Oscillatory lunatic fringe activity is crucial for segmentation of the anterior but not posterior skeleton

The segmentation of the vertebrate embryo is most obvious during the process of somitogenesis. Somites are the embryonic precursors to the axial skeleton, striated muscle and dermis of the back, and are formed by sequential budding from the anterior-most region of the presomitic mesoderm (PSM)(reviewed by Christ et al.,1998; Gossler and Hrabe de Angelis, 1998). This process is dynamic and complex. During gastrulation, cells enter the presomitic mesoderm via the primitive streak. Later in development (∼10.0 dpc) the tailbud forms, and further mesodermal cells arise from this structure (Gossler and Tam, 2002). Although this distinction between primary body formation (giving rise to cervical, thoracic and lumbar vertebrae) and secondary body formation (giving rise to post-anal structures) was originally proposed in 1925 (Holmdahl,1925), it remains unclear how, and to what extent, the genetic regulation of somitogenesis between these two processes may vary (reviewed by Handrigan, 2003).

Several models for the control of somitogenesis invoke a clock that provides a timing mechanism for segmentation(Cooke and Zeeman, 1976; Kerszberg and Wolpert, 2000; Meinhardt, 1986; Schnell and Maini, 2000). These models differ in their specifics, but all include an oscillating activity in the PSM with a period identical to the rate of somite formation. Molecular evidence for the segmentation clock came initially from the cyclic expression pattern of chicken c-hairy RNA(Palmeirim et al., 1997). Shortly thereafter, lunatic fringe (Lfng) RNA was found to have oscillatory expression patterns in the PSM, linking it to the clock as well(Aulehla and Johnson, 1999; Forsberg et al., 1998; McGrew et al., 1998).

The importance of Notch signaling during vertebrate segmentation is evident from the phenotypes associated with mutations in Notch pathway genes, many of which cause defects in embryonic segmentation. Furthermore, cyclic gene expression has been described in the presomitic mesoderm for many other genes linked to the Notch signaling pathway in mouse, zebrafish and chick (reviewed by Rida et al., 2004; Shifley and Cole, 2007). The Wnt pathway has also been linked to the clock. Both Axin2 and Nkd1 RNA levels oscillate in the PSM, and it has been suggested that the Wnt pathway lies upstream of oscillatory Notch signaling(Aulehla et al., 2003; Ishikawa et al., 2004). More recently, a large number of oscillatory genes have been identified, many of which are linked to the Notch, Wnt or FGF pathways(Dequeant et al., 2006),suggesting complex clock regulation involving multiple signaling pathways.

The analysis of Notch signaling in the segmentation clock mechanism is complicated by the fact that this pathway plays multiple roles during somitogenesis. The PSM can be divided into functionally distinct regions based on RNA expression patterns (reviewed by Saga and Takeda, 2001). In the posterior PSM (region I), cyclic expression of several genes reflects the function of the segmentation clock. In the anterior PSM (region II) the expression of the cycling genes is stabilized, and the pre-somites develop rostral and caudal compartments. These regions are demarcated by the graded expression of FGF8 in region I, which has been suggested to maintain the immature state of the cells (Dubrulle et al., 2001; Sawada et al.,2001). Several lines of evidence suggest that Notch signaling plays distinct roles in these two regions. In region I of the PSM, Notch activity levels oscillate, suggesting its function in this region is linked to the clock (Huppert et al.,2005; Morimoto et al.,2005). Some models suggest that this oscillatory activation may be achieved partially through the transitory inhibition of Notch signaling via its glycosylation by LFNG in the Golgi, and transcriptional feedback loops involving Hes7 (Dale et al.,2003). This oscillatory mechanism, however, clearly receives input from other members of the Notch pathway and from other signaling pathways,including Wnt and FGF. This complex network of interlocked oscillatory genes has been proposed to contribute to the robust nature of somitogenesis(Dequeant et al., 2006). Notch signaling also plays crucial roles in the patterning of the presumptive somites in region II of the PSM. It appears that interplay between the Mesp genes and the Notch pathway is required for the establishment of rostrocaudal polarity in the developing somites, with Mesp2 acting through the Notch pathway to downregulate Dll1 expression in the presumptive rostral somite compartment, while in the presumptive caudal compartment, Notch signaling upregulates Dll1 expression. Lfng is a direct target of Mesp2, and its stable expression in the rostral compartment may inhibit Notch signaling in this compartment(Morimoto et al., 2005; Takahashi et al., 2000).

We and others have defined genomic sequences sufficient to direct cyclic expression of Lfng in the PSM, and demonstrated that independent Lfng cis-acting regulatory regions drive stable RNA expression in the rostral compartment of the developing somites in the anterior PSM(Cole et al., 2002; Morales et al., 2002). Deletion of a conserved regulatory element termed fringe clock element 1(FCE1) from Lfng reporter transgenes eliminates cyclic expression in the caudal PSM, while maintaining expression in the anterior PSM, reflecting the distinct roles of Lfng in the segmentation clock and in R/C patterning of developing somites. Thus, the complex phenotypes of Lfng^-/- mice may arise from disruption of both of these roles, with variations in somite size perhaps resulting from impaired clock function, while the apparent mingling of somite compartments might be exacerbated by altered R/C patterning.

To dissect the functions of the Notch pathway during segmentation, we perturbed only one of the roles of Notch signaling, by disrupting oscillatory Lfng expression in region I of the PSM, while sparing its expression in region II of the PSM. We report here that the clock and patterning roles of Lfng during somitogenesis are functionally separable. Strikingly, we find that the loss of oscillatory Lfng expression and Notch1 activity in region I of the PSM has more severe effects during the segmentation of the thoracic and lumbar skeleton than the sacral and tail skeleton. This suggests that oscillatory Notch1 activation in the segmentation clock is much more important during primary body formation than during secondary body formation. By contrast, the specific localization of Notch activity to the presumptive caudal compartment of the pre-somite in region II of the PSM is important throughout development.

FCE1 and minimal flanking sequences were deleted from the Lfngfragment extending from the 5′XhoI site in the 5′ flank to the HindIII site in intron 1 and replaced with an EcoRV site (final allele: ggactttttccttgtcctGATATCaccaccatatcccactcc, upper case=EcoRV). This deleted fragment was the 5′flanking sequence for a floxed neo/testis cre cassette(Bunting et al., 1999). 3′ flanking sequences extended from the HindIII site to the XhoI site in intron 1. Linearized vector was electroporated into TC1 cells (Deng et al., 1996) and G418 resistant colonies were screened by Southern blot. Two independent ES cell lines were injected by the OSUCCC Transgenic/ES Core Facility, and transmitted through the germline. Results from the lines were identical and are combined. Mice were maintained on a mixed 129/Sv×C57BL/6J background. Lfng^tmRjo1/+ mice (R. Johnson), were maintained on a mixed 129/Sv×C57BL/6J background, or crossed one generation with FVB/J mice to increase the recovery of adult Lfng^{tmRjo1/tmRjo1} mice (referred to as Lfng^-/-). Mice were maintained under the care of the Ohio State University ILACUC.

Genomic DNA was prepared from tail clips via proteinase K saltout or from yolk sac fragments via the HOTSHOT procedure(Truett et al., 2000). Animals were genotyped by PCR. Lfng^tmRjo1 primers FNG322(5′-GAGCACCAGGAGACAAGCC-3′), FNG325(5′-AGAGTTCCTGAAGCGAGAG-3′) and PGK3(5′-CTTGTGTAGCGCCAAGTGC-3′) amplify a 170 bp wild-type product and a 200 bp mutant product. Lfng^ΔFCE1 primers SC284(5′-TTTGGTGGGAATGGATTAGC-3′) and SC285(5′-CTGGTCCATTTGCTCTGAGG-3′) produce a 340 bp wild-type and a 182 bp mutant bands, while SC286 (5′-TTGGGTCTATCTGGGAAACG-3′) and SC287 (5′-GCGACTCATCCAGACACAGA-3′) produce a 149 bp wild-type and a 250 bp mutant bands.

Embryos were collected from timed pregnancies (noon of the day of plug identification designated as 0.5 dpc). RNA in situ hybridization using digoxigenin-labeled probes was performed essentially as described(Riddle et al., 1993);however, embryos were blocked in a mixture of MABT +20% sheep serum +2%Boehringer blocking reagent, and all post-antibody washes were performed in MABT. Hes7 cDNA probes extend from the internal SmaI site to the stop codon. Hes7 intron probe was amplified using the primers 5′-GCTAGAGGCCATAGCTGGTG and 5′-CTGTGACCAGCGGGAAAG. Dll1intron probe was amplified using primers 5′-GTTGGCAGTGGGAAGAAGG and 5′-TGTGTTGTGCCAATGAAGGT. Nrarp probe was amplified using the primers 5′-GCGTGGTTATGGGAGAAAGA and 5′-TTCCTCCCACACTGGTTCAT. The Hesr1 probe comprises the coding region of the cDNA. Other probes were Lfng (Johnston et al.,1997), Mesp2 (Saga et al., 1997), Uncx4.1(Mansouri et al., 1997) and Mox2 (Candia et al.,1992).

Skeletal preparations of neonates or 18.5 dpc embryos were performed essentially as described (Kessel and Gruss, 1991). Neurofilament staining was performed using the 2H3 antibody (Developmental Studies Hybridoma Bank), using standard protocols. For histological analysis, embryos were fixed in Bouin's fixative and transferred to ethanol for storage. Embryos were embedded in paraffin and 10 μm sections were stained with Haemotoxylin and Eosin.

Embryos were fixed in fresh 4% PFA in PBS, then washed in PBS. After overnight incubation at 4°C in PBS containing 0.1% hydrogen peroxide, 1%Triton X-100 and 10% fetal calf serum (TS-PBS), embryos were transferred into 10 mM sodium citrate (pH 6.0), 0.1% Tween-20 (CT), boiled for 10 minutes and then transferred back to PBS. After washing in TS-PBS, embryos were incubated for 5 days in primary Cleaved Notch1 (Val1744) antibody (Cell Signaling Technology) in TS-PBS (1:250). After washing embryos were incubated overnight in AP-conjugated secondary antibody in MABT (1:500). After washing, embryos were transferred to NTMT and stained with BCIP/NBT as described(Riddle et al., 1993).

To specifically disrupt Lfng expression in the segmentation clock,we deleted FCE1 from the endogenous Lfng locus producing the allele Lfng^ΔFCE1 (Fig. 1A,B). We hypothesized that this mutation would disrupt expression of Lfng in the caudal PSM (region I), where the clock is active,while preserving Lfng expression in the anterior PSM (region II),where R/C somite patterning is initiated.

Lfng expression is perturbed in the PSM of Lfng^{ΔFCE1/ΔFCE1} mutant embryos. In wild-type embryos, three distinct phases of Lfng expression are seen in the PSM, reflecting cyclic expression (Fig. 1C, parts c-f). By contrast, Lfng^{ΔFCE1/ΔFCE1} embryos express LfngRNA in a single band in the anterior PSM, with no expression observed in the caudal PSM where the clock is active (Fig. 1C, parts g,h). Similar results were seen at stages between 8.5 and 11.5 dpc (data not shown). Although the anterior band of Lfngexpression in Lfng^{ΔFCE1/ΔFCE1} embryos is weaker than the anterior-most band of Lfng expression in wild-type embryos, these results demonstrate that the deletion of the FCE1 enhancer prevents oscillatory expression of Lfng in region I of the PSM, while sparing some level of expression in region II. In addition, we find that Lfng expression in region II of the PSM is largely confined to the presumptive rostral compartment of somite S-1 (data not shown), indicating that the endogenous Lfng expression pattern in the anterior PSM is preserved in the Lfng^ΔFCE1 allele.

Although the Lfng^-/- genotype is reported to be viable,we find that on a mixed 129/Sv×C57BL/6J background, only rare animals survive postnatally, and homozygous males are infertile. By contrast,homozygous Lfng^{ΔFCE1/ΔFCE1} animals survive to adulthood at Mendelian ratios, and homozygous animals of both sexes are fertile. Lfng^{ΔFCE1/ΔFCE1} animals have segmentation defects, including shortened body and variably kinked tails(Fig. 2A). In the anterior skeleton, both Lfng^-/- and Lfng^{ΔFCE1/ΔFCE1} animals are severely affected. Multiple rib fusions and bifurcations as well as severely disorganized vertebrae are observed (Fig. 2B). When defects in the thoracic region of the skeleton are quantified, we find similar levels of disorganization in Lfng^{ΔFCE1/ΔFCE1} and Lfng^-/- animals (Fig. 2C).

In the more posterior skeleton, however, Lfng^{ΔFCE1/ΔFCE1} animals are much less affected than Lfng^-/- animals(Fig. 2B,D). In the thoracic and lumbar region of the skeleton, vertebral condensations in both Lfng^{ΔFCE1/ΔFCE1} and Lfng^-/- animals are irregular and misaligned. Strikingly,this pattern is altered at the lumbo-sacral junction. In the sacral region of Lfng^{ΔFCE1/ΔFCE1} animals, normal vertebral condensations are seen in all animals, and the tail vertebrae appear relatively normal, though variable kinks in the tail are seen ranging from mild (0-1 in 40% of mice) to moderate (2-5 in 60% of mice). By contrast, in Lfng^-/- animals, vertebral condensations are abnormal throughout the sacral region and the tail appears truncated, a phenotype never seen in Lfng^{ΔFCE1/ΔFCE1} animals(Fig. 2B,D). Thus we find that the loss of oscillatory Lfng expression in region I of the PSM causes pronounced defects in the axial skeleton, but these defects are much more pronounced in the thoracic and lumbar regions, while the sacral and more caudal regions of the skeleton are less affected in comparison to the null allele. Interestingly, the lumbo-sacral junction, the point where skeletal morphology largely recovers in Lfng^{ΔFCE1/ΔFCE1}animals, represents the transition point between primary and secondary body formation, suggesting that oscillatory Lfng plays, at most, a minor role in secondary body formation.

To test the hypothesis that early and late somitogenesis are differentially affected by the loss of oscillatory Lfng expression, we examined somite morphology at different stages of embryonic development. Somites that contribute to the thoracic and lumbar regions of the skeleton are produced during primary body formation (Gossler and Tam, 2002). In Lfng^{ΔFCE1/ΔFCE1}embryos, these somites are irregularly sized and spaced with frequent fusions between neighboring somites (Fig. 3B), and the mature derivatives of these somites remain irregularly spaced and sized at 10.5 dpc(Fig. 3D). During secondary body formation, however, somite development recovers in Lfng^{ΔFCE1/ΔFCE1} embryos, producing relatively evenly sized and spaced epithelial somites (Fig. 3F). These data support the idea that the loss of oscillatory Lfng expression differentially affects primary and secondary body formation with thoracic and lumbar somites being more sensitive to the loss of cyclic Lfng activity than are more caudal somites.

Lfng^-/- embryos have severe defects in R/C somite patterning (Evrard et al.,1998; Zhang and Gridley,1998). To address whether the Lfng expression in region II of the PSM of Lfng^{ΔFCE1/ΔFCE1} embryos could rescue R/C patterning, we examined compartment formation in the anterior PSM and in mature somites of Lfng mutant mice. We examined R/C patterning in region II of the PSM by assessing the expression of Mesp2. Mesp2defines the presumptive rostral compartment of somite S-1, and interacts with Lfng and Notch1 signaling during the process of R/C patterning(Morimoto et al., 2005; Takahashi et al., 2000). During both primary and secondary body formation, we find that Mesp2is expressed in a single band of varying width in the anterior PSM of both wild-type and Lfng^{ΔFCE1/ΔFCE1} embryos,reflecting the early expression and subsequent refinement of Mesp2 in the presumptive rostral compartment. However, in Lfng^{ΔFCE1/ΔFCE1} embryos, we frequently see a less distinct rostral border, regardless of the stage of somitogenesis(Fig. 4A). These results demonstrate that the rostral compartment is being defined in the presomites in region II of Lfng^{ΔFCE1/ΔFCE1} embryos throughout somitogenesis but may suggest that this earliest marker of patterning is mildly disrupted. As this disruption is seen throughout somitogenesis, it may be due to the reduced dose of Lfng in the anterior PSM, rather than to differences in primary and secondary body formation.

We then examined patterning of the mature somites. Uncx4.1,marking the caudal compartment of epithelial and mature somites, is expressed in clear compartments in all somites of wild-type embryos during primary body formation (Fig. 4B, parts a,b). In Lfng^-/- embryos, little compartmentalization of somites is seen, with rostral and caudal cells appearing mixed in a `salt and pepper'pattern (Fig. 4B, parts i,j)(Evrard et al., 1998; Zhang and Gridley, 1998). In Lfng^{ΔFCE1/ΔFCE1} embryos, Uncx4.1expression in newly formed somites is largely compartmentalized, with stronger expression in the more caudal region of somites S1 and S2. Clearer compartmentalization is observed in more anterior somites, but compartments are frequently irregularly spaced (Fig. 4B, parts e,f). Compartmentalization of mature somites in the thoracic region of Lfng^{ΔFCE1/ΔFCE1} embryos is more distinct by 10.5 dpc, with clear bands of Uncx4.1 visible in the sclerotome. Again, compartments of Uncx4.1 expression are frequently irregularly spaced or shaped, presumably reflecting the irregularities in somite size and shape observed morphologically in the thoracic region of the embryo (Fig. 4B, part g). Similar results are seen when examining Mox1 at 10.5 dpc(Fig. 4C). By contrast, the Uncx4.1 signal in the thoracic region of Lfng^-/-embryos fails to compartmentalize, maintaining an unsegmented pattern(Fig. 4B, part k). The somites formed during secondary body formation in Lfng^{ΔFCE1/ΔFCE1} embryos are clearly compartmentalized with regular rostral and caudal segmentation, whereas in Lfng^-/- embryos, R/C patterning continues to be abnormal(Fig. 4B, parts h,l).

Functionality of R/C patterning was assessed by neurofilament staining with 2H3. In wild-type embryos, regular neurofilament staining is observed,representing the axonal trajectories of spinal neurons through the rostral somite compartment. In Lfng^{ΔFCE1/ΔFCE1}embryos, axonal projections are seen, but their spacing is irregular(Fig. 4D). Thus, although Lfng^{ΔFCE1/ΔFCE1} embryos produce irregular somites during primary body formation, the retention of Lfngexpression in the anterior PSM supports relatively normal R/C patterning, and somites formed during secondary body formation undergo normal R/C patterning. This supports the idea that the role of Lfng in R/C somite patterning is distinct and separable from its functions in the segmentation clock.

Several groups have suggested that oscillatory expression of Lfngis involved in interlocking feedback loops that regulate oscillatory Notch1 activation in the PSM. To examine the effects of the Lfng^ΔFCE1 allele on Notch1 signaling, we visualized Notch activation using an antibody specific for the Notch1 ICD (NICD). Notch signaling levels oscillate in the PSM of wild-type embryos during primary and secondary body formation (Fig. 5) (Morimoto et al.,2005), with different patterns of NICD staining found in different embryos. By contrast, in both Lfng^{ΔFCE1/ΔFCE1}(Fig. 5) and Lfng^-/- (Fig. 5) (Morimoto et al.,2005) embryos, a gradient of NICD is seen in the PSM, reflecting ubiquitous, non-oscillatory Notch signaling throughout the PSM. This confirms that the Lfng^ΔFCE1 allele inhibits oscillatory Notch signaling in region I of the PSM during both primary and secondary body formation, and indicates that oscillatory Notch activation in region I of the PSM is largely dispensable for segmentation during secondary body formation.

To assess the effects of the loss of cyclic Lfng expression on the transcription of other segmentation clock genes, we first examined the expression of Hes7 in Lfng^{ΔFCE1/ΔFCE1}embryos. Hes7 has been proposed to play a role in the segmentation clock mechanism as part of the feedback loops regulating oscillatory Lfng transcription and Notch1 activation (reviewed by Rida et al., 2004). One report has suggested that Hes7 expression is ubiquitous in the Lfng^{tm1Grid/tm1Grid} null background at 9.5 dpc(Chen et al., 2005), while more recent results suggest the Hes7 expression is affected but still dynamic in the absence of Lfng(Niwa et al., 2007). During primary body formation, we used a probe specific for Hes7 intronic sequences to show that Hes7 RNA is transcribed in a stable ubiquitous pattern in the PSMs of Lfng^{ΔFCE1/ΔFCE1} and Lfng^-/- embryos, distinct from the dynamic banding pattern seen in wild-type embryos (Fig. 6A). Thus, during primary body formation, the loss of Lfng prevents the cyclic transcription of Hes7. In sharp contrast, we find that during secondary body formation, Hes7transcription oscillates in the same way as wild-type expression patterns in both Lfng^{ΔFCE1/ΔFCE1} and Lfng^-/- embryos (Fig. 6B). Similar results were seen using a Hes7 mRNA probe,indicating that post-transcriptional regulation of Hes7 mRNA levels is also normal in these embryos (Fig. 6C). Hes7 cyclic expression was confirmed by half tail culture experiments. PSMs were bisected, with one half fixed immediately and the other half cultured before fixation. After 1 hour of culture, the Hes7 expression pattern in the cultured half is different from the uncultured half regardless of genotype(Fig. 6D). These results suggest that Hes7 transcription may be differentially controlled during different stages of somitogenesis, requiring Notch oscillations during primary body formation, but not during secondary body formation.

We confirmed and extended these observations by analyzing the expression of other oscillatory genes in Lfng^{ΔFCE1/ΔFCE1}embryos. Similar to our results with Hes7, we find that Nrarp expression is differentially affected during primary and secondary body formation. Before tailbud formation, distinctive banding patterns are observed in wild-type, but not in Lfng^{ΔFCE1/ΔFCE1} embryos(Fig. 7A). After tailbud formation, oscillatory Nrarp expression recovers in Lfng^{ΔFCE1/ΔFCE1} embryos, although the cyclic expression patterns are less distinct than those in wild-type embryos. Other Notch pathway genes also oscillate during secondary body formation in Lfng^{ΔFCE1/ΔFCE1} embryos, including Dll1 (Fig. 7B), but the expression of some Notch targets, including Hesr1 is perturbed at this stage (Fig. 7C). Thus,although multiple genes that may be involved in the segmentation clock mechanism exhibit oscillatory expression in the absence of cyclic Notch activation during secondary body formation in Lfng^{ΔFCE1/ΔFCE1} embryos, cyclic Notch1 activity is required for proper expression of some genes in the region at this stage.

Ascertaining the functions of Notch signaling in segmentation is complicated by the fact that the pathway plays multiple roles during somitogenesis. It is unclear what aspects of the Lfng^-/-phenotype can be ascribed to its role in R/C patterning as opposed to any role in clock function, as both aspects of expression are perturbed throughout development in the mouse knockout. Indeed, in zebrafish, Lfng is expressed solely in the anterior-most region of the PSM, indicating that in this organism Lfng plays no role in the segmentation clock(Prince et al., 2001). We therefore specifically disrupted the oscillatory expression of Lfngin region I of the PSM to examine the role it plays in the segmentation clock during mouse somitogenesis. Targeted deletion of FCE1 sequences eliminates expression of Lfng in the posterior region of the PSM, indicating that the enhancer is required for cyclic Lfng transcription in region I of the PSM. However, the anterior band of Lfng expression in Lfng^{ΔFCE1/ΔFCE1} embryos is weaker than that seen in wild-type embryos, perhaps supporting an additional role for FCE1 in enhancing expression of Lfng in the anterior PSM.

As predicted, the loss of oscillatory Lfng expression in region I of the PSM affects segmentation in Lfng^{ΔFCE1/ΔFCE1} embryos, but distinct effects are seen during primary and secondary body formation. In the thoracic and lumbar skeleton, malformed vertebral condensations and rib abnormalities were seen in Lfng^{ΔFCE1/ΔFCE1} skeletons(Fig. 2B). The appearance of the vertebrae resembles the phenotypes seen in some cases of autosomal recessive spondylocostal dysostosis caused by mutations in DLL3 or LFNG; both the thoracic and lumbar spine are affected and vertebral bodies are irregularly shaped and fitted together(Bulman et al., 2000; Sparrow et al., 2006). To our surprise, we found that the caudal skeletal regions (sacral and tail vertebrae) were invariably less severely affected in Lfng^{ΔFCE1/ΔFCE1} animals than in Lfng^-/- animals. Especially striking is the fact that in the sacral region of the spine, Lfng^{ΔFCE1/ΔFCE1} animals exhibit essentially normal vertebral formation, whereas irregularities are still seen at this level in Lfng^-/- skeletons. The point of phenotype recovery at the lumbo-sacral junction indicates that secondary body formation occurs relatively normally in Lfng^{ΔFCE1/ΔFCE1}embryos.

This differential severity is reflected in the process of somitogenesis throughout development. During primary body formation, somites are frequently abnormal in Lfng^{ΔFCE1/ΔFCE1} embryos(Fig. 3). The production of irregularly sized somites suggests that the loss of oscillatory Lfngexpression in region I of the PSM interferes with segmentation clock function during primary body formation. By contrast, during secondary body formation,somites formed in Lfng^{ΔFCE1/ΔFCE1} embryos are evenly spaced and of regular size, and the phenotypes in the sacral and caudal skeleton are correspondingly milder (Figs 2 and 3). Thus, our data suggest that segmentation of the embryo during primary body formation (contributing to the thoracic and lumbar skeleton) is more sensitive to the loss of cyclic Lfng expression than is segmentation during secondary body formation. This sheds new light on one of the classical issues of developmental biology:the extent to which primary and secondary body formation represent distinct mechanisms of development.

Dll3-null embryos exhibit similar Lfng expression patterns to those observed in Lfng^{ΔFCE1/ΔFCE1}mice, with expression observed only in the anterior PSM after 9.5 dpc(Dunwoodie et al., 2002; Kusumi et al., 2004). Interestingly, Dll3-null mice exhibit disordered somitogenesis along the length of the vertebral column, suggesting that Lfng expression in region II of the PSM is not, in and of itself, sufficient to rescue secondary body formation. This may reflect a requirement for Dll3expression in the anterior PSM during secondary body formation. Alternatively,it was recently shown that the loss of Dll3 in the PSM leads to a loss or reduction in NICD levels in region I of the PSM, in contrast to the ubiquitous Notch1 activation observed in Lfng^{ΔFCE1/ΔFCE1} embryos(Geffers et al., 2007). This raises the possibility that while oscillatory Notch1 activation in the posterior PSM is not required during secondary body formation, some level of Notch1 activation is still necessary during this process. This may be especially interesting in light of the observation that constitutive overexpression of Lfng in the mouse PSM, which might be predicted to repress Notch1 activation, also perturbs somitogenesis along the entire axial skeleton (Serth et al.,2003).

Lfng^-/- embryos have severe defects in R/C somite patterning (Evrard et al.,1998; Zhang and Gridley,1998). This could arise due to downstream effects of the loss of Lfng in the segmentation clock, or more directly due to the loss of Lfng expression in region II of the PSM. Analysis of R/C patterning in Lfng^{ΔFCE1/ΔFCE1} embryos directly assesses whether the retention of Lfng expression in the presumptive anterior compartment of the forming somite can rescue R/C patterning in the absence of oscillatory Notch activity in the clock. During secondary body formation, Lfng^{ΔFCE1/ΔFCE1} embryos produce regular pairs of somites, and these somites are properly patterned. More surprisingly, the irregular somites produced during primary body formation in Lfng^{ΔFCE1/ΔFCE1} embryos are also patterned into clear rostral and caudal compartments(Fig. 4B), although this patterning may be somewhat delayed. We propose that in Lfng^{ΔFCE1/ΔFCE1} embryos, Mesp2expression in the anterior compartment of the developing somite is able to stabilize the pattern of Notch activation in somites S0 and S-1, at least in part via its specific activation of Lfng transcription. This allows the Notch pathway to function in R/C patterning in Lfng^{ΔFCE1/ΔFCE1} embryos despite the loss of cyclic Lfng expression in region I of the PSM. It is not clear at this time whether the delay in robust R/C patterning of thoracic somites is due to some underlying disorganization of somites S-1 and S0 as a result of perturbed clock function, or whether it may be a result of the reduced Lfng levels seen in the anterior PSM in Lfng^{ΔFCE1/ΔFCE1} embryos. However, the successful patterning of irregularly sized somites during primary body formation in Lfng^{ΔFCE1/ΔFCE1} embryos suggests that the Notch-based processes involved in the segmentation clock can be largely divorced from its roles in R/C somite patterning, and that the processes regulated by oscillatory Notch signaling in region I of the PSM are not prerequisites for the patterning of the pre-somites in region II.

The loss of cyclic Notch1 activation has distinct effects during primary and secondary body formation. During early stages of somitogenesis, the loss of oscillatory Lfng expression interferes with oscillatory Notch activation (Fig. 5A) and causes phenotypes (irregular somite size and positioning, alteration of oscillatory gene expression) that suggest defects in segmentation clock function(Fig. 3, Fig. 6A). By contrast, during later stages of segmentation, despite the continued absence of oscillatory Lfng and the presence of ubiquitous Notch1 activation, somitogenesis proceeds relatively normally in Lfng^{ΔFCE1/ΔFCE1} embryos, and the oscillatory expression of several clock genes largely recovers at these stages(Fig. 6B-D, Fig. 7). Although expression of some Notch target genes is slightly affected during secondary body formation in Lfng^{ΔFCE1/ΔFCE1} embryos, the mild phenotypes observed in the caudal axial skeleton suggest that these perturbations are relatively unimportant. Thus, it appears that segmentation clock function is more sensitive to the loss of oscillatory Lfngexpression during primary body formation than during secondary body formation.

Differential regulation of somitogenesis at different axial levels of the embryo is not unprecedented. The first five or six somites are frequently spared in mutations that affect the Notch signaling pathway, although in zebrafish these segments can be affected by the simultaneous downregulation of several clock components (Oates et al.,2005) perhaps indicating multiple, parallel mechanisms of regulation. More recently, it has been shown that in zebrafish the anlagen of the anterior trunk, posterior trunk and tail are specified before somitogenesis begins, raising the possibility that different genetic pathways may affect these regions in distinct ways(Szeto and Kimelman,2006).

Our data expand on these observations, suggesting that in the mouse, somite formation during secondary body formation is controlled largely by pathways that do not require oscillatory Notch1 activation. The robust nature of somitogenesis may reflect the existence of multiple, overlapping and interacting feedback loops controlling the oscillation of numerous genes in the Notch, Wnt and FGF pathways (Dequeant et al., 2006). For example, recent findings suggest that FGF signaling is required for oscillatory function of both the Notch and Wnt pathways in the PSM (Wahl et al.,2007), though most observations were confined to primary body formation. Our data support the additional hypothesis that oscillation of any individual pathway or component may be more or less important during different stages of somitogenesis. Our finding that Hes7 oscillations recover during secondary body formation is especially interesting in light of recent findings that Hes7 oscillations are regulated in part by the FGF pathway, and that oscillatory HES7 protein regulates the expression of FGF pathway components (Niwa et al.,2007). It is clear that regulated crosstalk among these pathways is important; however, our results suggest that specific interactions may be differentially regulated during primary and secondary body formation.

Wnt activity may play especially important roles in the regulation of posterior somitogenesis. Reductions in Wnt signaling levels can preferentially affect segmentation of the posterior embryo: the Wnt3a^vthypomorphic allele develops segmentation defects in the lumbar, sacral and tail regions, and mutations in Lrp6, encoding a Wnt co-receptor,affect the caudal axial skeleton more severely than anterior skeletal regions(Kokubu et al., 2004; Pinson et al., 2000). These data may indicate that the caudal skeleton is more sensitive to perturbations in Wnt pathway activity. Conversely, based on our data, Notch oscillations may play a more important role during the development of the thoracic and lumbar skeleton. It will clearly be important to carefully dissect the interactions among these three pathways to clarify fully the possibility that the segmentation clock mechanism is differentially regulated during primary and secondary body formation.

We find that many aspects of clock function recover in both Lfng^-/- and Lfng^{ΔFCE1/ΔFCE1} embryos after tailbud formation; however, the posterior skeletal phenotypes of Lfng^-/- animals are much more severe than those seen in Lfng^{ΔFCE1/ΔFCE1} embryos. Therefore, we propose that the truncation of posterior skeletal structures in Lfng^-/- animals is caused by the perturbation of R/C patterning in these embryos, rather than the loss of oscillatory Notch activity in the clock. Several lines of evidence suggest that R/C somite patterning contributes to the proper segmentation of the posterior embryo. Targeted deletion of Mesp2, which is exclusively expressed in region II of the PSM, causes disrupted R/C somite patterning and truncation of the posterior skeleton (Saga et al.,1997). Furthermore, it appears that the total dosage of MESP activity (comprising the additive effects of MESP1 and MESP2) in region II of the PSM is important. Manipulating the levels of MESP proteins can both partially rescue the R/C patterning of the somites and mitigate the caudal truncation of the axial skeleton (Morimoto et al., 2006). In addition, in newly developed Mesp2knockout alleles, Mesp1 expression is elevated leading to partial rescue of somitogenesis during secondary body formation(Takahashi et al., 2007). Interestingly, an Mesp2 mutation found in spondylocostal dysostosis also has more severe effects on the thoracic vertebrae than the more caudal skeleton (Whittock et al.,2004).

One possible explanation for these results is that continued R/C somite patterning is necessary for segmentation to proceed normally during secondary body formation. This could suggest that information transfer in the PSM can occur from the anterior to the tailbud, and that the segmentation of the most caudal embryonic structures may be reliant on proper patterning of more anterior structures. We propose that the expression of Lfng in the anterior PSM and the subsequent amelioration of R/C patterning defects in Lfng^{ΔFCE1/ΔFCE1} embryos, permits posterior segmentation to proceed relatively normally, preventing the tail truncation seen in Lfng^-/- animals. This underscores the potential for the transfer of information between the anterior and posterior regions of the PSM, at least during secondary body formation. Thus, the work reported here uncovers new levels of complexity linking differential regulation of clock function and R/C somite patterning to the long-known but little-understood processes of primary and secondary body formation.

We thank J. Thompson for critical reading of the manuscript, R. Johnson for Lfng^tmRjo1 mice, and Y. Saga and P. Gruss for reagents. ES cell injections were performed by the OSUCCC ES cell core facility. This work was supported in part by Basil O'Connor grant #5-FY04-217 from the March of Dimes.