|Home | About | Journals | Submit | Contact Us | Français|
Somites are segmental units of the mesoderm in vertebrate embryos that give rise to the axial skeleton, muscle and dermis. Somitogenesis occurs in a periodic manner and is governed by a segmentation clock that causes cells to undergo repeated oscillations of gene expression. Here we present a detailed analysis of cis-regulatory elements that control oscillating expression of the zebrafish her1 gene in the anterior presomitic mesoderm. We identify binding sites for Her proteins and demonstrate that they are necessary for transcriptional repression. This result confirms that direct negative autoregulation of her gene expression constitutes part of the oscillator mechanism. We also characterize binding sites for fused somites/Tbx24 and Suppressor of Hairless proteins and show that they are required for activation of her1 expression. These data provide the foundation for a precise description of the regulatory grammar that defines oscillating gene expression in the zebrafish segmentation clock.
In vertebrates, the segmental structure of the vertebral column is prefigured by the metameric organization of the paraxial mesoderm in the embryo. This tissue forms as two columns flanking the midline (the notochord and neural tube) and is subdivided into a series of reiterated units known as somites. Somitogenesis is a sequential process with new segments forming from a region of unsegmented presomitic mesoderm (PSM) in the tail of the embryo. Groups of cells periodically bud off from the anterior end of the PSM to form a somite, while new cells are added to the PSM at the posterior end as a result of growth of the embryo. Thus, somitogenesis occurs in an anterior to posterior procession that reflects the overall progression of embryonic development. The total number of somites and the rate of somitogenesis are species-specific characteristics. For instance, the zebrafish forms 31 somites at the rate of one approximately every 23 minutes (Kimmel et al., 1995; Schröter et al., 2008), while somitogenesis occurs with a period of two hours in the mouse embryo, generating a total of 65 somites (Tam, 1981). Thus, the frequency of somite formation must be strictly regulated and coupled to overall growth of the embryo to ensure reproducibility in somite size and number.
Over the last 12 years, a substantial amount of molecular evidence has accumulated in support of the “clock and wavefront” model of somitogenesis, which was first proposed more than 30 years ago (Cooke and Zeeman, 1976). According to this model, the frequency of somite formation is defined by the period of the clock, which causes cells in the PSM to undergo repeated oscillations of gene expression. Cells are only competent to form a somite during a particular phase of oscillation. The position of a new somite boundary is determined by the wavefront, which moves posteriorly in association with growth of the embryo. A somite forms only when cells in the anterior PSM that have escaped the wavefront also experience the appropriate phase of the clock. Thus, the size of a somite is determined by the distance traveled by the wavefront during one cycle of the clock.
The Fgf and Wnt signaling pathways are important components of the wavefront. For both pathways, a gradient of activity exists across the PSM, with signaling highest in the posterior and absent in the anterior where segmentation occurs (Dubrulle et al., 2001; Sawada et al., 2001; Dubrulle and Pourquié, 2004; Aulehla et al., 2008; Dunty et al., 2008). Notch signaling has an important function in the segmentation clock and oscillating expression has been documented for several components and targets of the Notch pathway. This oscillation is manifest as reiterated waves of expression, which sweep through the PSM from posterior to anterior. Components of the Notch pathway expressed in this manner include deltaC in zebrafish and Lunatic fringe (Lfng) in mouse and chick (Forsberg et al., 1998; McGrew et al., 1998; Jiang et al., 2000). These genes encode a Notch ligand and a glycosyltransferase that modifies Notch activity, respectively. Particularly prominent among the targets of Notch signaling involved in the segmentation clock are members of the Hairy/Enhancer of Split-related family (abbreviated as Hes in mammals, her in zebrafish). Genes of this type showing oscillating expression are Hes1, Hes5, Hes7 and Hey2 in the mouse, cHairy1, cHairy2 and cHey2 in the chick and her1, her7, her11, her12 and her15 in zebrafish (Palmeirim et al., 1997; Holley et al., 2000; Jouve et al., 2000; Leimeister et al., 2000; Sawada et al., 2000; Bessho et al., 2001; Dunwoodie et al., 2002; Henry et al., 2002; Oates and Ho, 2002; Sieger et al., 2004; Shankaran et al., 2007). These genes encode proteins of the basic-Helix-Loop-Helix (bHLH) family that are thought to function as transcriptional repressors (Fischer and Gessler, 2007).
In zebrafish, a negative feedback loop involving Notch signaling and the her genes is thought to be the core mechanism of the oscillator (Holley et al., 2002; Oates and Ho, 2002; Lewis, 2003). Activation of the Notch receptor by ligand binding leads to cleavage of the Notch intracellular domain (NICD), which translocates to the nucleus and interacts with a DNA binding protein of the CSL family (for CBF-1/RBPJ-k in mammals, Suppressor of Hairless (Su(H) in flies, Lag-1 in nematodes) (Bray, 2006). CSL binds to specific sites in the cis-regulatory regions of targets, presumably including the her genes, and the CSL/NICD complex recruits further cofactors to activate transcription (Bray, 2006). As Her proteins subsequently accumulate, it is supposed that they bind directly to regulatory elements of their own genes and repress transcription. The Her proteins have short half-lives (Giudicelli et al., 2007) and are rapidly degraded, therefore transcriptional repression is quickly relieved and a new cycle of expression can be initiated in response to Notch signaling. In the mouse, transcriptional repression by Hes proteins is also likely to be an important component of the oscillator mechanism (Bessho et al., 2003). However, the situation is substantially more complex as components of the Wnt and Fgf signaling pathways also show cyclical expression (Aulehla et al., 2003; Ishikawa et al., 2004; Dale et al., 2006; Dequeant et al., 2006; Niwa et al., 2007). It is not yet clear how oscillations in these pathways interact with each other and with Notch signaling within the segmentation clock.
In order to obtain a more fundamental understanding of the mechanism of the segmentation clock (and to refine mathematical models of somitogenesis) it will be necessary to acquire detailed information on the cis-regulatory elements of oscillating genes. Specifically, it will be crucial to know how the number and arrangement of binding sites for activators and repressors of transcription defines the proper oscillation of gene expression. However, very little is currently known about this regulatory grammar in clock genes.
In vitro studies have identified binding sites for the mouse Hes7 protein in the promoter regions of Hes7 and Lfng (Chen et al., 2005). Moreover, chromatin immunoprecipitation (ChIP) from mouse PSM extracts demonstrated that Hes7 is bound to DNA upstream of the Hes7 and Lfng genes in vivo (Bessho et al., 2003).Therefore, it seems likely that Hes7 is directly regulating transcription of both genes via the motifs identified by Chen et al. (Chen et al., 2005). However, the role of specific binding sites has not yet been confirmed in vivo. A 2.5kb DNA fragment immediately upstream of the mouse Hes1 gene is sufficient to drive oscillating expression of a luciferase reporter in transgenic animals (Masamizu et al., 2006). Indeed, this construct was used to visualize real-time oscillations in live embryos. Four Hes1 binding sites have been identified close to the Hes1 transcription start site and three of them are required for repression of promoter activity by cotransfected Hes1 in cultured cells (Takebayashi et al., 1994). Furthermore, two CSL binding sites in the Hes1 promoter are required for the induction of reporter expression by Notch signaling in transfection experiments (Jarriault et al., 1995). However, the function of neither the Hes nor the CSL binding sites have yet been determined in respect to Hes1 expression in the segmentation clock.
Analysis of reporter gene expression in transgenic mice demonstrated that as little as 2.3kb of 5′ flanking sequence of the mouse Lfng gene is sufficient to recapitulate the complete cyclic expression pattern (Cole et al., 2002; Morales et al., 2002). A similar fragment (only 2kb) has also been used to image oscillating gene expression in real-time in live embryos (Aulehla et al., 2008). Within the Lfng upstream region are several short elements that are highly conserved between the mouse and human sequences. One of these, referred to as Fringe Clock Element 1 (FCE1) (Cole et al., 2002) or block A (Morales et al., 2002), is necessary for oscillating expression in the posterior PSM. Moreover, deletion of FCE1 from the genomic locus leads to the complete absence of Lfng expression in the posterior PSM and causes significant segmentation defects in the axial skeleton (Shifley et al., 2008). Several putative protein binding sites were identified within FCE1, including two possible targets for bHLH proteins (Cole et al., 2002). Mutation or deletion of these sites completely eliminated expression of reporter genes in the posterior PSM of transgenic embryos. However, binding of specific proteins to these sites has not yet been established and these motifs were not among the Hes7 binding sites identified in vitro (Chen et al., 2005). FCE1 also contains a single consensus binding site for CSL proteins and a second such motif is present in another of the conserved regions, block B (Morales et al., 2002). Mutation of both sites within one construct caused a dramatic reduction in gene expression in the posterior PSM, although the pattern was apparently still cyclical. Subsequently, CSL binding to the site within FCE1/block A was confirmed by an in vitro binding assay (Morales et al., 2002). Currently, this finding represents the only example of a confirmed binding site for a specific protein that has been shown to be important for the regulation of oscillating gene expression in the segmentation clock.
In zebrafish, cis-regulatory elements have only been described for one gene with oscillating expression in the PSM, her1 (Gajewski et al., 2003). The full cyclical pattern of the endogenous gene was recapitulated in transgenic zebrafish using a GFP reporter driven by the 8.6kb of DNA immediately upstream of her1. A smaller 3.3kb fragment did not direct expression in the most posterior region of the PSM but did produce a dynamic pattern of two or three stripes in the intermediate to anterior PSM. However, a further deletion down to −2.8kb led to reporter constructs that displayed only weak expression over a broad region of the PSM with no evidence of oscillation (Gajewski et al., 2003). These experiments defined the 3.3-2.8kb region as essential for her1 expression in the anterior PSM. We therefore refer to this fragment as the Anterior Stripe Element (ASE).
Here, we report a detailed analysis of the ASE. We identify two binding sites for Her proteins and show that they are required for transcriptional repression of her1. We also demonstrate that the ASE contains a single binding site for T-box proteins and that this motif is necessary for expression in the anterior PSM. Furthermore, we show that there are no binding sites for the zebrafish CSL proteins (Su(H)1 and Su(H)2) within the ASE but identify two sites elsewhere within the 3.3kb enhancer that are essential for her1 expression. These experiments provide evidence for direct negative autoregulation of her genes in vivo and confirm that her1 is a direct target of Notch signaling, both important features of the prevailing models of the segmentation clock.
In order to precisely define the ASE for our investigation, we mapped the primers previously used to generate transgenic constructs (Gajewski et al., 2003) to the genomic sequence. This determined that the exact limits of the ASE are −3282bp and −2879bp, relative to the her1 initiation codon, and are therefore correctly rounded to −3.3kb and −2.9kb (Fig.1A). In our analysis of the ASE, we focused on particular candidate transcription factors that are likely to be direct regulators of her1 expression, based on genetic evidence and their expression patterns. Her proteins, including Her1 itself, are thought to be transcriptional repressors and several members of this family are expressed in the zebrafish PSM. Furthermore, knockdown of these her genes generally affects the expression of oscillating genes in the segmentation clock, including her1 (Holley, 2007). In our binding assays we have used Her1, because autoregulation is thought to be an important feature of the segmentation clock, and Her12 because knockdown of this gene produces a strong perturbation of somitogenesis (Shankaran et al., 2007). Tbx24 is a T-box transcription factor expressed in the anterior PSM and disrupted by the fused somites (fss) mutation (Nikaido et al., 2002). The most anterior stripe of oscillating her1 expression is lost in fss embryos (van Eeden et al., 1998; Holley et al., 2000). Finally, her genes are targets of the Notch signaling pathway, the activity of which is modulated by the CSL proteins (Bray, 2006). There are two paralogous members of this family in zebrafish, Su(H)1 and Su(H)2. These proteins are partially redundant and morpholino-mediated knockdown of either disrupts her1 expression (Sieger et al., 2003; Echeverri and Oates, 2007). We synthesized each of the candidate proteins using an in vitro translation system and tested their ability to bind to oligonucleotide probes derived from the ASE sequence in electrophoretic mobility shift assays (EMSA). The probes represented putative binding sites for the candidate proteins that were identified by visual inspection of the ASE sequence (Fig.1B).
Her proteins are members of the basic-helix-loop-helix (bHLH) family, which generally bind to DNA motifs known as E-boxes (CANNTG) or N-boxes (CACNAG) (Fischer and Gessler, 2007). More specifically, members of the Hairy/Enhancer of Split class recognize class B or class C E-boxes, with the consensus sequence CACGYG. Both Her1 and Her12 bound a control probe containing a CACGTG site and showed a greater affinity for this sequence over other variations of the E-box or N-box (Fig.2 and data not shown). Consequently, no protein binding was observed with several variant E-boxes (CAGGTG, CAGGCG and CATGCG) located in the ASE (Fig.1B and data not shown). In contrast, two palindromic sites (CACGTG; Fig.1B) were specifically bound by Her12 or Her1 (Fig.2). bHLH proteins bind DNA as dimers, with each monomer making specific base contacts with nucleotides 1 and 2 of the binding site (i.e. CACGTG) on opposing strands (Massari and Murre, 2000). Therefore, we modified four nucleotides in our motifs (GTCGAC) and showed that Her protein binding is completely abolished, using the mutated forms of the control probe and the ASE sites E1 and E2 (Fig. 2).
Although the target sequence has not yet been determined for Tbx24 specifically, binding site selection experiments with several other T-box proteins suggest that all members of this family recognize the consensus sequence TCACACCT (Naiche et al., 2005). Consistent with this data, Tbx24 bound a control probe containing this sequence in EMSA experiments (Fig.3). Note that because full-length Tbx24 was almost impossible to synthesize in vitro under our conditions, we used a truncated version, Tbx24N. This protein consists of the N-terminal 405 amino acids (compared to 874 in the full-length protein) and includes the complete T-box DNA binding domain (residues 62 to 247). EMSA experiments confirmed that this truncated version of Tbx24 was able to bind a control probe containing a proven binding site for T-box proteins (Fig.3) (Goering et al., 2003). We identified five motifs within the ASE that are close matches for the T-box consensus but none were able to bind Tbx24N (Fig.1B and data not shown). However, there is one additional site that perfectly matches the consensus (Fig.1B) and a probe containing this site did specifically bind Tbx24N (Fig.3). A single nucleotide change (TCAGACCT) in the T-box site of the control probe is sufficient to abolish Tbx24N binding (Fig.3), as reported for other T-box proteins (Conlon et al., 2001; Goering et al., 2003). Introducing the equivalent mutation into the T1 probe from the ASE also completely abrogated protein binding (Fig.3).
CSL proteins recognize a consensus binding site of YGTGRGAA, which was largely determined by experiments in Drosophila (Nellesen et al., 1999). Although CSL can bind single sites, a common target sequence is the SPS motif, in which two sites are arranged head-to-head and separated by 16-17bp (Nam et al., 2007). Surprisingly, we were unable to identify any bona fide Su(H) binding sites within the ASE despite testing five putative sequences, including an SPS motif previously predicted by bioinformatics (Gajewski and Voolstra, 2002). Lack of binding was not due to technical shortcomings of our assay as we were able to identify Su(H) binding sites in other regions of the her1 upstream sequence (see below). Therefore, it seems likely that the Notch signaling pathway does not directly regulate her1 expression through the ASE.
In summary, our in vitro analysis of the ASE identified two binding sites for Her proteins (E1 and E2) and one for Tbx24 (T1) (Fig.1B). Moreover, we determined specific nucleotide changes that abolish protein binding to each site. In order to test the function of these motifs in the regulation of her1 expression in vivo, we introduced the appropriate mutations into reporter constructs, which were then used to generate transgenic fish.
The constructs previously used to determine the activity of the 3.3kb region of her1 did not include any her1 coding sequence or the 3′ untranslated region (UTR) (Gajewski et al., 2003). We wanted to replicate this design as closely as possible and, therefore, initially created constructs in which the 3.3kb region of her1 was coupled to the emerald variant of GFP (emGFP) and the SV40 polyadenylation (SV40 polyA) signal. However, in situ hybridization of embryos from four transgenic lines carrying this construct revealed that emGFP transcripts were present in a broad pattern in the anterior PSM, not the expected stripes, and were also maintained throughout the somites (data not shown). The most likely explanation for this pattern is that the SV40 polyA sequence causes excessive stabilization of transcripts, masking the oscillating expression pattern in the PSM and leading to retention of emGFP mRNA in cells well after they have been incorporated into somites. We note that Gajewski et al. (Gajewski et al., 2003) used a transcriptional terminator sequence from the Drosophila hsp70 gene. This sequence obviously did not significantly stabilize transcripts and allowed the authors to detect dynamic expression in the PSM. In order to resolve this issue, we exchanged the SV40 polyA sequence in our constructs for the her1 3′UTR and successfully restricted transgene transcripts to the anterior PSM (see below). This basic design was then used for all reporter constructs, although all modifications of the cis-regulatory sequence were performed in the context of a mCherry reporter instead of emGFP (Fig.4A).
Fluorescent proteins expressed under the control of her1 cis-regulatory elements do not provide an accurate reflection of oscillating expression due to their inherent stability ((Gajewski et al., 2003) and our data not shown). Therefore, fluorescence served as a useful means of screening for transgenesis but analysis of expression was performed by in situ hybridization. However, fluorescence also proved to be a more sensitive assay of expression levels. For almost all of our constructs we obtained transgenic lines for which we observed specific expression patterns by fluorescence but could not detect any signal by in situ hybridization (Fig.4A).
Although we identified three lines for the control construct Tg(−3.3her1:emGFP) by fluorescence, only one of these had a level of mRNA expression that was detectable by in situ hybridization (Fig.4A). As expected, expression was restricted to stripes in the anterior PSM (Fig.4B) in a pattern very similar to that previously attributed to the −3.3kb enhancer (Gajewski et al., 2003). However, we were never able to detect the expected third, most posterior stripe by the standard in situ hybridization method, and the two anterior stripes were generally more poorly defined than previously reported. Fluorescent in situ hybridization resolved the expression pattern of Tg(−3.3her1:emGFP) in greater detail, revealing that there is actually weak expression in the region of the most posterior stripe but confirming that all the stripes are poorly defined (Fig.4C).
Double in situ hybridization was previously used to demonstrate that the expression pattern driven by the −8.6kb her1 fragment is virtually identical to that of the endogenous gene (Gajewski et al., 2003). However, a similar analysis was not reported for the −3.3kb enhancer. Therefore, we performed double fluorescent in situ hybridization for her1 and emGFP expression in Tg(−3.3her1:emGFP) embryos (Fig.4C-E). This analysis demonstrated that although Tg(−3.3her1:emGFP) expression substantially overlaps with her1, the patterns are not identical. Most notably, the relative intensity of different stripes varies considerably. For Tg(−3.3her1:emGFP), stripe one is more intense than stripe two, while the opposite is true for her1. Moreover, remnants of an even more anterior stripe are detected for Tg(−3.3her1:emGFP) in a region where her1 expression is absent (Fig.4C-E). The expression of endogenous her1 was used to arrange embryos in a series that represents different phases in the progression of the somitogenesis cycle (Fig.5). Comparison of Tg(−3.3her1:emGFP) expression in the same embryos indicated that in the anterior PSM, downregulation of the transgene seems to lag behind that of the endogenous gene leading to the apparent differences in stripe intensity.
Overall, our results suggest that, although Tg(−3.3her1:emGFP) expression resembles that of the endogenous her1 gene in the anterior PSM, it is not as precise and proper temporal control of oscillation is lacking. As the −8.6kb fragment appears to perfectly recapitulate her1 expression, elements in the region between −8.6kb and −3.3kb must be required in conjunction with the ASE to coordinate proper cyclical expression in the anterior PSM.
Although our data suggest that the ASE is not sufficient to fully recapitulate her1 expression in the anterior PSM, it has been shown to be necessary (Gajewski et al., 2003). We also generated a construct (Tg(−2.9her1:mCherry)) in which reporter gene expression was driven by the −2.9kb her1 fragment (Fig.4A). This construct was intended to serve as a negative control for other constructs containing mutated binding sites by confirming the expression pattern associated with complete absence of the ASE. The 2.9kb fragment was previously reported to direct only weak, diffuse expression in the PSM (Gajewski et al., 2003). In contrast, we consistently saw a complete lack of expression in the PSM, coupled with ectopic transcription in the somites (Fig.4F). The reason for this discrepancy with previous data is not immediately obvious as our construct differs only in the fluorescent protein reporter gene used and the inclusion of the her1 3′utr. We cannot exclude the possibility that elements within the 3′utr drive the ectopic somitic expression because, as described above, constructs that contain the SV40 polyA instead generate excessively stable transcripts that persist throughout the somites anyway (data not shown). Nevertheless, our results suggest that the ASE is not simply necessary for expression of her1 in the PSM but is also required to repress inappropriate transcription within the somites.
The particular transgenic line shown in Fig.4F also displayed significant ectopic expression in the notochord. This is not specific to the Tg(−2.9her1:mCherry) construct as we recovered at least one line for each of our constructs in which the expression pattern included the notochord (for example, Fig.4K,L). Ectopic notochord expression was previously described for transgenic reporters containing the complete 8.6kb her1 enhancer but not the 3.3kb fragment, and attributed to a cryptic notochord enhancer in the 8.6 to 3.3kb region (Gajewski et al., 2003). The notochord expression in our constructs cannot be attributed to an additional enhancer in the her1 3′utr as we also saw this pattern frequently with transgenic reporter constructs that contained the SV40 polyA signal in place of the her1 sequence (data not shown). Therefore, our experiments indicate that the 3.3kb enhancer of her1 does contain sequence capable of directing expression in the notochord and another region at the genomic locus (outside of the −8.6kb fragment) must be responsible for repressing this activity during normal zebrafish development.
Although our results indicate that the −3.3her1:emGFP construct does not fully recapitulate her1 expression in the anterior PSM, it is clear that the ASE is necessary for expression in this region. Therefore, we were still able to use this reporter construct as a context for analyzing the function of the protein binding sites that we identified in vitro. Specific nucleotide changes that abrogate Her protein binding (Fig.2) were introduced in order to generate reporter constructs in which either the E1, E2 or both (E12) binding sites in the ASE were disrupted (Fig.4A). These constructs were then used to generate stable transgenic lines. As these are randomly integrated transgenes, the level of expression obviously depends on both the number of copies and the integration sites. However, as these are reporter constructs, the mutations should only affect the readout from the transgene and should not disrupt the segmentation clock in general. Thus, any effect of the mutations provides a direct measure of the role of these binding sites in her1 regulation and is not complicated by any indirect effect via clock components.
Mutation of the E1 binding site did not have a dramatic effect (Fig.4G). Expression in Tg(−3.3her1.E1:mCherry) embryos was still restricted to the anterior PSM and was not substantially different from that seen with the control construct (Fig.4B), although it could be slightly more disorganized. Mutation of the E2 binding site caused a more pronounced change (Fig.4H). Tg(−3.3her1.E2:mCherry) embryos displayed a splotchy, disorganized pattern with no evidence of stripes. Moreover, the domain of expression was slightly expanded with positive cells detected in more anterior positions than in embryos transgenic for the parent construct (Fig. 4B). In contrast, combined mutation of both E-boxes had a much more dramatic effect (Fig.4I). Relative to the control construct (Fig.4B), expression in Tg(−3.3her1.E12:mCherry) embryos was expanded throughout the intermediate to anterior PSM in an apparently even distribution with no evidence of stripes. Her proteins are thought to function as repressors and the observed pattern is consistent with a failure to repress transcription during appropriate phases of oscillation. Interestingly, the expression pattern in Tg(−3.3her1.E12:mCherry) embryos is very similar to that seen when her1 or her7 function was knocked down in transgenic fish carrying a construct with the unmodified −3.3kb enhancer (Gajewski et al., 2003). This result suggests that the binding sites we have identified are genuine targets for Her protein binding in vivo.
To further investigate the role of Her protein binding sites in the regulation of her1 expression, we determined the effects of increasing the number of Her target sequences. A DNA fragment containing four CACGTG sites, spaced at 10bp intervals, was added at the 5′ end of the −3.3kb her1 enhancer (Fig.4A). Four sites were chosen so that, in combination with the two present in the ASE, the construct was effectively 3x in respect to Her target sites. Injection of this construct, yielded an unusually high proportion (3/6) of non-expressing transgenic lines, suggesting that transcriptional activation is generally downregulated (Fig.4A). Moreover, in those lines with detectable expression, the in situ hybridization signal was very weak and mCherry mRNA was almost completely absent from the PSM, with only a few scattered cells detected in the most anterior unsegmented region (Fig.4J). Combined with the results of the binding site mutations, these data indicate an important role for E-box motifs in mediating transcriptional repression of her1 in the segmentation clock, likely through specific binding of Her proteins.
In addition to the Her binding sites, our in vitro analysis identified one Tbx protein target site (T1) within the ASE (Figs.(Figs.1,1, ,3).3). A single nucleotide change that abolishes Tbx24 binding to the T1 site (Fig.3) was introduced into a reporter construct that was subsequently used to generate transgenic zebrafish (Fig.4A). Tbx24 is thought to be an activator of her1 expression (Nikaido et al., 2002), so mutation of the T1 binding site might be expected to abolish expression of the reporter gene. However, all the transgenic lines that we identified by PCR also showed specific patterns of expression by fluorescence, although only one of these lines had a level of expression that was detectable by in situ hybridization. These Tg(−3.3her1.T1:mCherry) embryos displayed ectopic expression in the notochord plus some splotchy expression in the somites. As noted above for the Tg(−2.9her1:mCherry) construct, the notochord is a frequent site of ectopic expression and therefore this is unlikely to be a specific effect of mutating the Tbx site. The additional expression in the somites is reminiscent of the effect of deleting the entire ASE (Fig.4F) and could imply a role for Tbx proteins in repressing transcription of her1 in this region. However, additional transgenic lines will be required to confirm or refute that hypothesis. Strikingly, no expression was detected in the anterior PSM of Tg(−3.3her1.T1:mCherry) embryos (Fig.4K,L). However, a proportion of the embryos (15/38) showed additional patches of expression in central regions of the PSM (Fig.4L). It seems likely that this represents the most posterior stripe (stripe 3) driven by the 3.3kb enhancer and that the variable appearance of this band in Tg(−3.3her1.T1:mCherry) embryos reflects the dynamic expression pattern. Nevertheless, it is clear that the T-box binding site in the ASE is required for activation of her1 expression in the anterior PSM.
As we had done for Her binding sites, we sought to gain further insight into the regulation of her1 by Tbx proteins by increasing the number of available binding motifs. Employing the same rationale, we introduced two extra T-box sites so that, in addition to the one present in the ASE, the construct was 3x in respect to T-box protein targets. Therefore, a DNA fragment containing two TCACACCT sites, 10bp apart, was cloned upstream of the −3.3her1:mCherry reporter (Fig4.A). Tg(2T-3.3her1:mCherry) embryos showed only a modest increase in expression relative to the control construct (Fig.4B,M). Furthermore, although expression is expanded to a larger area of the PSM, some evidence of a striped pattern remains, indicating that the normal cyclical expression has not been completely over-ridden. This result suggests that these two extra T-box sites are not sufficient to mediate robust upregulation of the transgene.
As noted above, we were surprised by the lack of Su(H) binding sites in the ASE because her1 is known to be a target of the Notch signaling pathway. Therefore, we searched the rest of the −3.3kb fragment for potential Su(H) target sites and identified two motifs (S1 and S2) that perfectly matched the consensus CSL sequence (Fig.6A, B). Both of these were individual sites, rather than SPS motifs, and were located 2337bp (S1) and 1155bp (S2) upstream of the her1 initiation codon. EMSA experiments confirmed that both sites would bind either zebrafish Su(H) paralogue (Fig.6C). The specificity of binding was confirmed by competition assays. An oligonucleotide containing an SPS motif from the mouse Hes1 gene (Jarriault et al., 1995; Nam et al., 2007) effectively competed for Su(H) binding to either the S1 or S2 site (Fig.6C and data not shown). Furthermore, specific nucleotide changes that have been shown to prevent CSL binding (Nam et al., 2007), completely abolished the ability of the Hes1 SPS oligonucleotide to compete. Similarly, removing the putative 8bp Su(H) sites from the S1 and S2 probes renders them unable to compete for protein binding (Fig.6C). Data is shown for the S1 probe with Su(H)2 protein but identical results were obtained with both probes using either zebrafish Su(H) paralogue (data not shown).
We engineered two 8bp deletions in our basic reporter construct that removed both of the Su(H) binding sites and generated transgenic fish carrying the modified construct (Fig.6D). An unusually high proportion (4/7) of the lines that we recovered showed no fluorescence and could be identified by PCR only (Fig.6D), indicating that deletion of the Su(H) binding sites leads to a substantial downregulation of reporter gene activity. Moreover, of the three lines in which expression was detectable by in situ hybridization, two showed no expression of emGFP mRNA in the PSM (Fig.6E). In the one line that exhibited PSM expression, the pattern was clearly abnormal with only a few scattered positive cells and no evidence of stripes (Fig.6F). Thus, even at the maximum level of expression observed with this construct, oscillating expression in the anterior PSM is severely disrupted. Therefore, our data demonstrates that Su(H) sites located outside the ASE are required for normal activation of her1 transcription in the anterior PSM.
In the present study, we have defined binding sites for Her proteins that are essential in vivo for oscillating gene expression in the segmentation clock. These sites are important for transcriptional repression, thus confirming an important component of the oscillator model: direct negative autoregulation of her gene expression. Mathematical modeling predicts that the number of binding sites for Her/Hes repressors will be an important influence on the behavior of the oscillator (Zeiser et al., 2006). We identified two Her target sites within the ASE (E1 and E2) and observed that mutating either of them individually has a much less pronounced effect than mutating both. However, the level of resolution of our experiments does not permit us to draw any inferences about the role of Her binding site number in controlling the dynamics of her1 expression. Furthermore, a third CACGTG motif was present in all our constructs. This site is located outside the ASE at approximately 2.5kb upstream of the her1 coding region. However, we have not confirmed that this site binds Her proteins, although that can be expected based on our current data, or investigated whether it is important for the regulation of her1 expression.
The affinity of Her protein binding sites will likely also have a significant effect on the behavior of the oscillator (Zeiser et al., 2006). Our in vitro binding experiments indicate that the E1 motif is a higher affinity site than the E2, yet mutation of E2 alone produced a more obvious effect on gene expression than disruption of E1. However, we do not know definitively which members of the Her protein family bind these sites in vivo. Several her genes are expressed in the zebrafish PSM, and the potential exists for multiple homo- and hetero-dimer species. Modeling has shown that ascribing different properties to different dimer species can help to more accurately recapitulate the biological data (Cinquin, 2007). Our current data indicates that there are no differences in target site specificity between different Her proteins but it will be important to determine the relative affinities of different proteins for the binding sites in the ASE in order to fully understand the regulation of her1 expression. Finally, cooperative binding of Her proteins to multiple sites is also predicted to effect clock dynamics (Zeiser et al., 2006). We have not tested this prediction explicitly for E1 and E2 but the substantial space between them suggests that this may not be a factor in the regulation of her1.
Issues of affinity and cooperativity may underlie the slightly counterintuitive result obtained with our construct that included additional Her binding sites. As the transgene was expressed in the context of a functional endogenous segmentation clock, Her proteins should still only be available to repress transcription in a cyclical manner. Thus, we expected to see some residual transcription of the transgene rather than the complete loss of expression that was observed. However, the additional Her binding sites were all of the type present in the control probe used in our EMSA experiments. These sites contain optimized flanking residues (Jennings et al., 1999) and are apparently higher affinity sites than the endogenous sites in the ASE (Fig.2). Moreover, the extra sites were arranged in a closely spaced array (with 10bp between each site), while the two sites in the ASE are separated by 79bp. It seems plausible that this arrangement could also facilitate cooperative binding of Her proteins to adjacent motifs, further enhancing the affinity of these sites. Thus, this array of four E-boxes might mediate the assembly of a robust transcriptional repression complex that is not appropriately degraded or displaced during progression of the clock. Ultimately, analysis of Tg(4E-3.3her1:mCherry) embryos provided additional confirmation of the role of Her protein binding sites in mediating transcriptional repression but did not offer further significant insight into the mechanism of the oscillator. More instructive data may be obtained in the future by adding fewer and/or lower affinity Her protein binding sites or by modifying the affinity of the endogenous sites in the ASE. Furthermore, if cooperative binding of Her proteins does occur on adjacent sites, then the spacing between them may also prove to be relevant. It is also important to note that we have not formally excluded the possibility that the array of Her binding sites in this transgene also impacts the endogenous clock. If the pool of available Her proteins is limiting, the high affinity sites in the Tg(4E-3.3her1:mCherry) construct might sequester them away from regulating endogenous oscillating genes. However, we did not observe any segmentation defects in these embryos (or any of our transgenic lines), indicating that the clock is indeed functioning normally.
The correct balance between the number of binding sites for activators and repressors is likely to be critical for proper behavior of the oscillator. We identified a single Tbx protein binding site in the ASE, specifically confirming an interaction with Tbx24. Mutating this site completely abolishes reporter expression in the anterior PSM supporting a role for Tbx proteins in the activation of her1 transcription. In contrast, loss of tbx24 function in fss mutants results only in the loss of the most anterior stripe of her1 expression (van Eeden et al., 1998; Holley et al., 2000). This discrepancy might indicate that other Tbx proteins are regulating her1 through the same binding site. Indeed, recent studies suggest that her1 could be regulated by two other Tbx proteins, No tail (Ntl) and Spadetail (Spt/Tbx16) (Garnett et al., 2009; Morley et al., 2009). However, the expression domains of ntl and spt preclude a role in controlling her1 expression in the anterior PSM. Another family member, tbx6, is expressed in the posterior to intermediate PSM, overlapping slightly with tbx24 (Nikaido et al., 2002), but not in the more anterior region in which reporter expression is lost after mutation of our binding site. Thus, although it is possible that any or all of these other Tbx proteins regulate her1 expression via the site we have identified in other regions of the embryo, it is not yet clear why mutation of the Tbx site in the ASE causes a more pronounced effect than loss of tbx24 function.
Tbx24 is expressed throughout the anterior PSM but obviously only activates oscillating genes periodically. One explanation for this would be that binding of Her proteins and Tbx24 to cis-regulatory elements is mutually exclusive. We have not explicitly tested this hypothesis for the ASE but given the spacing of the sites we believe it is unlikely. It seems more plausible that Tbx24 can bind to target sites in cells throughout the anterior PSM but is only capable of activating transcription in synergy with periodic loss of Her repression. Accordingly, introducing two additional T-box binding sites to our reporter caused a very modest increase in expression at most and apparently did not override the normal oscillation in the anterior PSM.
Our Tbx24 experiments were also constrained by our relative ignorance of what constitutes a target sequence for Tbx24 binding. Although it seems that all Tbx proteins recognize sites very similar to the TCACACCT consensus, it is not yet clear what variations in that sequence are tolerated by specific Tbx proteins in vivo. Moreover, preferences for orientation and spacing of binding sites have been documented for some Tbx proteins (Conlon et al., 2001; Farin et al., 2007). Interestingly, the zebrafish Tbx24 protein binds motifs in the medaka mespb promoter that contain two adjacent sites in opposite orientations (Yasuhiko et al., 2008). Precisely defining the preferred binding motif for Tbx24 should allow us to more effectively manipulate oscillating expression in the future.
Notch signaling does not seem to regulate her1 expression directly through the ASE as we could not identify Su(H) binding sites within that region. Instead we confirmed two such motifs within the −2.9kb region. These sites are essential for transcriptional activation as expression is almost completely abolished in the anterior PSM when the sites are mutated. In contrast, all previous zebrafish studies in which Notch signaling was disrupted have demonstrated that her1 expression is not eliminated but present in a salt and pepper pattern in the anterior PSM (Holley, 2007). Since our transgene lacking Su(H) binding sites was analyzed in the context of a normal segmental pattern, we would not expect to see the salt and pepper expression pattern. Nevertheless, as some her1 expression is maintained in the absence of Notch signaling, we might expect to see some residual reporter gene transcription, perhaps even stripes. Furthermore, the salt and pepper expression of her1 in the anterior PSM after loss of Notch activity is dependent on fss/tbx24 (van Eeden et al., 1998; Holley et al., 2000; Sieger et al., 2003). We have shown that our constructs contain a site for Tbx24 binding, yet they show no expression after deletion of the Su(H) binding sites. Together, our result suggests that the 3.3kb enhancer of her1 lacks some other cis elements necessary for residual expression in the absence of input from the Notch pathway. These elements would likely also contribute to normal regulation of oscillating expression.
Our experiments suggest that the ASE as currently defined does not accurately recapitulate the expression of her1 in the anterior PSM. It remains possible that a more extensive deletion series through the 8.6kb of upstream sequence would define an element genuinely capable of driving precise oscillating expression in this region. However, it may also be true that the elements driving expression in the posterior and anterior PSM cannot actually be uncoupled. In this regard, it has not yet been shown that the 8.6kb to 3.3kb region, or any fragment of it, is sufficient for oscillating expression in the posterior PSM (although the whole region is clearly necessary). The activator and repressor binding sites required for oscillation may be spread throughout the full 8.6kb enhancer and might contribute to expression in both regions of the PSM. Resolving these issues will be key to determining the full significance of the binding sites that we have identified in this study. This is particularly evident in light of experiments demonstrating that temporary disruption of oscillating gene expression does not affect formation of somites by cells already in the anterior PSM (Giudicelli et al., 2007). However, it does affect cells in the posterior PSM, leading to delayed segmentation defects. Thus, the function of the segmentation clock is likely more dependent on proper regulation of oscillations in the posterior rather than the anterior PSM. Moreover, we do not yet know whether any of these cis-regulatory elements are shared with the her7 gene, which lies just 12kb away from her1, or if the two genes are regulated independently. Finally, our results suggest that the ASE is also responsible for suppressing expression of her1 in the somites.
In summary, we have performed a detailed characterization of the protein-DNA interactions that regulate expression of her1 in the anterior PSM. Further investigation of the her1 upstream sequence will reveal the full compliment of binding sites necessary for the complete her1 expression pattern throughout the PSM. Moreover, similar detailed analyses of the cis-regulatory regions of additional oscillating genes should lead to a comprehensive understanding of the regulatory logic that defines oscillating gene expression in the segmentation clock.
Constructs for protein synthesis were derived from the pAMB-CAT plasmid supplied with the ActivePro In Vitro Translation kit (Ambion). Initially, specific pairs of complementary, overlapping oligonucleotides, designed to encode standard epitope tags, were cloned into the pAMB vector, replacing the chloramphenicol acetyltransferase (CAT) gene. The genes encoding the proteins of interest were then cloned into the modified vectors to produce N-terminally tagged constructs. The truncated version of V5tbx24 was generated by digesting the full length construct with StuI and XhoI to remove the 3′ 1406 bp of coding sequence, blunting the ends and religating. Proteins were synthesized using the ActivePro In Vitro Translation kit (Ambion), according to the manufacturers instructions. Synthesis of proteins was confirmed by SDS-PAGE and western blotting. Epitope tags were detected with the following primary antibodies: anti-Flag M2 monoclonal (Sigma); anti c-myc 9E10 monoclonal (Covance); anti-V5 V5-10 monoclonal (Sigma). Blots were developed using an anti-mouse IgG peroxidase conjugated secondary antibody (Sigma) and ECL Plus Detection Reagents (GE Healthcare). Note: the ActivePro kit has been discontinued but we have successfully used our vectors with the Easy Xpress Protein Synthesis kit from Qiagen.
Complementary pairs of oligonucleotides were end-labeled by standard procedures using γ32P-dATP (370MBq/ml; MP Biomedicals) and T4 polynuceotide kinase (NEB). Labeled probes were purified using Mini Quick Spin Oligo columns (Roche). 20 μl binding reactions (see below for details of binding buffers) containing 1 μg poly[dI.dC] and 1-5 μl of synthesized protein (plus antibodies and/or competitor oligonucleotides as required) were incubated on ice for 20 minutes. 10 fmol of labeled probe was then added and reactions incubated for a further 20 minutes on ice. Reactions were run on 5% native polyacrylamide gels in 0.5x TBE at 200V at room temperature. Dried gels were exposed to Biomax MR film (Kodak) in the presence of a Biomax intensifying screen (Kodak). Her binding buffer (1x): 25 mM Tris-HCl (pH7.5), 200 mM KCl, 1 mM DTT, 10% glycerol. Tbx binding buffer (1x): 25 mM Tris-HCl (pH7.5), 100 mM KCl, 1 mM DTT, 10% glycerol. Su(H) binding buffer (1x): 10 mM Tris-HCl (pH8.0), 0.1 mM EDTA, 10 mM MgCl2, 2 mM DTT, 17.5% glycerol, 2 mg/ml BSA. The following oligonucleotides were used as probes or unlabelled competitors (the sequence of only one strand is shown for each complementary pair): Her control GGTGGCACGTGCCATT; Her control mutant GGTGGGTCGACCCATT; ASE-E1 GACGACACGTGCTCAC; ASE-E1 mutant GACGAGTCGACCTCAC; ASE-E2 GTCTCCACGTGCCCAT; ASE-E2 mutant GTCTCGTCGACCCCAT; Tbx control CCCCAATTTCACACCTGGGCTGCAG; Tbx control mutant CCCCAATTTCAGACCTGGGCTGAG; ASE-T1 TGTCTGATTCACACCTCTGCAACAG; ASE-T1 mutant TGTCTGATTCAGACCTCTGCAACAG; Her1-S1 TTATACTTTCCCACGGGATTTT; Her1-S1 mutant AACATTATACT—GGATTTTTCTT; Her1-S2 AATATTGTTCCCACGAGAAAAA; Her1-S2 mutant CCACAATATTG—AGAAAAATCCT; Hes1-SPS GGTTACTTGTGGGAAAGAAAGTTTGGGAAGTTTCACACGAGCCGTTCC; Hes1-SPS mutant GGTTACTGACGCTAAGAAAGTTTGGGAAGTTAGAGTCGAGCCGTTCC. The nucleotides altered in the mutant probes are underlined. For the S1 and S2 mutants, the position of the deleted binding site is indicated by a dash.
A custom DNA fragment was designed containing the rabbit β-globin 5′utr, the emGFP coding sequence (codon optimized for expression in zebrafish) and the SV40 polyadenylation signal, flanked by the minimal Tol2 transposon elements (Urasaki et al., 2006). This 2.1kb fragment was synthesized commercially (GeneArt) and cloned into a pBluescript backbone to create the pTol2übervector. An alternative version was created by cloning the mCherry gene in place of emGFP. PCR products representing the 3.3kb and 2.9kb enhancer regions of her1 were inserted upstream of the emGFP or mCherry genes, replacing the rabbit β-globin sequence. The her1 3′utr sequence was amplified by PCR and cloned into the required vectors, immediately downstream of fluorescent protein coding sequence and replacing the SV40polyA signal. Specific mutations of binding sites were generated by PCR. Additional protein binding sites were introduced by cloning specifically designed oligonucleotides into restriction sites immediately upstream of the 3.3kb enhancer sequence. The accuracy and integrity of all constructs was confirmed by sequencing.
Transposase mRNA was synthesized from NotI-linearised pCS-TP plasmid (a gift from Koichi Kawakami (Urasaki et al., 2006)) using the Sp6 mMESSAGEmMACHINE kit (Ambion). Plasmid DNA was purified using Qiagen Midiprep kits and injected into early one-cell stage embryos as circular DNA at 15 to 25 ng/μl, mixed with transposase mRNA at 25 ng/μl. Injected embryos were examined around 15hpf and healthy embryos exhibiting fluorescence were retained and raised to adulthood. Potentially transgenic fish were crossed to each other and their progeny screened for fluorescence. Positive embryos were collected and fixed at the 10-12 somite stage for analysis by in situ hybridization. Transgenic founder fish were isolated and assigned a specific line number. Founders were mated regularly and fluorescent embryos collected for in situ hybridization. During screening, clutches containing no fluorescent embryos were additionally analyzed by PCR. Extraction of genomic DNA was performed essentially as described (Gilmour et al., 2002), except that the volume of buffer was adjusted to the number of embryos in each sample (10 μl/embryo) and samples were not pooled for DNA precipitation. 200μl of each sample was precipitated and the pellets resuspended in 50 μl TE. 1 μl of genomic DNA was added to 20 μl PCR reactions using the Mastermix system (5 Prime). Specific primers were designed to amplify a 236 bp fragment of emGFP or a 275 bp fragment of mCherry (details below). Each reaction contained an additional pair of primers that amplify a 484 bp fragment of the keratin4 (krt4) promoter that serves as a positive control for the presence of genomic DNA. Cycling parameters were: 95°C 4min; 30 cycles of 95°C 30s, 59°C 30s, 72°C 30s; 72°C 5min. Primers: emGFPTg-F GTGAACGGACACAAGTTCAGC; emGFPTg-R AAGATGGTTCTCTCCTGCACG; mCherryTg-F GTTCATGCGCTTCAAGGTGC; mCherryTg-R CGTCCTCGAAGTTCATCACG; krt4-F ACACAAGACGGCACAAGACG; krt4-R GCATTTAGTTATTGCCCAGCAG. Fish identified as positive by PCR were carefully re-screened for transgene expression but only one line (of 42 total) was subsequently found to show weak fluorescence.
The mCherry sequence was cloned into pCS2+ and the resulting plasmid was linearized with BamHI for probe synthesis. The emGFP probe was synthesized from a purified 1kb linear fragment produced by SacII digestion of pCS2+itga5emGFP (provided by Dörthe Jülich). The her1 probe has been described previously (Holley et al., 2000). Probe synthesis and standard and fluorescent in situ hybridization were performed as previously described (Jülich et al., 2005; Brend and Holley, 2009). Images of fluorescent in situs were captured using a Zeiss LSM 510 confocal microscope.
We thank Andrew Collard for assistance in screening transgenic fish and Koichi Kawakami and Dörthe Jülich for providing plasmids. This work was supported by a grant from the National Institute of Child Health and Human Development (NIHCHD: R01 HD045738) to S.A.H.
Funded by: NICHD; Grant number: R01 HD045738