|Home | About | Journals | Submit | Contact Us | Français|
The spatial and temporal control of gene expression is key to generation of specific cellular fates during development. Studies of the transcriptional repressor REST/NRSF (RE1 Silencing Transcription Factor or Neural Restrictive Silencing Factor) have provided important insight into the role that epigenetic modifications play in differential gene expression. However, the precise function of REST during embryonic development is not well understood. We have discovered a novel interaction between zebrafish Rest and the Hedgehog (Hh) signaling pathway. We observed that Rest knockdown enhances or represses Hh signaling in a context-dependant manner. In wild-type embryos and embryos with elevated Hh signaling, Rest knockdown augments transcription of Hh target genes. Conversely, in contexts where Hh signaling is diminished, Rest knockdown has the opposite effect and Hh target gene expression is further attenuated. Epistatic analysis revealed that Rest interacts with the Hh pathway at a step downstream of Smo. Furthermore, we present evidence implicating the bifunctional, Hh signaling component Gli2a as key to the Rest modulation of the Hh response. The role of Rest as a regulator of Hh signaling has broad implications for many developmental contexts where REST and Hh signaling act.
The emergence of the diverse cell types that comprise the vertebrate nervous system is dependent on a carefully orchestrated program of gene expression. The transcriptional repressor REST (also known as NRSF) has been implicated as a major regulator of neural gene expression (Ballas et al., 2005; Lunyak et al., 2002; Westbrook et al., 2008). Rest is expressed in neuronal precursors and in vitro studies demonstrate that the degradation of REST protein is a key step in differentiation of neural progenitors in culture (Ballas et al., 2005; Westbrook et al., 2008). Recent studies have identified hundreds of REST target sites that potentially regulate an immense set of transcripts, including miRNAs (Ballas et al., 2005; Bruce et al., 2004; Conaco et al., 2006; Johnson et al., 2007; Mortazavi et al., 2006; Otto et al., 2007; Singh et al., 2008; Su et al., 2004; Watanabe et al., 2004).
However, only a fraction of putative RE1-regulated transcripts are upregulated when REST is degraded (Johnson et al., 2008). Many studies highlight the importance of developmental context on the activity of REST (Bergsland et al., 2006; Chen et al., 1998; Johnson et al., 2008; Jorgensen et al., 2009). For instance, while overexpression of REST blocks terminal differentiation in vitro (Ballas et al., 2001), mosaic expression of REST in chick did not prevent overt differentiation of neural precursors (Paquette et al., 2000). However, upregulation of terminal differentiation genes was observed upon inhibition of REST function in chick spinal cord (Chen et al., 1998). This upregulation was shown to require the presence of upstream activators of those genes (Bergsland et al., 2006). Most importantly, premature neurogenesis was not observed in the mouse knockout of Rest, and only one of several RE1 containing genes tested was inappropriately expressed (Chen et al., 1998). Rest −/− embryos are retarded in growth and do not survive past E11.5, precluding in-depth studies of REST function during maturation of the nervous system. However, early induction and patterning of the nervous system appeared normal in Rest knockouts. In contrast, interference with REST function in Xenopus embryos resulted in downregulation of some target genes, possibly due to early patterning defects attributed to involvement of REST in the BMP pathway (Olguin et al., 2006). The embryonic lethality of the mouse knockout demonstrates the necessity of REST, but a broader understanding of the requirement for Rest in regulation of specific developmental processes is lacking.
In this study, we demonstrate a novel role for zebrafish Rest in modulation of the Hedgehog (Hh) pathway. Hh signaling is involved in many aspects of development including regulation of cell type specification, neurogenesis, cell survival and proliferation (Briscoe and Novitch, 2008; Cayuso et al., 2006). In vertebrates, Sonic Hedgehog (Shh) has perhaps been best characterized as a morphogen that establishes dorsal-ventral patterning of the neural tube. Shh secreted from the ventral midline of the neural tube induces ventral cell fates in a dose dependent manner, generating distinct neural subtypes. The transcription factors expressed in response to the Hh gradient are categorized as class I genes (e.g. pax3. pax7, dbx1a) or class II genes (e.g. nkx2.2a, nkx6.1), which are repressed or induced, respectively, in response to Hh signaling (Briscoe and Novitch, 2008).
Members of the Gli family of transcription factors are key effectors of Hh signaling (Huangfu and Anderson, 2006; Jacob and Briscoe, 2003; Ruiz i Altaba, 1999; Stamataki et al., 2005). Like Drosophila Ci, vertebrate Gli2a and Gli3 are bifunctional and act as both activator and repressors of Hh target genes. In the absence of Hh signaling, protein kinase A (PKA)-dependent proteolytic cleavage produces a repressor protein (GliR), while activation of the Hh pathway allows full-length or near full-length Gli protein to function as an activator (GliA). Gli1, in contrast, lacks a repressor domain and is thought to function only as an activator (Dai et al., 1999; Ruiz i Altaba, 1999). In zebrafish, gli1 is transcriptionally regulated by Gli2a and Gli3, and is thought to amplify Hh signaling after the initial activation of Gli2a and Gli3 (Karlstrom et al., 2003; Tyurina et al., 2005). Although both Gli2a and Gli3 have early activator roles in zebrafish, they act chiefly as repressors during later stages as their expression becomes limited to cells outside the zones of strong Hh signaling. This downregulation of gli2a and gli3 is in part mediated by Hh signaling (Karlstrom et al., 2003; Tyurina et al., 2005). Recently, a second zebrafish Gli2 orthologue, Gli2b, which also functions in the nervous system, was identified (Ke et al., 2005; Ke et al., 2008).
Our in vivo studies demonstrate that Rest influences Hh signaling through regulation of Gli2a activity. We observed that when Rest levels are decreased, Hh signaling is enhanced and the response to ectopic Hh is elevated. Conversely, when Hh signaling is diminished, reduction of Rest levels leads to diminished expression of Hh target genes. Several lines of evidence support the hypothesis that this phenotype results from excess Gli2a activity. These include observations that gli2a expression is expanded in rest morphants and that disruption of Gli2a alters the consequences of Rest knockdown on Hh signaling. Regulation of gli2a transcription by Rest may be a wide-ranging mechanism to modulate the Hh response. These results reveal a novel requirement for Rest during zebrafish embryogenesis.
Adult zebrafish stocks were maintained at 28.5°C. Embryos were produced by natural matings, collected and stored at 28.5°C in embryo medium until desired stage according to Kimmel et al. (1995). Tg[islet:efp] are described by (Higashijima et al., 2000). The following mutant alleles were used in this study: gli1/dtrte370, gli2a/yotty17.
Embryos (10/tube for control, 15/tube for rest morphants) were collected at the appropriate stage and placed in TRIzol reagent (Invitrogen) for RNA extraction. cDNA was synthesized from .5–1 μg mRNA with the SuperScript First-Strand Synthesis System for RT-PCR (Invitrogen). Real-time PCR was performed and analyzed as previously described (Londin et al., 2005).
Full-length rest cDNA was isolated from 12hpf cDNA using the following primers: Forward TTTCAGTGGTCCAGCATGTC and Reverse ACATCTGACCCAGTTCGGTT. The PCR product was cloned into a pCS2+ vector, using the BD infusion method (BD Biosciences).
CLUSTAL W (Thompson et al., 1994) and BOXSHADE version 3.3.1 were used for protein sequence alignment.
Embryos were fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS) overnight at 4°C then stored in 100% methanol for storage at −20°C. In situ hybridizations were done as previously described (Thisse et al., 1993). Constructs used to synthesize the following probes have been described previously: nkx2.2a (Barth and Wilson, 1995), axial/foxA2 (Macdonald et al., 1997; Odenthal and Nusslein-Volhard, 1998), nkx6.1 and olig2 (Guner and Karlstrom, 2007), pax6a (Krauss et al., 1991) pax3a and pax7 (Seo et al., 1998), ptc1(Concordet et al., 1996), gli1 (Karlstrom et al., 2003), gli2a (Karlstrom et al., 1999), gli3 (Tyurina et al., 2005), phox2a (Guo et al., 1999), shh (Krauss et al., 1993)After in situ hybridization, embryos were mounted in 75% glycerol and photographed using a Zeiss Axiocam mounted on a Zeiss Axioplan microscope.
After whole mount RNA in situ hybridization, embryos were dehydrated in ethanol, infiltrated and embedded in JB-4 resin (Ted Pella) 10μM sections were obtained using a ultramicrotome (LKB 8800 ultratome III; Bromma)
Capped rest, shh (Krauss et al, 1993) and dnPKA (Ungar and Moon, 1996) mRNA was made using the mMESSAGE mMACHINE RNA synthesis kit (Ambion). One- to two-cell embryos were injected with 50–100pg (dnPKA, shh) or 500pg (rest) of mRNA diluted in 0.2 M KCl and phenol red. A splice inhibiting morpholino (MO) against the intron-exon boundary of zebrafish rest exon 3
(5′-GGCCTTTCACCTGTAAAATACAGAA-3′) and a translation blocking MO
(5′-AAACACCGGCTGAGACATGCTGGAC -3′) were synthesized by Gene Tools (Philomath, OR). Unless otherwise noted, the splice blocking MO was used for all described experiments. Prior to microinjections, embryos were dechorionated in 1 mg/ml pronase (Sigma-Aldrich). Morpholinos were diluted in .2M KCl and phenol red from a 34mg/ml stock to 8–10mg/ml. MO was injected at the one cell stage, using 4 ng for the shh mRNA combination experiments and 5 ng for all other experiments. Equivalent amounts of the standard control morpholinos provided by Gene Tools (5′-CCTCTTACCTCAGTTACAATTTATA-3′) were used in all experiments. For mRNA and MO combination injections, embryos were first injected with rest MO or control MO, then mRNA was injected into each of these. At the appropriate stage, embryos were fixed in 4% paraformaldehyde for in situ hybridization or placed in TRIzol reagent (Invitrogen) for RNA extraction.
Cyclopamine (CyA) (Calbiotech) was diluted in EM from a 10mM stock dissolved in DMSO. Embryos were incubated in the desired concentration of CyA media from 4hpf (Fig. 5) or at shield stage (Fig. 4) on, with control embryos in the equivalent concentration of DMSO in EM.
To examine the role of Rest in neurogenesis, we isolated a zebrafish rest cDNA using primers derived from the zebrafish genome sequence. The predicted zebrafish Rest protein (855 residues) is 39% identical and 54% similar to the human Rest protein (1097 residues). Like the mammalian REST proteins, the zebrafish protein is predicted to encode 9 zinc fingers (8 of which comprise the DNA binding domain) (Supplemental Fig. 1). In zinc fingers domains, the human and zebrafish proteins have a higher degree of similarity (89%) and identity (81%). This suggests the Rest homologues interact with similar DNA elements. Computer algorithms to identify RE1 sites predict over 1000 putative RE1 sites in the zebrafish genome (Mortazavi et al., 2006) and data not shown), which is comparable to the numbers of RE1 sites present in other vertebrate genomes (Bruce et al., 2004; Mortazavi et al., 2006). Zebrafish Rest is also highly similar to human REST within the domains that are key to interactions with the Sin3 and CoREST corepressor complexes (Nomura et al., 2005; Tapia-Ramirez et al., 1997). Phosphodegron motifs, which are required for post-translational regulation of REST levels, are also conserved (Guardavaccaro et al., 2008; Westbrook et al., 2008).
We have analyzed rest expression in zebrafish by whole mount RNA in situ hybridization. rest is present as a maternal transcript (data not shown), and is expressed ubiquitously until mid-somitigenesis (Fig 1A, B). rest expression then becomes increasingly confined to non-neural tissue and proliferative zones within the nervous system. For example, rest is expressed at low levels throughout the extent of the hindbrain at 16 hpf (Fig. 1D), a stage when cells are largely proliferating and few neurons have differentiated. Around 20hpf, (1E) rest transcript is downregulated in the ventrolateral domains where neuronal differentiation is occurring (arrowheads E–G).
During the accelerated period of neurogenesis from 24hpf to 48hpf (Lyons et al., 2003), rest persists in the undifferentiated dorsal rhombic lip (Fig 1G, K), and in proliferative midline (ventricular zone, vz) cells extending to the floorplate. Dorsoventral differences in expression likely reflect the pattern of differentiation in the neural tube, where maturation occurs ventrally before dorsally. This is most clearly demonstrated in the midbrain region during late day one development (Fig 1J). rest is expressed throughout the undifferentiated dorsal midbrain, (optic tectum, OT). Whereas the ventral midbrain (tegmentum, tg) has undergone extensive neurogenesis and rest transcript is largely restricted to the vz. rest is expressed in domains outside the neural tube, but is excluded from the mature somites (Fig 1K, L). In addition, rest is excluded from differentiated sensory cranial ganglia adjacent to the neural tube (Fig 1E, F, G, K asterisks).
The expression pattern of rest is largely coincident with proliferative and nonneural tissue, as seen in mouse (Chong et al., 1995)) and chick (Chen et al., 1998). However, the significance of rest transcript must be interpreted with caution. Rest levels are regulated posttranslationally (Guardavaccaro et al., 2008; Westbrook et al., 2008; Ballas et al., 2005). Therefore, the sites of rest expression may not equate with the sites of Rest activity. It is also likely that the availability of individual components of the co-repressor complexes will affect the repression mediated by Rest. Conversely, because the mechanism of Rest repression involves epigenetic changes to the chromatin environment of target genes, the effects of Rest activity may outlast the presence of Rest.
To investigate the role of Rest during development, we generated two independent rest morpholinos (MOs). The first targets the translation start site, while the second targets an intron-exon boundary to block mRNA splicing. While injection of both MOs led to similar defects, the splice blocking MO was more potent and was thus used for the following studies. The splice blocking morpholino binds to the intron-exon boundary of the third exon, and this is predicted to produce a truncated protein due to inclusion of the second intron in the mature mRNA (Supplemental Fig. 2A). The predicted protein produced from the misspliced transcript lacks zinc fingers 6–8 of the DNA binding domain and the C-terminal CoREST/HDAC interaction domain. Quantitative Real Time PCR was used to assay the amount of wild-type rest mRNA present following rest mo treatment (Supplemental Fig. 2B). By 6 hpf, only about 50% the amount of wild-type rest mRNA is present in rest morphants compared to controls. The amount of wild-type rest mRNA in rest mo treated embryos drops to less then 10% the control amount by 12 hpf. This demonstrates that the morpholino is effective in reducing the levels of wild-type rest transcript. However, the experiment also reveals that the morpholino does not entirely eliminate wild-type rest mRNA. In addition, the morpholino is less effective at disrupting mature mRNA production at early stages. This is presumably due to maternal rest mRNA, which is impervious to the splice blocking morpholino.
A second set of primers that detect the mis-spliced product containing intron sequences reveals that this product is significantly enriched in the rest morphant cDNA (Supplemental Fig. 2C). This provides additional proof that the predicted mis-splicing event occurred. Interestingly, wild-type cDNA also contains low levels of this product. This may result from trace amounts of immature, partially spliced mRNA in the wild-type sample. Alternatively, zebrafish may have a Rest splice variant akin to the Rest4 form that has been observed in mammals (Magin et al., 2002; Palm et al., 1998; Shimojo et al., 1999). However, we have not detected any activity of the truncated Rest form in overexpression assays (data not shown).
By gross morphological examination, rest morphants appear wild-type through 3 dpf, except that they tend to be smaller, and progress more slowly than control-injected embryos. Examination of neural tube markers by RNA in situ hybridization revealed that the neural tube of rest morphants is mildly ventralized (Fig. 2). Expression of the ventral marker, nkx2.2a, was expanded in rest morphants (Fig. 2D), while expression of the dorsal marker pax3a was reduced in hindbrain sections (Fig. 2B). nxk2.2a expression is induced by Hh signaling and pax3a expression is repressed by Hh (Guner et al., 2008; Liem et al., 1995). The polarity of the vertebrate neural tube is established by a gradient of Shh (Briscoe and Ericson, 1999), and we wondered if Rest knockdown enhanced expression of other Hh target genes within the developing neural tube. Therefore, we determined the expression of two direct Hh target genes ptc1 and foxA2 (Vokes et al., 2007) in Rest compromised embryos. As with nkx2.2a, the expression of foxA2 and ptc1 was stronger within their normal domain and expanded 1–2 cell widths laterally and dorsally in rest morphants (Fig. 2F, H).
To quantitatively examine the alterations in Hh target gene expression in these embryos, qPCR was performed on 29hpf control and stage-matched rest MO injected embryos. The levels of nkx2.2, foxA2 and ptc1 transcripts are enhanced in rest morphants (Fig. 2I). Conversely, the levels of pax3a transcript were slightly reduced, while a second class I gene, dbx1a/hlx1 (Fjose et al., 1994), which is expressed ventral to pax3a (Guner and Karlstrom, 2007; Hauptmann et al., 2002) was also reduced (Fig. 2I).
Regulation of Hh signaling by Rest is not limited to the neural tube, as Rest knockdown also enhanced expression of nkx2.2a in the developing pancreas (Supplemental Fig. 3) and foxA2 in the pharyngeal endoderm (Fig. 2H, arrows.) These results indicate the Rest function is involved not only in the proper dorso-ventral patterning of the neural tube, but in Hh signaling in general.
Enhancement of Hh target gene expression in the neural tube was strongest near the floor plate, which is a source of Hh. This suggested that rest morphants show increased senesitivity to high levels of Hedgehog. To determine whether Rest knockdown augmented the response to Hh, we examined the expression of markers in wild-type embryos and rest morphants injected with shh mRNA. Ectopic activation of Hh signaling causes a dorsal expansion of Hedgehog targets ptc1 and nkx2.2a, while the domain of the dorsal class 1 marker, pax3a is reduced (Dessaud et al., 2008; Guner and Karlstrom, 2007) and Fig. 3). Co-injection of rest MO and shh mRNA enhanced the ventralizing effects of Shh on the neural tube in a synergistic manner (Fig. 3, Table 1A). The expression domains of ptc1 and nkx2.2a were expanded compared to shh treatment alone, while pax3a expression was further suppressed. Other patterning genes in the neural tube that showed a synergistic response to Rest knockdown in conjunction with Shh activation include shh, foxA2, phox2a, nkx6.1, olig2, pax6a and pax7 (Supplemental Fig. 4). The increased domains of patterning genes such as nkx6.1 and phox2a are the likely cause of excess production of Islet:GFP positive cranial motor neurons seen in later staged embryos (Compare Figs. S4W and S4X). We conclude that Rest is required to limit Hh signaling and that Rest likely modulates a fundamental aspect of the Hh pathway.
Importantly, a similar enhancement of the effects of shh mRNA overexpression was observed with the second rest morpholino which blocks translation (rest ATG) (Supplemental Fig. 5A–D and Table 1B). To confirm that the effects of the rest morpholino treatment result from Rest knockdown, we assayed the ability of rest mRNA to rescue the effects on Shh signaling produced by the splice morpholino. rest mRNA was co-injected into rest morphants treated with shh mRNA and RNA in situ hybridization was used to assay pax3a expression. rest mRNA microinjection largely restored the pax3a expression (Supplemental Fig. 5H) compared to rest MO/shh injected embryos (Supplemental Fig. 5G, and Table 2). These experiments demonstrate that the effects of the morpholino treatment are produced by knockdown of Rest.
Our initial findings suggest that Rest is a negative regulator of the Hh pathway. We therefore treated wild-type embryos with low doses of the Hh antagonist cyclopamine (CyA) (Chen et al., 2002; Hirsinger et al., 2004; Wolff et al., 2003) to determine whether blockage of Hh signaling could be alleviated by Rest knockdown. Wild-type embryos were microinjected with control or rest morpholino and incubated in a low concentration of CyA (1.5μM) or control media from shield stage (6 hpf) onward. Treated embryos were fixed at 24 hpf and stained for the class II gene nkx2.2a (A–D) As expected, rest MO treatment led to an increase in nkx2.2a expression (Fig. 4B). Surprisingly, in the presence of CyA, Rest knockdown caused further reduction in Hh signaling as revealed by diminished nkx2.2a expression (Figure 4D). At this concentration of CyA, most control embryos showed a mild (33%, n=75) or moderate (47%, as in Fig. 4C) reduction of nkx2.2a expression in hindbrain and spinal cord, while retaining nkx2.2a expression in the basal forebrain and diencephalon. The remaining embryos (20%) had a severe reduction overall in nxk2.2a expression. rest morphants were more severely affected compared to control embryos treated with the same dose of CyA. Most morphants displayed a total absence of hindbrain expression and a strong loss in the head region (74% n=78, as in Fig. 4D, compared to 20% of control embryos). The remaining rest morphants showed a moderate reduction of nkx2.2a expression.
Since cyclopamine treatment also results in the upregulation of genes negatively regulated by Hh signaling (Hammond et al., 2007; Martin et al., 2007), embryos were also assayed for ventral expansion of class I gene pax3a. This concentration of CyA resulted in a mild ventral expansion of pax3a in control embryos (6/16 as in 4G, 10/16 comparable to wild-type) despite downregulation of nkx2.2a (compare sibling control embryos, in 4C, G). In contrast, rest morphants often had significant expansion of pax3a (8/18 as in 4H, the remainder were similar to control in 4G) Thus, Rest knockdown increases response to high levels of Hh signaling (Fig. 3), but further attenuated Hh signaling in the context of low levels of Hh signaling (Fig. 4). We conclude that regulation by Rest positively or negatively influences the Hh response depending on the state of Hh signaling.
RE1 sites are associated with hundreds of genes and many direct or indirect interactions with the Hh pathway can be postulated to account for the bimodal phenotype observed. To determine whether Rest interacts with the Hh pathway by modulating an intracellular signaling step in the cascade, we activated the Hh pathway cytoplasmically by overexpressing a dominant negative form of PKA (dnPKA)(Ungar and Moon, 1996). PKA has multiple roles, and in context of the Hh pathway, PKA activity is required to generate the repressor forms of Gli2a and Gli3. Overexpression of dnPKA strongly activates Hh target gene expression, presumably because Gli activator forms predominate in that condition (Hammerschmidt et al., 1996).
As expected, injection of dnPKA mRNA resulted in a dorsal expansion of nkx2.2a expression (Fig. 5C). As with shh overexpression, upregulation of nkx2.2a by dnPKA was further enhanced by simultaneous co-injection of rest MO (Fig. 5D). While this result suggests that Rest knockdown acts intracellularly on the Hh pathway, Hh signaling also results in transcription of Hh ligands and amplification of the signal (Blader et al., 1997; Neumann and Nuesslein-Volhard, 2000)Supplemental Figure 4B, C). It is therefore possible that some of the effects on Hh signaling of Rest knockdown is a result of heightened extracellular Hh ligand activity due to ligand production, processing or diffusion. To address this possibility we injected dnPKA mRNA to activate the pathway intracellularly, but subsequently blocked extracellular Hh signaling at the level of the transmembrane receptor Smo with a high dose of CyA (50 μM). If Rest regulates Hh signaling components downstream of Smo, then CyA treatment will not alter the enhancement of the effects of dnPKA overexpression resulting from Rest knockdown (as in Fig. 5D). CyA treatment largely eliminated nkx2.2a expression in embryos injected with control MO (Fig. 5B) or rest MO (Fig. 5D). Importantly, Rest knockdown enhanced the effects of dnPKA mRNA treatment on nkx2.2a expression in the presence of 50 μM CyA (compare Fig. 5H with 5D). Thus, blocking extracellular signaling events did not significantly alter the enhanced effects of dnPKA on Hh signaling produced by Rest knockdown. We conclude that Rest interacts with the Hh pathway at an intracellular step downstream of Smo.
Intracellular pathway components include the mediators of Hh signaling, the Gli family of transcription factors. To determine whether Rest regulates gli transcription, we examined expression of the four zebrafish gli genes (gli1, gli2a, gli2b and gli3) in rest morphants. gli1 is expressed in regions where Hh signaling is active and expression is enhanced in rest morphants (Fig. 6B). The stronger expression of gli1 in rest morphants is consistent with enhanced Hh signaling because gli1, like ptc1, is a Hh response gene. Gli1 is not thought to have repressor activity (Dai et al., 1999; Ruiz i Altaba, 1999), so upregulation of gli1 is unlikely to produce the negative effects we observed on Hh signaling in CyA treated embryos (Fig 4D, H).
Both gli2a and gli3 are initially expressed in ventral regions of the CNS and act as weak activators of the initial Hh response. As development proceeds, gli2a and gli3 are repressed by Hh signaling. Expression of both transcripts becomes restricted to regions distant from the ventral source of Hh (Karlstrom et al., 2003; Tyurina et al., 2005; Vanderlaan et al., 2005, and Figure 5). At these stages, both proteins function as repressors of the Hh response. We did not detect significant alterations of gli3 mRNA in rest morphants from 18–24hpf (Fig. 6D). gli2b expression was also unaltered by Rest knockdown (data not shown).
In contrast, gli2a expression was enhanced in rest morphants. In the hindbrain of wild-type embryos, gli2a is initially expressed across the dorso-ventral extent of the neural tube before 15hpf (Vanderlaan et al., 2005) and then is progressively reduced in the ventral and midline regions as development proceeds (Fig. 6E, G). In rest morphants, gli2a fails to be properly downregulated at 18hpf, and remained in the ventral CNS at 24 hpf (Fig. 6F, H). Thus, Rest knockdown results in inappropriate expression of gli2a in ventral domains. In these regions, cells are exposed to higher levels of Hh ligand and Gli2aA is presumably generated. gli2a expression is repressed by Hh (Supplemental fig. 6C), so the expansion of the gli2a domain is unusual among the Hh targets we have observed in rest morphants. In the experiments described earlier (Fig. 3 and Supplemental Fig. 6D′), Rest knockdown enhanced Hh signaling when ectopic shh mRNA was applied. However, double injection of rest MO with shh mRNA did not enhance the Hh-mediated repression of gli2a expression (Supplemental Fig. 6D). We conclude that Rest plays a role in repressing gli2a expression, and that this repression is independent of effects resulting as a consequence of enhanced Hh signaling.
These experiments do not establish whether regulation of gli2a by Rest is direct or indirect. Using an algorithm optimized to identify mammalian RE1 sites (Otto et al., 2007), we did not detect any canonical RE1 sites in the zebrafish gli2a locus, which is large and encompasses over 100kb. Recently, two groups have identified RE1 site variants termed split-RE1 sites (Johnson et al., 2007; Otto et al., 2007). The error rate for predicting split-RE1 sites is higher. We identified three split RE1-like sequences in the gli2a locus. One site is positioned at ~15kb upstream and two at about 80kb from the transcription start site. Functional data suggests that optimal repression from RE1s occurs when the site is within 2–3kb of the transcriptional start site (Johnson et al., 2008). Therefore, these sites are not strong candidates to mediate repression of gli2a by Rest. Rest may indirectly bind DNA and in addition, Rest appears to interact with sites that lack characterized RE1 motifs (Johnson et al., 2008).
The failure to downregulate gli2a transcription (Fig. 6F, H) combined with the context dependent response of rest morphants to manipulations of Hh signaling (Fig. 3, Fig. 4) led us to hypothesize that Gli2a activity was central to the interaction between Rest and the Hh pathway. To further investigate the role of Gli2a to the Rest-Hh interaction, we assayed the consequences of blocking Rest activity in yot/gli2a mutants. This mutation is a lesion in gli2a that results in the production of a truncated protein and acts as a dominant repressor of Hh signaling (Gli2aDR), most likely through interference of Gli activator function (Karlstrom et al., 1999; Karlstrom et al., 2003) While most nkx2.2a expression is lost in yot mutants, domains remain in the ventral forebrain and hindbrain. If enhanced Hh signaling in rest morphants is primarily dependent on Gli2aA, then restoration of nkx2.2a expression will not occur in yot mutants treated with rest mo. Alternatively, if the rest morphant phenotype was due to excess activity by Glis other than Gli2a, Rest knockdown in yot mutants might have either no effect, or restore Hh signaling by compensation or competition.
As expected, loss of Rest function led to an increase in ptc1 and nkx2.2a expression in wild-type siblings (Fig. 7B, F). However, in yot mutants, which are readily identified by characteristic loss of Hh target gene expression, Rest knockdown did not alleviate, but instead further reduced expression of ptc1 and nkx2.2a in the midbrain region (Fig. 7D, 7H arrows). This result suggests that Rest knockdown increases GliR activity in yot embryos, which is opposite to the effect of Rest knockdown in wild-type embryos. Because yot mutants only generate Gli2aDR and not Gli2aA, this finding is consistent with the model that misregulation of Gli2a accounts for the effects of Rest knockdown on the Hh pathway.
The yot/Gli2DR results above led us to hypothesize that excess Gli2a activity produced the observed phenotypes. We therefore sought to test the ability of Rest knockdown to alleviate defective Hh signaling due to loss of Gli1 function. Detour (dtr) mutants lack functional Gli1 and exhibit a loss of Hh target expression, including a complete loss of nkx2.2a in the hindbrain (Karlstrom et al., 2003). While Gli1 is the main activator of Hh target genes in zebrafish, Gli2a is a weak activator of Hh targets, and is partially redundant with Gli1 (Karlstrom et al., 2003; Park et al., 2000). Embryos from a dtr +/− intercross were injected with rest MO and assayed for nkx2.2a expression at 24hpf. rest MO injected dtr mutants resemble dtr mutants (reduced or absent nkx2.2a expression) throughout most of the embryo except, strikingly, in the hindbrain where some nxk2.2a expression is restored (23/33 in dtr−/− embryos, Fig. 8B). nkx2.2a expression was not detected in 26/26 in control injected dtr −/− embryos. Restoration of nkx2.2a expression in rest morphants indicates that while Gli1 function is required for the expansion of nkx2.2a in most tissues in rest morphants, there is partial compensation for loss of Gli1 in the hindbrain. This is significant because dtr mutants are largely refractory to exogenous shh mRNA (Fig. 8C, Karlstrom et al, 2003). However, injection of shh mRNA and rest MO into dtr mutants resulted in ectopic nkx2.2a in the dorsal hindbrain (6/16 dtr−/− embryos as in Fig. 8D) or partial restoration of ventral nkx2.2a expression (14/16 dtr −/− embryos) in the hindbrain. These results demonstrate that while the enhancement of Hh signaling in rest morphants largely depends on Gli1 function (possibly in response to Gli2aA), Rest also regulates Gli1-independent activities.
This is consistent with the hypothesis that derepression of gli2a accounts for the biphasic alterations of Hh signaling in rest morphants. In that case, simultaneous knockdown of Gli2a would negate the observed effects of Rest knockdown in dtr/gli1 mutants. To determine if the restoration of nkx2.2a in the hindbrain of dtr mutants treated with rest morpholinos was due to excess Gli2aA activity, rest MO and gli2a MO were injected into a dtr intercross. While dtr mutants injected with rest MO displayed nkx2.2a expression in the hindbrain (9/11 in dtr−/− embryos, as in Fig. 7B), this expression was suppressed by Gli2a knockdown (10/15 absent, (p= .0000011237) 5/15 reduced in dtr−/− embryos, Fig. 7F). Therefore, we conclude that restoration of Hh target gene expression in dtr by Rest knockdown requires Gli2a.
Rest plays a central role in regulation of gene expression required for cell proliferation and differentiation in a variety of contexts. Recent analysis of Rest function has led to proposals of a number of novel and seemingly contradictory roles for Rest. How the myriad of potential interactions mediated by Rest is translated into biologically relevant outcomes is poorly understood. We provide in vivo genetic evidence for an essential function for zebrafish Rest in regulation of the Hh pathway.
Rest has been best characterized as a transcriptional repressor, and we hypothesized that Rest directly or indirectly represses one or more Hh signaling components. Several lines of evidence implicate the Gli transcription factors, particularly Gli2a, as the key Rest target. First, Rest knockdown enhanced Hh target gene expression in multiple tissues implying that Rest regulates a fundamental aspect of the Hh pathway. Second, the effects of both positive and negative alterations in Hh signaling were enhanced with Rest loss of function. This is consistent with excess Gli activator function in contexts of strong Hh signaling, and excess Gli repressor activity when Hh signaling is reduced. Thus, excess Gli activity could account for the exaggerated response to both high and low levels of Hh signaling. In addition, our epistatic experiments reveal that Rest interacts with the Hh pathway downstream of Smo and upstream of Gli1 with the exception of the hindbrain, where Gli1-independent activities are present. Gli2a and Gli3 transduce Hh signaling chiefly through transcriptional regulation of gli1, but are also weak activators of other Hh targets (Fig. 9). We cannot rule out the possibility that Gli1 and/or Gli3 activity is also increased, independently of excess Gli2a activity.
The expansion of the gli2a domain into regions of active Hh signaling in rest morphants provides a mechanism to account for the effects on Hh signaling by Rest. The expression pattern of gli2a in rest morphants runs counter to alterations in other Hh pathway genes tested, including gli3 and gli2b, because enhanced Hh signaling normally results in decreased gli2a expression.
In addition, derepression of gli2a also accounts for enhanced repression of Hh targets by Rest knockdown in yot mutants, in which only the dominant repressor form of Gli2a is produced. Finally, restoration of Hh signaling in the hindbrain of dtr/gli1 mutants with compromised Rest function is blocked by Gli2a knockdown. This demonstrates a requirement for Gli2aA in Rest mediated enhancement of Hh signaling. Together, our results support a model in which Gli2a is the principal point of Rest interaction within the Hh pathway (Fig. 9). Following ectopic activation of the Hh pathway (shh or dnPKA mRNA treatment), conversion of Gli2a to Gli2aA predominates resulting in synergistic enhancement of Hh target gene expression. In CyA treated rest morphants, the excess Gli2a would be converted to Gli2aR, which would repress Hh target gene expression.
Ectopic gli2a expression is most evident in the midline and ventral domains in rest morphants (Fig. 6H), and we were not able to discern any consistent differences from control embryos in the dorsolateral domains. In a wild-type embryo, this repression may serve to dampen the response to high levels of signaling. While Rest function in the developing embryo tempers the cellular response to Hh, the modest alteration in target gene expression seen in rest morphants may be due to redundant regulation of Hh signaling (Dessaud et al., 2007; Jeong and McMahon, 2005). Subtle phenotypes in zebrafish arise from mutations in negative regulators of the pathway including ptc2, sufu and hip (Koudijs et al., 2008; Koudijs et al., 2005).
Although the absence of strong RE1 sites near or within the gli2a locus does not rule out direct regulation by Rest, the transcriptional upregulation of gli2a in response to reduced levels of Rest may indeed be indirect. In addition to Hh, other signaling pathways including the FGF, Notch and Wnt pathways also regulate Glis (Alvarez-Medina et al., 2008; Brewster et al., 2000; Ke et al., 2005). A model in which Rest knockdown activates one of these pathways, which in turn enhances transcription of one or more gli genes is consistent with our observations.
Here, we show that reduced levels of Rest during zebrafish development leads to alterations in the progenitor domains responsible for generation of distinct neural subtypes. In addition to a role for REST in repressing neural genes in non-neural cells, REST has been implicated in the control of neurogenesis at multiple steps (Ballas et al., 2005; Bergsland et al., 2006; Otto et al., 2007; Su et al., 2004). Outside the developing nervous system, Rest has been placed upstream of the network controlling pancreatic islet development (Johnson et al., 2007; Kemp et al., 2003). This is an interesting finding as the Hh pathway regulates both neural and pancreatic development. We also find that Rest knockdown enhances expression of nkx2.2a in the developing zebrafish pancreas (Fig. 7, Supplemental Fig. 3). The wide range of potential activities proposed for REST underscores the importance of considering the unique cellular environment in which REST is acting.
REST mutant mice undergo widespread apoptosis beginning at day E 9.0 and die by day E11.5 (Chen et al., 1998). In contrast, we observed that the rest MO treated zebrafish present with a much subtler phenotype, ventralization of the neural tube. The expansion of ventral cell types in rest morphants, is unlikely to be produced by increased apoptosis. However, there are key differences in the mouse and fish experiments. Primarily, our treatments produce a knockdown, not a knock-out of Rest. In addition, zebrafish rest is supplied as a maternal transcript, which may allow for adequate Rest activity during early stages.
REST has been identified as both a tumor suppressor (Coulson et al., 2000; Westbrook et al., 2008; Westbrook et al., 2005) and an oncogene (Lawinger et al., 2000; Su et al., 2006). It will be important to determine whether REST regulates Shh signaling in transformed cells. It is perplexing that REST downregulation results in differentiation in some cell populations and proliferation in others. For example, β-TRCP dependent degradation of REST allows differentiation in neural stem cell culture, but proliferation in human mammary epithelial cell culture (Westbrook et al., 2008). How loss of REST allows activation of such different pathways is not well understood, but it is clear that differential target regulation depends on cellular context. For example, in neural progenitors, cell cycle progression relies on degradation of REST during the G2 phase for optimal expression of mad2, a direct REST target (Guardavaccaro et al., 2008). Needless to say, the full repertoire of genes under the control of REST is not activated during the G2 phase of mitotic cells. Differential target regulation may depend on many factors, including preexisting epigenetic modifications and the unique combinations of co-repressors and/or transcriptional co-regulators present. Unique combinations of such factors determine and are determined by the unique cellular context, allowing a small number of signaling pathways to affect a wide array of transcriptional networks and produce diverse outcomes.
Our studies reveal a novel and unexpected interaction between Rest and the Hh pathway. We find that Rest acts as a modulator of the Hh signal by regulation of gli2a. Rest likely fine-tunes the response of cells to Hh signaling by controlling transcription levels of gli2a and possibly additional factors. Transcriptional repression by Rest may thus be an additional limiting factor for Hh signal transduction, independent of antagonistic pathway components such as Ptc and Hip. Hh signaling plays a key role in regulation of progenitor cell proliferation and differentiation in many places within the developing nervous system and other tissues. Regulation of Hh signaling by Rest may be critical in many of these domains. These findings have broad implications for regulation of signaling in the many places where Hh acts and provide an avenue for future studies into Hh-mediated cell fate decisions.
We thank Gail Mandel, Gil Levkowitz, Bernadette Holdener, Nurit Ballas and Fatma Kok for helpful suggestions and comments; Richard Grady and Andrew Taibi for Fish care; Sue Van Horn for help with histology. Special thanks to John Hover and Sean R. McCorkle for bioinformatics analysis of the zebrafish genome. We are also grateful to the many labs that provided reagents including fish stocks, constructs and probes. This work was supported by NIH grant RO1HD043998 (HS).
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.