Introduction

The grainyhead-like (Grhl) gene family, vertebrate orthologues of the antecedent Drosophila gene grainyhead (grh), are a highly conserved family of developmentally-critical transcription factors. Comprising three members in vertebrates (Grhl1-3), with a further sub-functionalisation of grhl2 in zebrafish (grhl2a and grhl2b respectively), this family regulates diverse processes which ensure correct cell and tissue patterning and development, including formation of the Drosophila cuticle1 and skin barrier2,3, polarity-dependent processes of epidermal migration and wound healing4,5, neuroblast proliferation and apoptosis6,7,8, convergence-extension (CE) mediated movements in neural tube closure and axial patterning9,10, and development of the craniofacial skeleton and palatal fusion11,12,13,14 in vertebrates. Importantly, almost all phenotypes described in Drosophila resulting from loss of grh have thus far also been characterised in vertebrate models, further highlighting the remarkable conservation of gene function over 700 million years of evolution2.

The three vertebrate orthologues vary in their functional importance during development, although all are predominantly expressed in epithelia15. Grhl1 is implicated in the development of both auditory16 and epidermal17,18,19 tissues, although Grhl1 −/− mice are viable and fertile18. Conversely, loss of either Grhl2 or Grhl3 in mice leads to severe defects in cell survival, migration, integration, fusion and patterning. Although the early embryonic lethality (between embryonic day [E] 9.5-E11.5 of development) of Grhl2 knockout animals has precluded analyses of skin barrier formation, craniofacial patterning and wound healing, Grhl2 −/− mutants present with severe exencephaly, fully-penetrant split-face and maxillary clefting, an open neural tube and defects in brain cellularity and neuronal survival8,10,20. Similarly, loss of Grhl3 leads to fully penetrant thoracolumbosacral spina bifida, occasional exencephaly, neurocognitive and behavioural disorders, defects in epithelial development of the bladder and intestine and convergence-extension defects9,21,22,23; although embryos survive to birth, Grhl3 −/− embryos die soon after due to rapid trans-epidermal water loss as a consequence of an impaired skin barrier and dehydration following exposure to a terrestrial environment24.

Although numerous studies have determined key developmental roles following loss of Grhl expression, the over-expression of Grhl factors has been difficult to model, largely limited to the generation of transgenic mice. Whereas transgenic expression of Grhl3 has been reported25, utilising a BAC to restore Grhl3 expression and rescue spina bifida-like defects in a hypomorphic Grhl3 model (the curly-tail mouse), this study did not report any over-expression-related phenotypic abnormalities in these mice, nor were over-expression studies performed using the BAC construct in WT mice. However, over-expression of Grhl2 (in the spontaneously occurring axial defects [axd] mouse) was reported to lead to an increase in cellular proliferation and neural tube defects26,27. These data suggest that tight homeostatic regulation of both Grhl2 and Grhl3 is required to ensure correct neural tube morphogenesis and axial development, although differences in gene function are likely to exist between disparate animal models.

Although virtually all grh-dependent phenotypes in Drosophila are conserved in vertebrates, the full spectrum of vertebrate-specific developmental processes in embryogenesis remains incompletely characterised. Utilising the highly genetically tractable zebrafish model, we have previously shown that loss of grhl2b (the fish orthologue most closely related to Grhl2) leads to aberrant midbrain-hindbrain boundary (MHB) morphogenesis and patterning28 whereas loss of grhl3 leads to impaired craniofacial cartilage growth and development12. Here, we utilised the zebrafish model to further uncover novel phenotypes following genetic deletion of grhl3, and knockdown and over-expression of zebrafish and mammalian grhl2b, Grhl2 and Grhl3, and ascribe novel cellular and morphogenetic functions regulated by this transcription factor family in vertebrates.

Methods

Zebrafish microinjection techniques

All animal experimentation was conducted under ethical approval granted by the Monash University, La Trobe University and Walter and Eliza Hall Animal Ethics Committees (AEC), and performed in accordance with all relevant guidelines and regulations. ATG-blocking morpholinos (MO; Gene-Tools) were designed to inhibit translation of arhgef19 (GAGAGATTCCATAGCCAGGAAGCAT) and spec1 (CAATCTTATGCCAGAACGCACTCAT), and these were injected at sub-phenotypic dosages with MOs targeting either grhl2b or grhl3 as reported by us previously12,28. Full-length cDNA isoforms of murine Grhl2 and Grhl3, and zebrafish grhl2b and grhl3 were cloned by PCR, inserted into the expression vector pCS2 + , and verified by sequencing. mRNA for injection was generated using the mMESSAGE mMACHINE Sp6 transcription kit (Ambion) and precipitated using Ethanol/LiCl. All in-situ hybridisation and imaging techniques were as reported previously28,29.

Validation of grhl2b and grhl3 morpholino specificity

Specific anti-sense oligonucleotides targeting both grhl2b and grhl3 used in the present study have been rigorously validated and reported by us previously. Pertaining to MOs used to inhibit grhl2b, we utilised both ATG- and splice-blocking MOs to show reproducibility of loss-of-grhl2b phenotype in the brain and otic vesicle, loss of midbrain-hindbrain boundary markers (eng2a, wnt1, pax2a and her5), confirmed presence of non-spliced grhl2b-transcript using the splice-blocking MO, and confirmed functional rescue of phenotypes using zebrafish grhl2b mRNA (with silent mutations introduced at the ATG site to prevent inhibition by the ATG-MO), murine Grhl2 mRNA, and target gene (eng2a) mRNA28. Pertaining to MOs used to inhibit grhl3, we again utilised both ATG- and splice-blocking MOs to show reproducibility of loss-of-grhl3 phenotype in the craniofacial skeleton, loss of neural crest and pharyngeal markers (edn1, dlx3, hand2), confirmed presence of non-spliced grhl3-transcript using the splice-blocking MO, and confirmed functional rescue using both murine Grhl3 mRNA and target gene (edn1) mRNA12. The negative 5-base pair mismatch grhl3 control morpholino sequence was TGAcAGgCTCAATgTCCTTcGTgAT.

Generation of grhl3 −/− fish via CRISPR/Cas9 mediated deletion

The target site for generation of grhl3 single guide (sg) RNAse was predicted using ZiFit (http://zifit.partners.org/ZiFiT/). Oligonucleotides (IDT) encoding the determined target site (GGTTGCCATGATTTCTGCGA) were cloned into the plasmid pDR27430, and sequence-verified plasmids were used as a template to transcribe sgRNA using the MEGAscript T7 kit (Ambion) as per manufactures recommendations. sgRNA, Cas9 nuclease (NEB M0386S), and a specific targeted STOP codon cassette31 (CTCTCCTGCCTCAACCGTCGGTCATGGCGTTTAAACCTTAATTAAGCTGTTGTAGCAGAAATCATGGCAATCGGA) were co-injected into the cell of single-cell stage WT Tuebingen zebrafish embryos. sgRNA was injected at concentrations ranging between 20–750 pg/nl, with greater efficiency of cutting observed at higher, as previously reported for other sgRNA32, and Cas9 protein was injected at a final concentration of 320 ng/μl. DNA sequencing using specific oligonucleotides flanking the target site (F: CAGCTCTTCCCCTGAACTTG, R: ACATAAATGCGGACCTCAGGTGT) were used to determine genomic changes in injected embryos. DNA was isolated from injected embryos in 20 μl 50 mM NaOH, and heated to 95 °C for 10 min. The solution was neutralised by adding 2 μl 1 M Tris-HCL, pH 8.0. Samples were pulse centrifuged to pellet debris, and supernatant used as a template for subsequent PCR reactions. Presumed nonsense-mediated decay was detected by firstly isolating mRNA from 50 embryos of each of control and grhl3 −/−(+14bp) mutant embryos at 80% epiboly; cDNA was synthesised using SuperscriptIII (Invitrogen) according to manufacturer’s instructions. One μl of cDNA was used to confirm comparable expression of β-actin by Q-RT-PCR, as well as expression of grhl3 transcript by standard PCR, using a forward primer (CACATATTTAACCACAAGCACACC) in Exon 1, and a reverse primer (AATGCTCGATGATGTTGTCG) in Exon 15 to determine relative abundance of grhl3 transcript.

Confocal Microscopy

Embryos were fixed in 2% Paraformaldehyde (PFA) overnight and mounted in 2% Low melting temperature agarose. Samples were stained for F-actin before imaging with Rhodamine-Phalloidin (Life Technologies) as per manufacturers’ recommendations. Prior to imaging, embryo heads were dissected from the body and embedded in 4% low melting point agarose under a coverslip. Serial confocal images were taken on a Nikon C1 confocal microscope with either a 20x or 40x lens. Figure preparation was performed using Illustrator CS5 (Adobe).

Cell area and number measurements

The area of EVL cells over the midbrain-hindbrain boundary was determined using automated quantitation and imaging software (FiJi) to manually outline all individual cells within a defined region of 2500 μm2, derived from confocal image stacks. Cell counts were also determined manually for all cells within the defined 2500 μm2 region.

Statistics

Quantitative data derived from at least three independent experiments; descriptive statistics are mean ± SEM of data for (n) individuals or (n) independent experiments. GraphPad prism was used for unpaired 2-tailed Student’s T-tests, with normally-distributed continuous variables. P values are as follows * = P ≤ 0.05, ** = P ≤ 0.01, *** = P ≤ 0.001.

Results

grhl3 deletion leads to severe epithelial disruption and embryo rupture prior to gastrulation

Although MO-mediated loss-of-function transient knockdown zebrafish models have been reported previously for grhl1 16,19, grhl2a 28, grhl2b 28,33 and grhl3 12, thus far, only one loss-of-function genetic deletion model (a line null for grhl2b)31, has thus far been reported. In order to further investigate the role of this family, particularly grhl3, in zebrafish tissue patterning, organ development and morphogenesis, we firstly examined the early expression of grhl3 by ISH (Fig. 1a–f). We found that grhl3 is not maternally deposited, but rather is strongly and specifically activated in all EVL cells at ~4 hpf, corresponding temporally to the onset of zygotic transcription at the maternal-zygotic transition (MZT)34, before being downregulated in the EVL, but remaining expressed at high levels at the migratory front (“leading edge” of cells) during epiboly at 5 hpf. We generated a grhl3 mutant fish via CRISPR/Cas9 mediated deletion and homologous integration, targeting Exon 4 (Fig. 1g) at a site which lay 3′ to seven putative in-frame ATG-initiation codons35, thereby minimising the chance of a 5′-truncated (yet still partially functional) protein being generated35. Our resultant mutant carries an insertion of a STOP codon cassette at codon 143 (Fig. 1h), that results in a net +14 bp gain within exon 4 of grhl3 (confirmed by genotyping; Fig. 1i and Figure S7) and a stop codon being introduced at amino acid 143, with evidence of subsequent presumed nonsense-mediated decay (Fig. 1j). Double grhl3 −/−(+14bp) mutant embryos displayed normal morphology until shield stage, and initiated epiboly at a timepoint equivalent to both WT and heterozygous siblings (Fig. 1k,l). However, we noted that epiboly then appeared slowed in grhl3 −/−(+14bp) embryos compared to WT, with an initial bulging of the yolk cell that is retracted as epiboly continues, and culminating in a failure to complete epiboly and rupturing at the vegetal pole between ~80% epiboly and tailbud stage (Fig. 1m,n). This embryonic rupture is presumably due to significant weakening of the EVL/YSL interface. As expected, Q-RT-PCR analyses of ~80% epiboly stage WT and grhl3 −/−(+14bp) embryos (prior to rupture) showed that transcript levels of grhl3 were greatly reduced in phenotypic grhl3 −/−(+14bp) embryos, with a concomitant decrease in the closely related family member grhl1 (Fig. 1o). Transcript levels of the other two grhl fish orthologues, grhl2a and grhl2b were either unchanged (grhl2a) or slightly elevated (grhl2b) in grhl3 −/−(+14bp) embryos, indicating firstly that grhl3 may also regulate the expression of grhl1 during epiboly and entry into gastrulation, recapitulating the rupture-phenotype reported previously following MO-mediated knockdown of both grhl1 and grhl3 36, and secondly, that grhl2b may partially ameliorate early loss of grhl3.

Figure 1
figure 1

CRISPR-Cas9 mediated generation of a grhl3 deletion zebrafish model. (a–f) ISH expression of grhl3 at 3–5 hpf, shown in lateral (a–c) and dorsal (d–f) views. grhl3 is not maternally deposited (a,d), but is strongly and specifically expressed within all EVL cells at ~4 hpf, corresponding temporally to the maternal-zygotic transition (MZT; b–e). grhl3 mRNA is rapidly downregulated, remaining primarily localised at the leading-edge margins (asterisk, c). (g) Sequence details of the exon 4 region of grhl3, showing a loss of 3 bp and insertion of a STOP codon cassette comprising 17 bases, resulting in a net gain (genomic lesion) of 14 bp in the grhl3 mutant. (h) Schematic of the zebrafish grhl3 protein (to scale; amino acids 1–557), showing trans-activation (TA), DNA-binding (DNA), and dimerisation (DIM) domains, indicating site of genomic lesion at codon 143. (i) Sequencing of WT and grhl3 −/−(+14bp) embryos confirming presence of the genomic lesion. The red underline indicates (in WT) the position of the three deleted nucleotides, green underline shows insertion of 17 bp of STOP-cassette, black underline shows the resultant in-frame stop codon, yellow arrowhead indicates the CRISPR/Cas9-cut site. (j) RT-PCR showing ~50% loss of grhl3 transcript in grhl3 −/−(+14bp) embryos relative to controls at 8 hpf; a full-length gel photo is presented in Supplementary Figure S8. (k–l) Epiboly is delayed in grhl3 −/−(+14bp) embryos (l) relative to WT (k), resulting in premature contraction of the leading-edge actin ring (arrows in L) at ~75% epiboly. (m,n) WT embryo at the onset of somitogenesis showing normal development (m), and grhl3 −/−(+14bp) embryo (n) showing rupture of the EVL resulting in lethality (arrow in n). (O) Q-RT-PCR showing decreased grhl3 expression, no change in grhl2a, and slightly elevated expression of grhl2b at ~75% epiboly in grhl3 −/−(+14bp) embryos relative to WT controls.

Full size image

grhl3 regulates midbrain-hindbrain neural tube folding

As genetic deletion of grhl3 in zebrafish was embryonic lethal, we further investigated the role of grhl3 using morpholino-mediated knockdown. As the Grhl family regulates both closure9,10 and morphogenesis8,28 of the neural tube, we injected 1–2 cell stage embryos with previously validated, specific antisense ATG- or splice blocking morpholinos (MO) directed against grhl3 mRNA12, to examine putative changes in morphogenesis of the neural tube (Supplementary Table S1). Consistent with the morphological defects we had seen previously following MO-mediated knockdown of grhl2b 28, we noted an overall shortening of the axis (Fig. 2a,b), and the presence of a curved tail at 96 hpf (Fig. 2a,b; insets), analogous to the murine hypomorphic Grhl3 mutant, curly-tail 25 which presents with posterior spinal cord defects. We also detected specific defects in the folding and morphogenesis of the midbrain-hindbrain boundary (MHB; Fig. 2c,d) by 24 hpf, without overt disruptions in the morphogenesis of other structures such as the eyes, heart, notochord or somites. These data were reminiscent of the MHB morphological changes we had observed previously following MO-mediated knockdown of grhl2b 28, indicating that the regulation of MHB morphogenesis by the grhl family may involve both multiple family members as well as conserved genetic mechanisms.

Figure 2
figure 2

Midbrain-Hindbrain Boundary (MHB) morphogenesis is disrupted following knockdown of grhl3-dependent transcriptional networks. (a,b) Axial length is reduced, neural morphology is disrupted, and fish often exhibit a “curved-tail” phenotype (inset; 96 hpf) in MO:grhl3 injected fish relative to controls. (c,d) The MHB defect in MO:grhl3 injected fish presents as aberrant folding (dorsal view), confirmed by horizontal sections of H&E stained MHB tissue (c,d; inset). (e,f) Although the MHB is mis-folded, MHB patterning markers (such as eng2a) are not differentially regulated. (g–j) Following sub-phenotypic co-knockdown of grhl3 and spec1 (MO:grhl3 + MO:spec1; g,i) the MHB is mis-folded when compared to phenotypically normal MO:spec1 + MO:control injected fish (h,j). (k–p) Following sub-phenotypic co-knockdown of grhl3 and arhgef19 (MO:grhl3 + MO:arhgef19; k), the fish are slightly shorter relative to phenotypically normal MO:arhgef19 + MO:control injected fish (l). Furthermore, the neural tube exhibits an “open” phenotype at the MHB region (dorsal view; m,n); this is highlighted by ISH for the MHB marker eng2a (posterior view of MHB; o). relative to the phenotypically normal MHB seen in MO:grhl3 + MO:control injected fish (p).

Full size image

As MO-mediated knockdown of grhl2b in zebrafish also leads to loss of the key MHB patterning marker, and direct grhl2b-target gene eng2a (with subsequent loss of other MHB patterning markers and MHB formation)28, we speculated that MO-mediated knockdown of grhl3 may similarly impact on MHB neural patterning. In order to determine whether this was the case, we investigated the expression of eng2a in grhl3-morphants. Somewhat surprisingly, we found no difference in the expression of eng2a (Fig. 2e,f) or expression of other MHB patterning markers, pax2a, wnt1 or her5 (Supplementary Figure S1) in grhl3 morphants at any key stage of development examined (9 hpf, 18 hpf and 24 hpf), indicating that although grhl3 function clearly regulates MHB morphogenesis, it does not impinge on neural patterning. These data are consistent with previous work showing that the processes of neural patterning and neural morphogenesis are mediated via distinct mechanisms37,38,39, and identify a novel specific regulator of MHB folding in the absence of aberrant neural patterning regulation.

grhl3 interacts with both spec1 and arhgef19 in MHB morphogenesis

Previous work from our groups had identified targets of the Grhl/grhl family involved in regulating both MHB morphogenesis in zebrafish (spec1)28 and cellular migration during epithelial wound-healing (ARHGEF19)5. To determine whether these downstream genes similarly co-operated with grhl3 in regulating MHB morphogenesis in zebrafish, we performed knockdown experiments, whereby sub-phenotypic dosages of MO:grhl3 were injected together with sub-phenotypic dosages of MOs targeting either spec1 (Fig. 2g–j) or arhgef19 (Fig. 2k–p). Consistent with previous experiments showing interaction between grhl2b and spec1 in MHB morphogenesis28, we found that sub-phenotypic knockdown of grhl3 and spec1 together also led to slight axial shortening and aberrant MHB folding (Fig. 2g,i); sub-phenotypic knockdown of either MO:grhl3 or MO:spec1 together with MO:control resulted in no visible phenotypic perturbations (Fig. 2h,j; Supplementary Table S2). Although the constriction of the MHB in the grhl3/spec1 double morphants appeared to initiate correctly, the bends did not fully form, possibly due to defective basal constriction38. Similarly, we found that sub-phenotypic knockdown of grhl3 and arhgef19 also led to mild axial shortening and aberrant MHB morphogenesis (Fig. 2k,m); sub-phenotypic knockdown of either MO:grhl3 or MO:arhgef19 together with MO:control resulted in no visible phenotypic perturbations (Fig. 2l,n; Supplementary Table S2).The grhl3/arhgef19 double morphants also presented with an open dorsal neural tube, highlighted by ISH for the MHB marker eng2a (Fig. 2o,p). As grhl3 is not expressed within the neural plate, rod or keel during embryogenesis12, but is expressed within the developing non-neural ectoderm/EVL (Supplementary Figure S2), this is a presumed secondary phenotype, most likely due to defects within the overlying formative epidermis. Furthermore, given the relatively small overlap between cells that would co-express both grhl3 and arhgef19 (within two discrete punctate regions of the EVL lateral to the MHB, and potentially in the EVL overlying the tip of the tail; Supplementary Figure S2), as well as grhl3 and spec1 (only within punctate regions of the EVL), significant differences in total transcript abundance by Q-RT-PCR analysis of both arhgef19 or spec1 in MO:grhl3 injected fish were not seen (Supplementary Figure S2). Taken together, these data indicate that down-regulation of grhl3 and grhl3-dependent signalling pathways leads to specific defects in neural morphogenesis, albeit not neural patterning.

MO-mediated knockdown of grhl3 signalling leads to differences in cell size and identity within the zebrafish EVL/periderm

In order to analyse the grhl3/arhgef19 phenotype in more detail, we performed confocal imaging of EVL cells at the level of the MHB in grhl3 morphants, grhl3/arhgef19 double morphants and control embryos at 24 hpf (Fig. 3a–f). In addition to confirming the MHB morphogenesis defects observed by low-power Brightfield microscopy, we extended these analyses to examine the EVL/periderm in greater detail. We characterised cellular morphology and cell size, cell fate, apoptosis, and marker gene expression (fibronectin, p63, ZO1 and E-Cadherin) to fully characterise EVL identity at 24 hpf in our grhl3-morphants. Morphometric quantitation of EVL/periderm cell size overlying the MHB showed that MO-mediated knockdown of either grhl3 alone, or co-knockdown of both grhl3 and arhgef19 led to a highly significant increase in cell size (Fig. 3g); knockdown of arhgef19 alone led to a smaller, yet still statistically significant increase in EVL/periderm cell size. These data indicate that grhl3-dependent pathways regulate EVL-developmental dynamics at the cellular level. In order to further investigate the characteristics of the EVL in MO:grhl3-injected fish, we performed immunohistochemical analysis of both actin (a marker of the cell-membrane of cells within the EVL) and fibronectin (a marker of the underlying basal keratinocytes) specifically within the outermost layer of the zebrafish EVL (Fig. 3h–m). As expected, actin was localised to the membrane of EVL cells in fish of both WT and grhl3-morphants. In WT fish, fibronectin is strongly expressed in basal keratinocyte cells at 24 hpf, but absent in the overlying EVL. MO-mediated knockdown of grhl3 led to ectopic expression of fibronectin within EVL cells at 24 hpf, as well as an expanded pool of fibronectin-positive cells several layers thick (Fig. 3n,o), demonstrating that these cells express markers of multiple cell types. However, MO:grhl3 EVL cells do not ectopically express the basal keratinocyte cell fate marker p63 (Supplementary Figure S3), indicating that these EVL cells have not lost periderm identity, or acquired basal keratinocyte cell identity. p63 is initially expressed in all epidermal progenitor cells at the 5 somite stage, but is later downregulated in EVL cells by 24 hpf19, indicating that grhl3 is not required for the downregulation of p63 in the EVL cells. Although we did detect a significant increase in apoptosis in MO:grhl3 injected embryos, this was restricted to the region of the EVL and closely apposed tissue (Supplementary Figure S3), indicating this is a direct result of grhl3-knockdown which is likely to contribute to the neural tube phenotype. No significant differences were detected in expression of ZO-1 at either 8 hpf (Supplementary Figure S4) or 24 hpf (Supplementary Figure S5), or epithelial integrity (cdh1 expression; Supplementary Figure S6). Taken together, our data show that MO-dependent inhibition of grhl3-signalling within the EVL leads to defects through production of fewer, larger cells, without concomitant disruption in cell fate, EVL integrity, cell death or establishment of junctional complexes. These data support the hypothesis that loss of grhl3 in EVL cells results in the expression of components found in other epithelial cells without a change in cell fate, suggesting that grhl3 functions to prevent EVL cells from expressing markers of other epithelial cell types.

Figure 3
figure 3

EVL cell size and identity is aberrant in MO:grhl3 and MO:arhgef19 injected zebrafish. (a–g) Rhodamine-phalloidin staining of F-actin in 24 hpf embryos injected with either MO:grhl3 + MO:arhgef19 MO:control (a,b) or MO:control (d,e) shows that periderm cells with abrogated grhl3/arhgef19 signalling are larger (boxed region in b,e). Morphometric quantitation of cell size using FiJi software (c,f) shows significant disparity in cell size in MO:grhl3 + MO: arhgef19 injected embryos (as well as singly-injected MO:grhl3 and MO: arhgef19 embryos), relative to controls (G). (h–o) The basal keratinocyte marker fibronectin (i) is ectopically expressed in EVL cells surrounding the MHB in MO:grhl3-injected embryos (merged image; j). Conversely, EVL cells (k) do not express fibronectin (l) in MO:control fish (merged image, m). Digital cross-section confirms that several dorso-ventral layers of fibronectin positive cells are present in MO:grhl3-injected fish (n), whereas fibronectin positive cells are distributed within a single, one-cell thick layer MO:control injected fish (o).

Full size image

MO-mediated knockdown of both grhl2b and grhl3 causes defects consistent with defective convergence-extension

To investigate the putative co-operative roles of grhl2b and grhl3 in early patterning and gastrulation, we performed co-knockdown experiments, utilising our previously published morpholinos targeting both orthologues12,28. Although our previous studies of grhl2b or grhl3 knockdown did not indicate that loss of either gene led to any appreciable defects in axial patterning, we found that co-injection of both MOs (each at phenotype-causing dosages) resulted in significant defects consistent with aberrant CE-mediated axial patterning (Fig. 4a,b; Supplementary Table S3). By quantitating these CE-like defects using the length to width ratio (LWR) measurement previously described for such analyses40, we found that these double morphants exhibited a ~40% decrease in LWR at 24 hpf (Fig. 4c), and furthermore, these fish presented with concomitant broadened somites and notochord (Fig. 4d,e), both classic hallmarks of the CE-mutant phenotype seen in models such as trilobite, knypek or silberblick 40; moreover, these double morphants also presented with significant defects in MHB morphogenesis (Fig. 4f,g). This phenotype was entirely consistent with the short, squat stature following Grhl3 deletion, and occasional cranial exencephaly at the MHB region previously observed in the Grhl3 −/− mouse mutant model9, and the CE-like phenotype common for animal models with impaired gastrulation movements during germinal layer formation. The CE-like defect was even more pronounced at 96 hpf (Fig. 4h,i), with fully-penetrant failure of posterior trunk extension in the most severely-affected double morphants. These data suggest that although MO-mediated knockdown of grhl2b or grhl3 in isolation does not affect CE-movements, loss of both these factors in fish appear to lead to severe impairments in CE-dependent migration post gastrulation.

Figure 4
figure 4

grhl2b/grhl3 double morphants present with phenotypes highly consistent with disrupted convergence-extension mediated migration, as well as MHB folding defects. (a–i) MO:grhl2b + MO:grhl3 injected double morphants (a) display a reduced axial length and decreased length:width ratio (LWR) relative to embryos injected with MO:control (b). These differences were quantitated (c), showing significantly decreased LWR in MO:grhl2b + MO:grhl3 injected double morphants (p < 0.0001 by Student’s T-test). (d–g) Dorsal views of MO:grhl2b + MO:grhl3 double morphants (d,f) relative to MO:control embryos (e,g) at 24 hpf highlighting the thicker trunk, broadened notochord (n) and splayed somites (s) characteristic of embryos with disrupted convergence-extension; as well as disruption of MHB morphogenesis (compare f with g). (h–i) MO:grhl2b + MO:grhl3 double morphants (f) relative to MO:control embryos at 96 hpf, highlighting severely reduced axial extension.

Full size image

Over-expression of grhl3 causes defects in axis extension and also leads to axial duplication

We complemented our loss-of-function analyses with an analysis of the consequences on embryogenesis of over-expression of grhl2b or grhl3. To further support the strong evolutionary conservation of these factors, we also over-expressed full-length mRNAs of the murine orthologues, Grhl2 and Grhl3 respectively, with identical phenotypic consequences (Supplementary Table S4; Supplementary Table S5). Our micro-injection experiments over-expressing full length grhl3/Grhl3 yielded three consistent, yet partially disparate phenotypes (dorsalisation, failure to extend axis and axial duplication; Fig. 5a–c). The first resulted in fish with normally developed anterior structures, including the MHB, but a severely shortened trunk and tail, which remained wrapped around the yolk rather than elongating, a phenotype we termed “failure to extend axis” (FTEA; Fig. 5a). The second phenotype was reminiscent of dorsalisation, with a severe reduction or absence of posterior structures (Fig. 5b). The third phenotype resulted in duplications of the trunk, reaching as far anteriorly as the MHB, and as far posteriorly as the mid-somite region (Fig. 5c), although interestingly, fish with either duplicated anterior (two heads) or posterior (two tails) regions were not seen. Molecular analyses on the axial duplication phenotype by ISH showed that both axes appeared to be patterned correctly (Fig. 5d–g), expressing markers of the posterior neural tube (shh), somites (MyoD), MHB (eng2a) and hindbrain (krox20), suggestive of either an axial duplication event or an exact symmetrical midline-bifurcation, comprising cells derived from all three germinal layers.

Figure 5
figure 5

Over-expression of both grhl3 and grhl2b results in defective patterning and cellular migration. (a–g) Over-expression of grhl3 results in three disparate phenotypes, notably a failure to extend axis (FTEA; a), dorsalisation (b), or mid-axial duplication (c). ISH showing duplication or bifurcation of the neural tube (shh; d), somites (myoD; E), MHB (eng2a, f) and hindbrain rhombomeres (krox20; g) (h–m) Over-expression of grhl2b mRNA (lateral view H; ventral view, i) results in cyclopia, a defect not seen following injection of control (LacZ) mRNA (lateral view J; ventral view, k). This defect is likely due to impaired anterior neural tube extension in embryos over-expressing grhl2b mRNA, as shown by expression of the neural tube marker shh (arrow) by ISH (l–m).

Full size image

Over-expression of grhl2b causes defects in eye-field separation and cyclopia

Fate mapping of cells involved in normal bilateral eye development indicate that the eye-field primordia undergoes significant rearrangement to accommodate the ventral diencephalon, putatively as a consequence of physical anterior migration of the neural tube, or through induction of localised diencephalic tissue to bisect the eye-field41. Over-expression of grhl2b/Grhl2 yielded a striking phenotype of defective forebrain formation with concomitant cyclopia (Fig. 5h–k). This phenotype is reminiscent of other models of cyclopia, particularly in fish with defective PCP/CE signalling40,42 which is caused by the aforementioned defective anterior migration of the neural tube. To investigate whether cyclopia in our model was a result of failed neural tube migration, we performed ISH for the neural tube marker, shh (Fig. 5l,m). We found that shh expression in the grhl2b-overexpressing fish did not extend as far anteriorly as in mRNA:LacZ injected controls, and the diencephalic anlagen demonstrated a bunched and clustered appearance, highly consistent with impaired migration, rather than lack of neural tube formation, resulting in subsequent failure of eye-field separation. These data indicate that tight homeostatic regulation of grhl2b is essential for anterior migration of the zebrafish neural tube, and suggesting that grhl2b may negatively regulate, or suppress, a factor critical for neural tube migratory dynamics.

Discussion

We present here several novel phenotypes associated with disrupted grhl-family signalling in the zebrafish, Danio rerio, following both MO-mediated down-regulation, and over-expression through injection of CpG-capped full-length mRNA. Both loss or gain of function models showed specific effects on cellular migration, cellular dynamics, morphogenesis, and axial patterning/extension, thereby greatly expanding upon the known roles of the Grhl family during vertebrate embryonic development.

When grhl3 is down-regulated, either alone or together with downstream targets spec1 or arhgef19, the folding (but critically not the neural induction or subsequent patterning) of the midbrain-hindbrain boundary (MHB) is disrupted. These data are consistent both with previous work from our laboratory showing disrupted neural tube closure in mice lacking Grhl2 or Grhl3 9,10, and aberrant MHB folding following loss of grhl2b 28, as well as the newly-emerging paradigm which separates the processes of neural tube morphogenesis from induction and patterning37,38,39. Although the grhl3/arhgef19 phenotype is consistent with the open posterior neural tube and occasional exencephaly of Grhl3 loss in mouse9, the mechanisms of neurulation in fish (evagination of a lumen from a solid rod of cells) differ from those in mouse, suggesting that this common phenotype is either “fated” during the convergence-extension movements of the pre-tubular neural plate, which is common to both species43, or more likely, is caused through distinct mechanisms.

In the context of neural tube closure, Grhl3 in mouse is expressed within the non-neural ectoderm (NNE) immediately abutting the neural plate border preceding neurulation9, and this NNE is critical for neurulation44. It is likely that loss of Grhl3 in this region in mouse leads to either mechanical, signalling, extra-cellular matrix (ECM) deposition or polarity defects in the NNE which impinge on the subsequent co-ordinated migration, apposition and closure of the neural tube. The role of the ECM between the neural plate and its surrounding tissue layers during neural tube morphogenesis is well-established. Knockdown of the ECM components, laminin and fibronectin disrupt neural tube morphogenesis in zebrafish by preventing tissue apposition and coupling movements between the neural plate and mesodermal tissues45. The expression of grhl3 in the developing zebrafish is more abundant than in mouse, present throughout the entire EVL/periderm overlying the neural plate (but importantly, not within the neural plate itself) up to and including neural tube closure, ventricle inflation and morphogenesis12,46. Given that our results identified an altered expression pattern of the ECM component fibronectin in grhl3-morphants, it is possible that changes in ECM composition (together with the concomitant increase in apoptosis we observed) in MO:grhl3 embryos cause a disruption and subsequent remodelling of the tissue coupling between the EVL and neural plate, resulting in a loss of anchoring, and may be responsible for the MHB defects observed. Consistent with our data showing aberrant EVL/periderm cell size and morphology in grhl3 and grhl3/arhgef19-morphants, and early rupture of the grhl3 −/−(+14bp) embryos, it is likely that the EVL in these fish is significantly weakened. As the grh/Grhl/grhl family are crucial regulators of both cell division and survival6,12,28,47, the increase in cell size is likely due to decreased proliferation or increased apoptosis of the EVL in early development, resulting in fewer cells forming the overlying EVL/periderm barrier during neurulation. This suggests that this tissue may possess insufficient tensile strength to efficiently contain the expanding neural tube during evagination – hence, a “rupture” of the neural tube at its dorsal-most point, resulting in an open-neural tube. This phenotype is reminiscent of the zebrafish mutant bloody-fingers; blf 48, whereby a defective overlying epidermal layer results in delamination of dorsal neural tube cells and a spina bifida like phenotype. We have also recently noted a similar impact of the NNE and/or ECM-remodelling defects in Grhl3 −/− mice at the level of the cranial neural tube, and in particular the influence of the overlying epidermis on the development of the skull; namely, these mice present with an increased apposition of the frontal and parietal bones during development in a phenotype analogous to the human condition craniosynostosis49.

Although our novel grhl3 −/−(+14bp) fish line does not allow analysis of tissue patterning and development, due to the very early embryonic lethality, it is likely to be an excellent model of the maternal-zygotic transition (MZT). Previous work in Drosophila indicates that grh regulates activation of the zygotic genome50, whereas unbiased large-scale molecular profiling of MZT-stage zebrafish embryos indicates that grhl3 is the fourth most-highly upregulated transcript of the earliest zygotically-transcribed genes51. Interestingly, blf is the sixth-most highly upregulated transcript, suggesting there may be commonalities between genetic regulation during the MZT, EVL development, and epidermal/neurulation defects.

Both our MO data and our striking mutant phenotype raise the question as to whether other grhl-orthologues may be functionally compensating for grhl3 in early development. Animal models in which Grhl1 is deleted or downregulated do not present with a neurulation phenotype17,18,19, moreover we show here that grhl1 is not upregulated to compensate for grhl3 loss (Fig. 1o). Although little is known about grhl2a, we show here that grhl2a expression is also largely unchanged (Fig. 1o). We have previously shown that the earliest expression of grhl2b is in the leading edge of the neural plate, and/or the polster, from approximately 8 hpf28; moreover both our previous work, plus another study33, do not show any role for grhl2b in epidermal/EVL development. Our data therefore allow us to conclude that loss of grhl3 in our mutant is not significantly functionally compensated for by up-regulation of other grhl-family members. Together, these finding suggest that grhl3 and its downstream target genes regulate neural tube morphogenesis as a secondary consequence of defective EVL/periderm/epidermis development and maintenance.

Grhl3 interacts genetically with mammalian Planar Cell Polarity (PCP) genes5, and correct planar cell-polarity establishment is required for subsequent cellular migration, especially during convergence-extension (CE) mediated cell movements during axial patterning and morphogenesis. Given this, an interesting observation is the fact that although grhl3-morphants are often shorter than controls, they do not display an obvious CE-defect, unlike morphants in which both grhl2b and grhl3 have been concomitantly down-regulated. Although this may appear to be at odds with mouse Grhl3-knockouts, which are noticeably shorter and broader than WT-littermates9, as MO-mediated knockdown in fish does not lead to full inhibition of gene function, one may surmise that residual grhl3 gene-function in zebrafish embryos is sufficient to preserve the normal polarity establishment and signalling required for correct CE migration. Unfortunately, CRISPR-mediated deletion of grhl3 in the grhl3 −/−(+14bp) embryos does not allow further examination of gastrulation phenotypes due to premature lethality during epiboly, which perhaps suggests a lower degree of redundancy between grhl-orthologues in zebrafish than in mouse. The grhl2b/grhl3 double morphants however display the classic zebrafish CE phenotype of a shortened trunk, decreased length-to-width ratio, broadened notochord and thickened somites, and further strongly suggest that one role of grhl2b and grhl3 interaction is to regulate convergence-extension in axial development. Furthermore, genetic co-operativity between these genes has previously been described in mammalian models - insertion of Grhl2 into the Grhl3 genomic locus led to rescue of a proportion of neural tube defects3. The recent finding that mutations in human Grhl3 are associated with spina bifida52, in itself a defect of neurulation possibly due to incorrect convergence-extension, suggests that zebrafish grhl2b/grhl3 double-morphants may ultimately prove to be a useful model of human CE-like defects.

Our over-expression data show further novel gastrulation phenotypes associated with disrupted Grhl-gene signalling. Increased grhl2b expression leads to failed anterior neural tube migration and cyclopia. These defects are reminiscent both of over-expression of the homeobox transcription factor, and direct grhl2 target gene engrailed 2a, which suppresses formation of anterior structures, including eyes and forebrain53, and over-expression of slit2, the ligand for roundabout (robo) which also results in failed neural tube migration and cyclopia54. Our Q-RT-PCR analyses, however showed that neither gene was significantly differentially regulated in grhl2b over-expressing morphants (data not shown). Other pathways which may be involved include nodal, shh, zic2 55 and the nodal-related protein squint 56, and future work will focus on analysing these, and other mechanisms of failed anterior neural tube migration.

Three further novel phenotypes were seen following over-expression of grhl3, although interestingly, no specific defects in MHB morphogenesis or neurulation were detected. These data support the hypothesis of a weakened EVL overlying the periderm; whereas loss of grhl3 leads to loss of EVL integrity, gain of grhl3 does not lead to any further increase in EVL tensile strength or integrity. All of dorsalisation, failure to extend axis (FTEA) and midline bifurcation are highly consistent with defective cellular migration. The FTEA phenotype in particular suggests a milder dorsalisation phenotype, as although the anterior structures are patterned correctly, ventral structures are under-represented, in this case due to an extension defect. The mechanism for these phenotypes is currently unknown, although an attractive possibility is the involvement of a newly-characterised Grhl-family target gene, E-Cadherin 27. Severely affected zebrafish mutants lacking e-cadherin exhibit a highly-similar FTEA phenotype57 to our grhl3 over-expressing fish, and it is a possibility that defective ECM scaffolding through repressed e-cadherin may contribute to failed cellular migration and subsequent tail extension. Furthermore, recent studies indicate that in addition to activating E-Cadherin, or serving to alter E-Cadherin localisation58, Grhl3 may also act as an E-cadherin suppressor59, indicative of a dichotomous relationship between these two factors, which together may regulate convergence-extension mediated movements during embryogenesis.

The third phenotype, that of midline bifurcation in our model is of the rarer variety that doesn’t involve either head duplication, (e.g. in Xenopus over-expressing wnt)60 or tail duplication (e.g. embryos over-expressing activin)60, but rather a specific duplication which occurs in the trunk. Interestingly, similar axial duplication is also seen in embryos overexpressing fgf8 61, itself a known upstream regulator of the grh/grhl family28. These data add further support to the theory that Fgf-signalling may be regulate the phosphorylation and resultant function of the grh/Grhl genes across broad developmental contexts, having been previously experimentally determined for grh in Drosophila tracheal branching62 and regulation of MHB morphogenesis and neuroblast survival in conjunction with grhl2b in zebrafish28.

Taken together, our study characterises numerous novel vertebrate phenotypes due to disrupted grhl-signalling, and highlights the remarkable diversity of phenotypic processes regulated by this gene family. Future work will focus on further elucidating the genetic pathways (both upstream and downstream of the grhl-family) by which the vertebrate grhl family regulates these myriad developmental regulatory effects.

Data Availability Statement

All data (including raw data and supplementary data), as well as the grhl3 −/−(+14bp) fish line will be made available upon request.