The formation of somites, reiterated structures that will give rise to vertebrae and muscles, is thought to be dependent upon a molecular oscillator that may involve the Notch pathway. hairy/Enhancer of split related [E(spl)]-related (her or hes) genes, potential targets of Notch signaling, have been implicated as an output of the molecular oscillator. We have isolated a zebrafish deficiency, b567, that deletes two linked her genes, her1 and her7. Homozygous b567 mutants have defective somites along the entire embryonic axis. Injection of a combination of her1 and her7 (her1+7) morpholino modified antisense oligonucleotides (MOs) phenocopies the b567 mutant somitic phenotype, indicating that her1 and her7 are necessary for normal somite formation and that defective somitogenesis in b567 mutant embryos is due to deletion of her1 and her7. Analysis at the cellular level indicates that somites in her1+7-deficient embryos are enlarged in the anterior-posterior dimension. Weak somite boundaries are often found within these enlarged somites which are delineated by stronger, but imperfect, boundaries. In addition, the anterior-posterior polarity of these enlarged somites is disorganized. Analysis of her1 MO-injected embryos and her7 MO-injected embryos indicates that although these genes have partially redundant functions in most of the trunk region, her1 is necessary for proper formation of the anteriormost somites and her7 is necessary for proper formation of somites posterior to somite 11. By following somite development over time, we demonstrate that her genes are necessary for the formation of alternating strong somite boundaries. Thus, even though two potential downstream components of Notch signaling are lacking in her1+7-deficient embryos, somite boundaries form, but do so with a one and a half to two segment periodicity.
Segmentation of the body is a feature of development common to many animals. There are at least two modes of segmentation: the division of an existing tissue and the sequential division of a continuously growing tissue. Within the insects, segmentation in long germ band insects involves the division of an existing tissue whereas segmentation in short germ band insects is an example of the latter mode. During vertebrate development, rhombomeres form from an existing field of cells in the hindbrain, but somites form from a continuously proliferating field of cells. As it is not yet clear if the segmentation of long germ band and short germ band insects reflects conserved mechanisms (reviewed by Davis and Patel, 2002), it is an outstanding question as to whether there is any conservation of segmentation strategies between insects and vertebrates.
The anterior to posterior formation of somites from the presomitic mesoderm (PSM) is a highly dynamic process that underlies much of the segmentation of the adult. Many models of somitogenesis propose the existence of oscillatory behavior in the PSM as one method of creating pattern from an equivalent field of cells. In the ‘clock and wavefront’ model, cells in the PSM cycle between permissive and nonpermissive states (the ‘clock’) (Cooke and Zeeman, 1976). When this ‘clock’ interacts with a ‘wavefront’ (a maturation signal that tells cells to segment), cells in the anterior PSM in the permissive state form a somite. Presumptive somites, as well as formed somites, comprise an anterior compartment and a posterior compartment (Keynes and Stern, 1988) (reviewed by Hirsinger et al., 2000). This has led to the suggestion that a somite boundary may be specified at the juxtaposition of anterior and posterior cell fates (Meinhardt, 1986; Durbin et al., 1998; Durbin et al., 2000). In Meinhardt’s model, cells oscillate between anterior and posterior cell fates (Meinhardt, 1986). A single cell expresses one cell fate and instructs neighboring cells to adopt the opposite fate. This gives rise to oscillations in the PSM that generate stable stripes of cells expressing anterior and posterior fates in the anteriormost PSM. A somite boundary would then form at the juxtaposition of anterior and posterior cells.
One pathway that has been implicated as playing a role in biochemical oscillations that underlie somitogenesis is the Notch signaling pathway. Homozygous null mice for the Notch pathway members Notch, Dll1, RBPJk, presenilin, and Lunatic fringe all display defects in somite formation (Hrabe de Angelis et al., 1997; Conlon et al., 1995; Oka et al., 1995; Zhang and Gridley, 1998; Evrard et al., 1998; Koizumi et al., 2001). In zebrafish, most of the fused somites class of mutations, including after eight/deltaD, that are neurogenic and thought to disrupt the Notch pathway, produce defects in somitogenesis (van Eeden et al., 1996; van Eeden et al., 1998; Holley et al., 2000; Gray et al., 2001).
The most compelling evidence for a molecular oscillator in the PSM was the discovery that a chick homolog of Drosophila hairy, called hairy1, was expressed in a dynamic pattern (Palmeirim et al., 1997). During somitogenesis, hairy1 is expressed in successive posterior-to-anterior waves of expression, with each wave having a periodicity of the time needed to make one somite. The dynamic expression of hairy1 is not dependent upon protein synthesis, suggesting that hairy1 expression is an output of the clock rather than part of the clock itself (Palmeirim et al., 1997). Since then, several other genes have been shown to have a similar ‘cycling’ expression pattern, including hes, her and hey family members in zebrafish, chick and mouse (Holley et al., 2000; Jouve et al., 2000; Leimeister et al., 2000; Bessho et al., 2001a; Bessho et al., 2001b), lunatic fringe in chick and mouse (McGrew et al., 1998; Forsberg et al., 1998; Aulehla and Johnson, 1999), and delta homologs (deltaC and deltaD) in zebrafish (Jiang et al., 2000).
We have undertaken a genetic approach to understanding the role of hairy/E(spl)-related (her) genes during somitogenesis in zebrafish. We have isolated a deficiency, b567, that deletes both her1 and her7 genes. Like other Notch pathway mutants in mouse and zebrafish (Evrard et al., 1998; Kusumi et al., 1998; Zhang and Gridley, 1998; Durbin et al., 2000; Holley et al., 2000; Jiang et al., 2000) and the mouse Hes7 knockout (Bessho et al., 2001b), b567 mutant embryos show a disruption in somite anterior-posterior (AP) polarity. This disruption is phenocopied by injection of MOs (Nasevicius and Ekker, 2000) against her1 + her7. The abnormal expression of deltaD and deltaC in her1+7 MO-injected embryos indicates that coordinated expression of delta cycling genes requires her genes. Thus, her1 and her7 may feed back into the clock as well as being a potential output of the clock. Injection of either her1 or her7 MOs indicates a partial functional redundancy for these genes. However, her1 is necessary for the formation of the most anterior somite boundaries and her7 is required for the normal formation of more posterior somite boundaries. In contrast to the fused somites-type mutants bea, des and aei/deltaD, where a number of anterior somites are spared, somitic defects in b567 mutant and MO-injected embryos span the entire body axis. In both b567 mutant embryos and her1+7 MO-injected embryos, all somites are larger in the AP dimension than wild-type somites. These enlarged somites are delineated by stronger boundaries with at least one weak boundary attempt within the large somite. Thus, her1+7-deficient embryos have an ‘alternating boundary’ phenotype of strong boundary/weak boundary/strong boundary, demonstrating that her1 and her7 are essential for normal segmentation in zebrafish.
MATERIALS AND METHODS
Zebrafish mutant alleles, stocks and husbandry
Zebrafish embryos were obtained from natural spawnings of adult fish kept at 28.5°C on a 14 hour light/10 hour dark cycle and were staged according to Kimmel et al. (Kimmel et al., 1995). The homozygous b567 mutation was isolated during a large scale mutagenesis screen of haploid progeny of F1 females derived from γ-ray mutagenized males (see Walker, 1999). At approximately 10-11 hours post-fertilization (hpf), 20 haploid embryos from each clutch were fixed overnight in 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS) and dechorionated. Embryos were processed by in situ hybridization to detect transcripts of six genes: hatching gland 1 (hgg1) (Thisse et al., 1994), floating head (flh) (Talbot et al., 1995), pax2.1 (Krauss et al., 1991), valentino (val) (Moens et al., 1998), fkh6 (Odenthal and Nüsslein-Volhard, 1998), and her1 (Müller et al., 1996). In situ hybridization was performed as described below, either in Eppendorf tubes or in BEEM capsule baskets (see Moens et al., 1996). We focused on mutations that specifically disrupted her1 gene expression and describe one mutation, b567.
Mapping the b567 deficiency
DNA samples were prepared from b567+ and b567– diploid and haploid embryos (Postlethwait et al., 1994). PCR using her1 mapping primers (forward 5′-CAATCCTCTCAACCACGGAC-3′ and reverse 5′-ACAGCAAAGACCCCAGAACA-3′) amplified the expected 638 bp product from 28 b567+ embryos and failed to amplify a product from 18 b567– embryos. Other Linkage Group 5 (LG5) markers were similarly tested to estimate the size of the deficiency. Primer sequences used for mapping (Fig. 1) can be retrieved from the Zebrafish Information Network (ZFIN), the Zebrafish International Resource Center, University of Oregon, Eugene, OR 97403-5274; World Wide Web URL: http://zfin.org/ (see Sprague et al., 2001) or from the authors.
Morpholino-modified antisense oligonucleotides (MOs) were designed and synthesized by Gene-Tools, LCC. Antisense sequences and locations relative to the start site are:
her1a MO, 5′-GACTTGCCATTTTTGGAGTAACCAT-3′ location: +1 to +25;
her1b MO, 5′-ACACCTTCAGTATTGTATTCCCGCT-3′ location: –49 to –24;
her7a MO, 5′-TCAATGAGGATATGATTCCAGAAAA-3′ location: –50 to –25;
her7b MO 5′-TTTCAGTCTGTGCCAGGATTTTCAT-3′ location: +1 to +25;
her4 MO 5′-AGGAGTCATTGCTGTGTGTCTTGTG-3′ location –16 to +9.
MOs were solubilized in water to a stock concentration of 50 mg/ml. The stock solution was diluted to a working concentration of 0.5-5 mg/ml in 1× Danieau solution as described (Nasevicius and Ekker, 2000) and supplemented with 0.1% Phenol Red (Sigma). In experiments where several MOs were injected together, the diluted single Danieau MO solutions were mixed in equimolar concentrations prior to microinjection. Zebrafish embryos were injected with 2-3 nl of the Danieau MO solution or Danieau control (no morpholino) at the 1-4 cell stages. Note that her1 MO-injected embryos refers to embryos injected with her1a and her1b MOs and her7 MO-injected embryos refers to embryos injected with her7a and her7b MOs.
In situ hybridization and immunocytochemistry
Whole-mount in situ hybridization was performed as previously described (Jowett, 1999). F59 was used to visualize myosin fibers as previously described (Crow and Stockdale, 1986; Devoto et al., 1996). β-catenin antibody staining was performed as described (Topczewska et al., 2001). The original micrographs of β-catenin staining were inverted in Adobe Photoshop to facilitate visualization of the pseudocoloring (also performed in Adobe Photoshop).
Analysis of somite development
Live 14- to 18-somite embryos were mounted on a double-bridged slide (a coverslip with 2 adhered 1 μm coverslips) in 0.004% Tricaine in embryo medium (Westerfield, 1995). The embryos were then viewed in succession under the microscope. An acetate sheet was taped to the monitor and the somites traced. Somite boundary formation was judged by focusing medially and laterally to gain a three dimensional view of boundary formation. Strong boundaries were denoted with solid lines and weak boundaries were denoted with dashed lines. New acetate sheets were used for additional time points. At the conclusion of the time lapse, these sheets were aligned with each other (the yolk plug and tail were also traced to facilitate alignment), taped on a computer monitor, and traced using Adobe Photoshop.
Somite formation is disrupted in a zebrafish deficiency mutant that deletes two her genes, her1 and her7
In a haploid-based screen designed to identify mutations affecting early embryonic gene expression (see Materials and Methods), we recovered a γ-ray-induced deficiency, b567, that specifically abolishes expression of a hairy-related gene, her1 (Fig. 1). The her1 gene is normally expressed in two or three stripes in the PSM of wild-type embryos (Muller et al., 1996), but hairy/E(spl)-related expression is not detected in b567 mutant embryos (Fig. 1A,B). Several other genes were included in our in situ hybridization cocktail (see Fig. 1 legend); these genes are expressed normally in b567 mutant embryos (Fig. 1A,B). Somite boundary defects are observed in b567 mutant embryos at all stages of somitogenesis (Fig. 1C,D). Overall morphology in b567 mutant embryos, however, is relatively normal during the first day of development, suggesting that the effect on somitic gene expression and somitogenesis is specific. b567 mutant embryos do twitch. Generalized neural degeneration begins after the first day of development in b567 mutant embryos, and mutant embryos die between 80-120 hours post-fertilization (data not shown). Later phenotypes have not yet been examined in detail.
Because her1 is not expressed in b567 mutant embryos, we suspected that the b567 deficiency might delete the her1 gene. Using a PCR strategy (see Materials and Methods), we confirmed that her1, and other closely linked markers on Linkage Group 5, were missing on the b567 deficiency chromosome (Fig. 1E). One of the additional deleted genes is her7, another hairy/E(spl)-related gene that is expressed in a highly overlapping pattern with her1 (M. U., unpublished; M. Gajewski, D. Sieger, B. Alt, C. Leve, S. Hans, C. Wolff, K. Rohr and D. Tautz, personal communication). The b567 deficiency also deletes the ndr3 gene and a number of ESTs for which gene expression data is not available (Fig. 1E). The resolution of the genetic map in the region is somewhat uncertain, but we have localized the proximal breakpoint of the deficiency to a 0.1-3.9 cM interval near her1 and her7. The distal breakpoint is located within a larger interval, but the mapping data together allows us to estimate that the deletion is not any larger than ∼22 cM. Because a large number of genes could be absent in a 22 cM interval, we next used an antisense morpholino approach to identify which genes were likely responsible for the mutant somitic phenotype.
The somitic phenotype of b567 is due to the deletion of her1 and her7
As shown in Fig. 1, b567 is a deficiency that results in the deletion of a number of genes. Both her1 and her7 are transcribed in the PSM, suggesting that these genes may function in somitogenesis and their loss may be responsible for the segmentation defects in b567 mutant embryos. To test this hypothesis, we inhibited her1 and her7 mRNA translation by injecting morpholino antisense oligonucleotides (MOs) into zebrafish embryos (Nasevicius and Ekker, 2000). Two non-overlapping MOs were designed for both her1 (her1a MO and her1b MO) and her7 (her7a MO and her7b MO). Injection of all four MOs (her1+her7) perturbed formation of all somites (Table 1) and mimicked the b567 mutant somitic phenotype. Injection of equivalent or higher dose of a her1 MO combination (her1a MO+her1b MO) or a her7 MO combination (her7a MO and her7b MO) failed to phenocopy the b567 mutant somitic phenotype (Table 1). Injection of her1 MOs produced slight morphological defects in anterior somites (somites 1-3), whereas injection of her7 MOs caused boundary defects in posterior somites caudal to somite 11. As a control, a morpholino targeted against her4, which is only weakly expressed in the PSM (Takke and Campos-Ortega, 1999), was also tested. Most embryos injected with her4 MOs displayed normal somite morphology. her4 MO-injected embryos that did show segmentation defects also had extensive generalized disruption suggestive of MO toxicity. These results indicate that the segmentation defects in b567 mutants are likely caused by the specific lack of her1 and her7. The observation that her1 and her7 MOs individually produce distinguishable somitic phenotypes, but together mimic the b567 mutant phenotype at lower doses than the single injections (Table 1) suggests that her1 and her7 have partially, but not completely, redundant functions.
Segmental expression of paraxial mesoderm genes is disrupted in b567 mutant and MO-injected embryos
Paraxial mesoderm specification in her1+7 MO-injected embryos appears normal, as expression of spt and tbx6 (Griffin et al., 1998; Hug et al., 1997) in her1+7 MO-injected embryos is indistinguishable from control-injected embryos (data not shown). As somitogenesis is disrupted in her1+7 deficient embryos, we next examined whether anterior-posterior (AP) somite polarity and expression of cycling genes was normal in her1+7-deficient embryos (Fig. 2). myoD and paraxial protocadherin (papc/pcdh8) are markers of posterior and anterior somite polarity, respectively (Weinberg et al., 1996; Yamamoto et al., 1998). Segmental myoD expression in the posterior half of formed somites is disrupted in both b567 mutant and her1+7 MO-injected embryos. In these embryos, myoD expression is not restricted to the posterior half of formed somites and is instead expressed in all paraxial cells (Fig. 2A-D). In wild-type embryos, papc is segmentally expressed in the presumptive anterior half of the next two somites that will form (Fig. 2E,G). In both her1+7 MO-injected and b567 mutant embryos, papc is expressed broadly throughout the anterior PSM (Fig. 2F,H). In wild-type embryos, ephrin receptor ephA4 and ligand ephrin B2 expression is refined to stripes denoting the anterior and posterior aspects, respectively, of mature somites and the next presumptive somite (Durbin et al., 1998). In contrast, both ephA4 (data not shown) and ephrin B2 (Fig. 2Q,R) are expressed throughout the paraxial mesoderm of her1+7 MO-injected embryos. Thus, the AP polarity of somites is disrupted in her1+7-deficient embryos.
Previous analysis of Notch signaling pathway mutants demonstrated that disruption of one component can affect the expression of other Notch pathway genes (Zhang and Gridley, 1998; Hrabe De Angelis et al., 1997; Holley et al., 2000). For example, Delta signaling is required for the dynamic expression of hairy/E(spl)-related genes in both mouse and zebrafish (Jouve et al., 2000; Holley et al., 2000). Interestingly, Hes7, but not Hes1, is required for dynamic expression of lunatic fringe (Bessho et al., 2001; Jouve et al., 2000). We examined whether expression of Notch pathway components was affected in b567 mutant and her1+7 MO-injected embryos. The expression of notch5, normally a posterior somite polarity marker in wild-type embryos, is much more uniformly expressed in her1+7 MO-injected embryos (Fig. 2M,N). In wild-type embryos, both deltaC and deltaD are dynamically expressed in the posterior PSM but the expression of these genes becomes fixed in the anterior PSM in the posterior or anterior half of the next presumptive somite, respectively (Jiang et al., 2000). In b567 mutant and her1+7 MO-injected embryos, deltaD is expressed in a broad band in the anterior PSM rather than in discrete stripes as in wild-type embryos (Fig. 2I,J, and data not shown). The expression of deltaC in the anterior PSM is also seen in a large band in her1+7 MO-injected embryos (Fig. 2K,L). The pattern of expression does not vary among her1+7-deficient embryos, suggesting that there is no coordinated dynamic expression of deltaC or deltaD. In mouse, Mesp2, a gene encoding a basic helix-loop-helix protein, is expressed in the presumptive rostral region of somite minus 1 (the somite that will form next) and has been shown to interact genetically with the Notch pathway (Takahashi et al., 2000). The expression of mespA in b567 mutant embryos resembles the mespA expression in both bea and mib mutant embryos: there is one broad domain of reduced expression rather than in 2 sharp bands as in wild-type embryos (Fig. 2O,P) (Sawada et al., 2000). Thus, her1 and her7 are required for segmental expression of both Notch pathway genes and the specification of AP polarity.
As hairy/E(spl)-related genes are transcriptional repressors that feedback on Notch signaling (reviewed by Davis and Turner, 2001), one prediction would be that outputs of Notch signaling other than her1 and her7 should be activated in her1+7 MO-injected embryos. Zebrafish focal adhesion kinase (fak), shows a specific response to activation of the Notch pathway via activated Suppressor of Hairless (X-Su(H)1/Ank) and not to inhibition of the Notch pathway via a dominant negative Suppressor of Hairless (X-Su(H1)DBM) (Henry et al., 2001; Wettstein et al., 1997). The expression of fak in her1+7 MO-injected embryos resembles that of embryos injected with activated Supressor of Hairless, suggesting that the Notch pathway may be activated in her1+7 MO-injected embryos (Fig. 2S,T).
Somites in her1+7 MO-injected and b567 mutant embryos are enlarged in the AP dimension
In wild-type embryos, somites form with a regular temporal and spatial periodicity (Fig. 3A). In b567 mutant embryos, somites are enlarged in the AP dimension (Fig. 3B). There appear to be weak (incomplete) boundaries within the stronger (though imperfect) boundaries that delineate the large somites (Fig. 3B, arrowheads). Immunohistochemistry using F59 to stain myosin fibers (Crow et al., 1986; Devoto et al., 1996) affirms that somites in b567 mutant embryos are enlarged in the AP dimension compared to wild-type embryos (Fig. 3C,D). In wild-type embryos, muscle fibers span the length of one somite and terminate at the boundary of adjacent somites. In b567 mutant and her1+7 MO-injected embryos, muscle fibers cross boundaries and terminate within the myotome (Fig. 3C-E). Expression of a titin homolog that labels mature somite boundaries also reveals that somite periodicity is altered in b567 mutant embryos. In wild-type embryos, chevron-shaped titin expression is regularly spaced along the AP axis (Yan et al., 2002). In b567 mutant embryos, titin staining delineates somites that are poorly formed and enlarged in the AP dimension (Fig. 3F,G).
We have observed that b567 mutant embryos have imperfect somites that are enlarged in the AP dimension. In addition, weak boundaries appear to form within the enlarged somites. In order to more carefully assess the function of her genes in somite morphogenesis, we analyzed somite morphology at the cellular level in her1+7 MO-injected embryos by using an antibody against β-catenin (Topczewska et al., 2001). We asked two questions: where are boundaries forming in these embryos, and what is the morphology of the boundaries that do form? Unlike wild-type boundaries, the boundaries that form in b567 mutant or her1+7 MO-injected embryos are not fully extended in either the dorsal-ventral or medial-lateral dimensions. We therefore defined a strong somite boundary in her-deficient embryos as one that extends at least 90% of the dorsal-ventral dimension and occupies one third of the mediolateral dimension. Somites in her1+7 MO-injected embryos are consistently larger in the AP dimension compared to wild-type embryos (Fig. 4). While it is sometimes observed that a somite of wild-type size does form in between large somites (data not shown), the average spatial periodicity of somites in her1+7 MO-injected embryos is approximately one and a half to two somite equivalents (Fig. 4). Weak boundaries in the middle of the large somites were observed in her1+7 MO-injected embryos (arrowheads, Fig. 4D,F). These weak boundaries are defined as such because they appear to be formed from a few cells lining up but do not extend in the dorsal-ventral or mediolateral dimension. The observation of weak boundaries within the large somites suggests a strong boundary/weak boundary/strong boundary segmentation pattern in these embryos. Thus, analyzing somite formation in her-deficient embryos at high resolution has revealed a consistent pattern in what might otherwise be characterized as merely disrupted somites. In wild-type embryos, somitic boundaries are formed via the alignment and epithelialization of presumptive border cells (Henry et al., 2000). A boundary is thus flanked by a neat row of rectangularly shaped border cells (Fig. 4B). Boundaries in her1+7 MO-injected embryos are also flanked by rectangular, aligned border cells (Fig. 4D,F). Although the border cells in b567 mutant embryos are slightly more disorganized, aligned rectangular cells do flank the strong boundaries. This indicates that her1 and her7 are not necessary for border cell morphogenesis. However, her1 and her7 may be necessary for the refinement and strengthening of alternate somite boundaries.
The stronger boundaries observed in her1+7 MO-injected embryos extend further in the dorsal-ventral dimension than those of b567 embryos. The difference could be explained by (1) incomplete inhibition of her mRNA translation in MO experiments or (2) the deletion of another gene in the b567 deficiency. If the first scenario accounts for the differences observed, we would postulate that alternate somite boundaries are more sensitive to a decrease in her1+7 activity. In the second scenario, we would postulate that an additional gene that is deleted in the b567 deficiency is required to facilitate alignment of border cells in the dorsal-ventral dimension.
her1 functions in anterior somite formation, while her7 functions in posterior somite formation
The observation that somites in her1+7 MO-injected and b567 mutant embryos are enlarged in the AP dimension led us to analyze the specific defects in embryos injected with either her1 or her7 MOs. Most somites in her1 MO-injected embryos are well formed with normal periodicity (Fig. 5A), however, the anterior-most somites are defective. There is a range of defects observed, from fusion of either somites 1 and 2 or somites 2 and 3, to disrupted boundary formation among these somites. This relatively subtle but consistent effect of her1 MOs strongly suggests there is an early role for her1 for which her7 cannot compensate. In her7 MO-injected embryos, the two most recently formed somites are clearly enlarged (Fig. 5B). Analysis of somite formation in her7 MO-injected embryos at later stages (18-24 somites) also indicated a clear trend of ‘strong boundary/weak boundary/strong boundary/weak boundary’ (blue and red boundaries in Fig. 5C, respectively). Somites anterior to somite 11 were normally formed. This indicates that her1 cannot compensate for her7 in the posterior trunk and tail.
In her1+7 MO-injected embryos, somites along the entire AP axis are enlarged (Fig. 4C,E) and deltaD expression is disrupted in early and late somitogenesis (Fig. 5G, and data not shown). Embryos injected with her7 MOs, however, only display morphological segmentation defects late in segmentation (Fig. 5B,C). We therefore tested the hypothesis that deltaD expression would be normal early in somitogenesis but disrupted later. Indeed, this was the case. At the 2-somite stage, two deltaD stripes (similar to those in wild-type embryos) were observed in the anterior PSM in her7 MO-injected embryos (Fig. 5D,E); whereas, at the 10-somite stage, one large deltaD band, instead of two smaller bands, is observed (Fig. 5I). Thus, inappropriate deltaD expression correlates well with the commencement of somite boundary defects in her7 MO-injected embryos. Interestingly, we also observed milder disruption of presomitic deltaD expression at the 10-somite stage in her1 MO-injected embryos (Fig. 5H). However, the somitic expression of deltaD in her1 MO-injected embryos is segmental.
Selective strengthening of boundaries underlies the large somites in her7 MO-injected embryos
The phenotype of her1+7 MO-injected embryos, alternating strong and weak boundaries, is quite intriguing. There are at least 2 possibilities to explain this observation. First, the strong boundaries could delineate large somites and then the weak boundaries could subdivide the large somites. Alternatively, weak boundaries could form and a subset of these boundaries could be strengthened. We analyzed somite formation over time and determined that the latter explanation is most likely. It has previously been shown that somite boundary formation proceeds via the alignment of clefts (Wood and Thorogood, 1994). The clefting of strong boundaries in b567 mutant and her1+7 MO-injected embryos, although much more evident than clefting of weak boundaries, does not occupy the entire mediolateral extent of somites and is best visualized in three dimensions by focusing through the entire mediolateral extent of the somite. We followed somite development by tracing the boundaries of live embryos approximately every 30-40 minutes (see Materials and Methods). We followed somite development in her7 MO-injected embryos because the posterior somites in these embryos resemble those in her1+7 MO-injected embryos but somite boundaries are easier to visualize.
In wild-type embryos, a new strong somite boundary was seen approximately every half hour (n=7 boundaries in 2 embryos) (Fig. 6A,D). Because tracings were made every 30-40 minutes, a new strong boundary that occupies the entire dorsal-ventral and mediolateral extents of the PSM had usually formed in wild-type embryos (Fig. 6A,D). We also sometimes observed weak clefting as a wild-type boundary was forming (data not shown). Strong boundary formation was observed in her7 MO-injected embryos (n=18 strong somite boundaries formed in 5 embryos), however, there was more variability in timing of boundary formation than in wild-type embryos as shown in Fig. 6B and C. Furthermore, somite boundary formation in these embryos was delayed relative to wild-type embryos (data not shown).
Somites in her7 MO-injected embryos are enlarged in the AP dimension, yet weak boundaries are observed within these large somites (Fig. 5). In order to understand the role of her genes in somite formation it is necessary to analyze when these weak boundaries are forming relative to the strong boundaries. As outlined above, one possibility is that after a large somite forms in her7 MO-injected embryos, it subsequently becomes subdivided. If this were true, we would visualize strong boundary formation (delineating a large somite) prior to weak boundary formation within the large somites. This was not observed in her7 MO-injected embryos. Instead, weak boundaries were visualized before strong boundaries formed. Furthermore, most strong boundaries (82%) were preceded by weak boundaries at the same or similar location. Thus, in her7 MO-injected embryos there are weak attempts at boundary formation that proceed in an AP progression, but only a subset of these attempts (approximately every other boundary) becomes strengthened.
We have isolated a deficiency that deletes the hairy/E(spl)-related genes her1 and her7. Injection of antisense morpholino oligonucleotides demonstrates that the somitic phenotype of b567 mutants is largely due to the deletion of her1 and her7. Both her1+7 MO-injected and b567 mutant embryos display disrupted gene expression in the paraxial mesoderm and imperfect somites display a one and a half to two segment periodicity relative to wild-type embryos. Thus, her1 and her7 are required for normal somite formation.
A delineation of anterior and posterior half-segments may not be necessary for boundary formation
It has been proposed that the juxtaposition of anterior and posterior half-segments may determine where a somite boundary forms (Meinhardt, 1986; Durbin et al., 1998; Durbin et al., 2000; Jen et al., 1999). In support of this hypothesis, it has been shown that transplantation of cells expressing ephrin A4, a marker of anterior somite polarity, can locally rescue boundary formation in fss mutant embryos. fss mutant embryos are unique among the known zebrafish segmentation mutants as they form no somitic boundaries (van Eeden et al., 1996). In addition, anterior somite polarity markers are not expressed in fss mutant embryos; instead, markers of posterior polarity are expressed throughout the paraxial mesoderm (Durbin et al., 2000). In other zebrafish mutants, such as bea, des, mib and aei/deltaD, cells expressing anterior and posterior markers are intermingled in the mutant paraxial mesoderm, and mutant embryos fail to form somites in the posterior trunk and tail (van Eeden et al., 1996; van Eeden et al., 1998; Durbin et al., 2000; Holley et al., 2000). As in this latter class of mutants, markers of AP somite polarity are also misexpressed in b567 mutant and her1+7 MO-injected embryos. We have tested four anterior markers (papc/pcdh8, deltaD, mespA, EphA4) and five posterior markers (myoD, deltaC, notch5, fak/ptk2, ephrin B2) and the expression of all of these genes is continuous within the PSM and somitic mesoderm (Fig. 3). However, boundaries do form in her1+7 MO-injected embryos (again, stronger boundaries extend at least 90% of the dorsal-ventral dimension and one third of the medial-lateral dimension). In her1 MO-injected embryos, deltaD is correctly expressed in the mature somitic mesdoerm, but is not expressed in a tight band in the anterior half of the PSM. Despite this disruption in AP polarity of the PSM, somites in these embryos (except for the most anterior somites) form just as in wild-type embryos (Fig. 5). These data suggest that the somite that is about to form does not have to be subdivided into prospective anterior and posterior halves at the mRNA level in order for a boundary to form. However, we have not tested all markers of AP polarity. In addition, we currently lack the tools to determine if post-transcriptional mechanisms allow the expression of these proteins in the A or P half of the somite. Alternatively, there may be genes that have not yet been isolated that are expressed normally in the anterior and posterior halves of presumptive somites in her1+7 MO-injected embryos.
The expression of papc and deltaD in wild-type embryos delineates the anterior half of a prospective somite (Holley et al., 2000; Yamamoto et al., 1998). A wide band of strong expression of both deltaD and papc is seen in b567 mutant and MO-injected embryos. It is possible that this entire band of deltaD and papc expression is perceived in the presomitic mesdoerm as anterior identity and somite borders thus form at the juxtaposition of papc/deltaD-expressing cells with cells that do not strongly express papc/deltaD. Therefore, although intrasegmental polarity may not be necessary for segment formation, it is possible that the driving force behind the formation of enlarged somites is the formation of a boundary between a large group of cells expressing papc and those that are not (Kim et al., 1998). The formation of a weak boundary in between the strong boundaries could be due to the normal expression of genes involved in AP differentiation that are not yet identified or were not surveyed in this study.
her1 and her7 function partially redundantly
The somitic phenotype of b567 mutants is phenocopied via injection of MOs against her1+her7. Single injection of either her1 MOs or her7 MOs does not phenocopy the somitic defects seen in b567 mutant embryos. The dose required to phenocopy the somitic defects of b567 mutant embryos throughout the embryonic axis in her1+7 MO-injected embryos is much less (3 fold) than the doses required to partially disrupt segmentation in embryos injected with either her1 or her7 MOs (Table 1). Thus, her1 and her7 appear to function redundantly. However, some defects in somitogenesis are observed in embryos singly injected with either her1 MOs or her7 MOs. In her1 MO-injected embryos, the anteriormost somites are frequently fused suggesting that her1 has an early function for which her7 cannot compensate (Fig. 5). In her7 MO-injected embryos, somites posterior to approximately somite 10-13 are enlarged in the AP dimension much as somites in b567 mutant and her1+7 MO-injected embryos are (Fig. 5). Thus, her7 appears to have a late function for which her1 cannot compensate. These data indicate that although her1 and her7 have partially redundant functions, they are both required for normal somitogenesis.
hairy/E(spl)-related genes encode transcriptional repressors that bind to DNA as either hetero- or homodimers (reviewed by Davis and Turner, 2001). The partial redundancy of her1 and her7 may be explained in light of the ability of these proteins to function as hetero- or homodimers. It is possible that either her1 or her7 homodimers may be sufficient for normal somite formation in much of the trunk but her1 homodimers are not sufficient for normal somite formation in the tail. Alternatively, it is possible that other her genes such as her4 or her6 (Takke et al., 1999; Pasini et al., 2001) are upregulated and heterodimers including new combinations of Her proteins are sufficient for segmentation in the trunk.
The dynamic expression of hairy1 in the PSM is not dependent upon protein synthesis (Palmeirim et al., 1997), suggesting that hairy1 does not negatively regulate its own expression. However, it has been demonstrated that Hes7 promoter activity is increased in a Hes7 knockout mouse and that Hes7 can repress transcription in vitro (Bessho et al., 2001a; Bessho et al., 2001b). Furthermore, HES1 has been shown to negatively autoregulate its own transcription (Takebayashi et al., 1994). We have been unable to assess the effects of her1 and her7 on their own transcription because b567 is a deficiency, and because our experiments suggest that her1 and her7 morpholinos stabilize the corresponding mRNA transcripts (data not shown). We have shown that her4 MOs do not perturb segmentation (Table 1), but this does not preclude the possibility that upregulation of additional her genes (or other genes altogether) could account for the delayed formation of boundaries in her1+7 MO-injected embryos.
Somites in MO-injected embryos form by selective strengthening of boundaries
We have shown that somite boundaries in her1+7-deficient embryos are enlarged in the AP dimension. The spatial periodicity of somite formation in these embryos is one and a half to two somites. By following somite development over time, we have observed that the stronger somite boundaries that do form in her7 MO-injected embryos are formed via selective strengthening of weak boundaries (Fig. 6). We predict that the enlarged somites that form in her-deficient embryos also form by selective strengthening of weak boundaries. Combined, these data suggest that the her1 and her7 genes are necessary to form strong boundaries on schedule. Furthermore, her1 and her7 are necessary for the strengthening of alternate weak boundaries.
her genes and the segmentation clock
It was first demonstrated in chick that hairy1 is expressed in repeating posterior to anterior progressing waves of expression (Palmeirim et al., 1997). Other Notch pathway genes are also expressed in waves, including lunatic fringe in the chick and mouse (Forsberg et al., 1998; McGrew et al., 1998; Aulehla and Johnson, 1999), Hes1 and Hes7 in the mouse (Jouve et al., 2000; Bessho et al., 2001a; Bessho et al., 2001b), deltaD, deltaC and her1 in the zebrafish (Holley et al., 2000; Jiang et al., 2000; Sawada et al., 2000). The dynamic expression of hairy1 and hairy2 is not dependent upon protein synthesis (Palmeirim et al., 1997; Jouve et al., 2000). This observation has led to the suggestion that hairy1 may not be a component of the clock but rather an output of the clock (Palmeirim et al., 1997; Jouve et al., 2000). her1 expression does not oscillate in aei/deltaD embryos, indicating that her1 may be downstream of Notch signaling (Holley et al., 2000). Thus, if her1 is downstream of Notch signaling and an output of the clock, Notch signaling may be a component of the segmentation clock. We show that deltaC and deltaD do not cycle in her1+7 deficient embryos. Thus, although her1+7 may be direct outputs of the segmentation clock they also could feed back into the clock. Alternatively, Notch signaling, as well as her1 expression, may also be an output of the segmentation clock. In this scenario, her1+7 MOs may either feedback into the clock to eliminate deltaC and deltaD cycling, or these MOs may eliminate deltaC and deltaD cycling via constitutively activating Notch signaling.
Recently, it has been shown that FGF signaling encodes a ‘determination front’, the position of which affects the size of somites in both chick and zebrafish (Dubrulle et al., 2001; Sawada et al., 2001). FGF signaling in the posterior PSM keeps cells in an immature state. At a critical level of FGF signaling in the more anterior PSM, cells appear to become allocated to a particular somite. Thus, ectopic FGF in the PSM generates smaller somites, presumably by slowing down the reception of the ‘wavefront’ signal such that fewer cells are allocated into each somite. Conversely, inhibition of FGF signaling generates larger somites, and it was hypothesized that this affect was due to an increase in the number of cells that received the ‘wavefront’ signal. We have observed that fgf8 expression in the tailbud of her1+7 MO-injected embryos resembles that of wild-type embryos (data not shown). It thus seems to be unlikely that the enlarged somites in b567 mutant and her1+7 MO-injected embryos reflect changes in the ‘determination front’. However, we have not investigated the expression of other fgfs or the activation of ERK in MO-injected embryos.
her genes and synchronization of the segmentation clock
It has been proposed that one role for Notch signaling in coordinating somite formation is to synchronize the oscillations between neighboring cells (Jiang et al., 2000). In her1+7-deficient embryos, a large band of upregulation of deltaD and deltaC is seen. This is presumably due to the lack of the transcriptional repressors encoded by her1 and her7. If there were additional outputs of Notch signaling (such as Enhancer of split-related proteins [ESRs] as found in Xenopus) (Jen et al., 1999) then these outputs would be constitutively activated in the presence of the ligands deltaC and deltaD. The nonsegmental expression of fak in her1+7 MO-injected embryos (Fig. 2) suggests that there are outputs of Notch signaling besides her1 and her7. Thus, a larger group of cells may be synchronized and respond to the ‘somite formation signal’, or ‘wavefront’, by making a larger somite. The presence of a weak boundary forming in the middle of enlarged somites in her1+7-deficient embryos could reflect subtle differences in synchronization.
Similarities and differences in segmentation strategies
One outstanding question in the field of evolutionary developmental biology is how segmentation in different phyla has evolved. If a general consensus could be defined, it would probably be that somite formation is not analogous to segment formation in Drosophila (reviewed by Davis and Patel, 1999). The formation of enlarged somites with a one and a half to two segment periodicity in her1+7 MO-injected embryos is thus a very interesting phenotype. It is unlikely that this phenotype reflects a strict pair rule function for her1 and her7 because it is sometimes observed that a normal segment is formed in between two large somites. Furthermore, not every large somite is the same size.
However, it is possible that her genes may play a role outside of Notch signaling in the formation of alternate boundaries during zebrafish segmentation. In Tribolium castaneum, a short germ band insect, mutants that display a pair rule phenotype have been isolated (Sulston and Anderson, 1996; Sulston and Anderson, 1998; Maderspacher et al., 1998). The Tribolium itchy mutant displays a clear strong boundary/weak boundary phenotype as assayed by engrailed expression and is missing odd thoracic segments and a variable number of abdominal segments. The ability to identify the missing thoracic segments, combined with the reduced engrailed expression and missing abdominal segments allows classification of itchy as a pair rule gene. The first report of the expression of zebrafish her1 showed that her1 is expressed in a metameric pattern during somitogenesis. Furthermore, stripes of her1 expression gave rise to alternating somitic primordia at the two stages examined (Müller et al., 1996), suggesting a potential role of her1 as a pair rule-type gene. This suggests that one alternative possibility for the formation of enlarged segments in her-deficient embryos may be that hairy-related genes have been co-opted to play a role in the formation of alternate segments in vertebrates. However, we currently do not have the tools to distinguish between zebrafish vertebrae, and are therefore unable to determine if alternate segments are missing in her1+7 MO-injected embryos.
In Meinhardt’s model of somite formation, he proposes that boundaries form at the juxtaposition of anterior and posterior cell states, recognizing that an additional mechanism must operate to prevent boundary formation in the middle of a somite. [One exception to this rule is von Ebner’s fissure, which only forms in the sclerotome (Keynes and Stern, 1985), and has not been observed in zebrafish.] As one mechanism to explain the inhibition of boundary formation in the middle of a somite, Meinhardt proposed the existence of a third cell state such that boundaries form between but not within somites. It has been shown in zebrafish that two cell states, anterior and posterior, are sufficient for boundary formation (Henry et al., 2000). The authors proposed that apical-basal polarity superimposed upon AP identify may specify where a boundary forms (Henry et al., 2000; Henry et al., 2001). However, in light of the ‘alternate segments strengthened’ phenotype of her1+7-deficient embryos, it is worth noting that Meinhardt also proposed that an odd-even mechanism superimposed upon AP identity could specify where a boundary forms. Clearly, her genes are necessary for the strengthening, on average, of every other boundary in zebrafish. Whether this ‘alternate segments strengthened’ phenotype reflects any conservation of segmentation between insects and vertebrates or merely reflects a crucial role for her genes in the segmentation clock requires further analysis.
Note added in proof
A recent paper (Holley et al., 2002) has reported that deltaD expression does not oscillate in zebrafish PSM. In addition, they report molecular expression analyses of her1 MO-injected embryos similar to those reported in this paper (although we observed differences in morphological phenotype).
We thank Elizabeth Pickett and Ramona Pufan (University of California, Berkeley) and the University of Oregon Zebrafish Facility staff for excellent zebrafish care. Jennifer Anderson and Eric Parkhill provided expert technical assistance. We gratefully acknowledge our many collegues at the University of Oregon who participated in mutagenesis screening. We thank Yi-Lin Yan and John Postlethwait for generously sharing the titin plasmid prior to publication and Jose Campos-Ortega and Diethard Tautz for sharing information and reagents. We thank all members of the laboratory, especially Jennifer Anderson and Tina Han, for helpful discussion. This work was supported by an NIH grant (HD22486) to C. B. K. M. K. U. was supported by a National Science Foundation Predoctoral Fellowship and C. A. H. is a Miller Fellow in the Department of Molecular and Cell Biology at the University of California, Berkeley.