Timing of Tissue-specific Cell Division Requires a Differential Onset of Zygotic Transcription during Metazoan Embryogenesis

Metazoandevelopmentdemandsnotonlyprecisecellfatedif-ferentiation but also accurate timing of cell division to ensure proper development. How cell divisions are temporally coordinated during development is poorly understood. Caenorhabditis elegans embryogenesis provides an excellent opportunity to study this coordination due to its invariant development and widespread division asynchronies. One of the most pronounced asynchronies is a significant delay of cell division in two endoderm progenitor cells, Ea and Ep, hereafter referred to as E2, relativetoitscousinsthatmainlydevelopintomesodermorgans and tissues. To unravel the genetic control over the endoderm-specificE2divisiontiming,atotalof822essentialandconserved genes were knocked down using RNAi followed by quantification of cell cycle lengths using in toto imaging of C. elegans embryogenesis and automated lineage. Intriguingly, knockdown of numerous genes encoding the components of general transcription pathway or its regulatory factors leads to a significant reduction in the E2 cell cycle length but an increase in cell cycle length of the remaining cells, indicating a differential requirement of transcription for division timing between the two. Analysis of lineage-specific RNA-seq data demonstrates an earlier onset of transcription in endoderm than in other germ layers, the timing of which coincides with the birth

Metazoan development demands not only precise cell fate differentiation but also accurate timing of cell division to ensure proper development. How cell divisions are temporally coordinated during development is poorly understood. Caenorhabditis elegans embryogenesis provides an excellent opportunity to study this coordination due to its invariant development and widespread division asynchronies. One of the most pronounced asynchronies is a significant delay of cell division in two endoderm progenitor cells, Ea and Ep, hereafter referred to as E2, relative to its cousins that mainly develop into mesoderm organs and tissues. To unravel the genetic control over the endodermspecific E2 division timing, a total of 822 essential and conserved genes were knocked down using RNAi followed by quantification of cell cycle lengths using in toto imaging of C. elegans embryogenesis and automated lineage. Intriguingly, knockdown of numerous genes encoding the components of general transcription pathway or its regulatory factors leads to a significant reduction in the E2 cell cycle length but an increase in cell cycle length of the remaining cells, indicating a differential requirement of transcription for division timing between the two. Analysis of lineage-specific RNA-seq data demonstrates an earlier onset of transcription in endoderm than in other germ layers, the timing of which coincides with the birth of E2, supporting the notion that the endoderm-specific delay in E2 division timing demands robust zygotic transcription. The reduction in E2 cell cycle length is frequently associated with cell migration defect and gastrulation failure. The results suggest that a tissue-specific transcriptional activation is required to coordinate fate differentiation, division timing, and cell migration to ensure proper development.
Proper development of metazoans depends not only on precise differentiation of cell fate but also on tight control over division timing or division pace between cells, which we refer to as temporal coordination. A normal fate specification without correct division timing may lead to catastrophes, for example, cancerous development (1). Therefore, metazoan development demonstrates stereotyped division timing (2,3). Despite intensive studies on the regulation of cell fate differentiation, genetic control over temporal coordination of cell division during metazoan development is poorly understood. Timing of cell division is particularly critical during early developmental stages such as embryogenesis when cells undergo rapid division and migration, which is concomitant with cell fate differentiation.
Because of technical challenges in quantifying cell division timing especially when an embryo undergoes rapid cell division, most of the studies on cell division timing focus on the earliest stage of development (4). For example, the first embryonic division in Caenorhabditis elegans produces two daughters, namely AB and P1, with differential developmental potential. The two cells also divide asynchronously with the former dividing approximately 2 min earlier than the latter (2). Maternally provided PAR proteins were found to regulate the asynchrony by controlling the timing of mitotic entry and the rate of DNA replication (5,6). Recent developments in live cell imaging and automated cell tracking have greatly facilitated systematic and accurate measurement of cell division timing during entire or partial embryogenesis (7)(8)(9), which allows for identification of genes that control temporal coordination of cell division during metazoan embryogenesis. We have recently performed a reverse genetic screening using the highly conserved and essential genes and looked for those whose perturbation produces a significant reduction in division asynchrony between sibling cells developing into the same or distinct cell types (10). The results demonstrate that both fate asymmetry and asynchrony are controlled by a similar genetic scheme, raising the possibility that division timing could be an integral part of cell fate differentiation. However, in addition to the asynchrony between sibling cells, cell division timing also demonstrates a lineal or tissue-specific dependence. How cell divisions are timed in a lineal or tissue-specific fashion remains poorly understood.
C. elegans embryogenesis provides an excellent opportunity for illustrating the genetic control over the lineal or tissue-specific division timing because of its invariant development. Unlike embryogenesis of vertebrate and Drosophila species where the initial rounds of cell division are synchronized (11), C. elegans embryogenesis demonstrates widespread asymmetries in cell division throughout embryogenesis (2,10). One of the most prominent asynchronies is a significant delay in division of intestine progenitors, E2 (Ea and Ep cells), relative to their cousins from other lineal origins during C. elegans embryogenesis ( Fig. 1). At this stage, cell division is believed mainly controlled by maternal factors (12,13) except for E (14). E is the only lineage that gives rise to endoderm, and the development is clonal, i.e. no other tissue types are derived from the E lineage (2). Average cell cycle length of E2 is ϳ40 min, while that for their cousins of MS lineage (MSa and MSp) is roughly 20 min at 20°C (Figs. 1 and 2A) (10). Previous studies on gastrulation demonstrated that an inhibition of RNA polymerase II activity by its antisense RNA (13) or temperature-sensitive mutation in an uncharacterized protein, GAD-1, resulted in gastrulation failure and precocious division of E2 (15). E fate specification also plays an important role in regulating the cell cycle length. Perturbation of E specification is frequently associated with extra rounds of cell divisions, which is due to an accelerated division pace and fate transformation from E to FIGURE 1. Genetic pathways controlling endoderm specification and cell cycle control. A, summary of current understanding of genetic pathways controlling fate specification (blue) and cell cycle length (brown) of E lineage. E fate specification is under combinatorial control by maternal factors and Wnt signaling pathway as denoted by POP-1, which drives the expression of two redundant E fate specifiers, END-1 and END-3. The two factors in turn up-regulate intestine-specific transcription factors, ELT-2 and ELT-7. Cell cycle lengths are highlighted in brown and green for E2 and MS2, respectively. B-D, differential interference contrast micrographs of a developing embryo before (B) or after gastrulation (C and D). Progeny of E are highlighted.
MS-or C-like fates (16 -19). Single molecular RNA imaging combined with automated lineage study suggested that expression of fate markers in endoderm is independent of its division timing (20). However, a systematic analysis of how the division of E2 cells is specifically delayed compared with those of the remaining cells is still lacking. To this end, we took advantage of a large collection of data on cell cycle length that we generated upon perturbation of 822 essential and conserved genes using automated lineage (10). We found that in contrast to the cell cycle of the other lineages, an E2-specific increase in cell cycle length requires a robust activation of zygotic transcription, which may also be important for proper cell migrations. [end-1 promoter::H1-mCherry ϩ unc-119(ϩ)]) and RW10481 (stIs10226 [his-72 promoter::HIS-24::mCherry::let-858 3Ј UTR ϩ unc-119(ϩ)]; itIs37 [pie-1 promoter::mCherry::H2B::pie-1 3ЈUTR ϩ unc-119(ϩ)]; stIs10436 [hlh-1 promoter::TGF(6.2B4):: GFP::TY1::3ϫFLAG]) as described previously (10). wee-1.1 deletion strain RB669 was used for determining the genetic interaction between wee-1.1 and other genes as stated below. All the animals were maintained on NGM plate seeded with Escherichia coli OP50 at room temperature (25°C).

Materials and Methods
RNA Interference-Gene prioritization was described previously (10). Briefly, only genes that have a clear human ortholog and produce lethal phenotypes upon perturbation were screened for defects in division timing. Genes whose perturbation produced early embryonic arrest were excluded from further analysis. Gene-specific primers were picked, and RNAi was performed through microinjection as described (21). All the primer sequences were deposited in Phenics database.
Imaging, Automated Lineaging, and Gene Expression Profiling-In toto imaging was performed with an inverted Leica SP5 confocal microscope equipped with two hybrid detectors at a constant ambient temperature of 20°C. Images were consecutively collected for both GFP and mCherry channels every 90 s from 41 focal planes for a total of 6 h. For each gene perturbation, images were collected for at least three replicate embryos. Automated lineage and gene expression profiling were performed as described (22). Wild-type embryos were curated up to ϳ350-cell stage for all wild-type embryos and to the equivalent stage or to the last editable stage for all the perturbed embryos by RNAi. At least two replicated embryos were curated.
Manual Measurement of Cell Cycle Length-Time-lapse differential interference contrast imaging was performed in the same way as that for fluorescence imaging. For strains carrying a temperature-sensitive allele, a young adult or L4 animal was raised at 25°C overnight followed by embryo retrieval and imaging. The differential interference contrast images were converted into StarryNite compatible format and fed into Ace-Tree for visualization and manual measurement of cell cycle length by manually tracing the E2 cell from its birth to division.
Statistical Analysis on Cell Cycle Length-Cell cycle lengths for all wild-type and perturbed embryos were computed with StarryNite (7) and deposited in the Phenics database. Average cell lengths for 91 wild-type embryos were previously calculated up to ϳ350-cell stage (10). To identify genes whose perturbation produces a significant deviation from wild-type average, D'Agostino's K 2 test was performed as described previously (10) to evaluate the distribution of individual E2 cell cycle lengths between 91 wild-type embryos. At least 75.8% of all examined lengths passed the normality test with an ␣ value of 0.05, which allowed us to assign the probability of E2 cell cycle length of a perturbed embryo outside the 95 and 99% confidence interval of the distribution of wild-type E2 cell cycle length as the p value of p Ͻ 0.05 and p Ͻ 0.01, respectively. Significant effects are declared only upon observation of at least two statistically significant (p Ͻ 0.05 and p Ͻ 0.01) repeats from perturbation of a single gene.
Hierarchical Clustering of E2 Cell Cycle Length-Genes whose perturbation produced a significant (p Ͻ 0.01) increase or decrease in E2 cell cycle were pooled together to build a matrix consisting of cell cycle lengths. Only genes whose perturbation produced a significant change in at least two embryos were retained for further analysis. Specifically, cell cycle lengths of E2 as well as 12 other cells that are present at the same generation as E2 were collected, including those from MS2 (MSa and MSp), C2 (Ca and Cp), and AB8 (ABala, ABalp, ABara, ABarp, ABpla, ABplp, ABpra, and ABprp). Average fold change in cell cycle length between replicate embryos versus average cell cycle length of 91 wild-type embryos were computed for each gene. The resulting matrix was used as an input for clustering analysis, which was implemented in R with "heatmap.2" function in "gplots" package by using the Ward D "Minkowski" distance measure (Fig. 4B).
Mining and Analysis of C. elegans Blastomere-specific RNA Sequencing Data-Blastomere cell-specific RNA-seq 2 data over developmental time were downloaded from Gene Expression Omnibus (GEO) with accession number GSE50548 produced previously (23). Normalized transcript abundance (number of transcripts per million reads) of each blastomere was retrieved for all genes, which were divided into three germ layer-specific categories based on the tissue-specific SAGE data (24). RNAseq from blastomere E, AB, and MS were used as a proxy for endoderm, ectoderm, and mesoderm, respectively. Developmental time was aligned against E, E2, E4, E8, when E and progeny are available. Alignment to P1 and EMS was used before E was born as described (23).
Data Analysis on Cell Migration-Statistical analysis of division angles and positions was performed as described previously (10). Division angles were computed against three reference planes defined by the following axes, i.e. anterior, posterior (AP)-left, right (LR), dorsal, ventral (AP-DV), and LR-DV, respectively. Normality of the division angles and cell positions in all 91 wild-type embryos were examined with D'agostino's K 2 test. For position, 72.2% of the values passed the test when an ␣ value of 0.05 was used. For division angle, the ratios of normally distributed angles were 93.6, 84.2, and 75.1% against planes defined by AP-LR, LR-DV, and AP-DV, respectively, with a cutoff ␣ value of 0.05. The means and standard deviations were computed for angle distribution of wild-type embryos. A significant deviation from wild-type distribution for the cell migration of a perturbed embryo was assigned a p value of 0.01 and 0.05, respectively, for the probability that falls outside the 99 and 95% confidence interval of the wild-type distribution. Visualization of three-dimensional projection of cell migration was generated as described previously (10).
Genetic Interaction between wee-1.1 and the Components of General Transcriptional Machinery-Allele wee-1.1 (ok418) was crossed with N2 five times to reduce background mutations. The presence or absence of the allele was examined by genotyping using single worm PCR with primers of the following sequences: ok418_ext2-L, ACCGATCTCATGTCCGAA-ATT, and ok418_ext2-R, ATGGCAGCTCACAAACTTGG; ok418_int2-L, GTGTCCATATGCTTCGCGATA, and ok418_ int2-R, ACATACATTCTCCGACGAAATGA. The allele was crossed into the lineage strain RW10226 (8) and rendered homozygous along with lineage markers to give rise to the strain ZZY505, which was used for measurement of cell cycle length with or without perturbation of components of general transcription machinery by automated lineage.

Genes with a Significantly Shortened E2 Cell Cycle Length upon Perturbation Are Predominantly the Components of General Transcription Pathway or Its
Regulators-To identify genes that regulate E2 cell cycle length, we have established a pipeline consisting of gene knockdown with RNAi through microinjection, in toto imaging of C. elegans embryogenesis, and automated lineaging (7,10). We first measured cell cycle length for all cells up to the 350-cell stage for a total of 91 wild-type embryos. We next prioritized a total of 822 genes to be included in the pipeline based on their conservation and reported phenotypes upon perturbation. We performed RNAi by injection on all prioritized genes and screened for those whose perturbation produced a significant decrease or increase (p Ͻ 0.01) in E2 cell cycle length against the average cell cycle length of the 91 wild-type embryos. Only genes with at least two embryos showing a reproducible change were included for the subsequent analysis. In most cases, the strain used for RNAi is RW10425 (25), which expresses PHA-4 as a cell fate marker. Lineaging results were curated up to the 350-cell stage for the wild-type and perturbed embryos. A subset of perturbed embryos was curated to the last editable time point because the perturbed embryos arrested before developing into the 350-cell stage. Cell cycle lengths for all wild-type and perturbed embryos were deposited into the Phenics database.
We classified the significant changes in E2 cell cycle length into two categories, i.e. increase and decrease in cell cycle length. Intriguingly, most genes that produced a significant reduction in E2 length upon perturbation encode the components of general transcription machinery or its regulators ( Fig.  2B and Table 1). For example, a total of 30 out of the 53 genes that show a significant reduction in E2 cell cycle length (p Ͻ 0.01) are those directly involved in general transcription, including two encoding subunits of RNA polymerase II, six involved in transcriptional initiation, one in transcriptional elongation, 17 in RNA splicing, three in poly(A) binding, and one in mRNA export and stability. For example, knockdown of rbpl-1, which encodes a protein predicted to be involved in mRNA poly(A) tailing, shortens E2 cell cycle length by 38.8% ( Fig. 2A). Most of the remaining genes encode regulatory proteins, including five components of Wnt signaling pathway, two transcription factors, and three chromatin modifiers. Involvement of the Wnt pathway in regulation of E2 cell cycle length is not surprising because knockdown of the pathway was found to cause fate transformation from E into MS-like fate, which is coupled with shortened cell cycle lengths (26). ada-2 and cbp-1 are predicted to encode proteins with histone acetyltransferase activity (27), which is important for initiation of gene expression. Also expected are maternal factors known to be required for E lineage specification, including SKN-1, PAR-2, and CUL-1. Depletion of SKN-1 was known to lead to fate transformation from E to a C-like fate (16), and knockdown of the latter two results in transformation from E to a MS-like cell (28).
Cell Cycle Lengths Are Differentially Regulated between E2 and Other Cells-Given a global coupling of developmental speed and transcription timing during C. elegans embryogenesis (20), we wondered whether an inhibition of gene activities leading to a significant reduction in E2 cell cycle length produces a similar effect in other cells. To this end, we first pooled cell cycle lengths for all cells roughly at the same generation as E2 cells for two categories of genes, i.e. those whose perturbation produced a significant decrease or increase (both p Ͻ 0.01) in average E2 cell cycle length compared with that of wild-type average. We next performed clustering analysis for all the cell cycle lengths with fold change of the cell cycle length versus the average of wild type as an input. Surprisingly, nearly all of the perturbations that produced a significant decrease in E2 cell cycle length led to an increased cell cycle length in the remaining cells (Fig. 3A). In contrast, perturbations that produced a significant increase in E2 cell cycle length also caused an increased cell cycle length in the remaining cells. It is intriguing that most of the latter genes are not directly involved in mRNA production but are involved in energy production, protein translation or trafficking, and transcriptional regulation such as transcription factor (supplemental Table S1). The remaining ones include those involved in cell adhesion, cytoskeleton organization, and chromatin modification. However, the scale of fold change in cell cycle length is not comparable between the categories of cell cycle length increase and decrease (Table 1  and supplemental Table S1). It is not clear why interference of zygotic transcription or protein translation produces an opposite effect on E2 cell cycle length.
To illustrate E2 cell cycle control at a systems level, we contrasted cell cycle lengths for all cells in both wild-type and perturbed embryos with two genes. One is rbpl-1 that encodes a putative mRNA poly(A)-binding protein, and the other is wee-1.1 that encodes a homolog of cell cycle-regulating Ser/Thr kinase Wee1 in yeast as well as wee1/2 in humans, where it functions as an integral component of the G 2 /M checkpoint by controlling the dynamics of the cell cycle oscillator (29). The latter is included because a previous study indicated its specific role in E2 cell cycle control (30) and is known to be a zygotic target of MED-1 and MED-2 in the C. elegans embryo (31). As stated above, knockdown of rbpl-1 led to an opposite change in cell cycle length between E progeny (but not E itself), and those from the remaining lineages, i.e. a significantly faster division pace of E progeny but an apparent slowdown in the progeny of remaining lineages compared with the wild type (Fig. 3B), sug-gesting a differential requirement of gene transcription between E progeny and the remaining cells as well as E itself. Such a change is traditionally described as a fate transformation from E into MS-like mostly based on division pattern. However, this

Endoderm-specific Cell Cycle Control in C. elegans
description is apparently not accurate in terms of cell fate (see below). Inhibition of wee-1.1 activities resulted in an E2-specific reduction in cell cycle length (Fig. 3C), which is consistent with the previous results (30). A tissue-specific reduction in cell cycle length upon wee-1.1 perturbation is surprising in that the protein is not only expressed in E at the 12-16-cell stage, but also in AB8 cells in C. elegans (32). This raises the possibility that wee-1.1 may provide some leverage for control of the E2-specific cell cycle, for example by E-specific fate specifiers, but itself may not specifically regulate E2 cell cycle length. Notably, except for wee-1.1, a significant reduction in the E2 cell cycle length was not observed in the genetic hypomorphic or conditional allele of the remaining tested genes whose perturbation gave rise to a shortened E2 cell cycle length by RNAi (supplemental Table S2). This is likely due to an incomplete depletion of their gene product in genetic allele compared with RNAi. Taken together, the results indicate distinct control over cell cycle length between E2 and other cells during C. elegans embryogenesis.

Endoderm Demonstrates an Earlier Activation of Robust Transcription than Other Germ Layers That Coincide with
Elongation of E2 Cell Cycle Length-Given a differential requirement of gene transcription for division timing between endoderm and other germ layers, we hypothesized that an elongated E2 cell cycle could be due to an earlier onset of zygotic transcription in endoderm than in other germ layers. To test the hypothesis, we mined the previous lineage-specific RNAseq data over nematode embryogenesis (23). As expected, the expression data demonstrated an earlier onset of zygotic transcription in E-derived endoderm than in the other two germ layers, i.e. MS-derived mesoderm and AB-derived ectoderm. To align the expression onset relative to E2 division timing, we plotted the average expression level for the genes that were shown to be germ layer-specific based on previous tissue-specific profiling data (33) over developmental time using divisions of E and its daughters as a reference whenever it is available (see "Materials and Methods") (Fig. 4). The data show a substantial increase of zygotic expression in E2 but not in E, suggesting a robust zygotic expression is not initiated in E until its division into E2. In contrast, zygotic expression in the other two germ layers does not reach a similar scale until E8 or E8ϩϩ stage for ectoderm and endoderm, respectively, suggesting a much later initiation of robust expression therein. It is worth noting that the magnitude of expression is much higher in E progeny than those in mesoderm and ectoderm, suggesting initiation of a strong and robust expression program is required for an extended E2 cell cycle in wild-type embryo (Fig. 4B), which is consistent with previous results (19,34).
Regulation of E2 Cell Cycle Length Can Be Coupled with or Independent of Intestine Fate Differentiation-Intestine fate specification and E2 cell cycle extension occur simultaneously during C. elegans embryogenesis. This raises the question of whether the two developmental processes are genetically coupled. Study of E2 cycle length in the presence and absence of wee-1.1 demonstrated that these events are at least partially separable from one another (30). This is based on the observation that a shortened E2 cell cycle length caused by a loss-offunction in wee-1.1 is not accompanied by a loss of expression of intestine markers. We previously showed that a gain-offunction in the cell cycle-promoting phosphatase, CDC-25.1, also produced a reduced E cell cycle length starting from E4 without changing the expression of the intestine marker (3), supporting a separate regulation between the E cell cycle and intestine fate specification.
However, as expected, knockdown of rbpl-1 along with other components of general transcription machinery not only reversed the extended cell cycle length observed in E progeny in a wild-type embryo but also erased fate differentiation as judged by a loss of expression of the fate marker, PHA-4, in both MS and E lineage and early embryonic arrest (Fig. 5) as well as a decrease in lineage marker expression (supplemental Fig. S1), suggesting the regulation of E2 cell cycle length and intestine fate differentiation converges at gene transcription. In contrast, a loss-of-function in Wnt signaling pathway components, for example LIT-1 or MOM-2 (Table 1), resulted in a similar change in cell cycle length of E progeny as that for rbpl-1 knockdown ( Figs. 2A and 3B), but the intestine fate was properly differentiated based on the END-1 expression. The results indicate that the intestine fate differentiation could be either separated from or coupled with the regulation of E2 cell cycle length.
wee-1.1 May Function Upstream of General Transcription Machinery in Regulating E2 Cell Cycle Length-Given that both a loss-of-function in wee-1.1 and inhibition of general transcription or its regulators could lead to a significant decrease in E2 cell cycle length, it would be essential to determine whether these genes function in the same or different pathways. To this end, we first quantified the E2 cell cycle length in the wild-type or wee-1.1 loss-of-function background by automated lineage. We next measured the E2 cell cycle lengths by individually knocking down two components of general transcription machinery in the wee-1.1 null background, including rbpl-1 and D1081.8, which are predicted to be involved in mRNA poly(A) tailing and pre-RNA splicing, respectively (27). We found that the E2 cell cycle lengths were comparable upon knockdown of the two components regardless of the presence or absence of the WEE-1.1 (Fig. 6). However, in both cases the phenotypes of wee-1.1 were masked by RNAi against either rbpl-1 or D1081.8 suggesting that wee-1.1 regulates the E2 cell cycle length upstream of RBPL-1 and D1081.8. CBP-1 is a homolog of the mammalian transcriptional cofactors CBP and p300, which have been shown to possess histone acetyltransferase activity. Knockdown of cbp-1 resulted in a reduction in E2 cell cycle length parallel to the depletion of many components involved in general transcription (Table 1). However, the E2 cell cycle length change caused by CBP-1 depletion was partially rescued by a loss-of-function in wee-1.1 (Fig. 6), suggesting that cbp-1 regulates E2 cell cycle length upstream of wee-1.1 or the two function in redundant pathways in regulating E2 cell cycle length.
Defective E2 Cell Cycle Length Is Commonly Associated with Cell Migration Defects and a Gastrulation Failure-A genetic screen for temperature-sensitive mutants with reversal of gonad handedness identified a gene, gad-1, with its name derived from the gastrulation defective mutation, which resulted in a failure of gastrulation (15). Perturbation of GAD-1 also led to precocious division of E2, suggesting a link between E2 division timing and gastrulation. To evaluate whether the defects in E2 cell cycle length picked up in our screening impair cell migration, we plotted the division angles of Ea and Ep for both wild-type and mutant embryos. Specifically, we first normalized the size and shape for all embryos (see under "Materials and Methods"). We then projected the division angles of all wild types and a subset of mutant embryos in a single three-dimensional space. We observed that for 46 out of the 76 genes with a significant decrease in E2 cell cycle length upon perturbation (p Ͻ 0.05), both Ea and Ep division angles were significantly deviated from the predominantly left-right (LR) axis as observed in 91 wild-type embryos (Fig. 7, A and B). Similarly, we  (Fig. 7E). The positional defects are also manifested as a displacement of Ea and Ep from its starting and ending time point (supplemental Fig. S2). As expected, a significant reduction in E2 cell cycle length is commonly associated with a failure in gastrulation (supplemental Fig. S3 and Movie S1 and S2), suggesting precise division timing is essential for proper E2 migration.

Discussion
How cell division pace is coordinated among one another during the proliferative stage of metazoan embryogenesis is a fundamental question in metazoan developmental biology. However, a molecular understanding of genetic control over embryonic cell division timing has only begun to emerge. Cell division during early embryogenesis has been thought to be mainly under maternal control. Consistent with this, a C. elegans embryo is able to achieve embryonic division up to 87 cells with few effects on asynchrony in the presence of the RNA polymerase II inhibitor, ␣-amanitin (12). However, inhibition of RNA polymerase II activities with its antisense RNA blocks the onset of C. elegans gastrulation and leads to a precocious division of E2 cells (13), suggesting a role of zygotic transcription in controlling division timing of embryonic cells. In agreement with this, activation of zygotic transcription is associated with a loss of cell cycle synchrony at the MBT stage of Xenopus embryogenesis (35). Unlike embryogenesis of Xenopus laevis, Danio rerio, and Drosophila melanogaster, where the initial rounds of cell division are synchronous and bulk activation of zygotic transcription only occurs after cell division begins to slow (11), C. elegans embryogenesis demonstrates abundant asynchronies between cells giving rise to the same or distinct cell type(s) from the very first cell division (2). A candidate approach identified ATL-1 and CHK-1, the C. elegans homologs of ATR (ATM and Rad3-related Protein) and Chk1 check-point kinases, respectively, which contribute to the observed asynchrony (36). We have recently performed a targeted reverse genetic screen with a focus on genetic control over asymmetry between sibling cells, and we demonstrated that a similar genetic architecture is in use for both division asynchrony and fate asymmetry during the proliferative stage of C. elegans embryogenesis (10). However, the asynchronies are not only found between sibling cells but are also observed for specific lineage or tissue. How asynchrony forms in a tissue-or lineage-specific way remains largely unknown.  Here, we provide evidence that tissue/lineage-specific regulation of cell cycle length requires a differential onset of gene transcription, which is supported by lineage-specific RNA-seq analysis (23). Identification of numerous components of the general transcription pathway or its regulators such as those involved in chromatin modification suggests that an earlier onset of transcription in E is triggered by the Wnt signaling pathway, which is likely to be followed by a combination of chromatin modification and clearance of maternal factors that inhibit transcriptional activation (Fig. 8). The requirement of activation of zygotic transcription for a delayed cell cycle is surprising because a general assumption is that active transcription is necessary for cell cycle progression. For example, division of the germ line precursor, P4, is substantially delayed compared with its cousins (2), because it remains transcriptionally quiescent until its division into Z2 and Z3 (37,38), suggesting a robust transcription is necessary for driving cell cycle progression.
A prevailing model on cell cycle control during embryogenesis is the volume ratio between nucleus and cytoplasm termed as the N/C ratio (39). Under this model, a decreasing N/C ratio is responsible for the slowing cell cycle during a later stage of embryogenesis (35,40). This is based on the observation that a decreased N/C ratio is accompanied by a titration of maternal factors that may inhibit the zygotic activation of transcription. Apparently, the model cannot account for the E2-specific cell cycle delay in C. elegans because the cell cycle length and cell volume follow a power law relationship in the AB, MS, C, and P lineages but not in E whose daughters exhibit an abrupt deviation from the common law obeyed by the remaining cells (41). One possible explanation is that delay is evolved to facilitate gastrulation. First, gastrulation occurs during the E2 stage, and E2 cells divide right after completion of ingression (Fig. 1, B and  C). Second, a substantial decrease in E2 cell cycle length in a perturbed embryo is commonly associated with gastrulation delay or failure. Third, mis-regulated E2 cell cycle length is also associated with abnormal E2 division angles (Fig. 7). A previous study on DNA replication and gene transcription demonstrated that the G2 gap stage was introduced during the E2 lifetime (42). We speculate that the specific activation of zygotic expression in E2 cells is probably used for introducing the gap stage, although this seems not the case in zebrafish (43). Consistent with this, many of our perturbed embryos demonstrated an E2 cell cycle length of around 20 min (Table 1), which is roughly the time used for DNA synthesis (42). Future work should investigate the exact relationship between the initiation of zygotic transcription and gap/checkpoint introduction. It is also interesting to study how regulatory pathways, such as Wnt signaling, interface with cell cycle regulation. For example, does the signaling pathway regulate E2 cell cycle length by controlling the cell cycle checkpoint or by regulating the frequency of DNA replication firing through chromatin remodeling or histone modification?
In contrast, evidence shows that cell cycle length also plays a role in zygotic gene activation. For example, chemical inhibition of either DNA replication or the cell cycle progression results in precocious expression of zygotic transcripts in Xenopus (44). Similarly, interphase arrest induces premature transcription at cell cycle 10 in D. melanogaster (45). Transcription and cell proliferation remain synchronized during C. elegans embryogenesis when the overall developmental rate is changed by a temperature shift or a clk-1 (qm30) background. However, expression of fate markers, including a gut marker ELT-2 and a muscle marker UNC-120 or HLH-1, is independent of cell cycle progression in the presence of emb-29 (g52) or div-1 (or148), respectively (20). Whether an increased E2 cell cycle length contributes to the activation of zygotic transcription in C. elegans remains to be determined. Given our criteria in prioritizing genes for perturbation, the genes that are involved in specific control of E2 cell cycle length but produce an otherwise normal embryo upon perturbation could be missed in our screening.
Study of genetic interaction between a component of the general transcription pathway, D1081.8, which is predicted to encode an RNA splicing factor, and a G 2 /M checkpoint component, wee-1.1 suggests that the latter regulates cell cycle by controlling gene transcription. However, the ability of a loss-offunction allele in wee-1.1 in rescuing E2 cell cycle defect pro- FIGURE 8. Model of genetic control on E2 cell cycle length. Wnt signaling along with maternal factors (not shown) triggers zygotic initiation of gene transcription in E through chromatin remodeling and histone modification as well as clearance of maternal inhibitors of transcription, which makes E transcriptionally ready and produces a few E fate specifiers, including END-1 and END-3. Robust gene transcription is up and running starting at E2, which is essential not only for intestine fate specification but also for an extension of E2 cell cycle length probably by introduction of G 2 gap stage, although control of cell cycle length in other cells is mainly under maternal control at the same stage. The controlled cell division timing is presumably important for proper cell migration, but the exact relationship between the two remains to be determined. Known cell cycle regulators involved in the E2 cell cycle control are shown.

Endoderm-specific Cell Cycle Control in C. elegans
duced by CBP-1 depletion indicates that CBP-1 regulates E2 cell cycle by controlling wee-1.1 activities. Given that the CBP-1 homolog in Drosophila was shown to respond to Wnt signaling for turning on gene expression (46), it is possible that the Wnt signaling pathway regulates the E2 cell cycle through CBP-1, which further controls wee-1.1 activities. In agreement with this, although wee-1.1 is expressed not only in daughters of E but also in daughters of AB (30,32), only E2 cells receive the signal. Taken together, we propose a model for E2-specific cell cycle control (Fig. 8), in which Wnt signaling activates E2-specific gene expression by changing chromatin modification, probably through ADA-2, CBP-1, and SWSN-1, which in turn initiates the expression of intestine-specific factors, such as END-1 and END-3 in E. Previous study also revealed a requirement of chromatin regulators for cell cycle progression (9). Expression of the cell cycle negative regulator wee-1.1 may function in synergy with degradation of the cell cycle-promoting phosphatase CDC-25.1 by the SCF (Skr-Cullin-F box) complex in elongating the E2 cell cycle length. Thus, the controlled E2 cell division timing may facilitate its migration in the context of a developing embryo. Although differential expression of F-box that is responsible for CDC-25.2 degradation appears to contribute to its intestine-specific degradation (47), how wee-1.1 plays an E-specific role in regulating cell cycle length remains unclear. It would be intriguing to investigate whether wee-1.1 is a direct target of end-1 and end-3. The exact relationship between E2 division pace and cell migration remains to be determined.
Author Contributions-Z. Z. conceived and coordinated the study and wrote the paper. M. K. W., V. W. S. Z., X. A., and X. R. performed the experiments. M. K. W. contributed to manuscript writing. D. G., R. L., and K. H. C. N. performed data analysis. All authors reviewed the results and approved the final version of the manuscript.