Early determination in the C. elegans embryo: a gene, cib-1, required to specify a set of stem-cell-like blastomeres

The development of a multicellular organism requires the generation of many different cell types in the correct spatial arrangement. One widely used strategy to solve this problem is a cell lineage mechanism where a cell’s fate is causally linked to its line of descent (for review see Stent, 1985). Cell lineage not only assures the cell type but also the right cell position to build the embryo. To understand how this is achieved, several problems have to be solved. What is the pathway and structure of the underlying logic of decision making? How are determination and timing of the cell divisions coordinated to assure the correct superimposition of the cell type and cell position in the embryo? What is the nature of the determinative events and the molecular inventory to achieve determination? The genetic ablation of genes involved in the determination of cell fate allows conclusions about the strategy of development and also opens the possibility of initiating a molecular analysis of the events.

We are using a genetic approach to analyze the embryonic development of the soil nematode Caenor-habditis elegans, which follows a lineage mechanism. The complete lineage from the zygote to the adult is known (Sulston and Horvitz, 1977; Deppe et al. 1978; Kimble and Hirsh, 1979; Sulston et al. 1980, 1983). For the work presented here, only the very early lineage is of interest. The somatic founder cells and the germ line of the embryo are produced during the first four stemcell-like unequal cleavages of the blastomeres of the P lineage (P_o-P₃) (see Fig. 5 A). The fertilized zygote P_o divides into the larger somatic founder cell AB and the next stem-cell-like cell P_t. Pi then gives rise to the blastomeres EMS and P₂. The EMS blastomere cleaves unequally into the larger founder cell MS and the smaller founder cell E. P₂ produces the somatic founder cell C and P₃. In the last stem-cell-like division, P₃ produces the somatic founder cell D and the germ line precursor P₄. The very early characteristic cleavages of the embryo can easily be observed using Nomarski light microscopy which has allowed us to screen for embryonic lethal mutations affecting the early lineage of the embryo.

We report here the characterization of 7 allelic mutations defining the gene cib-1 (changed identity of blastomeres) involved in the specification of the stemcell-like fate of the P₁-P₃ cells.

The basic methods of C. elegans culture and handling were as described previously (Brenner, 1974; Wood, 1988).

N2 Bristol was used as the standard wild-type strain. The following mutant alleles and strains were used in the genetic analysis of the cib-1 locus: bli-3 (e767); dpy-5 (e61); him-3 (ell47); unc-11 (e47); unc-13 (e450); unc-29 (el93); unc-32 (el89); unc-38 (e264); unc-73 (e936); dpy-5 (eól); unc-13 (e450); sDp2’, pha-1 (e2123); vab-7 (el562). The mutations are listed in Wood (1988); pha-1 (e2123), H.S., R.S., unpublished. Some strains were provided by the Caenorhabditis Genetics Center.

The mutation cib-1 (e2300) was fortuitously isolated in a screen designed to identify zygotic temperature-sensitive embryonic lethal mutants. The strain him-3 (ell47) was mutagenized with ethylmethane sulfonate following the general procedure of Brenner (1974). F_t progeny (L4 hermaphrodites) of mutagenized worms were placed individually in 24-well microtiter plates (Nunc), which were partially filled with agar. Plates were incubated at 15 °C for two days to allow egglaying. The F₁ worms were individually transferred to new microtiter plates and incubated for one day at 25 °C. The F_t animals were removed and the brood of each animal was analyzed the next day. If a quarter or more non-viable eggs were found at 25 °C and none or only a few at 15 °C, the strain was further analyzed to recover the zygotic temperaturesensitive mutation. The strict maternal-effect embryonic mutation e2300 segregated from such an isolate. The zygotic mutation in this strain was lost. The other mutations in cib-1 were isolated in a non-complementation screen designed to also isolate new alleles in another embryonic lethal gene, pha-1 (H.S., R.S., unpublished). L4 hermaphrodites of the genotype bli-3 (e767) cib-1 (e2300) I; vab-7 (el562) pha-1 (e2123) III were crossed with mutagenized (ethylmethane sulfonate) N2 males. Fi cross progeny (10512) were analyzed as described above. F1 animals producing live progeny at 15°C and not more than 10 viable larvae and many dead embryos at 25 °C were analyzed further for carrying a newly induced cib-1 mutation by a complementation test with homozygous e2300 males. Six cib-1 mutants (e2301-e2306) were recovered in this screen. The non-conditional allele e2303 was subsequently balanced with sDp2.

All cib-1 alleles were tested for complementation and each combination failed to complement. All alleles map within the une-73 dpy-5 interval. The alleles e2300 and e2303 were further mapped into the middle of the unc-38 dpy-5 interval. Map data have been submitted to the Caenorhabditis Genetics Center.

Light microscopy was performed with Zeiss photomicroscopes equipped either for interference contrast or epifluorescence. The DNA quantification was carried out with the Zeiss microscope photometer MPM 200. Embryos were stained with 0.06 % Hoechst 33258.

The procedures of Sulston et al. (1983) were followed for mounting. The temperature of the microscope stage was regulated by using a special stage through which water at 25 °C can be circulated. Since the phenotypes of the conditional alleles depend on the age of the hermaphrodite and temperature, the eggs were collected from hermaphrodites that were shifted to 25°C as L4 larvae about 15 h prior to observation. The temperature of the incubator was regulated to 25±0.2°C. To compare the lineages of cib-1 embryos to those of wildtype, the average slowdown of an embryo was calculated from the delays of the AB cleavage rounds. This was used to correct the other cleavages.

In wild-type embryos, the cleavage times of cells in the early embryo show very little variation, not more than ±3 min. As shown in the results section, the pauses occurring in the P₁ cell cycles of cib-1 embryos vary from very brief time periods (1 min) to the length of a full cell cycle (20 min) depending on the gene dosage. Similar variations occur in cell lineages of the daughters of an equal P cell cleavage. This variation was quantified as follows. After the equal P₁ cleavage, the subsequent divisions of the daughters occurred exactly at the times corresponding to a perfect transformation of P₁ into EMS in 11 embryos from e2303. In e2300, however, the daughters in 4 out of 11 embryos cleaved earlier than expected for a perfect transformation. In 4 out of 9 embryos showing an equal P₂ division, both daughters cleaved very closely to the time when the 2 to 4 C cell cleavage occurs, as expected for a complete transformation. In the other 5 embryos, the cleavage occurred too early for a complete transformation but well after normally the first C cell cleavage is due. In 5 out of 6 embryos that underwent an equal P₃ cleavage, these cleavages occurred within 5 min of the wild-type D cell cleavage time. In one, the equal P₃ division had no delay at all. The most variation is thus seen in alleles of intermediate strength when P₂ is affected. P₂ itself never paused for a full cell cycle, rather only for one third to half of a cycle. This could be explained if the dosage range between an equal P₁ and P₂ cleavage overlap and the dosage required for a very late P₂ cleavage would earliei lead to an equal P₁ cleavage. Usually the two daughters of a given P cell show the same relative length of cell cycle. However, if the daughters cleave at a time point close to the general arrest after the onset of gastrulation, the timing between sisters may vary or no cleavage may occur at all.

There are no discrete quanta! jumps in fates but all intermediates in timing of the events between the wild-type and the changed fate are observed.

Early eggs were collected by cutting gravid hermaphrodites. Terminal mutant embryos were obtained by plating out a large number of LI larvae so that after 48 h at 25 °C egg-laying ceased through starvation of the animals. After a further incubation of 12−15 h, eggs were collected and processed for immunofluorescence following the outline described in Wood (1988) except that eggs were fixed in methanol at 4°C for 15 min and not dried. Primary antibodies were incubated at 4-10°C overnight. Double stains were carried out by repeating the staining procedure using an FITC-conjugated secondary antibody in the first and a Rhodamine-conjugated secondary antibody in the second round. In controls with wild-type embryos, no cross reaction between the first primary and the second secondary antibody was observed.

Embryos were cleavage-blocked and cultured as described by Edgar and McGhee (1988) with the following modifications. Gravid hermaphrodites were collected and transferred into 30 ml of worm cleaner (1.8% basic hypochlorite, 0.25 N-KOH) and shaken for 5imin. The eggs were collected on a 10-mesh nylon filter sealed to an Eppendorf tube which had been cut 5 mm below the lid. The eggs were washed with 20ml egg buffer. The egg shell was removed with a 3 min chitinase chymotrypsin treatment. The enzyme treatment was stopped by washing with 1 ml of trypsin inhibitor dissolved in EGM (embryo growth medium). Eggs were washed with 1ml EGM+2.5μgml^-1 cytochalasin D to equilibrate them in this medium. Washes were carried out by touching a pad of filter paper to the lower side of the nylon filter and by adding fresh medium dropwise to the eggs. Eggs were transferred in 100 μl of medium with a mouth pipette into the siliconized lid of a 0.5 ml Eppendorf tube. Eggs were permeabilized by sucking them three times through a mouth pipette with the bore of a diameter corresponding to about 3 egg diameters. A few eggs were removed to monitor the cleavage block under the microscope. A moist chamber was created by closing the lid with an Eppendorf tube whose tip was filled with 100 μl of sterile water. Eggs were incubated at 25°C for 10 h. The eggs were fixed for 15 min by inverting the lid into an Eppendorf tube filled with 1 ml of ice-cold methanol. Eggs were washed twice with 50% methanol in Tween-TBS and five times in Tween-TBS to remove the cloudy precipitate from the EGM medium. The embryos were then double stained with the monoclonal antibodies used to uncover MS and E fate.

The reference allele e2300 of cib-1 was isolated during a series of screens to isolate EMS-induced embryonic lethal mutations involved in the regulation of development of C. elegans (unpublished). Six more alleles were isolated in a non-compiementation screen from 10,512 EMS-mutagenized chromosomes. The allele e2303 is a non-conditional embryonic lethal mutation; all other alleles, e2300, e2301,’e2302, e2304, e2305 and e2306, are heat-sensitive mutations. Mothers homozygous for any of the heat-sensitive mutations produce viable progeny at 15 °C, whereas at 25 °C all progeny arrest during embryogenesis. All mutations are recessive. They all fail to complement each other and map to the same location on LGI and thus define the gene cib-1, which maps 0.21 map units to the left of dpy-5 (I). All alleles are strict maternal effect lethal mutations. Mothers of the genotype cib-1 /+ produce normal embryos regardless of the genotype of the embryo (cib-1/cib-1, cib-1/ + or +/+). For each allele, more than 400 progeny were counted from cib-1/+ mothers, and between 0 and 0.6% nonviable embryos were found. From large numbers of homozygous hermaphrodites, no hatching progeny were obtained for any of the cib-1-alleles. When crossed with wild-type males, cib-1 hermaphrodites also lay only nonviable eggs. Thus, cib-1/ cib-1 homozygous mothers produce only mutant embryos regardless of the genotype (cib-1 /cib-1 or cib-1 / +) of the embryo. This is a strict maternal effect lethality, since a paternally contributed wild-type copy of the gene cannot rescue the embryonic lethality. In summary, the cib-1 gene product must be supplied by the mother and is required for embryogenesis.

Embryos from mothers homozygous for any of the alleles develop normally up to the 2-cell stage. The first division of the zygote (T=0min) is always unequal, producing the larger somatic founder cell AB and the stem-cell-like germline precursor Pj (Fig. 1 Al, Bl, and Fig. 5). In the wild-type embryo, the next round of divisions is initiated by the AB cell (T=14min), the P₁ cell following 3 min later (T=17min); (Fig. 1 A2). P₁ cleaves unequally to produce the larger EMS and the smaller P₂ cell (Fig. 1 A2, A3). In cib-1 embryos the AB cleavage is delayed up to 2 min (T=16min) (Fig. 1 B2). The first major deviation from wild-type development occurs when P, does not cleave at its normal time but pauses (Fig. 1B2). The length of this pause depends on the allele analyzed. In the two strongest alleles e2303 or e2300, P₁ pauses 20±2 or 19±4min. Thus a 3-cell instead of the normal 4-cell embryo occurs (Fig. 1 A3, B3). At this stage, a wild-type embryo contains AB_a, AB_P, EMS and P₂, the mutant embryo only AB_a, AB_p and Pp After the long pause, the P, cell always divides equally instead of unequally (Fig. 1 B4, Fig. 2 B2). Mothers homozygous for the alleles e2302, e2304 or e2305 produce embryos with two kinds of phenotype. One is the phenotype described above, the other is a shorter P₁ pause followed by an unequal division. The daughter P₂ then divides equally after a brief pause (Fig. 2 B3). Mothers homozygous for the two alleles e2301 or e2306 produce embryos with the two phenotypes already described and a third one where P₁ pauses very briefly. Pi and P₂ divide unequally but P₃ divides equally (Fig. 2 B4). Thus the mutations isolated form an allelic series where the strong alleles affect the P₃ cleavage and in the weaker alleles sometimes the P₂ or P₃ cleavages are affected instead. For a quantitative analysis see Fig. 3A.

In embryos in which Pi pauses for only 1−4 min, P₁ and P₂ will divide unequally and only the P₃ cell will divide equally. A pause of 5−14 min correlates with an equal P₂ division, longer pauses lead to an equal division of the P₁ cell. Thus there exist three discrete thresholds in cell-cycle timing of P_1; which are coupled to the determination of the fates of the P cells.

In addition to their specific effects on the P cells, cib-1 embryos also show pleiotropic effects. From a certain time well after the effect on P_1; the cell-cycle clocks of the whole embryo are slowed down by a factor of about two (for example e2300 f=2.5±0.7; e2303 f=2.3±0.2). The embryos do not divide progressively slower but at a constant rate until the embryos arrest after the onset of gastrulation with about 50 to 100 cells. The time when mutant development deviates from normal development can be estimated by extrapolating back from the later AB divisions to the point of deviation from wild-type development. This point is at about 10 min into the second AB cell cycle.

All processes during the cell cycle are slowed down. The mitoses are prolonged, the midbody persists for an extended period of time and the reformation of the nuclei is delayed and sometimes incorrect. This occasionally causes cells to fuse or to become polynucleate. Cells are often more flaccid than in a wildtype embryo. The morphology of the cells usually improves again as development proceeds.

All P cell cleavages (P_o-P₃) leading to the formation of a somatic blastomere are unequal. In cib-1 mutations, all P cells except for P_o cleave equally. The lack of asymmetry could be an indication that the P cells lost or never acquired a P-cell fate and instead behave like somatic cells. This can be tested using the so-called P granules as markers for P-cell fate.

Strome and Wood (1982, 1983) used monoclonal antibodies to stain the P granules and showed that they always segregate to the P cells and finally to the germline founder cell P₄ (Fig. 2 C1-C4). During the interphase of each P cell, the P granules move into the cell portion that will constitute the next P cell. Segregation is not perfect; some P granules are left behind in the sphere of the future somatic cell. These granules are degraded subsequently. Thus the degradation of P granules marks a somatic fate, while their maintenance marks a P or germline fate.

In cib-1 embryos, the P granules are always segregated into the P₁ cell as in wild-type embryos (Fig. 2 DI), which corroborates our observation that development up to the 2-cell stage is normal. Afterwards the P granules are not segregated into the future P₂-cell domain, and in the following equal cleavages, they are distributed randomly to both daughters (Fig. 2 D2-D4). The P granules are further distributed during the subsequent cleavages and are finally degraded (not shown). This is exactly the pattern expected for cells that have lost or never acquired their P-cell fate and instead behave like somatic cells. An additional clue is given by the fact that in wild-type development the P1-P3 cell cleavage directions are invariant and organized by an active orientation of the centrosomes prior to mitosis (Hyman and White, 1987). The equal cleavages of the P1-P3 cells occur in all possible directions, indicating that the spindles are not oriented as in wildtype development (not shown). This suggests that the P1-P3 cells also never assume this P-cell characteristic. We thus conclude that the cib-1 gene is required to specify P1-P3 cell fate.

Up to now there are no criteria to distinguish the mechanisms by which the different P cells are generated. Is there one in common or are there different ones for the different P cells? The mutations described here never interfere with the asymmetry of the P_o cell. This could either mean that all alleles only reduce the function of the gene and loss of function would also affect P_o or that the strongest alleles are loss-of-function alleles and P_o is specified by a mechanism different from that of the P₁-P₃ cells.

The following arguments support the notion that the two strongest alleles represent the loss-of-function phenotype.

(1) All weaker alleles produce embryos in which several different P cells are affected by a reduced activity of the cib-1 gene product. In the strong alleles, only the P, cell is affected, never P_o. So the spectrum of phenotypes ends abruptly with the P₁ phenotype (Fig. 3 A) suggesting that P_o cannot be affected.

(2) The question of whether a certain mutant allele of a gene causes loss of function can be tested by placing the allele in trans to a deficiency completely removing the gene. The phenotype of a loss-of-function allele should not become more severe when placed in trans to a deficiency (m/Df=m/m). The deficiency sDf4 (Howell et al. 1987) spans the cib-1 locus. The heterozygotes e2300/sDf4 and e2303/sDf4 fulfill the above criteria (Fig. 3B) and therefore the two strongest alleles are likely to represent loss-of-function alleles of cib-1. Embryos derived from mothers of both described genotypes develop normally up to the 2-cell stage, the unequal cleavage of P_o is not, but that of P₁ is affected, P₃ pauses and cleaves equally. That the gene is generally sensitive to gene dosage is shown by the fact that the weakest allele e2306 in trans to sDf4 shows a more severe phenotype, which is very close to that observed in the strongest alleles.

(3) The phenotypes of the combinations e2306/sDf4 and e2306/e2300 are indistinguishable. This is further evidence indicating that the strong allele e2300 behaves very much like a deficiency of the locus and thus should represent the loss-of-function phenotype.

We thus conclude that the strongest alleles of the allelic series very probably represent the loss-of-function phenotype, which is in strong favour of the second hypothesis that the gene product is only required to specify the P₁-P₃ cell fate.

Embryos derived from mothers heterozygous for the strong allele e2300 and the weaker conditional alleles show strong phenotypes. In contrast, the strong nonconditional allele e2303 in trans to all other alleles leads to a weaker phenotype than that of any allele by itself. For each heterozygous genotype, the lineages of 5−6 embryos were followed. In all combinations with e2303, development proceeds either normally up to the onset of gastrulation and then arrests or development slightly slows down after the onset of the fourth cleavage round to arrest after the onset of gastrulation. In the second case, the P₃ cell always pauses for a long period and divides equally. Thus, for example, the two strongest alleles e2300 and e2303 partially complement each other despite the fact that both by themselves behave like loss-of-function alleles. This is an indication of partial intragenic complementation, a genetically well understood phenomenon suggesting that the cib-1 gene product may interact with itself (Garen and Garen, 1963).

The temperature-sensitive period (TSP) of the alleles e2300, e2304, e2306 has been determined by shifting embryos from the permissive to the non-permissive temperature and vice versa (Fig. 4). The results for the other conditional alleles are identical (not shown). Once the oocyte has matured, a shiftdown from 25 to 15°C cannot rescue the embryo. Therefore, the cib-1 gene product must already be functional in the mature oocyte or the function cannot be restored when needed later. In the reverse experiment when the oocytes or embryos are shifted from the permissive to the non-permissive temperature, all stages up to those just initiating gastrulation (between 32 and 64 AB descendants) are affected. Thus the protein is needed only up to the point when all P cells are specified. A later shift to the non-permissive temperature has no deleterious effects, hence it is very unlikely that the gene has a second zygotic function.

The results so far show that the gene cib-1 is required for a brief period in the early development of the C. elegans embryo to specify the Pj-P₃ cells. The absence or reduction of the gene product causes a pause in the P cell cycle and the subsequent loss of P-cell fate. These cells behave like somatic cells. In the following section, the lineages of these somatic cells and their identity will be discussed.

The lineage alterations observed in embryos derived from mothers homozygous for the allele e2300 are considered first. The lineages in cib-1 embryos show a certain variability, as is frequently the case also for mutations affecting postembryonic development (Hor-vitz and Sulston, 1980). For a quantitative analysis, see experimental procedures. Therefore, a single example of a lineage of an e2300 embryo will be described below.

The timing of the cell cleavages of this embryo is compared in Fig. 5B to that observed during normal development (Fig. 5A). In this lineage representation, no cell fates were assigned to the cells after the P₁ division since, as already shown earlier, the P₁ cell does not execute the normal P-cell fate. As a result of the pause in the P₁ cell cycle, 8 of the 16 cells normally derived from P₁ are missing in the mutant embryo at the onset of gastrulation. This corresponds to the loss of one cell division round and shows that the daughters of P₁ resume the normal cell cycle rate relative to the AB cells after the pause of Pi.

Fig. 5C shows that the lineage of Pi after pausing strongly resembles that of its daughter EMS during normal development. This leads to the hypothesis that the mutant P₁ cell executes the fate of its daughter EMS. This hypothesis accounts for all of the following observations: (i) the P₁ cell executes a somatic fate; (ii) it divides at the time when normally its daughter EMS divides during wild-type development and (iii) the posterior two daughters of P₁ sometimes have the morphology typical for the two E cells in the wild-type embryo.

To verify this hypothesis, the following additional points have to be demonstrated, (i) The Pi cell truly pauses for one cell cycle and (ii) the daughters of Pi express markers characteristic for the EMS cell lineage.

2-cell embryos were stained with the anti-tubulin monoclonal antibody YL1/2 (Kilmartin et al. 1982). In wildtype 2-cell embryos, which are ready to undergo mitosis, spindles are visible in both the AB and the Pi cell. However, in cib-1 (e2300) embryos, P₁ is still clearly in interphase while AB has already set up a spindle. The centrioles in Pi are not yet separated and the tubulin staining still shows the typical interphase distribution (Fig. 6). The nuclear envelope does not break down and the chromosomes do not condense. In all these respects, the Pi cell is truly pausing for one cycle, whereas the AB cell proceeds normally.

To exclude the possibility that the cell is still undergoing a S phase, the DNA content (C-value) of P₁ was measured at the end of the pause and after the next cell cycle. The DNA was stained with Hoechst 33258 and the fluorescence was quantified using a microphotometer. The amount of DNA was normalized using P_o metaphase plates and the DNA content in AB cells that had just undergone mitosis and are thus diploid. In cib-1 (e2300) embryos just after AB mitosis and at the end of the P₁ pause, P₁ is still diploid (C=2.4±0.2) and might just have initiated the new S phase. Dividing Pi cells are tetrapioid (C=3.7±0.6). The measured DNA contents in P₁ exclude an S phase during the pause of P_1; which would cause the cell to be tetrapioid after the pause and octoploid before the next division. In summary, we conclude that P₁ is truly pausing in cib-1 (e2300) embryos. During this pause, the cell seems to be blocked in an artificial G₁ phase before DNA synthesis.

The EMS cell founds the MS lineage which produces mostly pharynx and muscle cells, and the E lineage, which gives rise solely to gut (Fig. 5A). Most cib-1 embryos arrest in a morphologically undifferentiated state but, nevertheless, the differentiation of cells can be evaluated in some embryos using antibodies against molecules expressed in the MS or the E lineage which are, at least with low frequency, expressed in terminal stage mutant embryos.

A marker for gut cell differentiation (E lineage) is provided by the monoclonal antibody ICB4 (Okamoto and Thomson, 1985). During early development, only gut cells are stained. Later, three gland cells and several neurons are also stained and can be distinguished from gut cells by their completely different morphology. A combination of markers allows one to test for all major fates characteristic of the MS lineage. The monoclonal antibody 3NB12 stains all pharynx muscle precursor cells and two neurons (Priess and Thomson, 1987). The antibody 48C7 stains all pharynx precursor cells and a subset of neurons, and the monoclonal antibody ICA4 stains only body wall muscle cells (H. Okamoto, unpublished).

Fig. 7 shows the result of double stainings with the antibody ICB4 indicative for E cell fate and with each of the three other antibodies. Terminal embryos from the non-conditional allele e2303 were stained. In the embryos shown, the gut marker specific for E-cell fate is expressed (Fig. 7 A2, B2, C2) and, in the close vicinity, cells express the two different markers indicative for pharynx (Fig. 7 Al, Bl) and also muscle differentiation (Fig. 7 Cl).

The combination of these markers demonstrates unequivocally the presence of the EMS fate in cib-1 (e2303) embryos. The expression of the pharynxspecific markers alone demonstrates the presence of MS fate in those embryos. The AB lineage is the only other lineage producing pharynx cells. However, the determination of pharynx cells in this lineage requires the presence of the MS pharynx precursor cells (Priess and Thomson, 1987).

Since only about 1 % of the terminal e2303 embryos express the E-and MS-specific markers simultaneously, it is necessary to exclude the possibility that those embryos resulted from a very low percentage of animals producing a true EMS cell by an unequal cleavage of P_t. In fact, we never observed a normal 4-cell embryo in hundreds of e2303 embryos but still this formal possibility exists. To confirm that the E and MS cells are directly derived from Pi, early e2300 embryos were cleavage-blocked with cytochalasin D and cultivated overnight to allow expression of the cell-specific markers (Laufer et al. 1980; Cowan and McIntosh, 1985; Edgar and McGhee, 1988). Fig. 7 D-F shows a 6-cell embryo where one cell expresses MS-and another E-cell-specific markers, so the composition of the embryo should be 4 AB descendants, MS, and E. If the MS and E cells were derived from a true EMS cell as in wild-type embryos, seven or eight cells should be present, 4 AB descendants, MS, E, and P₂ or its daughters C and P₃. This is clearly not the case. Furthermore, the MS and E cells in cib-1 embryos are much larger than in wild-type embryos (Fig. 1 A4, B4), since one division has been skipped. In the cleavage-blocked embryo, the cell sizes of MS and E are much larger than in the wild-type embryo and correspond to the cell sizes observed during the lineage of cib-1 embryos. Thus in this embryo the MS and E cell are clearly derived from P₁.

Fig. 7 C3 shows that terminal e2303 embryos also express the antigen for the antibody 48C7 in small cells that resemble nerve cells in the anterior of the embryo very distant from the MS cluster. All but two nerve cells outside the pharynx, which descend from C, are derived from the AB lineage (Sulston et al. 1983). Since the C lineage is completely absent in e2303 embryos, these stained cells show that the AB lineage can differentiate nerve cells and thus that the AB descendants are intact in cib-1 embryos.

Lineage changes in C. elegans have been taken as evidence for fate changes in postembryonic development (Sternberg and Horvitz, 1984). Here, the hypothesis, derived from the lineage analysis, that Pt acquires the fate of EMS after pausing for a full cell cycle has been verified using monoclonal antibodies as probes. As pointed out earlier, in weaker alleles of cib-1, P₂ and P₃ also cleave equally and behave like somatic cells. What are the fates of these cells?

After an equal P₂ cleavage, its daughters only divide at a time when the two C cells divide in a normal embryo. Furthermore, both cells divide equally, whereas in a wild-type embryo one of them, P₃, divides unequally (Fig. 5 D). An equal P₃ division occurs at a time when its daughter D should divide in a normal embryo. For a discussion of the observed lineage variability, see experimental procedures. By analogy to the transformation observed in the P₁ cell, we argue that P₂ in those embryos is likely to be transformed into C and P₃ into D. Thus the P₁-P₃ cells after losing their P-cell-like fate, pause and afterwards appear to execute the fates of their somatic daughters.

As already pointed out, in cib-1 embryos the P₁-P₃ cells lose their asymmetry and the active reorientation of their cleavage direction. It is of interest whether the P₁-P₃ cells are also important in establishing other asymmetries in the embryo. Schierenberg (1987) has suggested that the presence of P₂ is required for proper E-cell differentiation. In cib-1 embryos that completely lack a P₂ cell, gut-specific markers can be detected by monoclonal antibodies. This shows that differentiation of E cells is possible even in the absence of a P₂ cell.

However, we observed that in some embryos the posterior daughter of Pi behaved like MS and the anterior one like E, which indicates that the orientation of the MS/E decision might be randomized in absence of a P₂ cell. Thus, P₂ does play a role in the execution of EMS fate as suggested by Schierenberg (1987).

In cib-1 embryos executing an equal P₂ cleavage, EMS cleaves equally (MS:E=49.3: 50.7±1.8) in contrast to the unequal cleavage of EMS in wild-type embryos (MS:E=53.7:4fi.3±3.2). These observations suggest that P₂ is necessary to establish not only the interior polarity but also the physical asymmetry of EMS.

The described mutations of the gene cib-1 result in a recessive strict maternal effect, embryonic lethality, indicating that the expression of the cib-1 gene product is only necessary for embryogenesis. There is most probably no zygotic function for the gene since the allele e2300 complements the deficiency sDf4 at 15 °C and thus the non-complementation screen used to isolate new alleles should have picked up any loss-of-function allele. The finding that all conditional alleles have temperature-sensitive periods that extend from shortly before fertilization through the first six cleavage rounds of the embryo suggests that the gene product itself is incorporated into, and functions within, cells of the early embryo.

We propose that the cib-1 gene product is required for the specification of the P1-P3 cells in a dosagedependent fashion. The activity of the gene product is first required in the 2-cell stage when the Pi cell is bom and mutant development begins to deviate from that of wild-type embryos. The development of the whole embryo slows down by a factor of two until it arrests after the onset of gastrulation. This slowing down of development is very probably not the cause of the lineage alterations since other mutants we have isolated show the same slower development without or with very different lineage alterations to those in cib-1 (R.S., H.S., unpublished observations). These mutations, like cib-1, cause embryos to arrest around the onset of gastrulation. Thus this arrest seems to be a rather general phenomenon and could be explained by a general failure to activate zygotic transcription. Indeed it has been found that the transcriptional activation of the C. elegans genome occurs around the onset of gastrulation (Hecht et al. 1982) and that inhibition of transcription with alpha-amanitine causes an arrest of development at this time (L. Edgar, personal communication).

So far we and others (Denich et al. 1984) have not been able to isolate mutations affecting early development that do not also slow down development. The only exceptions are the 4 par genes (Kemphues et al. 1988) which were specially selected for not displaying the pleiotropies described. Because of the general occurence of the described effects, we tend to believe that the early C. elegans embryo is very sensitive to alterations of the early specification process and that this causes the pleiotropic effects. This working hypothesis can only be tested by characterizing genes like cib-1 at the molecular level to learn more about their functions in development.

We have postulated above that the effect of cib-1 mutations on the cell cycles of the AB-and the other lineages is an indirect one, since the AB cell is not affected at all in cib-1 mutants and the effects on the AB descendants only occur after Pi has been affected, and also since the capability of these lineages to differentiate into their terminal fates is not altered. The other possibility is that cib-1 has a function in all early cells, for example, it could be necessary to drive the embryonic cell cycle. If it were an essential cell cycle gene, the null phenotype should be a block of already the first cell cycle; this is not the case. There is a possibility that the gene is not essential for running the cell cycles, only for their precise regulation. In this case, the fate changes observed would be secondary alterations caused by changes in cell cycle timing. To our knowledge, there is no precedent for this. So far, it has only been shown in C. elegans that the lack of an S-phase can suppress the later zygotic expression of certain cell-specific markers after founder cell formation (Edgar and McGhee, 1988); this means that an S-phase is required for the establishment of the determined state, but not necessarily for the determination per se. That cib-1 primarily affects embryonic cell cycles cannot be ruled out completely, but the following arguments speak against this hypothesis, (i) As already mentioned above, the AB cell is never affected in cib-1 embryos, which could be a threshold problem, i.e. AB could have a lower requirement than P_k (ii) One would, however, expect that daughters of P cells, which paused because the gene product necessary for cell cycle is low or absent, should pause as well. This is not observed, if anything, cell cycles are too fast (see lineage variability), (iii) If the cell-cycle defect is the primary defect in cib-1 mutants, one would have to postulate a special sensitivity of the P1-P3 cells to explain the observed fate changes; this cannot be excluded, but seems rather unlikely.

Also, the very specific early temperature-sensitive phase of the conditional mutations shows that the gene has no general cell lethal function, e.g. in cell proliferation during or after embryogenesis.

The removal of the most posterior cytoplasm of P_o or P, causes Pi to behave like a somatic AB cell (Schieren-berg, 1988), a result consistent with the hypothesis that prelocalized molecules, so-called determinants (for review see Davidson, 1986) are involved in the specification of the P-cell fate. Mutations affecting the P-cell character may identify genes coding for such a determinant or for molecules involved in the interpretation or localization of such determinants. We have identified two more genes with phenotypes similar to that of cib-1 (unpublished).

The logic of decision-making during early development Mutations in the par-1 to -3 genes cause P_o to cleave equally and the posterior daughter, which normally becomes P|, behaves in some aspects like a somatic AB cell (Kemphues et al. 1988). These par embryos also fail to segregate the P granules, an indication that these genes affect the partitioning in P_o. In par-4 an unequal P_o cleavage occurs, but the smaller posterior cell often does not execute a P-cell fate. Instead it shows an equal cleavage pattern like a somatic AB cell. The lack of P-granule segregation in P_o of par-4 mutants shows that, as in the other pur-mutants, the mutation affects P_o. In contrast, cib-1 mutants have no effects on P_o, but only affect P1-P3. Since P_o is defective in par-4 mutations, all determinative inventory might be removed from P_t allowing the cell to behave like an AB cell. In cib-1 embryos, however, the inventory necessary for the determination of the somatic fates EMS, C and D is still present, repressing the acquisition of an AB fate. Only the inventory to establish the stem-cell-like P fate is incomplete preventing the initiation of a cell cycle and thus the production of the cells necessary to segregate somatic and stem-cell-like fates. Since a P fate is absent, the underlying somatic fate is executed (Fig. 8). Thus the P-cell identity is not a necessary intermediate state in the specification of somatic cells.

During the development of many organisms a series of synchronous proliferative cell cycles is followed by differential or asynchronous cell cycles, which mark the onset of differentiation. The early cleavages are run by a cytoplasmic free-running oscillator (for review see Kirschner et al. 1985). In C. elegans, cell cycles have different lengths from the 2-cell stage on. Manipulating the embryo with a laser microbeam, Schierenberg and Wood (1985) showed that there is a cell cycle clock located in the cytoplasm, which is independent of the cytoplasm:DNA ratio. Because in cib-1 mutants a subset of cells is able to pause, the cell cycle oscillator cannot be free running but is instead directly coupled to the cell fate.

Based on a number of experimental systems many mechanisms have been proposed for the correct timing of determination events.

• - Cytoplasm:DNA ratio (Kobayahawa and Kubota, 1981; Newport and Kirschner (1982a,b\Mita, 1983; Mita and Obata, 1984; Edgar et al. 1986).

• - Counting of replication rounds (Satoh, 1979; Satoh and Ikegami, 1981a,> Satoh, 1982a,b; Mita-Miyazawa et al. 1985; Schierenberg, 1985; Nishikata et al. 1987).

• - Quanta! cell cycle. Holtzer et al. (1983) propose that cells are determined by passing sequentially through special rounds of DNA replication.

• - Clock mechanism. Dan and Ikeda (1971) demonstrated in sea urchins a cell-cycle-independent developmental clock timing a morphogenetic event, the unequal cell divisions, independently of the cell cycle rounds that occurred previously. Gastrulation in amphibians also seems to be initiated by a developmental clock (for review see Kirschner et al. 1985).

The analysis of the cib-1 mutations suggests that only the last mechanism can be important in the early determination of the C. elegans embryo. Since the P₁ cell pauses for a cell cycle, all daughter cells are almost twice the normal size. This changes the cytoplasm: DNA ratio significantly without preventing the later determination of MS-and E-fate.

A counting mechanism is excluded, since one complete cell cycle is suppressed and the MS and E specification takes place one cycle too early. There cannot be a special quantal cell cycle in P₁ to switch fate to the EMS identity, since there is no S phase at all in cib-1. In other systems, it has also been shown earlier that quiescent cells can change fate (e.g. Blau et al. 1985), so the quantal cell cycle seems not to be a general phenomenon.

Our observations can best be explained by assuming that two independent clocks govern the events in the early nematode embryo. One cell cycle clock is responsible for the differential cell cycles observed in the embryo, and is controlled by the determinative state of a cell. A second developmental clock activates the determinative inventory, whatever its nature, at the right time. The first clock explains the pause after the failure to specify a given P cell and the reactivation of the sublineage-specific cell cycles after a determination takes place. The second developmental clock must be postulated to account for the precise activation of the underlying somatic daughter fate in the pausing P cell. Thus not only morphogenetic events but also determination can be timed by a cell-cycle-independent developmental clock. This developmental clock does not necessarily have to measure time in a temperaturedependent manner. It could also be related to cellular events that cannot be observed with the given methods.

Edgar and McGhee (1988) investigated the role of DNA synthesis during the establishment of tissuespecific transcription by blocking DNA synthesis with aphidicolin at specific stages in the early embryonic development of C. elegans. They conclude that neither the cytoplasm:DNA ratio nor the slowing down of the cell cycle are sufficient to activate zygotic expression. They could not strictly exclude a counting mechanism or a series of quantal cell cycles, which are excluded by our findings. They demonstrated that DNA synthesis must occur in a rather precise time window after the clonal establishment of a tissue and thus postulate the presence of a second cell-cycle-independent developmental clock to explain this result. This is in agreement with our findings. In addition, we find that the determination process also depends on a developmental clock.

We thank T. Bogaert, W. Driever, J. Priess, D. St. Johnston, J. Sulston, and O. Sundin for comments on the manuscript. We are indebted to L. Edgar for advice on the cleavage-block experiments and for a gift of EGM medium. We thank D. Miller and S. Strome for monoclonal antibodies and H. Scharf (Zeiss) for help with the microphotometer measurements. The authors were supported by Otto-Hahn-Fellowships of the Max Planck Gesellschaft and by EMBO Fellowships. We thank F. Bonhoeffer and Ch. Niisslein-Volhard for support. This work was supported by a grant of the Deutsche Forschungsgemeinschaft.