isolated P₁ cells produced muscle, gut and pharyngeal tissue, while isolated AB cells produced only hypodermis and neurons (Table 4). Inappropriate cell types were never produced from AB, indicating that cell fate potentials were not reversed in let-99 embryos.

From the examination of polarity markers and cell fate specification, we conclude that the let-99(it141) mutation affects nuclear positioning and spindle orientation, but not other asymmetries. Embryos from let-99(s1201) homozygous and let-99(it141)/sDf22 hermaphrodites showed similar phenotypes with respect to asymmetries in all cases examined (posterior pronuclear meeting, cell cycle asynchrony, P granule localization). Thus, we believe that the let-99 gene is required for controlling spindle orientations after polarity establishment.

let-99 embryos die as masses of differentiated cells, with little or no morphogenesis (not shown). Embryos examined for differentiated tissues had hypodermis and neurons, and many embryos produced body wall muscle (22/26), pharynx (11/26) and gut (17/26). The average number of gut cells in those embryos (20.3±3.9, n=6) was similar to that in wild type (20, Sulston et al., 1983), which was initially surprising. In wild-type embryos, the EMS cell inherits the potential to produce gut cells and pharyngeal tissue, and gut determination also requires induction from the P₂ cell (Laufer et al., 1980; Goldstein, 1993). We expected the failure of P₁ nuclear rotation in let-99 mutants to cause a symmetric division that produced either two EMS-like cells (which might form 40 gut cells) or two P₂-like cells (yielding 0 gut cells), as is seen in some cell fate determination mutants (Bowerman et al., 1992; Mello et al., 1992). In order to determine if the P₁ division in let-99 embryos produced hybrid daughters, both producing half of the intestinal cells, we carried out laser ablations. At the 4-to 6-cell stage in control and let-99 mutant embryos, we ablated the nuclei of all cells except either the posterior or anterior daughter of P₁. The embryos were then allowed to develop until differentiated tissues could be scored. As shown in Table 4, when gut cells were produced by the unablated blastomere in let-99 embryos, approximately normal numbers of gut cells were present. This suggests that all of the intestinal cells derived from one blastomere, rather than both daughters. In addition, there was a tendency towards the anterior P₁ daughter producing the gut cells, as it does in wild-type ablated controls.

Observations of other markers at the 4-cell stage indicated that, in many let-99 embryos, the P₁ division was not symmetric. In 4-cell let-99 embryos, 17.5% of embryos scored had P granules in only one of the P₁ daughters and 42.1% of the embryos appeared to have more P granules in one of the P₁ daughters (n=57). In embryos where one P₁ daughter was positioned more posteriorly, that cell was always the one with more P granules. Similarly, there was more PAR-1 protein on the periphery of the posterior P₁ daughter in 80% of 4-cell let-99 embryos (n=10). Furthermore, the anterior P₁ daughter clearly divided before the posterior daughter in 65% of let-99 embryos (n=37), similar to division of EMS before P2 in wild-type embryos. However, the cell cycle lengths of the two cells were usually much more similar than in wild type and, in the remaining let-99 embryos, appeared synchronous.

Thus, in let-99 embryos, the absence of nuclear rotation correlated with the mispartitioning of several factors, but the P₁ division retained some asymmetry in greater than 60% of let-99 embryos. Since rotation during prophase or metaphase occurred rarely in let-99 embryos (Table 3), these results suggest that the P₁ spindle became more aligned with the axis of polarity later in the cell cycle. A shift in the P₁ spindle position could be due to a late interaction with the normal anterior cortical rotation site, or could result from one spindle pole interacting with the ‘posterior determinant’ that normally positions the spindle asymmetrically, or both.

Given the remnant of asymmetry in the P₁ division in let-99 embryos, this mutant cannot be used to ascertain how an exactly transverse, symmetrical division of P1 would affect the fates of its daughters. However, our results do suggest that mixing some posterior cytoplasm, including P granules, with cytoplasm destined for the EMS cell is compatible with production of intestinal and pharyngeal cell types. This is consistent with previous results showing that the presence of P granules does not preclude the development of somatic tissues (Mello et al., 1992; Strome et al., 1995).

The par-3 and par-2 genes are needed for the establishment of polarity in the embryo, and play an important role in the control of spindle orientation. In wild-type embryos, PAR-3 protein is present around the entire periphery of the AB cell, and on the anterior periphery of P₁ (Etemad-Moghadam et al., 1995). In par-3 mutants, the AB and P₁ spindles are both longitudinally oriented. In contrast, in wild-type embryos, the PAR-2 protein is present on the posterior periphery of P₁, similar to PAR-1 (Boyd et al., 1996). In par-2 mutants, PAR-3 is around the periphery of both AB and P₁, neither cells’ nucleus undergoes rotation, and both spindles have a transverse orientation. Thus, it appears that PAR-3 inhibits nuclear rotation and PAR-2 restricts that activity to the anterior of the embryo so that, in wild type, the AB nucleus does not rotate, but the P₁ nucleus does. In other par mutants, a strong correlation exists between the presence of PAR-3 and absence of rotation (Cheng et al., 1995; Etemad-Moghadam et al., 1995).

To determine if the failure of nuclear rotation in the P₁ cells of let-99 embryos correlated with an abnormal PAR-3 or PAR-2 distribution, we stained let-99 mutant embryos with antibodies directed against these proteins. One-cell let-99 embryos showed the same distribution of PAR-3 as wild type (97%, n=32 versus 100%, n=21 for wild type). The distribution of PAR-3 in the AB and P₁ cells also appeared normal (Fig. 3C,F; 88%, n=16 versus 83%, n=12 for wild type). The distribution of PAR-2 in 1-cell and 2-cell let-99 embryos was also normal (98%, n=42 versus 100%, n=21 for wild type). Mutant embryos did show abnormal distributions of the PAR proteins after second cleavage, in patterns that were consistent with the altered P₁ division plane and P granule inheritance. These results show that the reversed pattern of spindle orientations in let-99 mutants does not correlate with a change in the asymmetry of PAR-3 or PAR-2 and are consistent with the view that let-99 mutations do not alter overall polarity.

To further explore the relationship between let-99 and the par genes, we constructed double mutants using putative null alleles of par-3 and par-2. Embryos produced by par-3; let-99 mothers exhibited characteristics of both single mutants. At first cleavage, the spindle often formed at an angle as in let-99 embryos, but the cleavage was equal as in par-3 embryos (not shown). Nuclear-centrosome complexes in the double mutant embryos exhibited the rocking phenotype characteristic of let-99 embryos. Second cleavage in double mutant embryos resembled that in par-3 embryos: Cleavage was synchronous and both the AB and P₁ spindles oriented longitudinally in most embryos (Table 5). Since the spindle orientations of both let-99 and par-3 single mutants are altered when embryos are ‘flattened’ on agar pads, we also analyzed the double mutants under these conditions. Surprisingly, the second cleavage spindle orientations in the double mutants did not resemble those of par-3 embryos under these conditions; longitudinal spindle orientations were reduced much more in let-99 par-3 embryos than in par-3 single mutants (Table 5). In let-99 par-2 double mutant embryos, cleavages were symmetric and synchronous as in par-2 embryos, and showed the let-99 nuclear rocking phenotype. Interestingly, however, many double mutants embryos exhibited longitudinally oriented spindles in both the AB and P₁ cells, a phenotype seen infrequently in let-99 mutants and never in par-2 mutants (Table 5). In summary, there is no simple epistatic relationship between par-3 and let-99 or par-2 and let-99 and the phenotype of the double mutants is best described as additive.

One explanation for the reversed spindle orientations in let-99 mutant embryos is that some factor required for orienting the spindle is mislocalized in these mutants. We showed above that PAR-3 is localized normally in 2-cell let-99 embryos. Two other proteins that show an asymmetric distribution in the 2-cell embryo and which have been postulated to play a role in spindle orientation are actin capping protein (CP) and actin. CP and actin accumulate in a ‘dot’ on the anterior cortex of P₁ near or at the midbody, as well as over the entire cortex of both the AB and P₁ cells (Waddle et al., 1994). The CP/actin dot marks the cortical site where microtubules appear to attach during nuclear rotation, but a role for CP in rotation remains to be shown. We stained wild-type and let-99 mutant embryos with anti-CP and anti-actin antibodies to determine the location of the cortical dot. The observed staining patterns in both wild-type and let-99 embryos fell into three general classes: asymmetric accumulation, symmetric accumulation and no increased accumulation at the midbody region (Table 6). Embryos were scored as showing asymmetric accumulation when there was a dot or plaque of CP/actin clearly in the P₁ or AB cell. Embryos were scored as having symmetric accumulation when there appeared to be equal amounts of CP/actin on both sides of the membranes between AB and P₁; this usually appeared as a plaque on both sides, rather than as dots protruding into both cells.

Most wild-type embryos in prophase through metaphase showed an asymmetric accumulation of CP or actin in P₁ (Fig. 4A). Many let-99 mutant embryos also had an asymmetric accumulation of actin or CP on the anterior cortex of P₁ (Fig. 4B, Table 6); of those embryos, when spindle orientation could be determined, it was always transverse (n=12). Relative to wild type, an increased proportion of let-99 embryos showed asymmetric AB accumulation (Fig. 4D), symmetric accumulation or no accumulation (Table 6). Of those let-99 embryos in which AB was dividing longitudinally (n=8/28), one embryo had AB asymmetric accumulation, one had symmetric, two had P₁ asymmetric, and four showed no accumulation.

Thus, we conclude that, although the distribution of CP was somewhat abnormal in let-99 mutant embryos, the assembly of CP at the cortical rotation site was not simply reversed, and there was no clear correlation between the accumulation of CP/actin complexes and nuclear rotation in either AB or P₁. If CP/actin accumulation is involved in rotation, then in let-99 mutant embryos it appears that the P₁ CP/actin complex is either inactive, or the microtubules are otherwise prevented from a making a productive interaction with the rotation site in P₁. In addition, since the symmetric accumulations in both wild-type and let-99 embryos did not appear as dots protruding into both cells, this class may actually reflect the instability of CP/complexes under the fixation conditions used (see Methods). In this case, the increase in frequency of let-99 embryos with symmetric or no accumulation of CP/actin could indicate that CP/actin complexes are less stable in this mutant. Otherwise, there was no difference in the size of the plaques and dots seen in let-99 embryos compared to wild type.

Maternal effect lethal mutations in the let-99 gene of C. elegans have dramatic effects on nuclear position and spindle orientation during early cleavage. Spindle orientations in let-99 mutants are abnormal and, in some cases, are reversed with respect to the wild-type pattern, even though several other aspects of embryonic polarity are normal. The orientation defects lead to abnormal cellular arrangements and disrupt cytoplasmic partitioning, both of which probably contribute to the inviability of the embryos. We propose that the let-99 gene functions to control spindle orientation, but not to establish other asymmetries. These characteristics indicate that let-99 is different from the previously characterized par genes and let-99 mutations are among the first mutations that specifically affect spindle orientation to be described in an animal system. The C. elegans gpb-1 gene, which encodes a G protein β subunit, is required for normal spindle orientation in the embryo, as well as for other processes. The cleavage pattern defects of gpb-1 -depleted embryos (flattened on agar) bear some similarity to let-99 mutant embryos (on agar) and P granule localization is initially normal (Zwaal et al., 1996). The similarity in phenotypes raises the possibility that gpb-1 and let-99 function in a common mechanism to control spindle orientation, but a more detailed comparison of the mutant phenotypes is needed.

We speculate that the role of let-99 is to mediate or regulate connections between microtubules and the cortical cytoskeleton. In the P₀, P₁ and AB cells of let-99 mutants, the nuclear-centrosome complex exhibits a rocking motion. This phenotype and the altered spindle orientations could each reflect abnormal interactions between astral microtubules and the cortex of the cell. For example, LET-99 protein could function as a structural component of the cortical cytoskeleton or as part of the linkage between microtubules and the cortex, or LET-99 could regulate such structural components. The observation that ectopic rotations occur in let-99 embryos argues against a role for LET-99 in the function of the cortical rotation site per se, but LET-99 could modify its location or activity.

The par-3 gene has been proposed to play a role in stabilizing spindle position in the early embryo (Cheng et al., 1995; Etemad-Moghadam et al., 1995). PAR-3 acts to inhibit rotation in the AB cell in wild-type embryos. In both let-99 and par-3 mutants, the AB spindle can be longitudinal, but flattening embryos reduces the frequency of longitudinal spindle orientations. Unlike let-99 mutants, however, par-3 mutants usually have nuclear rotation in P₁ and the nuclear-centrosome complexes do not rock. Finally, there is no clear epistasis of the let-99 and par-3 mutations. We conclude that par-3 and let-99 perform similar but not identical roles in stabilization and that they do not appear to function in a simple linear pathway for synthesis or function. However, our observations do not rule out the possibility that LET-99 is downstream of PAR-3 in a localization pathway.

We suggest that let-99 carries out its function in all cells of the early embryo. Why then, does the loss of let-99 activity have different consequences in the anterior versus the posterior cells? The model that we favor is that the asymmetric distributions of PAR-3 and PAR-2 (and/or other factors) in the AB and P₁ cells lead to different consequences upon removal of let-99(+) function. The double mutant results are consistent with this model. In both the let-99 par-3 and let-99 par-2 double mutant embryos, the AB and P₁ cells behave more similarly to each other than in let-99 single mutants. This could be explained in part by the observation that, in par-2 mutants, PAR-3 is symmetrically distributed at the 2-cell stage and, in par-3 mutants, PAR-2 is symmetrically distributed (Etemad-Moghadam et al., 1995; Boyd et al., 1996). The unique phenotype of each of the double mutants is presumably because the absences of PAR-3 and PAR-2 have different effects on the distribution of other factors that influence spindle orientation.

Some evidence argues against alternative models for the let-99 phenotype. For example, one explanation for the let-99 second cleavage phenotype is that spindle orientation is really variable but, because the P₁ cell is much smaller, or because it divides after AB, the spindle orients transversely. This model is inconsistent with the lack of correlation between cell size and spindle orientation in let-99 and the data from let-99 par-1 embryos. Another alternative is that let-99(+) normally localizes a factor required to inhibit spindle orientation to the anterior pole (or vice versa). In let-99 embryos, this factor is mislocalized to the posterior pole in some embryos, producing the reversal phenotype, or is uniformly distributed in some embryos, producing transverse spindles in both AB and P₁. (The let-99 alleles act as loss-of-function mutations, but mislocalization could occur through the action of cytoplasmic streaming or other asymmetries.) However, PAR-3 and PAR-2 were distributed normally in early let-99 embryos. In addition, the accumulation of CP and actin on the cortex, which correlates with nuclear rotation in wild-type embryos, was not reversed in let-99 embryos. The strongest argument against the mislocalization model is that it does not explain the observation that nuclear-centrosome rocking occurs in both the AB and P₁ cells of let-99 embryos.

How are the wild-type patterns of spindle orientation controlled? The simplest model would be that factors required for rotation are localized exclusively to the P₁ cell. Conversely, factors that inhibit rotation, like PAR-3, could be localized to the AB lineage. An alternative hypothesis, alluded to above, is that each cell contains multiple factors that influence spindle orientation. We speculate that a specific combination of factors (and/or their distribution) in P₁ facilitates interactions between microtubules and the cortex that allow for efficient nuclear rotation during prophase and then stabilize the spindle in the longitudinal position. A different combination of factors in AB may mediate very stable interactions which prevent nuclear rotation. This combinatorial model is consistent not only with our observations of let-99, but also with the fact that PAR-3 is present all around the periphery of EMS, yet that cell shows nuclear rotation (Etemad-Moghadam et al., 1995). The molecular characterization of let-99 and the immunolocalization of the LET-99 protein, as well as the study of other spindle orientation genes identified in our screen, will be important for testing these ideas.

We would like to thank Melissa Lamb and Cathy Caron for help with screening for mutations, and Chris Coup, Brenda Cartland, and Mona Hassab for media preparation. We thank the Baillie lab for strains, Margaret Fuller and Susan Strome for antibodies, and Jim Waddle, Su Guo, Lynn Boyd and Bijan Etemad-Moghadam for providing antibodies, advice on immunolocalization, and for communicating unpublished results. We are grateful to Drs Scott Hawley, Mark Peifer, Alan Rose, Jodi Nunnari and the reviewers for helpful comments on the manuscript. This research was funded by a Life Sciences Research Foundation-Burroughs Wellcome Fund Postdoctoral Fellowship and an NIH Postdoctoral Fellowship to L. Rose, and NIH grant HD27689 to K. Kemphues.