Cell cycle progression is required for zebrafish somite morphogenesis but not segmentation clock function

Somites are the segmented precursors to the axial skeleton and musculature created as the trunk and tail elongate. The periodic formation of somites is governed by the segmentation clock, which creates oscillations in gene expression in the presomitic mesoderm (PSM)(Pourquié, 2003). In zebrafish, the segmentation clock requires Notch signaling, while the amniote clocks also incorporate Wnt and Fgf signaling(Holley, 2007). It is debated whether the Notch, Wnt or Fgf pathways constitute core components of the clock, or whether they are a readout of a global clock that governs all of embryonic development (Aulehla et al.,2003; Dequeant et al.,2006; Niwa et al.,2007; Wahl et al.,2007). For instance, some models link the segmentation clock to the cell cycle oscillator (Collier et al.,2000; McInerney et al.,2004; Primmett et al.,1989; Primmett et al.,1988).

Somite morphogenesis occurs as the segment boundary cells undergo a mesenchymal to epithelial transition (MET), forming a ball of cells with an epithelial surface and a core of mesenchyme(Holley, 2007). Zebrafish somite morphogenesis requires the transcription factor fused somites(fss; also known as tbx24), Eph/Ephrin signaling and integrin α5/fibronectin function(Barrios et al., 2003; Durbin et al., 1998; Durbin et al., 2000; Jülich et al., 2005a; Koshida et al., 2005; Nikaido et al., 2002; van Eeden et al., 1996). fss links the segmentation clock and somite morphogenesis(Holley et al., 2000). fss mutants fail to maintain the segmentation clock in the anterior PSM, lack segment polarity and ephA4 (epha4a - Zebrafish Information Network) expression, and exhibit a complete loss of MET in the paraxial (somitic) mesoderm (Durbin et al.,2000; Holley et al.,2000; Oates et al.,2005b; van Eeden et al.,1998). However, exogenous expression of ephA4 in genetic mosaics can induce boundaries in fss^-/- embryos(Barrios et al., 2003). Similarly in mouse genetic mosaics, EphA4 expression correlates with boundary formation (Nakajima et al.,2006). Integrin α5-GFP clusters along the basal side of nascent somite boundary cells, and integrin α5 and its ligand fibronectin are required for the maintenance and full maturation of the boundary in zebrafish, mice and Xenopus(Georges-Labouesse et al.,1996; Goh et al.,1997; Jülich et al.,2005a; Koshida et al.,2005; Kragtorp and Miller,2007; Yang et al.,1993). Double mutants between integrin α5and the Notch pathway lead to a complete loss of MET in the paraxial mesoderm(Jülich et al., 2005a). Simultaneous loss of ephrin B2a, a ligand for ephA4, and integrin α5 leads to a synergistic defect in somite boundary morphogenesis (Koshida et al.,2005). Ena/Vasp and Fak, which function in Integrin signaling, are necessary for somite formation in Xenopus(Kragtorp and Miller, 2006). Chick somite morphogenesis is regulated by Snail2 and Cdc42, which promote mesenchymal cell morphology, and Rac1, which fosters epithelial cell morphology (Dale et al., 2006; Nakaya et al., 2004).

Emi1 is a negative regulator of the Anaphase Promoting Complex (APC) and is required for entry into mitosis in Xenopus embryos(Reimann et al., 2001). APC,an E3 ubiquitin ligase, also functions in post-mitotic cells. In Drosophila and C. elegans neurons, APC localizes to the synapse and regulates the turnover of glutamate receptors(Juo and Kaplan, 2004; van Roessel et al., 2004). In vertebrate neurons, inhibition of APC by RNA interference or overexpression of Emi1 increases axonal growth and overcomes much of the growth-inhibitory effects of myelin. In contrast to the synapse studies, virtually all of the APC is located in the nuclei of these neurons, and the axon growth phenotype appears to be due to stabilization of Id2 and SnoN(Lasorella et al., 2006; Stegmuller et al., 2006).

Here, we identify a zebrafish mutant for emi1 that ceases mitosis at the beginning of gastrulation. Using this mutant, we find that normal cell cycle progression is not required for segmentation clock function, but rather that mitosis is a modest source of noise for the clock. Finally, we show that the cell cycle defect leads to hyper-epithelialization of the somites after the initiation of morphological segmentation.

Breeding and meiotic mapping followed standard protocols(Geisler, 2002; Nüsslein-Volhard and Dahm,2002). The coding sequence of emi1 (GenBank NM_001003869)was isolated via RT-PCR and cloned into pCS2+. This clone was used to generate sense mRNA using the Ambion SP6 mMessage Machine kit, and an antisense riboprobe using the Roche digoxigenin-labeling mix. For allele sequencing, we used an emi1 template from two independently derived tiy121RT-PCRs. Wild-type embryos were injected with 0.5 mM emi1 morpholino targeting the splice donor of the second intron(5′-TGATTGTCGTTTCACCTCATCATCT-3′).

Fibronectin, phalloidin, S58 staining(Jülich et al., 2005a),and fluorescent in situ hybridization with β-catenin immunohistochemistry(Jülich et al., 2005b),were performed as previously described. All in situ hybridizations were performed with digoxigenin-labeled riboprobes. her1 and deltaC antisense probes were made from plasmid clones, as previously described (Holley et al.,2000; Holley et al.,2002). The tbx18, mesogenin, mespb and ripply1coding sequences were isolated via RT-PCR and subjected to an additional round of PCR in which a T7 promoter was added in the antisense orientation. Antisense riboprobes were then created using T7 RNA polymerase (NEB). Integrinα5-GFP (Jülich et al.,2005a) and YFP-Emi1 were visualized with rabbit anti-GFP (1:1000,Invitrogen) and anti-rabbit Alexa 488 (1:200, Invitrogen). Goat anti-EphrinB2(1:500, R&D Systems) was paired with anti-goat Alexa 647 (1:200,Invitrogen). Rabbit anti-Phospho-Histone H3 (PHH3) antibody (1:1000, Sigma)was used with goat anti-rabbit-HRP (1:400, Invitrogen) and Fluorescein TSA(Perkin Elmer).

We analyzed her1 expression in emi1 mutant and sibling embryos injected with translation-blocking morpholinos against either deltaC or deltaD (Holley et al., 2002). Three independent trials were performed with embryos derived from different parents and injected on different days. deltaC morpholino-injected and deltaD morpholino-injected embryos were fixed in 4% paraformaldehyde (PFA) at the ∼2-somite and∼5-somite stages, respectively. Embryos were co-stained for her1expression with NBT/BCIP and for PHH3 by immunofluorescence. Absence of PHH3 staining was used to sort emi1^-/- from sibling embryos.

Embryos were incubated in 150 μM aphidicolin and 20 mM hydroxyurea(Sigma) in 4% DMSO, from the germ ring/early shield stage until fixation(Harris and Hartenstein, 1991; Lyons et al., 2005). Drug treatment at this stage blocked mitosis by the late shield stage, mimicking the onset of the emi1^-/- phenotype. To assay for DNA synthesis, 10 mM BrdU was injected into the yolk just after the shield stage,at the 1-somite stage or at the 8-somite stage. Embryos were fixed in 4% PFA at the 14- to 15-somite stage. BrdU incorporation was visualized using a mouse anti-BrdU antibody (1:200, Sigma) and an Alexa 647-labeled goat anti-mouse antibody (1:200, Invitrogen). Embryos injected at each stage showed BrdU incorporation, indicating that endoreplication occurs continuously during late gastrulation and trunk segmentation in emi1 mutants.

We identified a zebrafish mutant, tiy121, which exhibits a mitotic block (Fig. 1A-D). By the shield stage, mutant embryos cease all mitosis, as visualized by immunostaining for phosphorylated Histone H3. Despite the mitotic arrest,mutant embryos undergo gastrulation and axis elongation(Fig. 1E,F). Measurement of the distance from the otic vesicle to the tip of the tail indicates that tiy121 embryos (n=15) are on average 22% (s.d. ±3.2%)shorter than their wild-type siblings (n=17). After the mitotic block, mutant embryos continue endoreplication, as indicated by BrdU labeling(Fig. 1G,H). tiy121embryos ultimately develop a pericardial edema and extensive necrosis in the head, and die 2-3 days post-fertilization. The relatively normal progression of early development in tiy121 embryos parallels the finding that early Xenopus development is unperturbed by the chemical inhibition of mitosis (Cooke, 1973; Harris and Hartenstein, 1991; Rollins and Andrews,1991).

We mapped tiy121, via meiotic recombination, to chromosome 13 between the simple sequence length polymorphisms (SSLP) z24268 and z55656,near a zebrafish homolog of early mitotic inhibitor 1 (emi1)(Fig. 2A). Determination of the emi1 coding sequence in the single mutant allele revealed a premature stop codon that truncates the protein prior to the F-box domain, which is likely to create an amorphic allele (Fig. 2B). Injection of a splice-blocking morpholino against emi1 recapitulates the mitotic defect through gastrulation. However,the antisense inhibition declines by the tailbud stages, and mitosis is normal by the 5-somite stage (Fig. 2C). In morpholino-injected embryos, phosphorylated Histone H3 staining (PHH3) was absent (89%) or reduced (9%) at the shield stage(n=101), was reduced in 79% of embryos at the tailbud stage(n=39), and was indistinguishable from controls at the 5-somite stage(n=37). Injection of in vitro synthesized mRNA for YFP-emi1 rescues the mitotic defect through gastrulation,but the rescue declines by the tailbud stage and is absent by the 18-somite stage (Fig. 2C). YFP-emi1 mRNA (5 ng/μl) was injected into clutches from tiy121^+/- parents. Normal PHH3 was seen in 99%(n=75) and 98% (n=141) of embryos at the shield and tailbud stage, respectively. By the 5-somite stage, 16% (n=110) showed reduced PHH3, and at the 18-somite stage, 27% (n=153) showed no PPH3. tiy121 fails to complement hi2648, a hypomorphic, retroviral allele of emi1 (Fig. 2D) (Amsterdam et al.,2004). Together, these data indicate that the tiy121phenotype is due to perturbation of emi1.

emi1 mRNA is maternally deposited(Fig. 2E), and later ubiquitously expressed in the blastula and gastrula stages(Fig. 2F). Emi1-YFP protein localizes to the nucleus but is diminished in cells undergoing mitosis(Fig. 2H-J). During segmentation, emi1 is broadly expressed, including within the somites(Fig. 2G).

Although tiy121 embryos are short, the mutant phenotype indicates that cell proliferation is not absolutely required for trunk and tail extension. However, the mutants display irregularly sized and partially fused somites and myotomes (Fig. 3A-D; see also Fig. S1 in the supplementary material). The segmentation clock creates oscillations in transcription that manifest as stripes of expression sweeping through the cells of the PSM in a wave-like fashion. We examined the expression of three oscillating genes, her1,her7 and deltaC, at the 3-, 8- and 15-somite stage and found no appreciable defect in their expression in emi1^-/-(Fig. 3E-J; data not shown). Note that, at the 8- and 15-somite stage, the tailbud of emi1^-/- embryos is smaller than normal (compare Fig. 3F and G to 3I and J, respectively). This decrease is reflected in the reduction of the domain of mesogeninexpression (see Fig. S1 in the supplementary material). Although emi1mutants undergo mitosis at the beginning of gastrulation when oscillations are first seen (Riedel-Kruse et al.,2007), our data indicate that continued oscillation of the segmentation clock is not dependent upon the cell cycle.

In contrast to models that link the cell cycle to the segmentation clock,it has been postulated that mitosis is actually a source of noise for the clock (Horikawa et al., 2006). To test this hypothesis, we examined the effect of inhibiting mitosis in embryos lacking either of the Notch ligands deltaC or deltaD. The deltaC and deltaD mutants form the first 3-5 and 7-9 somites, respectively, as the oscillating pattern of gene expression gradually breaks down, leading to the segmentation defect(Fig. 3K-N; see also Fig. S1 in the supplementary material) (Holley et al., 2000; Jiang et al.,2000; Jülich et al.,2005b; Oates et al.,2005a; van Eeden et al.,1996; van Eeden et al.,1998). This breakdown may be accelerated because of noise. Thus,if mitosis is a source of noise in the segmentation program, one would predict that the breakdown would decelerate in the absence of cell division. We assayed the expression of her1 mRNA in deltaD or deltaC morpholino-injected embryos that were either wild type or mutant for emi1 (Fig. 3N, Fig. S1 in the supplementary material). The difference between the mutants and siblings was not immediately apparent. However, upon careful categorization of the expression patterns, we found a subtle improvement in the integrity of the her1 stripes in embryos lacking emi1compared with sibling embryos. For each trial, the more organized stripe patterns are biased towards the injected emi1 mutants, and the two more disorganized expression categories are biased towards the injected siblings. In summary, these results are consistent with mitosis being a modest source of noise in the segmentation clock.

Further examination of the segmentation defect in emi1^-/- embryos revealed profound abnormalities in somite morphology. Although emi1^-/- somites initially contain internal mesenchymal cells, these cells leave the core of the somite and at least some integrate into the epithelial somite boundary(Fig. 4A,B,D,E). We have observed other cells migrating to the lateral surface of the paraxial mesoderm. The somite boundary cells then appear to elongate and meet in the middle of each segment, creating somites solely consisting of two rows of boundary cells (Fig. 4B,E,G,I). These hyper-epithelialized somites, having no internal mesenchyme and abnormally elongated epithelial border cells, often fuse to create irregularly sized segments. The nuclei of the boundary cells show a basal localization, as does Integrin α5-GFP clustering (Fig. 4A,B). Fibronectin matrix is also assembled along the somite boundaries (Fig. 4F,G). This maintenance of border cell polarity distinguishes the emi1^-/- phenotype from that of integrin α5 and fibronectin1a mutants(Jülich et al., 2005a; Koshida et al., 2005). Ephrin B2 is localized to the cortex of the somite cells, with slightly higher levels in the posterior somite cells, and this pattern appears largely intact in emi1^-/- (Fig. 4H,I; see also Fig. S1 in the supplementary material). Expression of mespb, ripply1 and tbx18, myod and deltaC is clearly segmental, although there is some aberrant expression of deltaC in the mutant embryos (Fig. 4J-M; Fig. S1 in the supplementary material). The segment polarity alterations observed in emi1 mutants are slight in comparison to those defects seen in fss and the Notch pathway mutants, and seem unlikely to be the cause of the morphological phenotype.

The morphological analysis suggests that the polarity of the somite boundary cells is maintained. In addition, the somite phenotype does not seem to follow from a defect in anteroposterior patterning of the somites. Rather,it appears that the somites in emi1^-/- mutant embryos are hyper-epithelialized. The hyper-epithelialization could be due to elevated APC activity, which may affect the stability of proteins involved in regulating cell morphology independently of the cell cycle(Juo and Kaplan, 2004; Konishi et al., 2004; Lasorella et al., 2006; Stegmuller et al., 2006; van Roessel et al., 2004). To test this hypothesis, we blocked mitosis using a combination of hydroxyurea and aphidicolin (Harris and Hartenstein,1991; Lyons et al.,2005). Addition of the compounds at the germ ring/early shield stage blocked all mitosis by the late shield stage and resulted in embryos lacking internal mesenchymal cells in their somites, strongly phenocopying emi1^-/- (Fig. 4C; see also Fig. S1 in the supplementary material). These data suggest that the segmentation defect in emi1^-/- mutant embryos is primarily due to the lack of normal cell cycle progression and not to a cell cycle-independent function of emi1 or APC. Note that in both mutant and drug-treated embryos, the cells and nuclei are larger than in wild type (Fig. 4). The increase in cell size, along with the decrease in cell number, might also be causally linked to the somite morphogenesis defect.

The mitotic defect in emi1^-/- embryos arises after the midblastula transition (MBT). MBT initiates during the tenth cell cycle [3 hours post-fertilization (hpf)], when divisions become asynchronous and zygotic transcription commences (Kane,1999; Kane and Kimmel,1993). During cycles 11 and 12, the blastula forms three domains,the extra-embryonic yolk syncytial layer and enveloping layer, and the deep cells that give rise to the embryo proper(Kane, 1999). At 5.5 hpf,gastrulation starts, as most of the deep cells are in cell cycle 14(Kane, 1999; Kane et al., 1992). emi1^-/- embryos cease cell division around this time. In wild-type embryos, the cell cycle lengthens during this period, with the thirteenth, fourteenth, fifteenth and sixteenth cycles averaging 54, 78, 151 and 240 minutes, respectively. During segmentation, most cells are in either cell cycle 16 or 17 (Kane,1999). The mild elongation defect in emi1^-/-is likely to be due to the fact that mitosis is normally not a great contributor to axial growth during the segmentation period. This conclusion was also reached by examining the elongation of clonal strings of cells in the CNS: the exponential lengthening of the string suggested that it was largely due to cell intercalation and not cell division(Kimmel et al., 1994). The relatively normal differentiation in emi1^-/- embryos can be explained by the fact that many cells undergo a terminal differentiation during cell cycle 15, 8-10 hpf, and a major wave of differentiation occurs during cycle 16 (Kane, 1999; Kimmel et al., 1994; Kimmel and Warga, 1987). Thus,for many cell lineages, the mitotic defect in emi1 embryos does not reduce dramatically the number of cell cycles that these cells would normally undergo.

The cell cycle could in principle serve as a clock to regulate developmental timing (Johnson and Day,2000). Experiments in ascidians have suggested that the timing of myogenesis may depend upon the number of cycles of DNA synthesis that a myogenic progenitor experiences (Satoh,1987). However, previous cell labeling experiments indicate that zebrafish myofiber differentiation is not regulated by such a cell cycle-counting mechanism (Kimmel and Warga, 1987). Similarly, the expression of differentiation markers in the C. elegans gut occurs independently of cell cycle counting(Edgar and McGhee, 1988). The segmentation clock has been suggested to be linked to the cell cycle oscillator. Reiterated segmentation defects are seen in chick embryos after treatment with cell-cycle inhibitors, and the periodicity of this defect is equal to the cell cycle length at that stage of development(Primmett et al., 1989). Similar periodic defects were seen after a single heat shock(Primmett et al., 1988). More recently, this cell cycle model has been formalized mathematically(Collier et al., 2000; McInerney et al., 2004). In the zebrafish, a single heat shock can produce reiterated segmentation defects, but the periodicity of the defect does not correlate with the length of the cell cycle during segmentation (Roy et al., 1999). Additionally, there is no organized pattern of cell proliferation in the zebrafish tailbud(Kanki and Ho, 1997). Our analysis of the emi1 mutant indicates that cell cycle progression is not required for zebrafish segmentation clock function. Conversely, our data are consistent with the hypothesis that mitosis is a modest source of noise for the segmentation clock (Horikawa et al., 2006).

We thank Hiroyuki Takeda and members of the lab for comments on the manuscript. The tiy121 mutant was isolated in the Tübingen 2000 Screen, and we acknowledge all participants of the consortium: F. Van Bebber,E. Busch-Nentwich, R. Dahm, O. Frank, H.-G. Fronhöfer, H. Geiger, D. Gilmour, S. Holley, J. Hooge, D. Jülich, H. Knaut, F. Maderspacher, H.-M. Maischein, C. Neumann, C. Nüsslein-Volhard, H. Roehl, U. Schönberger, C. Seiler, S. Sidi, M. Sonawane, A. Wehner, P. Erker, H. Habeck, U. Hagner, C. E. Hennen Kaps, A. Kirchner, T. Koblizek, U. Langheinrich, C. Loeschke, C. Metzger, R. Nordin, J. Odenthal, M. Pezzuti, K. Schlombs, J. de Santana-Stamm, T. Trowe, G. Vacun, B. Walderich, A. Walker and C. Weiler. We thank Nancy Hopkins for the hi2648 allele. We thank Joseph Wolenski for assistance with the confocal microscopy. This work was supported by the March of Dimes Foundation, research grant number 1-FY07-424,and the NICHD, R01 HD045738.