Hairy/E(spl)-related (Her) genes are central components of the segmentation oscillator and display redundancy with the Delta/Notch signaling pathway in the formation of anterior segmental boundaries in the zebrafish

Metamerism of the body plan is a feature of many animal phyla. The vertebral column and ribs are the most obvious segmental structures of the adult vertebrate, and these mesodermal derivatives, as well as associated muscle and dermis, develop from segmental units of embryonic mesoderm called somites. Somites are clusters of cells that form in an anterior-to-posterior order along both sides of the embryonic body axis by a serial mesenchymal-to-epithelial transition of the presomitic, paraxial mesoderm. Each somite possesses a distinct rostral and caudal half, and the epithelial boundaries of somites, the intersomitic furrows, are symmetrical across the midline. The presomitic mesoderm (PSM), despite its unsegmented appearance, possesses a segmental prepattern mirrored by segmentally restricted and polarized gene expression in the anterior of the PSM, including cell adhesion molecules, muscle differentiation factors and genes of the Delta/Notch signaling pathway. Some Delta/Notch genes are also expressed in dynamic striped patterns that lack segmental polarity in more posterior regions of the PSM and tailbud. Indeed, Delta/Notch signaling appears to have an important role in controlling vertebrate somitogenesis. In mouse and zebrafish, mutations in genes coding for cell surface ligands of the Delta family (Bulman et al., 2000; Holley et al., 2000; Hrabe de Angelis et al., 1997; Kusumi et al., 1998), in transmembrane receptors of the Notch family (Conlon et al., 1995; Krebs et al., 2000) or in components of the signal transduction pathway from the Notch receptor (Donoviel et al., 1999; Evrard et al., 1998; Oka et al., 1995; Shen et al., 1997; Wong et al., 1997; Zhang and Gridley, 1998) result in aberrant somitogenesis. These defects are visible as irregularly shaped somites with bilaterally asymmetrical boundaries and a loss of rostrocaudal polarity within each somite. An important feature of these mutant phenotypes is that as increasingly posterior somites are made, the severity of the defects increases.

In Drosophila and vertebrates, Delta/Notch signaling is modulated by the glycosyltransferase activity encoded by the Lunatic fringe (Lfng) gene and results in expression of members of the Her bHLH transcription factor family. In mouse and chick, mRNA from these genes displays dynamic patterns resembling a wave of expression in the PSM that appears to initiate in the posterior and move anteriorly, coming to rest at a position which predicts the site of a future somitic furrow (Aulehla and Johnson, 1999; Bessho et al., 2001; Forsberg et al., 1998; Jouve et al., 2000; Leimeister et al., 2000; McGrew et al., 1998; Palmeirim et al., 1997). The wave-like expression patterns are thought to arise because the level of Her or Lunatic fringe mRNA oscillates in individual cells of the PSM with a period equal to the interval between the formation of successive somites, and these oscillations are temporally coordinated between neighboring cells into a spatially dynamic wave-like expression domain (Palmeirim et al., 1997). The correlation of the spatial and temporal aspects of these dynamic expression domains in the PSM with the process of somitogenesis suggests that they, or underlying cellular oscillations, may have a causal role in segmentation, although direct evidence for this is lacking. A conserved role for these dynamic processes in vertebrate somitogenesis is supported by the existence of cyclic genes in teleosts. In the PSM of the zebrafish embryo, mRNAs from the hairy1 (her1) and the deltaC (dlc) genes are expressed in patterns nearly identical to those seen for the cyclic Lfng and Her genes in mouse and chick (Holley et al., 2000; Jiang et al., 2000; Sawada et al., 2000).

The Delta/Notch signaling system appears to be required for maintenance of cyclic gene expression domain coherence in mouse and zebrafish. In mouse, loss of Delta-like 1, Csl, or Notch1 function leads to a severe downregulation of Lfng expression and a loss of cyclic Lfng patterns (Barrantes et al., 1999), and loss of Delta-like 1 or Notch1 has similar effects on the cyclic Her genes Hes1 and Hey2 (Jouve et al., 2000; Leimeister et al., 2000). In the zebrafish, loss of deltaD (dld) function in the after eight (aei) mutant also leads to disruption of cyclic gene expression, although, in contrast to the mouse, expression of her1 and dlc is not severely downregulated (Durbin et al., 2000; Holley et al., 2000; Jiang et al., 2000; van Eeden et al., 1998). Instead, at stages with comparable somite number to the mice described above, her1 and dlc expression is found in the anterior PSM in a wide stripe with diffuse boundaries and a speckled or ‘salt and pepper’ composition. Although dld is expressed in a dynamic striped pattern in the PSM throughout all stages of somitogenesis (Dornseifer et al., 1997), the loss of dld function results in defective somites only posterior to somite 7 or 8 (Holley et al., 2000). Importantly, in these animals, the cyclic expression pattern of dlc and her1 is initially normal, but as development proceeds, the boundaries of the dynamic expression domains become more diffuse and PSM cells begin to express the cyclic genes out of synchrony with their immediate neighbors, leading to a gradual loss of coherence in the cyclic expression pattern (Jiang et al., 2000). Similar results are seen in the beamter, deadly seven and white tail mutants, which display very similar somitogenic phenotypes (van Eeden et al., 1996), and are also thought to result from lesions in Delta/Notch signaling pathway components or modifiers (Appel et al., 1999; Riley, 1999; Gray et al., 2001; Jiang et al., 2000; Jiang et al., 1996; Lawson et al., 2001). Thus, the posterior onset of somitogenic defects in Delta/Notch mutants is paralleled by a posterior loss of cyclic gene expression domain coherence.

What is the role of cyclic genes themselves in the generation of oscillations and/or their coordination? A cyclic gene whose function is required for the integrity of the segmentation oscillator has been defined as a component of the oscillator (i.e. in the absence of a component, coherent wave-like expression patterns of other cyclic genes are lost) (Palmeirim et al., 1997). If loss of function in a cyclic gene resulted in an immediate disruption of oscillation of all cyclic gene expression, it would indicate that the gene was a central component of the oscillatory mechanism. At present, this would be measured by determining that the dynamic domains of gene expression were immediately perturbed, or did not initiate. Note that genes without cyclic mRNA levels could still be central components of the oscillator by virtue of periodic modulation of protein activity, for example. If loss of function of a cyclic gene leads to a gradual loss of cyclic domain coherence, we might assign to this component the role of coordinating the oscillations (Jiang et al., 2000). By contrast, if loss of function of a cyclic gene did not result in a change in the cyclic expression of other genes, even after multiple cycles, the gene would instead be an output of the oscillator (Palmeirim et al., 1997). Such mutations could nevertheless give rise to somitogenic phenotypes. Mice that lack the function of the Hes1 gene display neither perturbation of cyclic Lfng expression nor defects in segmentation (Ishibashi et al., 1995; Jensen et al., 2000; Jouve et al., 2000). These data suggest that if Hes1 has a role in segmentation, it may be compensated for by other cyclic Her genes, such as Hey2 or Hes7. Alternatively, Hes1 in particular, and Her genes in general may have no function in segmentation and their cyclic expression patterns may be by-products of their functional linkage to Notch signaling in other tissues such as the CNS.

We wished to test whether members of the Her family play wholly or partially redundant roles in segmentation, whether some Her family members might be central or coordinating components of oscillatory mechanisms and whether the Delta/Notch signaling system interacts with Her gene function. To this end, we show that expression of the zebrafish Her gene, her7, cycles in the zebrafish PSM, and determine the spatial relationship of her7 expression to the other cyclic genes. To test the role of cyclic Her genes in segmentation, we analyzed the segmental phenotypes that resulted from the reduction of function of her7 alone or in combination with another zebrafish cyclic Her gene, her1, and with mutations in Delta/Notch pathway genes. We present evidence that her7 and her1 are partially redundant components of a Delta/Notch-linked oscillator, and that, in the absence of both her1 and her7 function, there are no dynamic domains of cyclic gene expression at any stage of development, giving rise to an animal that displays a profound disorganization of somitogenesis along the entire length of its axis. Our data suggest that redundancy between the Her genes and genes of the Delta/Notch pathway underlies the robust formation of anterior somites in vertebrates.

Fish were kept on a 14 hour light/10 hour dark cycle following standard culture methods and staged (Kimmel et al., 1995). Mutant strains after eight (aei^tr233), beamter (bea^tm98), deadly seven (des^tp37) and white tail (wit^ta52b, also known as mindbomb) have been described previously (van Eeden et al., 1996).

A partial her7 cDNA was amplified using PCR from a 15-19 hpf embryonic cDNA library (B. Appel, Vanderbilt University) with the primer pair her7-1 (catatcctcattgatatcaac) and her7-2R (agagattacacaaggcccatca), designed using Accession Number AF240772 (M. Gajewski, C. Leve and D. Tautz, Universitaet zu Koeln) and subcloned to create pGEM TEasy-her7cds. To test the efficacy of morpholino action, the 5′ UTR of her7 was amplified from embryonic cDNA with the primers her7-3 (ccggatccctcgatgaaagacctccacct) and her7-5R (ccggatcctgcacgtgtactccaatagtt) and subcloned into the BamHI site of pCS2+ upstream of the coding sequence of eGFP (Clontech, Palo Alto, CA), which had been amplified with the primers eGFPATGB (ccggatcccaccatggtgagcaagggcgag) and eGFPX (ccgctcgagctacttgtacagctcgtccatgccga), and subcloned into BamHI/XhoI sites to create pCS2⁺her7 5′UTR-eGFP. All constructs were verified by sequencing both strands by the SynSeq facility at Princeton University.

In situ hybridization was essentially as described (Oates et al., 2000), and two color reactions were performed as previously described (Prince et al., 1998). Some of the riboprobes were generated from PCR-amplified cDNA templates subcloned into pGEM-TEasy (Promega, Madison, WI) after restriction and transcription with appropriate enzymes. The primers are as follows: for mespa (Sawada et al., 2000), mespa-1 (cagccatggacgcctccacgt) and mespa-2R (tggtcagcactgtccatggaa); for mespb (Sawada et al., 2000), mespb-1 (cgacatgcaaacctcaagcaaga) and mespb-2R (tccgtcatctccagtaagtctga); and for dlc (Smithers et al., 2000), delC-3 (gtctgctatcgttcagtagcaga) and delC-4r (gtgctccagattgaagaattct). The remaining riboprobes were synthesized from plasmids as described: myod (Weinberg et al., 1996); paraxial protocadherin (Yamamoto et al., 1998); deltaD (Dornseifer et al., 1997); hairy1 (Muller et al., 1996); notch1a, notch5 (Westin and Lardelli, 1997), lunatic fringe (Prince et al., 2001), twist (Morin-Kensicki and Eisen, 1997). The twitchin plasmid was a gift from Y-L. Yan and J. H. Postlethwaite (University of Oregon).

Synthesis and injection of mRNA was essentially as described (Oates et al., 2000). Briefly, capped mRNA was generated from linearized plasmid DNA using the mMessage mMachine kit (Ambion, TX), frozen in small aliquots and diluted to experimental concentrations immediately before injection with 0.2 M KCl. In order to control the injection volume, a fixed proportion (one-third of a volume) of 0.2 mg/ml Fast Green (Sigma, St. Louis, MO) was used to dilute the RNA preparations so that the injected bolus of approximately 0.2-0.5 nl could be visualized. Depending on the experiment, the mRNA was introduced to one cell of a two- to eight-cell stage embryo by pressure injection under a Zeiss Axioskop compound microscope (Carl Zeiss, NY). Location and translation of test gfp mRNA was followed during development on a Leica MZFLIII fluorescent dissection microscope (Leica, NY).

To ensure that the effect of antisense oligonucleotides would not be confounded by sequence polymorphism, the 5′ UTR of her7 was amplified using primers her7-3 and her7-5R (above) from cDNA isolated from three zebrafish strains held in our laboratory (*AB, TLF, and GH) and sequenced. Antisense morpholino oligonucleotides complementary to the 5′ regions of the her7 and her1 cDNAs were designed and synthesized by GeneTools LLC (Philomath, OR): her7m1, cagtctgtgccaggattttcattgc; her7m2, gaggatatgattccagaaaatgtcc; her1m1, ttcgacttgccatttttggagtaac; and unrelated control oligo, gcaaaacagctatcattagtcgtcc. Morpholinos were resuspended from lyophilized powder, then diluted to 10 ng/μl in 1× Daniaeu’s solution and stored at –20°C. Immediately before microinjection into the yolk streaming of two- to 16-cell embryos, the morpholino solutions were diluted to the appropriate concentrations into 0.2 mg/ml Fast Green.

The monophyletic Her family of basic helix-loop-helix (bHLH) transcription factors [group E according to Ledent and Vorwoort (Ledent and Vorwoort, 2001)] contains several members that display cyclic expression patterns in the PSM of vertebrates. While three Her genes with cyclic expression in the PSM are known in both the mouse and chick embryo (Bessho et al., 2001; Jouve et al., 2000; Leimeister et al., 2000; Palmeirim et al., 1997), only one zebrafish Her gene hairy1 (her1) is known to cycle (Holley et al., 2000; Sawada et al., 2000). To test whether zebrafish also possess more than one cyclic Her gene, we re-examined the expression of known Her genes. The her7 cDNA was amplified by PCR from embryonic cDNA and sequenced, confirming the amino acid sequence previously reported (Leve et al., 2001). Below, we show that the her7 gene of zebrafish exhibits cyclic expression in the PSM.

We examined the expression pattern of the her7 gene by whole-mount in situ hybridization during the period of embryonic development when somites are patterned and formed. Between 50% and 85% epiboly, her7 is expressed in a ring in the hypoblast (not shown). At 85% epiboly, the expression domain of her7 undergoes a transition, with bilaterally symmetrical stripes apparently separating from the dorsolateral margin and moving rostrally (Fig. 1A). This dynamic expression pattern is maintained throughout the segmentation period (Fig. 1B-F) and is highly similar to the cyclic patterns of her1 and dlc (Holley et al., 2000; Jiang et al., 2000; Sawada et al., 2000). In particular, because her7 is not expressed in any tissue outside the PSM, its expression most resembles that of her1 (Muller et al., 1996). To confirm that her7 expression is genuinely cyclic and to determine precisely the anterior-most extent of the her7 expression domain within the PSM, we compared her7 expression with that of myod (Weinberg et al., 1996), which is expressed in two stripes marking the presumptive caudal half-somite in the anterior of the PSM (Fig. 1G). In carefully staged embryos, cyclic her7 expression domains move rostrally relative to the static myod stripes, and arrest immediately caudal to the posterior-most myod stripe, suggesting that at this point they mark the presumptive rostral somite half. This arrest position is one segment more caudal than that seen with her1 (Holley et al., 2000; Sawada et al., 2000).

To determine the phase relationship between the cyclic expression domains of her7, her1 and dlc, we compared their expression in the same animal. Within the PSM, the domains of her7 and her1 shared the same anterior boundary, but the posterior boundary of her7 lay rostral to that of her1 (Fig. 1H). The anterior boundary of the her7 domain lay caudal to that of the dlc domain, and the posterior boundary lay rostral to that of dlc, meaning that the her7 expression domain was entirely nested within the dlc domain (Fig. 1I). We also compared the anterior boundary of the her7 expression domain with that of dld, which, although not thought to display a genuine cyclic expression pattern, nonetheless exhibits dynamic refinement of striped boundaries (Dornseifer et al., 1997; Holley et al., 2000; Jiang et al., 2000; Topczewska et al., 2001). The anterior boundary of the dld stripe was rostral to that of the her7 domain (Fig. 1J). In the tailbud, the anterior boundaries of the expression domains appeared to overlap. Thus, the cyclic mRNA expression domains of the three zebrafish cyclic genes are broadly in phase, but display different anterior and posterior boundaries, as illustrated in Fig. 1K. Consequently, a cell in the PSM would experience a succession of gene expression states as illustrated in Fig. 1L.

To examine the effect on segmentation of a loss of her7 function, we used two antisense morpholino-containing oligonucleotides (her7m1 and her7m2) targeted to independent regions of the 5′ end of the her7 mRNA (Fig. 2A). In the absence of a specific antibody that recognizes the Her7 protein, we tested the efficacy of the antisense approach by inhibiting the translation of a gfp mRNA carrying the 5′ UTR from her7. Injection of gfp mRNA alone resulted in brightly fluorescent embryos (Fig. 2B,C), and co-injection of her7m2 did not affect this expression (Fig. 2D,E). Co-injection of the target her7 5′UTR-gfp mRNA and an unrelated control morpholino left fluorescence unaffected (Fig. 2F,G), whereas co-injection of the target mRNA and her7m2 into embryos completely abolished fluorescence, but did not adversely affect gfp mRNA levels (Fig. 2H,I), consistent with a specific block in GFP translation.

We noticed that in embryos co-injected with her7 5′UTR-gfp mRNA and her7m2 morpholino, gfp RNA was more robustly detected than any of the other combinations (compare Fig. 2I with 2C,E,G). We therefore investigated whether a specific morpholino affected the stability of its target mRNA. Injection of synthetic her7 mRNA resulted in the appearance of low levels of ectopic her7 transcript in the CNS of embryos (Fig. 2K) in addition to the endogenous PSM expression pattern (Fig. 2J). Co-injection of her7m1 caused a dramatic increase in the level of her7 transcript throughout the embryo (Fig. 2L), indicating that the her7 mRNA had been stabilized in some manner.

In experiments described below, the effects of her7m2 were indistinguishable from those of her7m1, although the effective dose differed and the injection of morpholino of unrelated sequence did not produce any of the observed segmental defects. These data suggest that the morpholino treatment is capable of significantly and specifically reducing Her7 protein levels; however, without direct measurement of endogenous Her7 protein, we cannot be sure that it is entirely eliminated. We therefore refer in what follows to a decrease or reduction in Her7 function resulting from targeted morpholino treatment.

Injection of embryos with either her7m1 or her7m2 antisense morpholino resulted in a distinctive segmental disruption in the 26 hpf embryo (Fig. 3A-D, Table 1). These segmental defects were found only in the posterior trunk and tail, and close examination revealed an anterior limit at approximately myotome eight, with the mode at somite 10 (Fig. 3E). The most rostral defect was often a loss of bilateral symmetry in positioning of several myotome boundaries, which was followed caudally by disruption of boundary morphology (Fig. 3B). The defects increased in severity and number with increasing dose of injected antisense morpholino (Fig. 3B-D; Table 1). The dose had no significant effect on the distribution of anterior-most defects, and the anterior trunk segments were not significantly affected at any dose tested. Comparison of the anterior limit of defects and segment morphology of embryos with reduced her7 function with the phenotype of the segmentation mutants bea, aei, des revealed a similar distribution and appearance (see Fig. 6).

To test that the defects noted at 26 hpf were due to problems in segmentation of the PSM, we examined markers of segment polarity in the anterior PSM at the 10 somite stage. In her7m1- and her7m2-injected embryos we observed defective myod stripe formation in the anterior PSM, and also in formed somites more posterior than somite seven [22/28 embryos (79%), Fig. 3F-I]. Additionally, the expression of the dynamically expressed dld gene [16/17 embryos (94%), Fig. 3J,K], caudal segment polarity marker notch5 [8/9 (89%), Fig. 3L,M], and rostral segment polarity markers papc [6/9 (67%), Fig. 3N,O], mespa [13/14 (93%), Fig. 3P,Q] and mespb [17/19 (89%), Fig. 3R,S] were disrupted, resulting in a loss of stripe formation. The expression levels of notch1a were not perturbed in the PSM (0/10 embryos), although a lack of polarity was observed in formed somites (Fig. 3T,U). Thus, reduction of her7 function disturbs segmental gene expression patterning, but does not prevent the formation of mature paraxial mesoderm cells with either rostral or caudal half-segment identity. Combined, these results indicate that her7 has an essential, non-redundant role in patterning the segments of the posterior trunk and tail.

We wanted to know whether the segmental defects due to a reduction of her7 function arose because of a problem with the cyclic expression domains in the PSM, or whether they were attributable to a developmentally later event, perhaps a perturbation of segment polarity in the anterior PSM.

We examined the expression of dlc, her1 and her7 in her7m1- and her7m2-injected embryos, and observed a distinctive, dose-dependant loss of cyclic expression domain integrity. At lower doses, the normally sharp anterior boundary of the domains became diffuse, and cells expressing dlc, her1 and her7 were evident in the interstripe regions; this type of defect was termed ‘fuzzy stripes’ (Fig. 4B,H,N). At higher doses, the anterior boundaries became uneven and/or displayed breaks, and the interstripe regions were so densely populated with dlc-, her1- and her7-expressing cells that successive domains were often partially joined together; this was termed ‘disrupted stripes’ (Fig. 4C,I,O). At the highest doses, the cyclic expression domain structure was effectively destroyed, and a single diffuse region of expression in the anterior half of the PSM was observed; this effect was termed ‘salt and pepper’ (Fig. 4D,J,P). This class of expression pattern is strikingly similar to that seen for the cyclic genes in the aei, bea, des and wit segmentation mutants at comparable developmental stages (see Fig. 6K-Y) (Durbin et al., 2000; Holley et al., 2000; Jiang et al., 2000; van Eeden et al., 1998). These three phenotypic classes represent easily scored stages in what is likely to be a continuum of increasing cyclic expression domain perturbation, as indicated by the occurrence of increasing proportions of the more severe classes across the concentration series (Fig. 4E,K,Q).

Although in normal embryos the expression patterns of dlc, her1 and her7 cycle between states in which the posterior PSM and tailbud exhibits alternating high levels of transcript (embryos on left, Fig. 4A,G,M) and low levels (embryos on right), expression levels in the posterior PSM and tailbud of morpholino-injected embryos were always elevated (Fig. 4B-D,H-J,N-P). This suggests that a consequence of a reduction in her7 function is the de-repression of cyclic gene expression. Only in the case of her7 expression can we attribute some of this elevation to the effect of the her7 morpholino on endogenous her7 mRNA stability demonstrated above (Fig. 2J-L). Nevertheless, the three classes of disrupted gene expression were also observed for her7 expression (Fig. 4N-P). Thus, reduction in Her7 levels causes defects in the dynamic expression domains of all known cyclic genes, indicating that her7 encodes a component of the oscillator.

To determine the correspondence between loss of cyclic expression domain coherence and the onset of the somitogenic phenotype as determined by myotome boundary structure at 1 dpf, we measured the expression of dlc, her1 and her7 at various developmental stages up to and including those at which somitic development was perturbed. We observed a gradual loss of cyclic expression domain integrity over time that resembled the gradual loss seen in the dose response series (Fig. 4F,L,R). The first defects, consisting mostly of the fuzzy stripe class, were observed at bud stage, which occurs at 10 hpf, and by the three-somite stage (11 hpf) less than 50% of the embryos were normal in appearance. From seven- to 10-somite stages (12-14 hpf), there were essentially no normal-looking embryos, and the majority of defects shifted in severity from the fuzzy to the disrupted class. By the 13-somite stage (15.5 hpf), the majority of the embryos displayed defects in the salt and pepper class. Thus, the cyclic expression domain structure degrades with time, and the disruption of segmental pattern in the expression domains correlates well with the timing of somitogenic perturbation. Furthermore, the gradual degradation of domain coherence suggests that her7 alone does not encode a central component of the oscillator, but instead is likely to encode a component of a coordinating or coupling mechanism, as has been proposed for the Delta/Notch system (Jiang et al., 2000). We next investigated the relationship between her7 and Delta/Notch signaling.

Embryos carrying mutations in the beamter (bea), after eight (aei), deadly seven (des) or whitetail (wit) are thought to possess defective Delta/Notch signaling (Appel et al., 1999; Riley, 1999; Gray et al., 2001; Jiang et al., 2000; Jiang et al., 1996; Lawson et al., 2001). These mutants display segmentation defects similar to those seen in mice with Delta/Notch signaling system mutations: there is a progressive worsening of the orientation and spacing of epithelial furrows in posterior tissue concomitant with a loss of segment polarity. The bea, aei and des mutations produce a relatively distinct transition from well formed to abnormal somites, which occurs in bea after approximately the fourth somite stage, and in aei and des at the seventh somite stage (van Eeden et al., 1996). There is a more gradual degradation of somite integrity in the wit mutant. We examined her7 expression in these mutants and found a gradual degradation in the integrity of the cyclic expression domain boundaries that presaged the appearance of the defective morphological boundaries (Fig. 5B-D). In embryos staged after the overt segmentation phenotype was visible, there was a strong reduction in her7 levels accompanied by a complete loss of cyclic expression domain initiation or propagation (Fig. 5F-I). Thus, Delta/Notch signaling is required for the maintenance of coherent cyclic her7 gene expression domains in the PSM.

When intercrossed, the zebrafish mutants bea, des and aei do not produce an enhanced segmentation phenotype (Jiang et al., 2000; van Eeden et al., 1996; van Eeden et al., 1998), suggesting that they act in a similar genetic pathway. To test whether the her7 gene was involved with similar segmentation mechanisms to those disrupted by the bea, aei, des and wit mutations, we assessed whether a loss of her7 function induced by antisense morpholino injection was capable of enhancing or suppressing these mutant phenotypes (Fig. 6). No change in the severity of boundary defects was detected at 26 hpf in any combination of mutant and her7m1 or her7m2, indicating that in this respect, a reduction of her7 function behaves like aei, des and bea. However, reduction of her7 function in aei or des mutants enhanced the onset of the boundary phenotype, shifting the anterior limit of boundary defects from segment 7 to segment 5 (Fig. 6A,B,D,E,H,I). The bea mutant phenotype was more strongly enhanced by a decrease in her7 function, shifting the anterior limit of boundary defects from segment 4 to the extreme anterior end of the paraxial mesoderm (Fig. 6C,F,J). The boundary phenotype of uninjected wit mutants at 26 hpf is severe, possibly reflecting additional degeneration subsequent to initial somitogenic defects, and could not be meaningfully scored for alterations after injection with her7 morpholino (data not shown).

We compared the cyclic expression domain defects of the bea, aei, des and wit mutants at the 10-somite stage in the presence or absence of morpholino-induced her7 deficiency to determine if the observed rostral shift in the anterior limit of defects was accompanied by a novel cyclic expression domain defect in the PSM, such as loss of expression, or persistent expression in the mature paraxial mesoderm. A slight increase in expression levels in the posterior PSM and tail bud was observed in dlc, her1 or her7 expression in every mutant background after injection of either her7m1 or her7m2 (Fig. 6K-Y). However, no significant difference in dlc, her1 or her7 expression patterns was observed in the anterior PSM, suggesting that the change in anterior limit of boundary defects was not due to a qualitatively different type of defect in cyclic gene expression, but rather an acceleration of the degradative process. Combined, the data above indicate that a reduction in her7 function is not equivalent to a loss in either aei, des or bea, despite the similarities in phenotype, and suggests that her7 functions in a linked, but separable mechanism in segmentation.

Although her7 deficiency is sufficient to produce posterior segmental defects in wild-type zebrafish, the finding that anterior segments were affected in embryos sensitized by loss of Delta/Notch function indicates that her7 has functions in the anterior segments that are normally compensated by other molecules. To test the possibility that her1 shares some of the functions of her7, we first used antisense morpholinos to create a her1 reduction of function phenotype in zebrafish embryos (Fig. 7). Animals injected with her1m1 (Fig. 7A) at the highest concentration that did not induce nonspecific abnormalities (5 ng/μl) displayed a weakly penetrant myotome boundary phenotype at 26 hpf, involving isolated boundary defects and local regions of register defects in the anterior as well as posterior trunk (Fig. 7B,C, Table 1). This effect was preceded in mid-segmentation stages by correspondingly weak alterations in myod expression (9/24 embryos (37%), Fig. 7D,E), indicating that the defects were the result of events occurring during somitogenesis. Expression of dlc, her1 and her7 in the PSM of her1m1 morpholino-injected animals exhibited mild cyclic expression domain defects conforming to the fuzzy boundaries class [9/12 (75%), 7/13 (53%), 13/19 (81%), respectively; Fig. 7F-K]. While expression of her1 appeared elevated after her1m1 injection (Fig. 7H,I), this is likely to be due in part to the effect of specific morpholino on target mRNA stability (see Fig. 2). By contrast, her7 levels were reduced in the PSM of her1m1-injected embryos, although their cyclic domain structure was relatively intact (Fig. 7J,K), suggesting that her1 may play a role in her7 expression. These results indicate that like her7, her1 is a component of the oscillator, although one with a weak linkage.

The anterior defects seen in her1m1-injected embryos were surprising given the reported exclusion of her1 expression from cells thought to form the first four somites (Muller et al., 1996). We therefore examined the expression of her1 and mespa, a marker of the presumptive S0 somite (Durbin et al., 2000; Sawada et al., 2000), in embryos at bud stage (10 hpf) before the formation of the first somite. As shown in Fig. 7L,M, both mespa and her1 are expressed at the equator of the embryo, the site of the future first somitic furrow (Kimmel et al., 1995), and the cyclic her1 expression domains extend rostrally to the mespa stripe (Fig. 7N). These data indicate that her1 mRNA cycles in the presumptive cells of the first four somites as well as more posterior tissue, and are consistent with the anterior segmental defects reported above.

We next determined the effect of co-injecting her1m1 and her7m2. At the three- to four-somite stage, affected embryos possessed no epithelial furrows in the paraxial mesoderm but were otherwise normal (31/31 embryos; Fig. 8A,B). At the 10-somite stage, shallow disorganized furrows could be seen only in the anterior trunk, approximately to the level of the fifth or sixth somite (69/70 embryos; Fig. 8C,D). By the 12-somite stage, the furrows could be detected at the seventh or eighth somite level (36/36 embryos; data not shown). The furrows in her1m1/her7m2 morpholino-injected embryos differed from normal somite boundary furrows in important ways: (1) they did not show a regular segmental spacing; (2) their orientation was not normal to the AP axis of the embryo; (3) they often did not extend all the way through the paraxial mesoderm; and (4) their formation was delayed relative to normal somite boundary formation. Nevertheless, these furrows possessed nuclei aligned along their borders, similar to those in normal animals, consistent with an epithelial character (Fig. 8E,F). Thus, epithelial furrow formation was delayed by some 2 hours relative to control embryos, and had lost spatial organization.

At 26 hpf, embryos with her1 and her7 deficiency displayed a highly penetrant loss of boundary integrity and a large variation in apparent segment size throughout the paraxial mesoderm (Fig. 9A,B; Table 1). The number of segments affected and the frequency and severity of boundary defects was higher than seen in embryos injected with either of the morpholinos alone (Fig. 9C,D; Table 1), or by adding the effects of each morpholino together, indicating a synergistic effect of removing both her1 and her7 function. Furthermore, the severity of the boundary defects was greater than that seen in the segmental mutants bea, aei and des. The affected embryos expressed myod at 14 hpf (30/30 embryos; Fig. 9G) and possessed twitching striated muscle fibers at 26 hpf (32/32 embryos), indicating that muscle differentiation was unaffected. The expression of twist in sclerotome progenitors (Morin-Kensicki and Eisen, 1997) was not segmentally patterned, but was nevertheless robust, indicating that this derivative of the somite was also present (15/15 embryos, Fig. 9E,F).

We wanted to determine whether segment polarity had been affected by the reduction in her7 and her1 function, so we assayed the expression patterns of markers of rostral and caudal half-segment identity. myod, notch5 and dlc expression marks the caudal half-segment in the very anterior PSM and somites of normal embryos. In embryos with reduction of her7 and her1 function, the expression of all three markers was severely perturbed. myod-expressing cells were mixed together with non-expressing cells throughout the paraxial mesoderm resulting in a loss of stripes (29/30 embryos; Fig. 9G). notch5 expression was similarly disorganized and was also reduced in level (27/27 embryos; Fig. 9H), and dlc expression was severely reduced in the mature paraxial mesoderm (28/28 embryos; Fig. 9U). The expression of markers of the rostral half-segment such as fgf8, lfng and dld was likewise thoroughly disorganized, leaving little or no evidence of segmental pattern. While fgf8 was expressed in a rostrally tapering, diffuse band in the posterior of the mature paraxial mesoderm (28/28 embryos; Fig. 9I), lfng-expressing cells were arranged in clumps and patches, particularly in the more anterior paraxial mesoderm (22/22, Fig. 9J) and the expression of dld was severely reduced in mature paraxial mesoderm (27/27; Fig. 9V). By contrast, no change in notch1a expression could be detected (29/29 embryos; Fig. 9K). In the anterior-most PSM, the expression of papc, mespa and mespb is usually restricted to cells that appear to lie in the presumptive rostral segment-half. However, after injection of the anti-Her morpholinos, this restriction was lost and expression patterns instead displayed a salt and pepper-like intermingling of expressing and non-expressing cells (papc 35/35, mespb 18/18 embryos; Fig. 9L-N). In addition, mespa levels were reduced (24/24; Fig. 9M). Thus, although markers of both rostral and caudal half-segment identity are expressed in the paraxial mesoderm of embryos with reduced her7 and her1 function, there is a profound loss of rostrocaudal polarity.

Finally, we wanted to determine the integrity of cyclic gene expression domains in these embryos. At the 10-somite stage, embryos with reduced her1 and her7 function display a fully penetrant salt and pepper class cyclic domain defect (her1 25/25, her7 23/23, dlc 28/28 embryos, Fig. 9O,P,U, respectively). A similar phenotype was observed for dld expression in the PSM (27/27 embryos; Fig. 9V). In affected embryos, expression of the cyclic genes in the PSM is stronger and more widespread than in her7m2-injected aei, des, bea or wit mutant embryos (compare Fig. 9O,P,U to Fig. 6K-Y), suggesting a more severe loss of repression. To determine the time course of cyclic gene expression domain degradation, we analyzed the expression of dlc, her1 and her7 in injected embryos at 80% epiboly, 90% epiboly, bud and three-somite stages. There was no evidence of organized cyclic domain expression for any of these genes over this period (80% epiboly 0/13, 90% 0/16, bud 0/28, three somite 0/27 embryos, Fig. 9Q-T, respectively; data not shown for her1 and her7), indicating that segmental defects at the anterior end of the paraxial mesoderm correlated with a failure from the beginning of the somitogeneic stage to initiate dynamic gene expression domains. Thus, her7 and her1 together define a central component of the cellular oscillator.

To test whether the severe segmental defects seen with a reduction in function of her1 and her7 could be enhanced by the additional removal of Delta/Notch signaling, we assessed the 26 hpf phenotype after injection of both morpholinos into bea mutant embryos. The boundary defects in these mutant animals were indistinguishable from those seen after her7m2 and her1m1 co-injection into wild type (Fig. 9W,X), and the anterior limit of segmental defects was not significantly different (Fig. 9Y), indicating that loss of Delta/Notch function does not enhance the segmental disruption further.

In this report, we have investigated the function in segmentation of two zebrafish cyclic Her transcription factors, her7 and her1. Our results indicate that while each has a distinct and non-redundant role in controlling segmentation, together they constitute a central component of a somitic oscillator without which coordinated, dynamic domains of cyclic gene expression are never established in the fish.

her7 mRNA is expressed in the hypoblast, tailbud and paraxial mesoderm throughout the segmentation period in rostrally moving gene expression domains characteristic of a cyclic gene. Comparison of her7 expression with other cyclic genes dlc and her1 demonstrates that the dynamic expression domains possess a complex internal structure. Although the expression domains are broadly in phase with each other, the Her genes are asymmetrically nested within the delta expression stripe. These findings raise the possibility that cross regulation between cyclic genes may be responsible for some aspects of the cyclic phenomenon. A reduction of Delta/Notch signaling gradually destroys the her7 cyclic expression domains, with a close temporal correlation between the loss of domain integrity and the onset of posterior somite defects, in agreement with previous findings on the effect of Delta/Notch mutations on cyclic genes.

her7 is required for the correct segmentation of posterior trunk and tail, giving a reduction of function phenotype remarkably similar to the Delta/Notch mutants aei (deltaD), des and bea. However, in these genetic backgrounds, her7 is also required for the correct segmentation of more anterior somites, indicating a redundant role for her7 with Delta/Notch signaling in the anterior trunk. Underlying the disruption of overt segmentation in these animals are profound defects in the coherence and maintenance of the cyclic expression domains of dlc, her1 and her7 itself, indicating that her7 is a component of the oscillator. Furthermore, without her7, the level of her1 and dlc expression is elevated in the PSM, suggesting that her7 normally functions to repress their expression. A reduction in her1 function, in contrast, produces mild, isolated segmental disturbances without A/P restriction.

Combining a reduction in her7 and her1 function produces a striking loss of segmental organization along the entire anteroposterior length of the paraxial mesoderm, including the anterior trunk, which cannot be enhanced by a further loss of Delta/Notch signaling. This indicates that her7 also has redundant functions with her1 in the anterior trunk. Cyclic gene expression does not exhibit coordination into dynamic domains at any stage in animals with this extreme phenotype, indicating that the Her genes constitute a critical component of the segmentation clock. As these embryos form epithelial furrows in, and develop twitching muscles and sclerotome from the paraxial mesoderm, these results demonstrate that the function of cyclic expression domains is limited to the positioning of the segmental boundaries and is not required for the morphological processes of epithelial boundary formation or the differentiation of somitic derivatives. Furthermore, our findings show a clear role for genetic redundancy in the robust development of the somites of the anterior trunk.

As both her7 and her1 are expressed and appear to cycle with equivalent periodicity throughout segmentation stages, it was surprising that they generated such distinct reduction-of-function phenotypes. In both cases, defects in expression domain coherence were observed, indicating that each Her gene can be considered a component of the segmentation oscillator. The restriction of segmental defects to the posterior trunk and tail in her7-deficient embryos and concomitant degradation of expression domain coherence indicates that the function of her7 involves coordinating the oscillations of neighboring cells. The similarity of this phenotype in morphology and effects on gene expression patterns to that of the aei and des mutations strongly suggests that her7 functions in concert with these genes in the posterior body. By contrast, the lack of AP restriction and spatially isolated nature of the segmentation defects seen after reduction of her1 suggest that her1 function may be required in some stochastic manner along the AP axis. The observation that higher doses of her1 morpholino led to nonspecific defects and embryo death, along with the complete penetrance of the combined her7/her1 reduction-of-function phenotype argue against a simple failure to reduce levels of Her1 translation in a spatially homogeneous manner. We speculate that her1 may perform an ‘error-correcting’ role in segmentation that is required only when external perturbation or locally aberrant cell mixing produce potential cyclic gene expression domain defects.

A reduction in function of her7 in the genetic background of a loss of function in Delta/Notch signaling created a synergistic phenotype in which more anterior defects were observed than in either individual lesion. In the case of her7 reduction in an aei or des mutant background, the rostral shift was modest but highly reproducible, while in the bea mutant background, simultaneous reduction of her7 function led to a larger rostral shift that included defects at the anterior-most end of the paraxial mesoderm. Thus, in the absence of bea function, her7 is sufficient to mediate segmentation in the anterior trunk, and vice versa, indicating that both her7 and bea possess functions in segmental patterning of the entire axis that are usually masked by each other. These observations imply that Delta/Notch signaling defined by these mutants and Her7 are involved in partially overlapping mechanisms leading to correct segmentation. What might be the molecular basis for this redundancy? There are two known Notch receptors (notch1a and notch6), and two Delta ligands (dlc and dld) expressed throughout the zebrafish PSM (Bierkamp and Campos-Ortega, 1993; Dornseifer et al., 1997; Smithers et al., 2000; Westin and Lardelli, 1997). The relative specificity of these Delta proteins for the Notch receptors is unknown, but the possibility exists that several distinct Notch signaling pathways are active in the PSM. If her7 were to mediate input from more than one signaling pathway in segmentation, the requirement for signals beyond the aei/des/bea pathway would only become evident when her7 function was additionally compromised. The correct patterning of anterior segments in the absence of Her7 indicates that these signals in turn are not restricted to Her7 as a choice for mediating their output. The similar expression pattern and putative biochemical properties of her1 suggest it as a plausible alternative.

Combining a reduction in her7 and her1 function produces a dramatic segmental phenotype that affects the anterior-most paraxial mesoderm and causes boundary defects more severe than any Delta/Notch mutant in zebrafish, and, to our knowledge, at an earlier stage in the process of segmentation than mutants in other species. The zebrafish fused somites mutant possesses segmental disruption extending from the anterior-most paraxial mesoderm, but this phenotype is characterized by a relatively late block in segment polarity downstream of cyclic gene expression domain formation and hence oscillator function (Holley et al., 2000; Sawada et al., 2000; van Eeden et al., 1998). The anterior somitic defects seen in zebrafish embryos with a morpholino-induced foxc1a reduction of function, and in Foxc1/Foxc2 compound homozygote mice, are likewise caused by a failure to generate rostrocaudal segment polarity in the anterior PSM, and are accompanied by an intact oscillator (Kume et al., 2001; Topczewska et al., 2001). The double presenilin homozygote (Psen1/Psen2) mutant mouse embryo shows no somite formation or apparent segment polarity, but the severely pleiotropic phenotype and the lack of data on cyclic genes in these animals confound an accurate assessment of their segmental state at present (Donoviel et al., 1999). Thus, previously described mutants or mutant combinations with anterior somite defects exhibit a failure in the maturation of presumptive somites downstream of apparently intact oscillator function.

By contrast, the phenotype of a reduction in her7 and her1 function combines an apparently disrupted oscillator with aberrant segmentation along the AP axis in an otherwise normal embryo. Those aspects of segmentation that are affected in her1/her7 reduction of function embryos reveal the biological roles of the dynamic domains of cyclic gene expression. Most obviously, rostrocaudal segment polarization is profoundly disrupted, indicating that the organization of cyclic gene expression into coherent, dynamic domains is required for the establishment of segment polarity. However, as the production of cells with both rostral and caudal segment identity occurs, the dynamic domains of cyclic gene expression are not required for the production of either identity in particular. Similarly, the presence of differentiated muscle and sclerotome indicates that production of somite derivatives is not hindered in the absence of coordinated oscillations. Finally, the observation that epithelial furrow formation is present, if delayed, demonstrates that the process of furrow morphogenesis itself is not a direct output of the dynamic expression domains of cyclic genes. The appearance of clusters of lfng-expressing cells in the anterior paraxial mesoderm in the affected embryos may reflect a sorting out of cells that have taken rostral or caudal fates in the anterior PSM, and the delayed epithelialization may follow this delayed segregation (Durbin et al., 2000). Alternatively, the final position of the furrows may be constrained by physical properties of the epithelia, or may be essentially random. Thus, we hypothesize that the role of the coordination of oscillations into dynamic expression domains is restricted to the placement of the segmental boundaries, presumably through the spatial control of rostral/caudal segment polarity. This conclusion, drawn from the most extreme phenotype is well supported from analysis of the milder phenotypes seen at low doses of her7 or her7 and her1 morpholinos in which the most common defects are the failure of boundaries on either side of the midline to remain bilaterally symmetrical. The correlation of these register defects with a predominance of the fuzzy boundaries defect in cyclic gene expression indicates that the distance between successive boundaries is the first parameter to change as oscillator function is compromised. These conclusions do not imply that there are no other central components to the segmentation oscillator and the redundancy that we have demonstrated is in accordance with previous speculations on a network of HER proteins involved in segmentation (Leimeister et al., 2000).

The segments of the vertebrate anterior trunk appear to be formed in a more robust manner than those of the posterior trunk and tail, as mutations and treatments leading to somitic patterning defects usually do not affect the anterior body (Conlon et al., 1995; Donoviel et al., 1999; Evrard et al., 1998; Holley et al., 2000; Hrabe de Angelis et al., 1997; Krebs et al., 2000; Saga et al., 1997; Takada et al., 1994; van Eeden et al., 1996; Wong et al., 1997; Yoon and Wold, 2000; Zhang and Gridley, 1998). These data have been plausibly interpreted to mean the existence of distinct mechanisms for segmentation along the AP axis, perhaps reflecting an ancient division of the body plan that was secondarily modified (Holley et al., 2000). When intercrossed to produce double homozygotes, the segmentation mutants bea, aei and des display a simple epistasis; bea;des or bea;aei resemble bea, and aei;des cannot be distinguished from either aei or des (Jiang et al., 2000; van Eeden et al., 1996; van Eeden et al., 1998). These results suggest that aei, des and bea act in the same pathway or are elements of the same mechanism. The simple epistatic interaction of the zebrafish aei, des and bea mutants is consistent with a strategy for segmentation that is, in some respect, modular along the AP axis. Our results showing posterior segmental defects after a reduction of her7 function appear to support this notion, and are consistent with a role for her7 as a target gene of a Delta/Notch pathway involving DeltaD and the protein products of the des and bea loci. However, the synergistic interaction of her7 function with Delta/Notch signaling and with her1 described above, suggests that genetic redundancy plays an important role in the apparent separation of anterior and posterior segmentation.

In support of this idea, we note that several mutations of Delta/Notch signaling genes in mouse show synergistic effects when trans-homozygous, in contrast to the zebrafish bea/aei/des series. Notch4 mouse mutants do not exhibit somitogenic defects, but Notch1/Notch4 double mutant embryos show synergistic effects on several tissues, including a more rostral onset of somite defects than the Notch1 mutant embryo (Conlon et al., 1995; Krebs et al., 2000). Similarly, although the Presenilin 2 (Psen2) mutant shows no somitogenic abnormalities, the Psen1/Psen2 double homozygote, as mentioned above, as well as the Psen1^–/–Psen2^+/– genotype give rise to more severe segmentation phenotypes than the Psen1 mutant alone, extending into the anterior paraxial mesoderm (Donoviel et al., 1999; Shen et al., 1997; Wong et al., 1997). The redundancy evident in vertebrate Notch and Psen function may be a general property of Delta/Notch signaling, as synergistic phenotypes are also observed by combining mutations in C. elegans Notch or Psen homologs (Lambie and Kimble, 1991; Li and Greenwald, 1997; Westlund et al., 1999). Of course, these observations do not rule out the existence of a gene or genes whose loss of function might duplicate the severe effects on anterior segmentation found to date only in double or triple loss of function experiments.

Thus, redundancy between some of the components of the segmental machinery seems to be generally responsible for the robust formation of anterior segments in vertebrates. While removal of two elements with overlapping function generates anterior segmental defects (e.g. her7 and bea), elimination of each element individually does not. This redundancy among components is not observed in the posterior trunk and tail regions, where removal of a single element is sufficient to cause segmental abnormalities (e.g. her7 or bea). In the absence of data about the interaction of her1 and Delta/Notch signaling, the weak effects of her1 morpholinos on anterior boundaries could be interpreted to indicate a HER-dependant, Delta/Notch signaling-independent mechanism in anterior segmentation. We are currently investigating this possibility.

An important question, first raised by Palmeirim et al. (Palmeirim et al., 1997), and still the subject of debate, is whether the visible dynamic expression domains are the output of a still-hidden oscillator or whether interactions between the cyclic genes themselves may be responsible for the cyclic behavior. Although our findings cannot exclude the existence of an independent underlying oscillator, and leave open the possibility that individual cells in the PSM continue to oscillate in some manner in the absence of HER and Delta/Notch function, they demonstrate that the coordination of oscillations into dynamic expression domains is controlled throughout segmentation at the level of Delta/Notch-linked HER function, and suggest that any putative underlying oscillator has, in isolation, little or no effect on somite boundary placement. Below, we integrate our findings with previous work on the interaction of Her and Delta/Notch gene function in the PSM of vertebrates and argue for a role for the cyclic genes themselves in the generation of cyclic phenomena.

Based on the overlap of cyclic gene expression measured in the PSM of chick embryos, the domains of the various cyclic genes are thought to be in phase and thus form a combined expression domain (Jouve et al., 2000; McGrew et al., 1998). Our findings agree generally with this view, but reveal a complex interrelationship between the gene expression domains within this combined cyclic expression domain. The relatively narrow AP dimension of the her7 domain and the spatial resolution afforded by two-color in situ hybridization in zebrafish embryos allowed us to determine that the anterior boundary of the her7 expression domain is shared with her1, but that it is nested entirely within the dlc domain (illustrated in Fig. 1K). Thus, a given cell in the PSM of a zebrafish embryo would experience a stereotypical cycle of gene expression each time a dynamic domain passed through (illustrated in Fig. 1J). First it expresses dld and dlc, then, second, simultaneously upregulates her7 and her1, before, third, downregulating her7 and dld, and, finally, downregulating her1 and dlc. A cell would repeat this cycle once every 30 minutes (the interval between the formation of successive boundaries in zebrafish), and would undergo at least six cycles, but potentially many more (A. O., unpublished), before ceasing oscillatory behavior and then becoming incorporated into a somite. These observations raise the possibility that sequential interactions between the various components that constitute the dynamic expression domains could play a role in their generation.

Previous work has shown that Notch signaling can activate the transcription of Her genes in vertebrate cells (de la Pompa et al., 1997; Jarriault et al., 1995; Jarriault et al., 1998; Maier and Gessler, 2000), and indeed an activated notch1a receptor causes ubiquitous expression of her1 in the PSM of zebrafish embryos (Takke and Campos-Ortega, 1999). Additionally, Her proteins act as transcriptional repressors in vertebrate cells on their own promoters and those of other genes (Nakagawa et al., 2000) and in zebrafish, overexpression of her1 leads to a decreased level of dld and dlc expression in the PSM (Takke and Campos-Ortega, 1999). Consistent with this notion, we have now shown that a reduction of her7, or both her7 and her1 function causes the widespread, elevated expression of dlc and her1 in zebrafish PSM (Figs 4, 7, 9). Indeed, these findings suggest that spatially controlled repression has an important role in the emergence of dynamic expression domains, and imply that the promoters of the cyclic genes are otherwise constitutively active in the PSM. Combined, these results suggest that the transcriptional response to Notch signaling in the zebrafish PSM can be negatively regulated after a short time lag by the action of induced HER proteins. What might activate the Notch signaling?

Although injection of synthetic dld or dlc mRNA into zebrafish embryos perturbs segmentation, it does not appear to disrupt the striped pattern of her1 expression (Dornseifer et al., 1997; Takke and Campos-Ortega, 1999), leading to the hypothesis that the delta genes act only at the transition to epithelial somites. However, we do not yet understand the dynamics of proteins from the cyclic genes, and it remains a possibility that levels of Delta proteins fluctuate markedly in the PSM, only stabilizing at the anterior end, and thus reducing the efficacy of delta gene overexpression in disrupting cyclic behavior. This limitation does not appear to hinder the overexpression of the nuclear Her7 (A. O., unpublished) or Her1 (Takke and Campos-Ortega, 1999). Our results that show a strong reduction in her7 cyclic domain coherence and expression levels in the dld mutant aei (Fig. 5) provide genetic evidence that at least one delta gene acts on cyclic behavior in the PSM, consistent with the findings of others showing perturbed her1 and dlc expression in aei (Durbin et al., 2000; Holley et al., 2000; Jiang et al., 2000; van Eeden et al., 1998). Thus, there appears to be the potential for a negative feedback loop involving Delta/Notch signaling and HER repressors in the zebrafish PSM that would be capable in principle of generating oscillatory behavior.

The short-term behavior of cyclic expression domains has been studied in chick using surgical explant and culture techniques. For periods of the order of the interval between the formation of successive somites (90 minutes in chick), the propagation of cyclic domains does not require direct propagation of a signal from the posterior of the PSM or signals from neighboring tissues such as epidermis, notochord or Hensen’s node (Aulehla and Johnson, 1999; Forsberg et al., 1998; Palmeirim et al., 1997). Thus, the PSM can be thought of as an autonomously oscillating tissue. It is unclear to what extent the individual cells of the PSM are strictly autonomous oscillators, however, as the oscillations of isolated PSM cells have not been measured. The possibility that over longer time scales, signaling between PSM cells might influence the phase of a given cell cannot therefore be excluded. If this were the case, the cells would be acting as coupled oscillators and Delta/Notch signaling has been proposed to fulfil this specific communication role (Jiang et al., 2000). The notion that the Delta/Notch system might perform a more central role in the oscillator has also been discussed previously (Jouve et al., 2000). Combining our findings with the work of others discussed above, we propose that an oscillator consisting of HER and Delta/Notch components provides the fundamental periodicity in addition to the local phase synchronization of cells.

One objection to this proposal is the observation that in chick, the dynamic expression domains of the three cyclic Her genes, hairy1, hairy2 and hey2 appear to propagate relatively intact for a time equivalent to the interval of somite formation before showing disruption after treatment with the protein synthesis inhibitor cycloheximide (Jouve et al., 2000; Leimeister et al., 2000; Palmeirim et al., 1997). This suggests that continued oscillations do not require the translation of new proteins, which is a crucial part of the feedback loop that generates the oscillations in the above proposal. We note that, first, in these animals, complete inhibition of protein synthesis was not achieved (Palmeirim et al., 1997); and second, in contrast to the Her genes, the Lfng domain ceases propagation immediately (McGrew et al., 1998), indicating that the oscillations of at least one cycling gene are dependant on sufficient levels of protein synthesis. Finally, from our work, dynamic expression domains of the cyclic genes do not form if the translation of only her1 and her7 is inhibited (Fig. 9). Thus, we argue that translation of new proteins is an essential part of the generation of the dynamic expression domains of the cyclic genes and hence in oscillator function as well.

Our finding that cyclic genes are broadly expressed in the PSM in the absence of HER function, and that the Delta genes are expressed alone on the leading edge of combined dynamic expression domains with Her genes nested internally, suggests a simple model for the organization of oscillations into dynamic wave-like expression domains based on a simple negative feedback loop created by HER-linked Delta/Notch signaling. We make the following assumptions: (1) a sharp, sigmoidal repression curve for the HER proteins on target promoters; (2) the dlc gene is more sensitive in response to Notch signaling than either her1 or her7, but the her genes are more sensitive to HER-mediated repression; and (3) the mRNAs and proteins of all the cyclic genes possess very short half-lives, such that in the absence of continued transcriptional activation they quickly fall to undetectable levels. Combining our findings with those of others, we propose that the following series of events, diagrammed in Fig. 10, comprise a core oscillator capable of giving rise to the complex dynamic expression domains seen in the zebrafish PSM.

Cells in the tailbud, perhaps under the influence of factors controlling mesodermal fate during gastrulation, transcribe a constant elevated level of dld and a basal level of dlc (Fig. 10A). Correct presentation of the Delta proteins activates Notch signaling both within the same cell and on neighboring cells. The dlc gene is an immediate target of Notch signaling, providing a rapid amplification of basal dlc expression (Fig. 10B). After a short lag time, Her target gene mRNA levels are also amplified by Notch signaling in these cells (Fig. 10C), then translated, and finally the HER proteins act on the promoters of their own and the dlc genes, switching off the loop (Fig. 10D). After degradation of HER proteins, the target promoters are released from repression and Delta-activated Notch signaling starts the cycle again (Fig. 10A). These steps give rise to and maintain a population of oscillating cells that are highly synchronized, and establish a periodicity of 30 minutes determined largely by the half-life of the proteins involved. As the feedback loop may exist within single cells, the cells can behave as autonomous oscillators (Fig. 10A-D). Because the feedback loop is also influenced by Delta proteins on adjacent cells, the cells can act as coupled oscillators, adjusting their expression phase to match their neighbors (Fig. 10E). The constant expression of dld may serve to entrain the new cells entering the tailbud to the period of the oscillator.

As the tailbud recedes because of convergent extension and other cell movements (Kanki and Ho, 1997), cells enter the PSM and exit the influence of gastrulation-associated factors and the domain of constant dld expression. The feedback loop between dlc and HER proteins established in the tailbud continues to operate. Cells correctly presenting DeltaC protein on their surface can now phase advance the activation of Notch signaling in their rostral neighbors, forming the anterior boundary of an emerging wave-like expression domain. When HER levels are sufficient they first repress their own promoters, then that of dlc, giving rise to the posterior boundary of the dynamic expression domain. This domain does not ultimately require intercellular signaling to propagate, because cells (1) were locally synchronized when they left the tailbud, (2) oscillate with a period according to their internal levels of HER protein and (3) use the PSM basal dlc (and/or dld) expression to start the cycle over again. Hence, physical transection of the PSM would not stop an expression domain propagating with this mechanism. Cells intermingling and intercalating as the embryo undergoes convergent extension are continually phase adjusted by the level of expression of Delta protein on their new neighbors. Thus, the combined action of cell-autonomous oscillation and adjustment of the phase of neighboring cells gives rise to a coherent wave-like expression domain with a sharp anterior boundary. Note that in this model there is no hidden oscillator – the HER-linked Delta/Notch signaling pathway feedback loop is the oscillator.

Currently, the favored schema for understanding segmentation in vertebrates is the recent incarnation of the Clock and Wavefront model (Cooke, 1998; Cooke and Zeemann, 1975; Palmeirim et al., 1997). In this model, a wavefront sweeping from anterior to posterior along the paraxial mesoderm interacts with a molecular oscillator (clock) and freezes the phase of the oscillator in the cell as it passes, thereby generating a succession of evenly spaced cell states. The HER-linked Delta/Notch feedback loop and resulting dynamic expression domains we have discussed above closely fits the expectations of the molecular clock in the revised Clock and Wavefront model and in the more general case of flow distributed oscillators (Kaern et al., 2000). However, the molecular mechanism whereby the wave-like expression domains slow down and produce segment polarity in the anterior PSM, thereby positioning the boundary between segments, remains unclear, and our findings do not currently address this problem.

The level of FGF signaling in cells in the anterior PSM of chicks and zebrafish appears to play a role in controlling the position at which the dynamic expression domains stop, and hence where the next boundary forms (Dubrulle et al., 2001; Sawada et al., 2001). In this regard, it conforms closely to the wavefront in the Clock and Wavefront model. In the zebrafish PSM, blocking FGF signaling causes her1 expression levels to rise prematurely and stops the dynamic domain of expression short of its normal anterior limit (Sawada et al., 2001). This suggests that the function of FGF signaling in the posterior PSM is to reduce the expression levels of the cyclic genes and allow the dynamic domains to propagate rostrally. In the light of our proposal, this would suggest that FGF signaling and HER-linked Delta/Notch signaling interact in the PSM. We offer the speculative hypothesis that when FGF signaling decreases below a certain threshold in cells that become located in a sufficiently anterior position in the PSM, the resulting increased levels of cyclic HER repressor proteins act to shut down the transcription of the Her genes permanently, thus ending cyclic behavior. As both dld and dlc continue to be expressed at lower levels in stable stripes in formed somites, their expression must become independent of oscillatory dynamics at this time. This model, while speculative, is easily testable and such experiments are ongoing in our laboratory.

We acknowledge Susan Cole and Mark Mandel for discussion; Ashley Bruce, Nipam Patel, Laurel Rohde and Isaac Skromne for comments on the manuscript; Michael Hunter, and Cristin Howley for technical advice; and Victoria Prince, in whose laboratory this work was completed. A. O. was supported by a Ludwig Institute for Cancer Research Postdoctoral Fellowship. This work was supported by grants from the NIH, NSF and March of Dimes Foundation.