The isthmic organiser located at the midbrain-hindbrain boundary (MHB) is the crucial developmental signalling centre responsible for patterning mesencephalic and metencephalic regions of the vertebrate brain. Formation and maintenance of the MHB is characterised by a hierarchical program of gene expression initiated by fibroblast growth factor 8 (Fgf8), coupled with cellular morphogenesis, culminating in the formation of the tectal-isthmo-cerebellar structures. Here, we show in zebrafish that one orthologue of the transcription factor grainy head-like 2 (Grhl2), zebrafish grhl2b plays a central role in both MHB maintenance and folding by regulating two distinct, non-linear pathways. Loss of grhl2b expression induces neural apoptosis and extinction of MHB markers, which are rescued by re-expression of engrailed 2a (eng2a), an evolutionarily conserved target of the Grhl family. Co-injection of sub-phenotypic doses of grhl2b and eng2a morpholinos reproduces the apoptosis and MHB marker loss, but fails to substantially disrupt formation of the isthmic constriction. By contrast, a novel direct grhl2b target, spec1, identified by phylogenetic analysis and confirmed by ChIP, functionally cooperates with grhl2b to induce MHB morphogenesis, but plays no role in apoptosis or maintenance of MHB markers. Collectively, these data show that MHB maintenance and morphogenesis are dissociable events regulated by grhl2b through diverse transcriptional targets.
The grainy head-like (Grhl) family of transcription factors are important regulators of neural development in diverse species. In Drosophila, grainy head (grh) is a key determinant of neuroblast specification, proliferation and apoptosis (Bray and Kafatos, 1991; Uv et al., 1997; Almeida and Bray, 2005; Cenci and Gould, 2005; Maurange et al., 2008), whereas, in mammals, the Grhl homologues functionally demarcate distinct regions of neural tube closure, failure of which leads to neural tube defects (NTD) (Ting et al., 2003; Rifat et al., 2010). Mice lacking the Grhl3 gene survive to birth and exhibit thoraco-lumbo-sacral spina bifida, and, rarely, exencephaly (Ting et al., 2003). By contrast, Grhl2-null mice die at embryonic day (E) 11.5, and display severe cranial neurulation defects, including a fully-penetrant ‘split-face’ malformation associated with cranioschisis, and also an open posterior neuropore (Rifat et al., 2010). The molecular bases for the profound defects in the Grhl2-null embryos are yet to be elucidated.
Clues to the roles of the mammalian Grhl members in both neural and non-neural developmental processes have emerged from studies of grh function in Drosophila embryogenesis. Of the known homologs of established grh target genes that have been implicated in vertebrate neural development, engrailed1 and engrailed2 (En1, En2), homologs of the segment polarity gene engrailed, represent potential candidate downstream effectors of the Grhl2 gene (Attardi and Tjian, 1993). The vertebrate engrailed genes are important for the molecular and cellular control of mesencephalon and metencephalon development (Joyner et al., 1991; Wurst et al., 1994; Joyner, 1996), contributing to both the establishment and maintenance of the midbrain/hindbrain boundary (MHB) (Hidalgo-Sanchez et al., 2005; Sato et al., 2004). The biochemical pathways mediating this are highly conserved phylogenetically, with the Drosophila engrailed gene capable of substituting for mouse En1 function in the mid-hindbrain (Hanks et al., 1998). The DNA consensus binding site of grh and the mammalian Grhl factors (AACCGGTT) is absolutely conserved across 700 million years of evolution (Wilanowski et al., 2008; Caddy et al., 2010), and we have previously shown that En1 is a putative mammalian Grhl target (Wilanowski et al., 2002). To explore the role of the Grhl2 gene in vertebrate neurulation, and to determine the functional relationship between Grhl2 and the engrailed genes, we elected to use the genetically tractable zebrafish model, with the aim of defining conserved developmental pathways.
MATERIALS AND METHODS
Cloning of zebrafish grhl2b
Zebrafish grhl2b was cloned from 18-24 hpf wild-type AB-strain fish cDNA by PCR amplification using Phusion polymerase (Finnzymes, Espoo, Finland), to minimise the chance of capricious mutations. Full-length PCR products were cloned into pCRII-TOPO (Invitrogen, Carlsbad, CA), and confirmed by sequencing. Verified coding DNAs were sub-cloned into pCS2+ (Turner and Weintraub, 1994) to generate the vector pCS2+grhl2b, and used for subsequent transcription of CpG-capped mRNA using the mMESSAGE mMACHINE kit (Ambion, Austin, TX).
Morpholino oligonucleotides (MO) and zebrafish micro-injection
Zebrafish embryos were micro-injected at the one- to two-cell stage with a bolus of approximately 2 nl. Dose-titration experiments were performed for all mRNAs/MO to determine optimum experimental dosages (supplementary material Table S1; data not shown). Murine Grhl2 and zebrafish eng2a mRNAs were injected at a concentration of 75-125 pg/bolus, whereas MO for grhl2b were routinely injected at concentrations of approximately 200 mM (ATG-blocking) and 250 mM (splice blocking and mismatch control). MOs targeting spec1 were both injected at a concentration of 250 mM. The ATG-blocking MO targeting grhl2a was injected at a concentration of 200 mM. The sequences for ATG and splice blocking and control MO for grhl2b, grhl2a and spec1 are listed in supplementary material Table S3. The sequences and doses of the ATG-blocking MOs targeting eng2a or fgf8 were as previously published (Araki and Brand, 2001; Scholpp and Brand, 2001). The sub-phenotypic doses of the grhl2b, eng2a and fgf8 MO were 100 mM, and the spec1 dose was 125 mM. Zebrafish protocols were pre-approved by the Walter and Eliza Hall Institute Animal Ethics Committee.
In situ hybridisation and imaging
Two independent grhl2b probes were cloned from 24 hpf AB-strain zebrafish [corresponding to nucleotides 219-759 and 1329-1779 of grhl2b (NM_001083072)]. Both probes were verified by sequencing before use and showed an identical expression pattern. All in situ hybridisation and imaging procedures were conducted as reported previously (Dworkin et al., 2007).
Reporter gene assays
A 3 kb fragment of the zebrafish eng2a promoter corresponding to positions –3029 to –29 relative to the transcriptional start site was amplified by PCR from wild-type AB 24 hpf genomic DNA and cloned into the pGL3basic-Luciferase vector to generate the vector pGL3eng2aPROM. Regulation of the eng2a promoter was assessed in vivo using the dual Luciferase reporter assay (Promega, Fitschburg, WI), following methods previously reported (Alcaraz-Perez et al., 2008).
Generation of MO-resistant grhl2b
The 5′ region of pCS2+grhl2b mRNA was amplified by PCR to generate a 7 bp mutation within the MO-recognition site without altering the amino acids synthesised. Making use of a 5′EcoRI site introduced into the product by the forward primer, and an existing AleI site towards the 3′ end of the amplicon, the PCR product was digested with EcoRI and AleI, and cloned into similarly digested pCS2+grhl2b, resulting in a change from ATGTCACAGGAGACAGACAGTAAG to ATGTCtCAaGAaACtGAtAGcAAa. The presence of the new 7 bp mismatch MO-binding region was verified by sequencing.
Cell death analysis assays
TUNEL staining to detect apoptosis was performed using the TMR Red In Situ Cell Death Detection Kit (Roche, Basel, Switzerland) according to the manufacturer’s instructions. Embryos were imaged by fluorescence microscopy on an Olympus FV1000 single-photon confocal microscope.
Laser capture microdissection and mRNA amplification
Paraformaldehyde-fixed embryos (11-12 hpf) were sequentially embedded in agarose and paraffin and 8 mm sections were cut onto PEN membrane slides. The sections were stained with the Histogene laser capture microdissection staining kit immediately prior to laser capture microdissection. The MHB region from four embryos was microdissected using a Veritas microdissection instrument (Arcturus) according to standard protocols and the tissues were captured onto Capsure HS laser capture microdissection caps. RNA extraction and amplification were performed according to the manufacturers’ instructions (Arcturus).
All Q-RT-PCR analysis was performed using GoTaq SyBR Green mastermix (Promega) and a Roche LC480 LightCycler (Roche, Basel, CH). Data were collated from a minimum of three samples per group, and expression levels were normalised to the gene encoding the housekeeping protein β-actin.
Generation of c-terminal FLAG fusion to grhl2b
The 3′ coding sequence of pCS2+grhl2b was mutagenised using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA), resulting in a single minor amino acid change of isoleucine to leucine at codon 600 (immediately prior to the terminal stop codon), thereby introducing a SacI site (GAAATA to GAGCTC). Complementary oligonucleotides consisting of sequence coding for a 5′ SacI site-FLAG-tag-stop codon-3′ XhoI site were generated and annealed. Making use of the novel SacI site in pCS2+grhl2b and an existing XhoI site at position 2147, SacI/XhoI double digestion facilitated cloning of the FLAG-tag into pCS2+grhl2b to generate the vector pCS2+grhl2b-FLAG. Following sequencing to confirm in-frame fusion of the FLAG-tag, pCS2+grhl2b-FLAG was transfected into HEK-293T cells and FLAG-tagged grhl2b protein was detected by western blotting using monoclonal anti-FLAG M2 antibody (Sigma-Aldrich, St Louis, MO) and standard methods.
Chromatin immunoprecipitation (ChIP) assay
ChIP assays were performed using the ‘microChIP’ protocol (Dahl and Collas, 2008), optimised for low starting cell number. Chromatin was derived from 50,000-100,000 cells, extracted from five to eight individual (de-yolked) zebrafish embryos per group. grhl2b-FLAG was immunoprecipitated using the anti-FLAG M2 antibody; negative controls were performed using anti-mouse IgG antibody (Sigma-Aldrich). As an additional specificity control, ChIP assays using anti-FLAG antibody were performed on zebrafish overexpressing non-FLAG-tagged grhl2b mRNA. Specific oligonucleotides flanking the predicted grhl2b-binding sites in the promoters of both eng2a and spec1 were used to amplify the enriched loci; primers ∼3 kb upstream of the predicted promoter binding sites and also ∼1 kb downstream of the transcription start site (within the coding regions) for both eng2a and spec1 were used as negative controls for binding.
Characterisation of the zebrafish grhl2 family
Previous work had identified two orthologues of mammalian Grhl2 in zebrafish – termed zgrhl2a and zgrhl2b, respectively (Janicke et al., 2010). Our analyses suggest that, based on homology with mammalian Grhl2, a previously deposited sequence (termed ‘Danio rerio hypothetical LOC558268, transcript variant 2’, GenBank Accession Number XM_701793) was more likely to be the correct full-length isoform of zgrhl2b. This sequence included 47 additional amino acids at the N terminus, but was otherwise identical with the sequence deposited as ‘zgrhl2b’ (GenBank Accession Number NM_001083072). Of these additional 47 amino acids, 36 were identical to both human and mouse Grhl2, and 34/47 nucleotides were also conserved in zgrhl2a. These homology analyses are shown in supplementary material Fig. S1A,B. Further confirmation of this N-terminal exon has been recently provided by others using 5′ RACE (Han et al., 2011). As grhl2b exhibits a higher degree of homology to mammalian Grhl2, our initial studies focused on characterising this orthologue.
Loss of grhl2b induces MHB and otic vesicle defects
The early expression patterns of grhl2b in the developing fish are shown in supplementary material Fig. S1C-G. In our initial functional experiments, an ATG-directed antisense MO that blocked grhl2b mRNA translation was used to examine the developmental consequences of grhl2b loss of function. Morphants lacking grhl2b developed normally until ∼21 hours post fertilisation (hpf), at which time the MHB region failed to undergo its characteristic horseshoe-shaped folding. By 24 hpf, this phenotype became more pronounced, and was accompanied by defective ventricle formation, and marked neural cell death (arrowed) in the midbrain and hindbrain regions (Fig. 1A,B, inset). The neural cell death was shown to be apoptotic by TUNEL assay (Fig. 1C,D). The severity of the phenotype correlated with the concentration of the grhl2b-MO injected (supplementary material Table S1), and an identical phenotype was observed utilising an exon 1/intron 1 (E1/I1) splice-blocking MO (supplementary material Fig. S2A-C). Some apoptosis was also visible in the trunk and tail with this MO (supplementary material Fig. S2D).
Recent studies of a mutant zebrafish line generated by Tol2 transposon-based gene trapping of an EGFP reporter into the first intron of the grhl2b gene demonstrated otic vesicle defects (Han et al., 2011). We confirmed that otolith formation was also disrupted in our ATG-blocking MO-injected morphants (supplementary material Fig. S3A), and that this defect could be rescued by injection of murine Grhl2 mRNA (supplementary material Fig. S4B). Otic expression of both claudin b and epcam was lost in our morphants, phenocopying the expression changes observed in the grhl2b mutant line (supplementary material Fig. S3C-F) (Han et al., 2011). However, no MHB defects were noted in this line, which we attribute to persistence of tissue-specific expression of grhl2b in the MHB region or, alternatively, to the presence of an alternate spliced isoform of grhl2b in the MHB (supplementary material Fig. S3G,H), a frequent event in gene trapping-based mutants (Clark et al., 2011).
Ontogeny of the MHB occurs in three stages: Fgf8/Wnt1-mediated positioning of the MHB boundary at the interface of otx2 (midbrain) and gbx2 (hindbrain) expression in the neural plate (∼6 hpf) (otx2/gbx1 interface in zebrafish) (Rhinn et al., 2009); induction of the MHB differentiation program, characterised by expression of genes including pax2a, wnt1, her5 and eng2a (∼9 hpf); and finally, maintenance of the MHB involving both feedback regulation and transcriptional interdependence of these genes (∼16-28 hpf) (Lun and Brand, 1998; Liu and Joyner, 2001; Martinez, 2001; Rhinn and Brand, 2001; Wurst and Bally-Cuif, 2001; Chi et al., 2003; Hidalgo-Sanchez et al., 2005). To analyse MHB development, we examined the expression of these crucial genes in the grhl2b loss-of-function and control MO-injected embryos. To ensure precise temporal comparison of gene expression, controls and grhl2b morphants were developmentally staged by expression of pax2a, which is sequentially expressed in the MHB, pronephric mesoderm and otic vesicles between 9 and 10 hpf (Thisse and Thisse, 2004) (data not shown). Neither otx2 nor gbx1 expression was altered in grhl2b morphants during the positioning/induction phases of MHB formation (Fig. 1E-H), suggesting that grhl2b does not play a role in regionalising the presumptive neural territories within the neural plate, or in positioning the midbrain/hindbrain at the junction of these regions. At the onset of the MHB induction signalling cascade, expression of the key patterning genes pax2a, wnt1 and her5 was unchanged between grhl2b morphants and controls, whereas eng2a expression was markedly reduced in the grhl2b-MO-injected embryos (Fig. 1I,J; supplementary material Fig. S4A-C′). This exclusive downregulation of eng2a was also apparent during the early stages of neural tube morphogenesis (14 hpf) and after morphological establishment of the MHB (23 hpf; Fig. 1K,L; data not shown). Expression of other MHB markers was not altered at these time points (supplementary material Fig. S4D-E′; data not shown). However, by 26 hpf, expression of MHB markers was substantially reduced or lost in 56/79 embryos examined [eng2a (22/27), wnt1 (8/11), pax2a (21/32) and her5 (5/9); supplementary material Table S1], indicating that grhl2b impacts on MHB maintenance at both structural and molecular levels (Fig. 1M-V). Interestingly, pax2a expression was also lost in the ventral hindbrain neurons of grhl2b morphants (Fig. 1Q), although whether this is due to direct regulation, loss of eng2a-dependent feedback loops or as a secondary consequence of increased apoptosis/decreased organiser fidelity remains to be determined. As specific staining of pax2a was observed both anteriorly (in the optic stalk) and posteriorly (in the otic vesicles) in grhl2b-morphants (Fig. 1P,Q), grhl2b does not appear to have a role in regulating patterning or gene expression at neural regions outside the brain. Consistent with this, expression of several well characterised patterning and specification genes, including fgf8, dlx3 (margin and neural plate), snail2 (neural crest), hgg, gsc, ntl (anterior notochord and polster), isl1 (neuron precursors), sox9b (neural crest) and shh (neural tube floor plate) at 8 hpf, 10 hpf or 18 hpf was largely unchanged, except for subtle variations attributable to alterations in grhl2b-morphant size and shape (supplementary material Fig. S4F-N′). Taken together, our experiments suggested that specific early loss of eng2a in grhl2b-morphants led to defective MHB maintenance between 23 hpf and 28h pf, due to either increased cell death in the MHB region or to impaired regulatory feedback between eng2a and other MHB patterning genes.
To determine whether grhl2a may also function in MHB patterning, we injected embryos with a specific grhl2a ATG-blocking MO. As shown in supplementary material Fig. S5, knockdown of grhl2a led to a specific defect in body patterning, with severe shortening of the tail reminiscent of convergence-extension (CE) defects in zebrafish (Heisenberg et al., 2000), and consistent with the known roles of the Grhl family in regulating CE in the mouse (Ting et al., 2003). MHB development, however, was not significantly affected morphologically, and no loss of MHB markers was observed in these morphants. When grhl2b and 2a were both downregulated through MO co-injection, a stronger trunk patterning defect was seen, concomitant with disrupted notochord formation. A slight defect was also seen in formation of the MHB basal constriction, suggesting that grhl2a may play a minor cooperative role with grhl2b in regulating MHB formation. An apparent sub-functionalisation of Grhl2 in the zebrafish has occurred, whereby grhl2a primarily regulates axial patterning and grhl2b regulates MHB patterning and neural tube closure.
Loss of MHB markers is not due to apoptosis in grhl2b morphants
To evaluate the relationship between the increased cell death and loss of MHB markers in grhl2b morphants, we initially performed rescue experiments using MO-resistant murine Grhl2 (mGrhl2) mRNA (Fig. 2A-E, Table 1). The abundant apoptotic cell death observed by bright-field microscopy and TUNEL staining in grhl2b morphants (51/81 embryos examined) was substantially reduced in Grhl2-rescued morphants (20/56 embryos examined; P=0.0017), emphasising the remarkable evolutionary conservation of function in the Grhl family. Importantly, the expression of MHB patterning markers, including eng2a and pax2a was also restored in Grhl2-rescued grhl2b morphants at 28 hpf (Fig. 2F-I; data not shown). These rescue experiments were subsequently repeated with a MO:resistant grhl2b mRNA, showing near-identical results for rescue of both apoptosis and MHB marker gene expression (data not shown). Despite this, the MHB folding defect was not rescued (Fig. 2B, inset), which we hypothesise to be due to the need for tightly controlled spatio-temporal expression of grhl2b for precise regulation of cellular polarity, consistent with known roles of grh in regulating cell polarity, and with previous work from our laboratory showing incomplete rescue of polarity-dependent phenotypes through simple restoration of gene expression (Caddy et al., 2010).
To determine whether prevention of apoptosis alone could also restore expression of MHB markers, the grhl2b-ATG blocking MO was injected into apoptosis-deficient, p53–/– fish (Berghmans et al., 2005) (Fig. 2J-O). Although apoptosis following MO-mediated grhl2b knockdown was substantially ameliorated in this line (Fig. 2J, inset), the number of fish exhibiting loss or reduction of MHB marker expression at 26 hpf was not significantly different between p53–/– fish [24/45; eng2a (96/12), wnt1 (7/7), pax2a (7/18) and her5 (4/8)] and wild-type fish [51/79; eng2a (16/24), wnt1 (18/23), pax2a (6/14) and her5 (10/18); P=0.219 by chi-square test; Fig. 2L-O; supplementary material Table S2). The defect in MHB folding was also not prevented. These findings indicate that p53-mediated neural apoptosis does not underpin MHB disruption in grhl2b morphants, and that the loss of MHB markers in these mutants may instead reflect perturbations of eng2a controlled homeostatic feedback mechanisms.
Eng2a rescues neural apoptosis and MHB marker loss in grhl2b morphants
The Drosophila orthologue of eng2a, engrailed, is a direct target of grh (Dynlacht et al., 1989; Attardi and Tjian, 1993), and the murine eng2a orthologue, engrailed 1, can be transcriptionally activated by the Grhl family (Wilanowski et al., 2002). Functionally, En1 and En2 prevent apoptosis of neural cells in the mouse midbrain (Sgado et al., 2006; Alavian et al., 2009), and vertebrate engrailed genes are crucial feedback molecules within the MHB maintenance cascade (Picker et al., 2002; Scholpp et al., 2003; Erickson et al., 2007). Furthermore, we observed elevated apoptosis in the MHB of eng2a morphants (data not shown). These data support the hypothesis that grhl2b-mediated loss of engrailed function leads to both neural apoptosis and loss of MHB markers. To test this, we injected grhl2b morphants with eng2a mRNA and examined the embryos by both bright-field microscopy (Fig. 3A-C) and TUNEL staining (Fig. 3D-F; Table 2). Of the non-rescued grhl2b morphants, 68/85 showed strong neural cell death, compared with 26/55 eng2a-rescued embryos (P<0.001), demonstrating a significant reduction in apoptosis. Crucially, expression of MHB markers was also restored in eng2a-rescued grhl2b morphants at the 26 hpf time point (Fig. 3G-I), but the MHB folding defect persisted (Fig. 3J,K). To confirm further the relationship between eng2a and grhl2b, we performed co-knockdown experiments, in which grhl2b and eng2a MOs were co-injected at sub-phenotype-causing doses (100 mM per MO). We observed loss of pax2a expression and increased cell death in approximately 50% of embryos (52/105), compared with 5-7% (4/72 and 8/112 for MO:grhl2b and MO:eng2a, respectively) when the MOs were injected individually (with MO:control) at the same concentration (Fig. 3L,M; Table 3). However, folding at the MHB still occurred in the co-injected embryos (Fig. 3M), suggesting that eng2a is crucial for grhl2b-dependent inhibition of apoptosis and maintenance of MHB markers, but not for morphogenesis. Previous experiments had shown an upregulation of the pan-neurotrophin receptor p75NTR following disrupted engrailed signalling in mouse neurons (Alavian et al., 2009). Q-RT-PCR of the MHB regions of grhl2b morphants likewise showed a significant upregulation of this gene (P=0.021; Fig. 3N), consistent with increased apoptosis and loss of engrailed signalling in our model.
Direct regulation of eng2a by grhl2b in the MHB requires the two genes to be temporally and spatially co-expressed during development. To examine this we performed in situ hybridisation with eng2a and grhl2b probes on coronal sections of the MHB from embryos at 12 hpf (Fig. 4A,B). Although robust expression of eng2a was observed in all cells in this region, expression of grhl2b was observed strongly in only the neural ventral tissue. Close examination of these sections suggested that low levels of grhl2b expression may be evident more broadly throughout the MHB, and supporting this were recent data that also showed low levels of expression in the midbrain region (supplementary material Fig. S3H). To circumvent the problem of low sensitivity with in situ hybridisation, a common issue with analysis of transcription factors in general and with the grhl factors specifically (Janicke et al., 2010), we performed laser-capture microdissection (Fig. 4C,D) on coronal sections of MHB from wild-type embryos, and analysed expression of grhl2b and eng2a in the isolated cells by Q-RT-PCR. cDNA generated from one-cell embryos (pre-zygotic transcription) served as the negative control. Both grhl2b (Fig. 4E) and eng2a (Fig. 4F) were expressed in these cells, and given the near-ubiquitous eng2a expression in the dissected region (Fig. 4A), it is likely that these cells co-express both factors.
To confirm that eng2a is a direct transcriptional target of grhl2b, we examined the promoter of the eng2a gene for potential grhl2b-binding sites. The DNA consensus-binding site for all grhl family members is highly conserved from fly to human (Ting et al., 2005; Wilanowski et al., 2008; Caddy et al., 2010). Two putative binding sites were identified, at positions –1949 to –1942 (AACCAGTT) and –835 to –828 (AACCGTTT) relative to the transcriptional start site (TSS). As no antibodies are available to detect endogenous zebrafish grhl2b, we generated a C-terminal FLAG-tagged grhl2b construct (grhl2b-FLAG). This fusion construct was detectable by western blotting (Fig. 4G), and when overexpressed in one- to two-cell embryos, generated an identical eye defect phenotype to that observed with untagged grhl2b mRNA at 24 hpf (data not shown). ChIP on grhl2b-FLAG-expressing embryos demonstrated enrichment of both predicted sites within the eng2a promoter (Fig. 4H), indicating that eng2a is a direct target gene of grhl2b. Consistent with both our ChIP and expression data, in vivo reporter gene studies showed that MO-induced loss of grhl2b led to a sixfold reduction in expression from an eng2a promoter-luciferase reporter construct in embryos (Fig. 4I). Our data indicate that grhl2b-mediated regulation of eng2a is a novel, direct mechanism for regulating MHB patterning.
spec1 is a novel grhl2b target gene crucial for MHB morphogenesis
Both our results, and previously published data (Scholpp et al., 2004), suggest that eng2a is not required for establishing the characteristic neural tube morphology at the level of the MHB. Furthermore, recent evidence suggests that the molecular events that regulate MHB induction and patterning are distinct from those that regulate MHB morphogenesis (Gutzman et al., 2008) and ventricle development. To define the molecular mechanisms underpinning grhl2b-mediated regulation of MHB folding, we used the Grhl DNA consensus binding site to interrogate a customised dataset of genomic regions located within 10 kb of gene transcriptional start sites that are highly phylogenetically conserved. We have successfully used this approach previously to identify direct target genes of other Grhl factors (Caddy et al., 2010; Darido et al., 2011). One promising candidate target that emerged was spec1 (small protein effector of cdc42, protein 1), also known as cdc42se1 (cdc42 small effector protein 1), a factor that binds to cdc42 and influences actin accumulation in polarised T cells (Pirone et al., 2000; Ching et al., 2005; Ching et al., 2007). Data from migrating cells (which exhibit cell polarity) suggest that SPECs may generally be involved in cdc42-mediated polarity establishment in other cell types, besides lymphocytes, and may also regulate cell shape (Pirone et al., 2000).
The spec1 promoter contained four putative grhl2b-binding sites within 2 kb of the TSS at positions: –1837 to –1830 (site 1, AACTAGTT), –714 to –707 (site 2, AACAGTTT), –291 to –284 (site 3, AACTGTTT) and –17 to –10 (site 4, AACCGGTG). ChIP assays using grhl2b-FLAG-expressing embryos demonstrated binding to three of these predicted sites (Fig. 5A), confirming that spec1 is a direct target gene of grhl2b. ATG-MO-mediated knockdown of spec1 led to complete abolition of MHB folding, phenocopying the non-folded MHB seen in grhl2b morphants (Fig. 5B,C). This phenotype was also seen with a splice-blocking-MO directed against the exon 1/intron 1 boundary of spec1 (supplementary material Fig. S6A-C). Consistent with separate mechanisms being responsible for the loss of MHB patterning and MHB morphogenesis, spec1 ATG or splice block morphants did not exhibit enhanced neural apoptosis (supplementary material Fig. S6D; data not shown) or loss of MHB-patterning markers (Fig. 5D,E; data not shown). spec1 is maternally deposited and exhibits ubiquitous expression throughout zebrafish development (Thisse and Thisse, 2004) (data not shown), rendering in situ hybridisation of little value for examining spec1 expression in grhl2b morphants. To counter this, we dissected out the MHB region from MO:grhl2b- and MO:control-injected embryos at 26 hpf and demonstrated reduced expression of spec1 mRNA by Q-RT-PCR (Fig. 5F). To determine whether grhl2b and spec1 functioned cooperatively in regulating MHB morphogenesis, we again performed co-knockdown experiments, in which grhl2b and spec1 MOs were co-injected at sub-phenotype-causing doses (MO:grhl2b 100 mM, MO:spec1 125 mM). We observed severe folding (Fig. 5G) and ventricle formation (Fig. 5H) defects, which phenocopied the grhl2b knockdown, in ∼40% of embryos (70/178), compared with 3-4% when the MOs were injected individually at the same concentration (6/154 for grhl2b and 7/168 for spec1, respectively; Table 4). Last, we reasoned that if spec1 were a direct transcriptional target of grhl2b in the MHB, then expression of spec1 mRNA should be increased following rescue of the grhl2b morphants with injection of Grhl2 mRNA. Q-RT-PCR confirmed this hypothesis (Fig. 5I). As an additional control, spec1 mRNA was also significantly increased in the MHB of wild-type embryos overexpressing grhl2b mRNA (Fig. 5J). Taken together, these data indicate that grhl2b directly regulates spec1 to control MHB morphogenesis.
grhl2b acts with fgf8 in MHB patterning and morphogenesis
Fgf8 functions as the main organising molecule of the isthmus, as it is required for development of both the tectum and cerebellum (Chi et al., 2003), and is sufficient to induce both structures in gain-of-function studies (Crossley et al., 1996; Martinez et al., 1999; Shamim et al., 1999). It also plays a central role in MHB maintenance, with En1/2, Pax2/5 and Fgf8 forming an inter-dependent feedback loop that is necessary for persistent expression of these genes (Nakamura, 2001). Consistent with studies in Drosophila, which show that the transcriptional activity of the Grh protein is enhanced by FGF signalling through a post-translational mechanism (Hemphala et al., 2003), we were unable to detect any change in either the timing or extent of grhl2b mRNA expression by either in situ hybridisation or Q-RT-PCR in the MHB region of fgf8 morphants generated using an established MO (Araki and Brand, 2001) (supplementary material Fig. S7). We therefore examined whether fgf8 and grhl2b interacted functionally by performing co-knockdown experiments with grhl2b and fgf8 MOs co-injected at sub-phenotypic doses (100 mM each MO; Fig. 6A-L). Control (phenotypic) doses of MO:fgf8 induced the expected ace phenotype, with concomitant apoptosis (Fig. 6A-C); MHB loss was also seen in 32% (11/34) of sub-phenotypically co-injected grhl2b/fgf8 morphants (Fig. 6D-F). A further 56% (19/34) of these morphants (Fig. 6G-I) exhibited an even stronger phenotype, that of MHB loss and increased apoptosis, indicative of MO dose dependency. In total, 88% (30/34) of sub-phenotypically co-injected morphants displayed defective MHB/cerebellum development, compared with 7% of embryos injected with sub-phenotypic doses of MO:fgf8/MO:control (2/27; Fig. 6J-L; Table 5) or with 6% of embryos injected with sub-phenotypic doses of MO:grhl2b/MO:control (3/47; Table 5). These data indicate that fgf8 and grhl2b operate within the same linear pathway to regulate MHB formation.
To determine whether these two genes could act cooperatively to regulate eng2a transcription, we performed sub-phenotypic injections of both MOs together with pGL3-eng2a-PROM, and assessed subsequent luciferase production. We found that eng2a-PROM-responsive luciferase production was significantly downregulated when both MOs were co-injected at sub-phenotypic doses, but not when either MO was injected at sub-phenotypic dose (together with MO:control), indicative of functional pathway cooperativity (Fig. 6M). Consistent with grhl2b being post-translationally, but not transcriptionally, modified by fgf8 signalling, enforced expression of wild-type grhl2b mRNA was incapable of rescuing the ace phenotype in MO:fgf8-injected embryos (supplementary material Fig. S7). As eng2a is already known to be lost in fgf8-null (ace) mutants (Reifers et al., 1998), we examined whether spec1 was also downregulated in fgf8-deficient morphants, as predicted if loss of fgf8 led to abrogated grhl2b function. We isolated regions of the neural tube at the level of the MHB in 18-19 hpf fgf8-morphants (pre-MHB morphogenesis), and Q-RT-PCR analysis showed that spec1 expression was significantly downregulated (P<0.005) in fgf8-morphants at the formative stages of MHB morphogenesis (Fig. 6N). Taken together, our data, together with the previous experiments in Drosophila (Hemphala et al., 2003), suggest grhl2b and fgf8 cooperate to regulate MHB patterning and morphogenesis in zebrafish, acting through eng2a and spec1 (Fig. 6O). We hypothesise that fgf8 may regulate grhl2 post-transcriptionally, as is evident in the corresponding relationship in Drosophila (Hemphala et al., 2003), although proof of this hypothesis would require experiments to define and mutate the putative fgf8-induced phosphorylation sites in grhl2b.
In this paper, we provide evidence that the zebrafish transcription factor grhl2b plays key roles, putatively downstream of fgf8, in the morphogenesis and patterning of the mesencephalic-metencephalic regions of the developing vertebrate brain. This data expands on the known functions of the Grhl family in neuroblast specification, proliferation and apoptosis in Drosophila (Bray and Kafatos, 1991; Uv et al., 1997; Almeida and Bray, 2005; Cenci and Gould, 2005; Maurange et al., 2008), and neural tube morphogenesis in mammals (Ting et al., 2003; Rifat et al., 2010), and highlights the evolutionary continuity of grh/Grhl gene function. We show that expression of MHB markers is initiated, but not maintained, in grhl2b-deficient embryos, and that this phenotype is not rescued by amelioration of apoptosis. Re-expression of eng2a in grhl2b morphants restores expression of pax2a and other MHB markers, and this, coupled with functional and biochemical data, identifies eng2a as a direct downstream target of grhl2b. However, the genetic interaction between grhl2b and eng2a does not influence MHB morphogenesis, suggesting that this process is separable from patterning. Consistent with this, we identified a second direct grhl2b target, spec1, which disrupts folding at the MHB, but does not alter expression of MHB markers.
The grhl-eng axis is conserved in evolution
Engrailed is an important factor for establishing and maintaining cellular compartments during Drosophila development (Hidalgo, 1996). Vertebrate homologues of engrailed also contribute to lineage-restricted compartments, as evidenced by the loss-of-function phenotypes in the mesencephalon (Wurst et al., 1994; Scholpp and Brand, 2001). As shown here in the fish, and previously in both fly and vertebrate systems, the Grhl factors are important regulators of engrailed expression (Dynlacht et al., 1989; Attardi and Tjian, 1993; Wilanowski et al., 2002), and expand the phylogenetically conserved mechanisms involved in regulation of the engrailed family. In the mouse, Pax2, Pax5 and Pax8, homologues of the Drosophila pair-rule gene paired (a key transcriptional regulator of engrailed), are directly involved in maintenance of En2 brain expression in a feedback loop (Song et al., 1996; Liu and Joyner, 2001). In the fish, the paired homologue pax2a is also necessary for normal eng2 brain expression (Krauss et al., 1991). Taken together, these findings suggest that the genetic pathways involving regulation of the engrailed factors are highly conserved, despite evolving to control divergent developmental processes in flies and vertebrates.
grhl2b acts within the fgf8 signalling pathway
Fgf8 signalling plays a central role in MHB positioning through coordinated expression of otx2 and gbx2 (Rhinn and Brand, 2001; Hidalgo-Sanchez et al., 2005). It is also pivotal for maintenance of the MHB, which is achieved via a complex series of feedback loops with factors including En1/2 and Pax2/5 (Joyner et al., 2000; Sato et al., 2004). Our data suggest that grhl2b is a crucial factor for fgf8-mediated induction of En expression. This effect does not occur at the level of transcription, as expression of grhl2b is not altered in fgf8 morphants and enforced expression of grhl2b in fgf8 morphants does not ameliorate the ace phenotype. However, the two genes act cooperatively, as shown by the high frequency of morphants with MHB defects observed when embryos are co-injected with sub-phenotypic doses of both grhl2b and fgf8 MOs. Studies in Drosophila have shown that FGF signalling upregulates grh activity post-translationally (Hemphala et al., 2003). This is thought to be due to FGF-induced phosphorylation of grh, most probably by activated MAPK (ERK2), as grh is a known substrate of this kinase (Liaw et al., 1995). Obvious parallels exist between these findings and the mechanism underpinning fgf8 function in the MHB, which is also mediated through ERK signalling (Sato and Nakamura, 2004). Expression of a dominant-negative form of Ras induces repression of Pax2/5 and En1/2 expression (Sato and Nakamura, 2004), and these findings would be consistent with grhl2b functioning as the key intermediary in this pathway. Thus, post-translational activation of grhl2b by fgf8 could ensure that the transcriptional activation of eng2a is confined to the MHB, and does not extend into adjacent regions, where grhl2b is also expressed.
Loss of grhl2b induces apoptosis
A prominent aspect of our grhl2b loss-of-function studies was apoptosis throughout the brain and nervous system, especially at the level of the MHB, which correlated with a loss of marker expression and MHB morphology. MHB apoptosis and marker loss is a feature of the fish mutant ace, which harbours a mutation in fgf8 (Reifers et al., 1998; Jaszai et al., 2003), and in other models of MHB-marker loss (Brand et al., 1996; Chi et al., 2003). Our experiments also show MHB apoptosis following MO-mediated downregulation of eng2a (data not shown). We observed rescue of apoptosis in the grhl2b morphants with loss of p53, or with re-expression of grhl2b and eng2a. This result does not implicate eng2a as the sole downstream effector of grhl2b for the prevention of apoptosis, however, as Wnt1, En and Pax genes display marked transcriptional interdependence. Despite the substantial rescue of apoptosis in the grhl2b morphants on a p53–/– genetic background, restoration of MHB marker expression was not observed, indicating that grhl2b provides instructive as well as anti-death signals. Consistent with this, rescue of grhl2b-morphants through the re-introduction of eng2a mRNA was both neuroprotective within the MHB, and restored the expression of MHB markers.
Defective neural tube folding in both mouse and fish Grhl2/grhl2b mutants
In addition to increased apoptosis and loss of MHB marker expression, we also observed defective MHB folding and ventricle development in the grhl2b morphants. This appears not to involve eng2 as a downstream effector, despite both grhl2b and fgf8 being implicated. Recent data suggests that MHB morphogenesis and subsequent ventricle inflation are regulated by non-patterning genes at the level of the MHB (Lowery and Sive, 2009). Zebrafish mutants sleepy and grumpy (laminin γ1 and β1, respectively) (Gutzman et al., 2008), snakehead (Atp1a1), and nagie oko (Mpp5) (Lowery and Sive, 2005) show various neural tube/MHB folding defects, and demarcate the process of MHB patterning from MHB morphogenesis. We have identified a further novel regulator of neural tube morphogenesis, spec1, which is a known regulator of the small RhoGTPase cdc42 and a direct target of grhl2b. As spec1 regulates cellular morphogenesis (Pirone et al., 2000), and changes in cell shape (in particular basal constriction) appear to be crucial for correct MHB folding in zebrafish (Gutzman et al., 2008), we propose that spec1-mediated regulation of cdc42 is important for correct neural tube folding at the MHB. Consistent with this predicted mechanism, our data show that spec1 has no effect in regulating neural cellular survival or MHB patterning. The involvement of grhl2b in both patterning and morphogenesis may explain why the phenotype of Grhl2-null mice is far more severe than most of the other MHB patterning mutants. A hallmark of the cranial defects in Grhl2-null mouse embryos is the absence of dorsolateral hinge point (DLHP) formation, which is central to neural plate folding in mammals (Rifat et al., 2010). Although the mechanism underpinning Grhl2-mediated DLHP formation is not known, a recent study suggests that defective formation of junctional complexes, mediated by E-cadherin and claudin 4 loss in Grhl2-deficient mice, may be responsible for failed neural tube closure (Werth et al., 2010).
Our study identifies two novel targets of grhl2b (eng2a and spec1), which regulate MHB patterning and morphogenesis. Through our use of combinatorial MO knockdown and mRNA rescue technologies, we have shown that these disparate processes are controlled by grhl2b through direct regulation of two different transcriptional targets. As the full spectrum of grhl2b target genes remains to be elucidated, future work will further refine these, and other genetic pathways influenced by grhl2b in neurulation.
We thank Dr Heather Verkade, Prof. Michael Brand and Dr Andrew Waszkiewicz for providing in-situ hybridisation probes and/or full-length cDNA plasmids; Dr Quan Zhao and Dr Chris Slape for assistance with ChIP assays; Mr Cameron Nowell for assistance with confocal microscopy; Dr Matthew McCormack for assistance with statistical analysis; and Dr Stewart Fabb, Dr Andrew Badrock and John Hayman for technical assistance.
S.M.J. is a Principal Research Fellow of the Australian National Health and Medical Research Council (NHMRC). The work was partially supported by project grants from the NHMRC [435112 and 544304], and by a grant from the March of Dimes Foundation [6-FY07-293].
Competing interests statement
The authors declare no competing financial interests.