|Home | About | Journals | Submit | Contact Us | Français|
Delta-like 3 (Dll3) is a divergent ligand and modulator of the Notch signaling pathway only identified so far in mammals. Null mutations of Dll3 disrupt cycling expression of Notch targets Hes1, Hes5, and Lfng, but not of Hes7. Compared with Dll1 or Notch1, the effects of Dll3 mutations are less severe for gene expression in the presomitic mesoderm, yet severe segmentation phenotypes and vertebral defects result in both human and mouse. Reasoning that Dll3 specifically disrupts key regulators of somite cycling, we carried out functional analysis to identify targets accounting for the segmental phenotype. Using microdissected embryonic tissue from somitic and presomitic mesodermal tissue, we identified new genes enriched in these tissues, including Limch1, Rphn2, and A130022J15Rik. Surprisingly, we only identified a small number of genes disrupted by the Dll3 mutation. These include Uncx, a somite gene required for rib and vertebral patterning, and Nrarp, a regulator of Notch/Wnt signaling in zebrafish and a cycling gene in mouse. To determine the effects of Dll3 mutation on Nrarp, we characterized the cycling expression of this gene from early (8.5 dpc) to late (10.5 dpc) somitogenesis. Nrarp displays a distinct pattern of cycling phases when compared to Lfng and Axin2 (a Wnt pathway gene) at 9.5 dpc but appears to be in phase with Lfng by 10.5 dpc. Nrarp cycling appears to require Dll3 but not Lfng modulation. In Dll3 null embryos, Nrarp displayed static patterns. However, in Lfng null embryos, Nrarp appeared static at 8.5 dpc but resumed cycling expression by 9.5 and dynamic expression at 10.5 dpc stages. By contrast, in Wnt3a null embryos, Nrarp expression was completely absent in the presomitic mesoderm. Towards identifying the role of Dll3 in regulating somitogenesis, Nrarp emerges as a potentially important regulator that requires Dll3 but not Lfng for normal function.
The vertebrate body is shaped from segmental units called somites, which are formed in a regular, repeated fashion during embryonic development. Somites are produced at the anterior end of the unsegmented presomitic mesoderm (PSM), where oscillatory waves of gene expression regulate prepatterning of these segmental units (reviewed in Dequéant and Pourquié, 2008; Kageyama et al., 2007; Kulesa et al., 2007). To date, many genes have been identified that demonstrate such oscillatory expression, including the Notch pathway genes Lfng, Hes1, Hes5, and Hes7 and the Wnt pathway genes Axin2, Nkd1, Dact1 and Dkk1 (Aulehla et al., 2003; Bessho et al., 2001; Dequéant et al., 2006; Forsberg et al., 1998; Ishikawa et al., 2004; Jouve et al., 2000).
The Wnt pathway plays a key role in the presomitic mesoderm. Small, irregular somites have been observed in beta-catenin null embryos and lengthened presomitic mesoderm observed in beta-catenin gain-of-function mutants, suggesting that the Wnt pathway regulates somitogenesis by activating target genes such as Dll1 and positioning boundary determination genes in the anterior presomitic mesoderm (Dunty et al., 2007; Hofmann et al., 2004). Wnt3a mutants disrupt the expression of a number of Notch pathway genes including Lfng, whereas the Wnt pathway gene Axin2 has been observed to display cycling expression in the Notch pathway Dll1 mutant (Aulehla et al., 2003). These observations and others have been used to support the view that the Wnt pathway is upstream of Notch signaling in the PSM.
During somitogenesis, Notch signaling has been proposed to be essential for one or more of the following functions; generation of oscillatory gene expression in PSM cells (Holley et al., 2002; Jouve et al., 2000; Morales et al., 2002), establishment of somite compartment polarity (Barrantes et al., 1999; Saga, 2007; Takahashi et al., 2000), and communication between neighboring cells to synchronize oscillations (Horikawa et al., 2006; Jiang et al., 2000; Özbudak and Lewis, 2008). Recently, pharmacological blockade of the Notch pathway in zebrafish exhibited somite defects only after long developmental delays, suggesting Notch signaling is essential for synchronizing oscillations of neighboring cells in the posterior PSM but not for somite border formation (Mara et al., 2007; Özbudak and Lewis, 2008). Furthermore, Feller et al (2008) suggested a similar role for the Notch pathway in the caudal PSM in mice as well as demonstrating a requirement for Notch signaling in somite compartmentalization and not border formation in the anterior PSM. In mouse, defects in Notch signaling disrupt somite segmentation and oscillatory expression of Notch pathway genes in the PSM (Bessho et al., 2001; Conlon et al., 1995; Evrard et al., 1998; Feller et al., 2008; Hrabě de Angelis et al., 1997; Kusumi et al., 1998, 2004;). In PSM S-1, i.e., somite minus one, the region from which the next somite will form (Pourquié and Tam, 2001), the transcription factor Mesp2, a direct target of Notch signaling, appears to regulate segmental border formation through activation of Epha4 and rostral-caudal compartmentalization through Uncx/Uncx4.1 (reviewed in Saga, 2007). However, the mechanisms by which Notch signaling directs expression of downstream genes necessary for paraxial mesoderm segmentation is still not well understood.
Notch signaling activity can be modified in a number of ways (reviewed in Bray, 2006). Two modifiers of Notch signaling, Lfng and Dll3, are noteworthy given their disruption in humans causes a severe, autosomal recessive vertebral disorder, spondylocostal dysostosis (SCD; Bulman et al., 2000; Sparrow et al., 2006). Disruptions of Lfng and Dll3 in the mouse result in somitic and vertebral phenotypes that are morphologically similar to each other and to SCD (reviewed in Turnpenny et al., 2007). Lfng is a modifier of Notch signaling. It encodes a glycosyltransferase that modifies Notch in the Golgi, and modulates the ability of Notch to bind to DSL ligands. Loss of Lfng function results in severe rostrocaudal patterning defects (Evrard et al., 1998; Shifley et al., 2008; Zhang and Gridley, 1998). In contrast to Lfng, Dll3 has only recently been identified as a modifier of Notch activity (Geffers et al., 2007). Dll3 encodes a highly divergent delta-type DSL ligand that, unlike the other DSL ligands, does not appear to bind Notch receptors at the cell surface, and instead regulates Notch signaling, perhaps within the Golgi (Dunwoodie et al., 1997; Geffers et al., 2007; Ladi et al., 2005). Null alleles of Dll3 disrupt transcriptional oscillation of some Notch pathway genes expressed in the PSM, as well as genes involved in determining the rostrocaudal polarity of the somite (Dunwoodie et al., 2002; Kusumi et al., 1998, 2004).
Loss of Notch signaling can lead to failure of Notch pathway gene expression during somitogenesis, as demonstrated in Dll1 mutations which lead to severely decreased expression of most reported Notch pathway genes (Barrantes et al., 1999; Kusumi et al., 2004). By contrast, loss of Dll3 results in decreased levels of gene expression of only some Notch pathway genes including Hes5 and Hes1; but the dynamic expression of the cycling gene Hes7 is unaffected (Dunwoodie et al., 2002; Kusumi et al., 2004). Dll3 mutations also lead to the loss of Lfng cycling expression (Kusumi et al., 2004) and disruption of cycling patterns of activated Notch1 (Geffers et al., 2007). For comparison, loss of Lfng expression (Morimoto et al., 2005) or its cyclical component (Shifley et al., 2008) both disrupt cyclical activation of Notch1 and Hes7 expression. In both humans and mice, DLL3 mutations produce vertebral disruptions as severe as those observed in disruptions of LFNG or HES7 (Bessho et al., 2001; Bulman et al., 2000; Dunwoodie et al., 2002; Kusumi et al., 1998; Sparrow et al., 2006, 2008).
Why does loss of Dll3 produce such a severe phenotype? One possibility is that the disruptions in Lfng cycling expression and activated Notch1 patterns account for the severe phenotype. An alternate possibility is that as yet unidentified genes are critically required for somitogenesis, and that these are disrupted in Dll3 mutants. Analysis of Dll3 mutant embryos by microarray studies could identify such factors. Our understanding of somitogenesis continues to be advanced by the identification of new genes. This includes genes that are highly expressed in the presomitic mesoderm, as well as genes that have altered expression in segmental mutants. Functional genomic approaches can be particularly useful to find such genes, since they can screen the transcriptome, and microarray studies have identified genes with oscillatory expression in mouse somitogenesis (Dequéant et al., 2006) and in cell culture models of this process (William et al., 2007). Others have used this approach to examine Dll1 (Machka et al (2005) and Hes7 (Niwa et al. (2007) mutant embryos. Previously we reported expression differences in whole 9.5 dpc Notch1 and Dll3 mutant embryos (Loomes et al., 2007). However we did not identify somitogenesis genes, probably because the PSM is only a minor portion of the entire embryo. Here, we aimed to identify genes enriched in the presomitic mesoderm, and we have identified and characterized 3 genes expressed in the caudal PSM. In addition, we examined Dll3 mutant PSM and somite level tissues, and identified a limited number of genes with disrupted expression. Nrarp (Notch regulated ankyrin repeat protein), which is normally expressed in a cycling manner during somitogenesis, emerged from our investigation as a gene that may contribute to the severe Dll3 mutant phenotype.
Dll3tm1Rbe (referred to as Dll3neo), Lfngtm1Rjo, and Wnt3atm1Amc mutations were crossed into the C57BL/6J background by backcross matings for over 10 generations (Dunwoodie et al., 2002; Kusumi et al., 1998; Takada et al., 1994). All animals were maintained according to Institutional Animal Care and Use Committee guidelines. Dll3 mutants were generated by an intercross mating of Dll3neo/+ mice, Lfng mutants were generated by an intercross mating of Lfngtm1Rjo/+ mice, and Wnt3a mutants were produced by an intercross mating of Wnt3atm1Amc/+ mice. Genotypes for Dll3neo, Lfngtm1Rjo and Wnt3atm1Amc were determined by PCR assay as described previously (Dunwoodie et al., 2002; Evrard et al., 1997; Kusumi et al., 1998).
Embryos were collected at day 9.5 of gestation, and dissected from decidua in cold M2 medium (Nagy et al., 2003) containing 10% fetal calf serum. All remnants of the allantois were carefully removed from intact embryos, which were then cut at the boundary of the PSM and most recently-formed somite. The released tissue (“PSM level” containing tissue from all three germ layers) was placed in cold RNA lysis solution and frozen at −80°C for later RNA extraction. The remainder of the embryo was cut at the level of the otic placode, and the heart and endodermal tissues removed to generate the “somite level” samples. Total RNA was extracted from 9.5 dpc embryos using a SuperScript RT II Kit (Invitrogen) and subsequently used to synthesize double stranded cDNA. Biotinylated cRNA targets were prepared with a Bioarray HighYield RNA Transcript Labeling Kit (ENZO). Targets were hybridized to the Affymetrix MOE430A array according to manufacturer’s protocol. Arrays were subsequently scanned using the following parameters: t = 0.015, a1 = 0.05, a2 = 0.065, median intensity value = 150. CEL files were normalized by the Robust Multichip Average method in order to compare the levels of expression between different samples and minimize technical variation accompanying use of the Genespring GX 7.3.2 module (Agilent).
Quantitative PCR was used to confirm the expression of selected candidate genes, using cDNAs assayed by Taqman® Gene Expression Assays (Applied Biosystems). Gapdh was used to normalize Q-PCR results, using assay Mm99999915_g1. Assays for selected genes examined were: Dll3 exons 2–3, Mm00432856_g1; Dll3 exons 5–6, Mm00432853_g1; Nrarp, Mm00482529_s1; Bcat2, Mm00802196_g1; and Hnrnpl, Mm01172981_g1. Cycling conditions used for the Taqman Real-time PCR were: 95°C 10 minutes followed by 40 cycles of 95°C 15 seconds and 60°C 1 minute. Statistical analysis on expression data was carried out using a t-Test (two sample assuming unequal variance), with a threshold of p<0.05 on a two-tailed test on microarray data and at least a one-tailed test for validation by Q-PCR.
Whole mount in situ hybridizations were performed on 8.5, 9.5 and 10.5 dpc embryos as described previously (Harrison et al., 1995; Wilkinson, 1992). DNA templates for RNA probes were generated by RT-PCR from a 9.5 dpc embryo cDNA template. In brief, chimeric PCR templates were generated using 3′ end primers with the T7 RNA polymerase-binding site added. Digoxigenin-UTP labeled (Roche) RNA probes were generated by in vitro transcription (MAXIscript™, Ambion). RNA probes were purified using MC Free filtration units (Millipore). To compare the expression of two genes within the PSM, we bisected the caudal paraxial mesoderm of fixed whole 9.5 dpc embryos prior to in situ hybridization. Hybridized embryos were photographed using a SMZ1000 stereodissecting microscope (Nikon) with a Retiga CCD digital camera (Q-Imaging).
To examine cycling gene expression, we bisected the caudal paraxial mesoderm of 9.5 dpc embryos into axial halves, fixing one half in 4% paraformaldehyde while culturing the remaining half for 60, 120 or 180 min. in DMEM/50% FBS prior to fixation (adapted from Forsberg et al., 1998; Kusumi et al., 2004).
Data sets from Affymetrix microarray analysis of microdissected embryonic tissues (MOE430A) are deposited at the Gene Expression Omnibus (www.ncbi.nlm.nih.gov/geo) with accession numbers (to be submitted before publication).
In carrying out microarray analysis of 9.5 dpc embryonic PSM and somite level tissues, we sought to 1.) identify genes that are enriched in PSM and somite tissues, and 2.) identify genes that were up- and down-regulated due to Dll3 mutation. We microdissected tissue from 9.5 dpc embryos to generate somite and PSM level samples (Fig. 1A). Since the amount of tissue collected from the PSM of one embryo was insufficient for microarray analysis using a single round of target amplification, we pooled dissected samples from ten embryos. We sought to avoid double amplification protocols that can lead to distortion of gene expression levels. Microarray analysis using Affymetrix MOE430A arrays was carried out on the biological pool triplicates, and genes with greater than two-fold differences were identified (Supplemental Tables S1 and S2).
Expression analysis identified 66 genes with greater than two-fold increased expression in somite when compared to PSM fractions. Using the Gene Expression Database (GXD; http://www.informatics.jax.org/expression.shtml) and the published literature, we identified 15 genes that have been reported to be expressed in somites, including Hes5, Meox1, Meox2, Pax1, Uncx/Uncx4.1, and Zic1 (Fig. 1B and Supplemental Table S1). We found a further 48 genes that were reported to be expressed in the other embryonic tissues, including the neural tube, surface ectoderm, or endoderm. Expression in the somites was not reported for these 48 genes and remains to be further characterized. Finally, there are also 3 genes whose expression has not been reported in embryos. We characterized the expression of these genes by in situ hybridization and observed that Pkdcc (protein kinase domain containing, cytoplasmic) was expressed in the endoderm, Elavl4 (embryonic lethal abnormal vision-like 4) demonstrated a salt-and-pepper expression in the neuroepithelium of the midbrain and patches of expression dorsal to the second pharyngeal arch (data not shown), and Itih5 (inter-alpha globulin inhibitor H5) expression was not detectable (Table 1 and Supplemental Fig. S1).
We further identified 87 genes with increased expression (≥ 2 fold) in the nascent PSM when compared to somite fractions (Figure 1C and Supplemental Table S2). As described above, we confirmed that 48 of these genes had been published as being expressed in PSM, including the somitogenesis genes Dll1, Dll3, Fgf8, Hes7, and Tbx6. 31 of these genes were reported to be expressed in the embryo in tissues including neural plate or surface ectoderm, but remain to be further examined in the paraxial mesoderm. This list includes the gene Phlda2, which interestingly is a paralogue of the Fas pathway gene Phlda1 that cycles in phase with Wnt pathway genes (Dequéant et al., 2006). Finally, there are 8 genes whose expression has not been described in embryos (Table 1). We characterized these genes by in situ hybridization and noted high levels of PSM expression of three of them, Limch1 (LIM & calponin homology domains 1), Rhpn2 (rhophilin 2), and the cDNA A130022J15Rik (Fig. 1D–I). Gene Ontogeny predicts that Limch1 encodes an actin binding protein and that A130022J15Rik encodes a glycosyltransferase, like Lfng (www.informatics.jax.org). Interestingly, Rhpn2 encodes a rho GTPase binding protein that could play a role in endocytosis (Behrends et al., 2005). Endocytosis of Notch1 extracellular domain-ligand heterodimers and of Notch1 receptors play a key role in trans- and cis-regulation, respectively. A targeted mutation of Rhpn2 has been generated and does not report any segmental defects, but the mutant was generated to examine thyroid function and segmentation may not have been examined in detail (Behrends et al., 2005). Targeted mutations or human mutations of Limch1 and A130022J15Rik have not been reported. Another cDNA, 2610528A11Rik was found to be localized to the caudal endoderm (Supplemental Fig. S1).
To enrich for genes in the paraxial mesoderm that are specifically disrupted by Dll3 mutation, we compared microdissected tissues from wild-type and Dll3 embryos. We generated biological replicate pools from Dll3+/+ (wild-type) or Dll3neo/neo embryos for a total of twelve pools. Microarray analysis using Affymetrix MOE430A arrays was carried out on the biological pool triplicates, and genes with significant differences of greater than two-fold were selected. Findings were validated by Q-PCR and all expression differences were confirmed as statistically significant. Dll3 itself displayed 3.6 fold decreased expression Dll3neo/neo embryos (Fig. 2A). This demonstrates that our microarray approach was capable of detecting relevant differences in gene expression between wild-type and null mutant embryos. A complete absence of Dll3 transcripts might be expected in the Dll3 null pools, and the Affymetrix probe set 1449236_at comprises oligonucleotides distributed over all of the Dll3 exons, including exons 1–4 that were not removed in the targeted mutation (Netaffyx, www.affymetrix.com; Dunwoodie et al., 2002). We therefore used quantitative PCR to examine expression levels of Dll3 exons 2–3 (not targeted) and exons 5–6 (deleted in the mutant; Fig. 2A). As expected, neither were detectable in Dll3neo/neo embryos, indicating that the targeted transcript was likely to have been subject to nonsense-mediated decay. Thus, levels of Dll3 detected by microarray in Dll3neo/neo mutants appear to represent baseline noise of the microarray system.
Surprisingly, few other genes were identified at an initial 2-fold cut-off level. We next examined a group of genes identified at a threshold of 1.5 fold changes (Fig. 2). In addition to Dll3, only Nrarp, which has been previously identified as an inhibitor of the Notch signaling pathway, was downregulated in the Dll3neo/neo PSM fractions. Nrarp will be described in detail below. The only upregulated genes in these fractions were Bcat2 (Fig. 2C), which encodes mitochondrial branched chain aminotransferase 2 that produces a growth and metabolic defect when knocked out (She et al., 2007) and Hnrnpl (Fig. 2D, E), which encodes heterogeneous nuclear riboprotein L factor that plays a role in mRNA processing (Griffith et al., 2006). We found that Bcat2 is highly expressed in the rostral first pharyngeal arch and also expressed in trunk mesenchyme and that Hnrnpl is expressed ubiquitously at 9.5 dpc (Supplemental Fig. S1). Since mRNA splicing is a key regulatory step for the dynamically expressed Notch pathway cycling genes, the upregulation of Hnrnpl may represent a compensatory change in response to disruption of the segmentation clock. However, the altered expression of both of these genes may be secondary to the disruption in Notch signaling.
In tissue collected from somite levels, Hnrnpl was the only gene that displayed increased expression in Dll3 mutants, and given the ubiquitous expression that we observed at 9.5 dpc, this is not surprising. We found that Uncx/Uncx4.1 was decreased in expression (Fig. 2E). Uncx is expressed in the caudal compartment of somites, and we have previously shown that Dll3 mutant embryos display disrupted expression in rostral-caudal compartments of somites (Dunwoodie et al., 2002). Uncx expression is also severely down-regulated in Dll1 and RBPjk null embryos when Notch1 signaling is absent (Barrantes et al., 1999; Takahashi et al., 2003). Uncx is required for the formation of the pedicles and proximal ribs, structures which are malformed in Dll3 mutant animals (Kusumi et al., 1998; Mansouri et al., 2000). However, the vertebral malformations in Uncx mutants are distinct from those observed in Dll3 and much less severe. Therefore, the malformations observed due to Dll3 mutation are not likely to be mostly accounted for by Uncx-mediated effects.
Evidence suggests that Nrarp may regulate both the Notch and Wnt signaling pathways. Nrarp was originally identified in an expression screen in Xenopus embryos as a member of the Delta-Notch pathway (Gawantka et al., 1998). In Xenopus, activation of the Notch pathway resulted in elevated levels of Nrarp expression, demonstrating Nrarp is a target of the Notch pathway (Lamar et al., 2001). Once expressed, Nrarp functions as a negative feedback regulator of Notch signaling by binding the activated form of Notch (Notch-ICD) and promoting a decrease in Notch-ICD levels and a reduction in Notch target gene expression. Nrarp has also been found to be a positive regulator of the Wnt pathway in zebrafish, by stabilizing the LEF1 protein (Ishitani et al., 2005). Furthermore, the activation of LEF1 was not found to affect Notch signaling in zebrafish, so the actions of Nrarp appear to act independently on the canonical Wnt and Notch signaling pathways.
In the mouse, Nrarp was described as being expressed within the paraxial mesodermal in a rostral band and a broad caudal domain in the PSM; however, in this study Nrarp was not noted as a cycling gene (Krebs et al., 2001). The rostral band was localized to the caudal S0 (the newly forming somite). The broad caudal domain of Nrarp in the PSM was defined by a rostral boundary at the second presumptive somite (S-2). These were spatially verified by comparison with Uncx, a marker for the somite caudal compartment, and Mesp2, a marker for the anterior half of the second presumptive somite in the PSM (Krebs et al., 2001). More recently, Nrarp was described to display cycling expression based on identification in a microarray-based screen for genes with oscillatory expression in mouse PSM (Dequéant et al., 2006); and two expression phases for Nrarp have been described (Dequéant et al., 2006, Shifley et al., 2008).
While Nrarp expression in PSM has been reported at particularly stages, we sought to fully characterize the cycling expression of Nrarp in the PSM from early to late-somitogenesis at 8.5, 9.5, and 10.5 dpc. First we identified that distinct caudal-to-rostral shifts, characteristic of cycling genes, occurred at each stage (Fig. 3). Rostral bands of Nrarp expression displayed anterograde shifts and rostrocaudal contraction characteristic of cycling genes at all three stages. At 8.5 and 9.5 dpc, Nrarp expression in the caudal PSM extended over a large proportion of the PSM, but the most intense areas of expression displayed anterograde shifts. In addition, there was strong expression of Nrarp remaining in the primitive streak in 8.5 dpc embryos, similar to other Notch pathway genes such as Lfng (Barrantes et al., 1999; Dunwoodie et al., 1997). At 9.5 dpc, there was a residual level of Nrarp expression remaining throughout the PSM. At 10.5 dpc, cycling patterns of Nrarp caudal bands were more clearly evident due to loss of this residual expression in the PSM and the tailbud.
The phase of a cycling gene is defined by its spatial expression along the caudal-to-rostral axis within the embryonic PSM at a particular point in time during a segmentation cycle (Pourquié and Tam, 2001). Using the phase descriptions used for other cycling genes, we defined three phases for Nrarp expression (Fig. 3G–I and M–O). In phase I, caudal expression of Nrarp remained in S1 at 8.5 and 9.5 dpc, but not at 10.5 dpc. Rostral bands of Nrarp expression in S0/S1 contracted along the rostrocaudal axis between phase I and III. Expression of Nrarp in the PSM was localized to the tailbud and caudal region in phase I, shifted rostrally in phase II, and condensed towards S-2 in phase III. Thorough description of these phases was a necessary step prior to analysis of effects of Dll3, Lfng, and Wnt3a mutations on Nrarp cycling expression.
Mouse embryos were bisected at 9.5 dpc, with left halves cultured for 15 minutes and then fixed, and right halves cultured for an additional 1 hour (Fig. 4A, n=8), 2 hours (Fig. 4B, n=7), or 3 hours (Fig. 4C, n=7) prior to fixation. Culturing led to a slight contraction of embryonic explant tissues, therefore halves were aligned with the first morphologically apparent forming somite. At a 1 hour time difference, anterograde shifts of Nrarp caudal expression were observed (Fig. 4A). At a 2 hour time difference, cultured halves displayed similar expression patterns of Nrarp, as shown in (Fig. 4B). At 3 hour time differences, cultured halves again differed in Nrarp caudal expression (Fig. 4C). The observation that cultured halves were in the same phase at 2 hours and out of phase at 1 and 3 hours is consistent with the 2 hour periodicity of cycling genes during mouse somitogenesis (Hirata et al., 2002).
Cycling genes tend to fall into two classes – those which cycle in concert with Notch and FGF pathway genes, and Wnt pathway-associated genes that do not. Comparing expression patterns of cycling genes can help give clues about the dynamics within and between gene pathways. Genes have been described as being “in phase” if similar expression patterns from phase I to III are observed for the two genes. Previous models have proposed that cycling genes within the same pathway will display more similar phases than genes in different pathways (Aulehla et al., 2003; Dequéant and Pourquié, 2008). When assayed by microarray analysis of caudal PSM tissues, levels of Nrarp expression were shown to peak in correlation with Lfng but not with Axin2 (Dequéant et al., 2006). Some genes such as Nkd1 have been shown to display dependency on both Notch and Wnt signaling, with Nkd1 transcription requiring Wnt3a and Notch signaling for oscillatory expression (Ishikawa et al., 2004).
We used bisected embryos to directly compare Nrarp cycling phases I, II, and III with patterns for Lfng and Axin2 in the other half (Fig. 5). We observed that the caudal-to-rostral progression and boundaries of Nrarp within the PSM were more comparable to that of the Notch modulator Lfng at 10.5 dpc (Fig. 5A–C; phases for Nrarp as diagrammed in Fig. 2) than at 9.5 dpc (Supplemental Fig. S2, A–C). In comparison, Nrarp patterns were distinct from the caudal-to-rostral progression of Axin2 at 10.5 dpc (Fig. 5D–F) and 9.5 dpc (Supplemental Fig. S2, D–F). Previous reports have described Lfng and Axin2 expression as being out of phase with each other (Aulehla et al., 2003), and we also observed out of phase patterns at 10.5 dpc (Fig. 5G–I). In contrast, the phase differences between Lfng and Axin2 were more difficult to detect at 9.5 dpc, suggesting that these genes may grow increasingly out of phase as somitogenesis progresses (Supplemental Fig. S2 G–I, n=9).
To further investigate these observations, we categorized 9.5 and 10.5 dpc embryos that were probed for Nrarp, Lfng and Axin2 (Table 2). Expression phases I, II, and III have often been envisioned as consisting of equal representation of embryos during cycling, and by inference, relatively equal time periods. We observed for Lfng and Axin2, that departure from equal distribution between phases I–III was not significant by X2 analysis (Table 2). In contrast, Nrarp expression departed significantly from equal distribution between the three phases. For Nrarp expression at 9.5, this may be partially accounted for by the greater region of Nrarp expression along rostrocaudal PSM axis (Fig. 3), but this effect is also observed for Nrarp at 10.5 dpc. Therefore, the assumption that the amount of time that cycling genes spend in phases I, II, or III may not be equal for all genes. It is possible that Nrarp “rushes” through phase III, making it less likely to be represented.
Given that our microarray analysis of Dll3neo null embryos identified Nrarp as one of a limited number of genes with decreased expression, we examined the spatiotemporal pattern of Nrarp by whole mount in situ hybridization. In 9.5 dpc Dll3neo homozygous embryos, Nrarp expression appeared to be fixed in phase I (n=9; Fig. 6A), and by 10.5 dpc, Nrarp showed decreased expression in the caudal PSM and in the newly forming somite, S0 (Fig. 6E). Nrarp cycling expression was observed in Dll3neo heterozygous embryos (Fig. 6B–D, F–H) and wild-type embryos (data not shown). The static pattern of Nrarp expression in Dll3 null mutants fixed in phase I bares a strong resemblance to the static pattern of activated Notch1 expression in Dll3 null mutants at 10.5 dpc (Geffers et al., 2007). Static patterns of activated Notch1 expression have also been observed in Lfng null mutants at 11.5 dpc (Morimoto et al., 2005). In Dll3 mutants, the Notch pathway cycling genes Hes1, Hes5, and Lfng are also frozen in a single phase (Dunwoodie et al., 2002; Kusumi et al., 2004). Interestingly, only Hes7 manages to maintain oscillation in Dll3 null embryos, in contrast to findings in embryos constitutively expressing Notch1-ICD (Feller et al., 2008), suggesting that the autoregulatory feedback loop for this bHLH factor is less dependent on Notch signaling to cycle (Kusumi et al., 2004).
To compare the effects of mutations in Lfng with those in Dll3 on Nrarp cycling expression, we examined Lfng mutant embryos at 9.5 dpc. We observed that Nrarp expression appeared to be cycling in Lfngtm1Rjo null mutants at 9.5 dpc (Fig. 7E–G) and dynamic in the caudal PSM at 10.5 dpc (Fig. 7K–M) without the characteristic caudal-to-rostral “wave” seen in wild-type embryos (Fig. 7N–P). Nrarp expression did not appear to be highly dynamic early in somitogenesis at 8.5 dpc (Fig. 7A–B). It is interesting that, despite reports that activated Notch1 does not cycle in Lfng null mutants at 11.5 dpc (Morimoto et al., 2005), Nrarp appears to somehow be “kick-started” to display cycling expression later in somitogenesis. In contrast, Nrarp does not cycle in Dll3 mutants, thus suggesting that factors in addition to activated Notch1 cycling are required for somitogenesis.
Nrarp has been reported to display stable expression at 8.5 dpc in Lfng mutants with a specific disruption of a regulatory element required for oscillatory expression (Shifley et al., 2008). However, by 10.5 dpc in embryos lacking cycling Lfng, Nrarp cycling recovers (Shifley et al., 2008). Before Nrarp was reported to display cycling expression, the effects of Notch1, Dll1, Lfng and Dll3pu mutations on Nrarp expression had been described for the paraxial mesoderm (Krebs et al., 2001). Both Notch1 and Dll1 null mutants result in severe down-regulation of Nrarp in the PSM. Thus, the Dll1 and Dll3 ligands have very different effects on Nrarp expression. Differential functions for delta proteins and splice variants have also been described in zebrafish (Mara et al., 2007, 2008). Nrarp expression in null Lfng embryos was reported to be severely down-regulated for rostral PSM but not caudal PSM expression. This does not agree with the findings of Shifley et al (2008) and our observations; however, this observation was made before Nrarp was known to be a cycling gene and may be describing different normal phases.
Wnt signaling is required for the expression of many Notch and Wnt pathway genes in the PSM (Aulehla et al., 2003, 2008). Wnt signaling acts to restrict boundary determination genes to the anterior PSM where somite size, morphology and polarity are defined (Dunty et al., 2008). Genes in other pathways such as Cdx2 and Cdx4 do not require Wnt3a for expression within the PSM (Ikeya and Takada, 2001). Given the role of Wnt signaling in regulating PSM formation and maintenance, we examined Nrarp expression in mutants disrupted for Wnt signaling (Wnt3atm1Amc mutants; Fig. 8). Wnt3atm1Amc/tm1Amc embryos lack caudal somites and a tail bud and exhibit axial truncation (Takada et al., 1994). At 8.5 dpc, Wnt3atm1Amc/tm1Amc embryos expressed Cdx4 at comparable levels to +/+ (Fig. 8D, E), suggesting the presence of paraxial and lateral mesoderm. However, Nrarp expression was severely down-regulated (Fig. 8A), compared to heterozygous and +/+ embryos (Fig. 8B–C). At 9.5 dpc, Wnt3atm1Amc/tm1Amc embryos exhibited severely malformed and decreased amounts of PSM, confirmed by Cdx4 hybridization (Fig. 8I), and the remaining tissue also did not show any expression of Nrarp (Fig. 8F), compared to wild-type controls (Fig. 8G–H). Thus, it appears that Nrarp requires Wnt signaling to initiate expression within the PSM.
There are two paralogues of Nrarp in the zebrafish. Morpholino-based knockdown of Nrarp-a expression in zebrafish has been reported to result in slightly smaller animals with altered pigmentation and curly and shortened tail phenotypes (Ishitani et al., 2005). These phenotypes likely result from the destabilization of LEF1 protein by removing the ubiquitylation-blocking function of Nrarp, resulting in altered neural crest development. The shortened tail phenotype does not appear to be due to major disruptions in somite patterning, but more subtle malformations remain possible (M. Itoh, personal communication). In contrast, Nrarp-b morphants did not display any phenotype. However, knockdown of Nrarp-a and –b enhanced expression of her1 indicating they function redundantly as negative regulators of Notch signaling during somitogenesis in zebrafish (Ishitani et al., 2005). The phenotypes of Nrarp a/b morphants were not described, therefore, the presence of two Nrarp homologues raises the possible that overlapping function that might make it difficult to detect vertebral phenotypes.
To date, there are no reports of targeted mutations in the mouse or human genetic disorders associated with NRARP alleles. However, the lack of a segmentation phenotype in a targeted mutation of Nrarp alone would not necessarily eliminate a role in somitogenesis. Knockouts of the cycling genes Hes1 and Axin2 do not have any reported somite phenotypes, despite their key roles in the segmentation regulatory mechanism (Ishibashi et al., 1995; Lustig et al., 2002; Tomita et al., 1996). Within the Notch pathway, overlapping function of homologues can make phenotypes difficult to examine in single gene mutations, e.g., Mfng and Rfng mutations do not have an observable phenotype but display liver phenotypes when crossed with mutations in the paralogous gene Lfng (Ryan et al., 2008). Paralogues of Nrarp have not been described in either the mouse or human genomes.
Human mutations in DLL3 have been identified in spondylocostal dysostosis type I (SCD1), which has been distinguishable by radiologists from SCD3 caused by mutations in LFNG, and more recently in SCD4 due to mutations in HES7 (Bulman et al., 2000; Sparrow et al., 2006, 2008; Turnpenny et al., 2007). Mouse osteological changes have not been examined so far at comparable detail. With the observation that Dll3 mutation leads to disruptions of Uncx and Nrarp expression, it is possible that the unique SCD1 phenotypes due to DLL3 mutation may be accounted for by disruptions in regulation of these two genes. Uncx plays a key role in shaping the proximal ribs and the pedicles, the portions of the vertebrae that are between the vertebral body and the bilateral transverse processes (Mansouri et al., 2000). The Dll3 pudgy and targeted mutation alleles both display proximal rib fusions and bifurcations, and dysregulation of Uncx may play a role in these anomalies. Identification of Uncx from these studies also warrants more careful examination of human and mouse pedicle osteology in mutants. As described above, Nrarp mutations in mouse or human have not yet been identified. Given its cycling expression in somitogenesis and identification as down-regulated in Dll3 null mutants, information about the phenotype in Nrarp mutants would allow us to evaluate this gene in contributing to DLL3-related defects and as a candidate for human disorders of the spine.
Expression of selected genes identified in microarray screens. Whole mount in situ hybridization analysis was carried out for Pkdcc (n=5; A); 2610528A11Rik (n=5; B); Bcat2 (n=5; C); Hnrnpl (n=5; D). Pkdcc was expressed 4.8 fold enriched in the somite level tissues compared to PSM (Supplemental Table S1), and was found to be expressed in trunk mesenchyme (A). 2610528A11Rik displayed 7.4-fold greater expression in PSM compared to somite tissues (Supplemental Table S2) and was localized to the caudal endoderm (B). Bcat2 exhibited increased expression in PSM of Dll3neo/neo mutant tissues compared to wild type and displayed expression in the rostral pharyngeal arch 1 and in the truck mesenchyme (C; 1.6 fold in PSM shown in Fig. 2). Hnrnpl displayed increased expression in Dll3neo/neo mutant tissues compared to wild type (D; 2.1 fold in PSM and 1.9 fold in somites shown in Fig. 2).
Nrarp cycling expression phases compared with phases of Lfng and Axin2 at 9.5 dpc. Mouse embryos were bisected and halves were analyzed with two different probes by in situ hybridization (A–I). Nrarp expression was compared to Lfng (n=17; A–C) and the wnt pathway cycling gene Axin2 (n=30; D–F); and Lfng and Axin2 were compared to each other (n=9; G–I). Areas of peak expression are indicated (arrowheads). Although the caudal to rostral shifts in Nrarp expression are distinct when compared to Lfng and Axin2, they generally corresponded to equivalent phases of Lfng (A–C) but less so for Axin2 (D–F). Furthermore, Lfng and Axin2 have been described being expressed out of phase with each other (Aulehla et al., 2003), but this is less evident at 9.5 dpc (G–I). All half embryos are aligned at S0 and oriented with rostral at the top.
We thank Joshua Gibson, Walter Eckalbar and Michael Chacon for technical assistance. We thank Stephen Pratt for helpful discussions. This work was supported by NIH RO1 AR050687 (KK) and a Hitchings-Elion Fellowship of the Burroughs Wellcome Fund (KK), National Health and Medical Research Council (NHMRC) Project Grant 404804 (SLD and DBS), a Westfield-Belconnen Fellowship (DBS), a Pfizer Foundation Australia Senior Research Fellowship (SLD) and a NHMRC Senior Research Fellowship (SLD).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.