|Home | About | Journals | Submit | Contact Us | Français|
The Forkhead Box H1 (FoxH1) protein is a co-transcription factor recruited by phosphorylated Smad2 downstream of several TGFβs, including Nodal-related proteins. We have reassessed the function of zebrafish FoxH1 using antisense morpholino oligonucleotides (MOs). MOs targeting translation of foxH1 disrupt embryonic epiboly movements during gastrulation and cause death on the first day of development. The FoxH1 morphant phenotype is much more severe than that of zebrafish carrying foxh1/schmalspur (sur) DNA-binding domain mutations, FoxH1 splice-blocking morphants or other Nodal pathway mutants, and it cannot be altered by concomitant perturbations in Nodal signaling. Apart from disrupting epiboly, FoxH1 MO treatment disrupts convergence and internalization movements. Late gastrula-stage FoxH1 morphants exhibit delayed mesoderm and endoderm marker gene expression and failed patterning of the central nervous system. Probing FoxH1 morphant RNA by microarray, we identified a cohort of five keratin genes – cyt1, cyt2, krt4, krt8 and krt18 - that are normally transcribed in the embryo’s enveloping layer (EVL) and which have significantly reduced expression in FoxH1-depleted embryos. Simultaneously disrupting these keratins with a mixture of MOs reproduces the FoxH1 morphant phenotype. Our studies thus point to an essential role for maternal FoxH1 and downstream keratins during gastrulation that is epistatic to Nodal signaling.
During gastrulation, radially symmetric blastula-stage embryos comprised mostly of pluripotent cells are transformed into embryos with an emerging vertebrate body plan and various committed cell types. This transformation involves an elaborate program of cellular induction, differentiation and movement. A growing number of molecules are implicated in the control of gastrulation. Some of these molecules’ functions remain elusive, due to disparate experimental results. One example is the Forkhead transcription factor FoxH1, previously known as Fast1, for which distinct loss-of-function phenotypes have been reported (Hoodless et al., 2001; Howell et al., 2002; Kofron et al., 2004; Pogoda et al., 2000; Sirotkin et al., 2000; Watanabe and Whitman, 1999; Yamamoto et al., 2001).
FoxH1 is recruited in response to the Nodal, Activin and Vg1 subclasses of the TGFβ-like superfamily of secreted ligands (Shi and Massague, 2003; Whitman, 2001). Receptor activation by these ligands stimulates the formation of phospho-Smad2/Smad4 complexes that are bound by FoxH1. FoxH1/Smad complexes localize to the nucleus where they drive transcription of genes bearing cis-regulatory FoxH1 and Smad DNA-binding motifs (Hart et al., 2005; Norris et al., 2002; Osada et al., 2000; Saijoh et al., 1999; Watanabe et al., 2002).
The FoxH1 gene is expressed throughout gastrulation in mouse, fish and frogs, and maternally in fish and frogs (Chen et al., 1996; Pogoda et al., 2000; Sirotkin et al., 2000; Weisberg et al., 1998). Various studies indicate that FoxH1 is active in the early patterning of vertebrate embryos. For instance, essential FoxH1 and Smad binding sites are upstream of genes with key roles in germ layer formation and axis establishment, such as nodal, lefty2, lim1 and mix1l (Hart et al., 2005; Norris et al., 2002; Osada et al., 2000; Saijoh et al., 1999; Watanabe et al., 2002). There is also evidence for Smad-independent roles of FoxH1, seen in its co-regulation of the WNT/β-catenin target Xnr3 and in its repression of the androgen receptor via direct binding (Chen et al., 2005; Kofron et al., 2004).
To assess functional requirements for FoxH1 in whole organism development, a number of in vivo perturbations have been performed in frogs, fish and mice, yielding mixed results (Hoodless et al., 2001; Howell et al., 2002; Kofron et al., 2004; Pogoda et al., 2000; Sirotkin et al., 2000; Watanabe and Whitman, 1999; Yamamoto et al., 2001). In some instances, loss of FoxH1 leads to failures in gastrulation movements and/or developmental arrest during gastrulation stages. This is seen in the most severe classes of mouse FoxH1 knockout embryos, which fail to form a primitive streak or an anterior-posterior axis (Hoodless et al., 2001; Yamamoto et al., 2001). Gastrulation defects are also seen in Xenopus embryos injected with FoxH1 antisense oligonucleotides or a FoxH1 antibody (Kofron et al., 2004; Watanabe and Whitman, 1999). At the other end of the spectrum, abnormalities of the mildest classes of murine FoxH1 knockout embryos are restricted to midline deficits, with embryos surviving well beyond gastrulation (Yamamoto et al., 2001). A similar midline deficit is seen in zebrafish schmalspur (sur) mutants, which carry homozygous point mutations in FoxH1’s DNA binding domain (Pogoda et al., 2000; Sirotkin et al., 2000). Still other models have intermediate phenotypes. For example, frogs or fish injected with a repressor form of FoxH1 show reductions in mesoderm and endoderm, and a third group of mouse knockout embryos have defects in the anterior primitive streak, leading to anterior body truncations (Hoodless et al., 2001; Pogoda et al., 2000; Watanabe and Whitman, 1999; Yamamoto et al., 2001).
There are four prominent cell movements during the blastula and gastrula stages of zebrafish development: epiboly, internalization, convergence and extension (Solnica-Krezel, 2005). Epiboly describes the spreading of embryonic cells around the yolk towards the vegetal pole. Internalization describes the ingression of mesoderm and endoderm precursor cells to form an inner cell layer. Convergence is the dorsal migration of outer and inner cells towards the midline. Extension is the elongation of the axis, particularly along the dorsal midline.
It is noteworthy that gastrulation movement defects are common in mouse and frog FoxH1 loss-of-function models, whereas zebrafish sur embryos display only midline defects, indicating relatively normal gastrulation movements. We have addressed this discrepancy by reassessing the role of FoxH1 in zebrafish, using antisense morpholino oligonucleotides (MOs) to disrupt the entire FoxH1 protein. We found that blocking translation of foxH1 mRNA does indeed interfere with all four gastrulation movements, leading to severe dysmorphology and early death. Our molecular marker and microarray analysis shows a delayed pattern of mesoderm marker expression, a loss of transcripts specific to the endoderm and regionally-patterned CNS and a cohort of five down-regulated keratin genes, which are normally expressed in the EVL, an epithelial cell layer that surrounds the developing embryo. Simultaneous disruption of these keratins in wild type (WT) embryos produces a phenocopy of the FoxH1 morphant phenotype, indicating that the down-regulation of multiple keratins underlies the FoxH1 morphant defects.
Adult zebrafish were used for embryo production by natural matings. Embryos were staged according to Kimmel (Kimmel et al., 1995). Morpholinos, synthesized capped mRNA, plasmid DNA and proteins were dissolved into injection buffer (5 mM HEPES, 200 mM KCl, 1 mg/ml phenol red; pH7.0), and injected into the yolk of one- to four-cell-stage embryos. For histology, embryos were fixed in 4% formaldehyde when controls had reached the 70%-epiboly stage, then paraffin-embedded and sectioned at 5 micrometers per section (American Histolabs, Bethesda).
All morpholinos were from Gene Tools, LLC (Philomath, OR). The sequence for the foxh1 splice disrupting MO was 5’ TACTTAACCCTACCTCTGATAAAGT 3’, which targets the exon 1 splice donor signal. The control MO (5’TAGTTAAGCCTAGCTCTCATAAACT 3’) was originally designed as a 5 base mismatch for the foxh1 splice disrupting MO. It has no significant targets in the zebrafish genome (BLAST search of Ensembl’s Danio rerio assembly 40) and was used as the control MO throughout this study. The sequences for the two non-overlapping foxH1 translation blocking MOs were: FoxH1 MO1: 5’ TGCTTTGTCATGCTGATGTAGTGGG 3’ and FoxH1 MO2: 5’ GGAGGTGGAAGGTATGGTCGCTCCT 3’. The sequences for MOs targeting translation of keratins are as follows:
Time lapse images were obtained at low magnification using OpenLab Software and a high-resolution Jenoptix ProgRes C-14 digital camera, with frames captured every 10 minutes. For the time-lapse series in Fig. 1 and Fig. S1, embryos were dechorionated and soft-mounted in agar/agarose, as has been described (Karlstrom and Kane, 1996). For the series in Fig. 6, embryos were kept in their chorions and placed into 2.2 mm wide × 2.5 mm deep conical wells formed into in a petri dish containing 2% agarose/0.3X Danieau’s. These cones were formed from Sylgard 184 silicone elastomer casts of a custom-milled aluminum mold custom-fabricated at the NIH Mechanical Instrumentation Design and Fabrication branch, and allow for minimally invasive time-lapse monitoring of an ordered array of 56 specimens. Temperature was maintained at 28 °C by placing embryos underneath a Leica MATS thermocontrol heating stage system on a Leica MZ16 or MZ12.5 stereomicroscope.
To eliminate binding to FoxH1 MO1, wild type foxH1 mRNA from plasmid pCS2+/FoxH1 was modified (Sirotkin et al., 2000). This was done by PCR, using T3 as the downstream primer and the following upstream primer designed to introduce two silent mutations: 5’ ATGACGAAACATTGGGGGGGTCCAGGC 3’ (start-codon ATG, italic, silent mutations, bold). The amplicon was cloned into the pGEM vector (Promega, Milwaukee, WI) downstream of the SP6 promoter. Capped mRNA synthesis was done with the eMESSAGE eMACHINE RNA Transcription Kit (Ambion Inc., Austin, TX)
RNA probes were synthesized as previously described except template DNA was generated by PCR (Divjak et al., 2002; Thisse and Thisse, 1998). Whole-mount in situ hybridizations were performed as described previously (Thisse and Thisse, 1998). For stages later than 40% epiboly, more severely affected FoxH1 morphants tended to disintegrate during processing, limiting our documentation to less severely affected specimens.
Plasmids pGL3-3ARE and pRL-CMV (w/w 10:1, 25 pg/embryo) were co-injected with 8 ng of the indicated MO and 10 pg of squint mRNA into WT embryos. Pools of six embryos were collected in triplicate at dome stage and lysed in 20 µl of buffer for each data point. Luciferase activity was measured by the Dual-luciferase reporter system from Promega. Firefly luciferase activity was normalized to Renilla luciferase activity. Average values and standard deviations of triplicate experiments are shown.
To measure internalization, recombinant His-tagged Kaede protein (Ando et al., 2002) was expressed in E. coli and sequentially purified by Ni-NTA agarose affinity, sizing and hydroxyapatite columns (ProteinOne Inc., College Park, MD) and co-injected with MOs into WT embryos at a dose of 4 ng. Injected embryos were incubated at 28°C until the 40%-epiboly stage, dechorionated and mounted laterally in 3% methyl cellulose/ 0.3X Danieau’s buffer (Nasevicius and Ekker, 2000). Two small groups of cells, one along the margin and one four tiers above the margin, were labeled red by 405 nm laser-photoactivation of Kaede protein (optimal photoconversion occurs at 385 nm) with a confocal microscope (LSM 510 Meta Carl Zeiss, Thornwood, NY).
To assess convergence, photoactivatable green fluorescent protein (PA-GFP) (Patterson and Lippincott-Schwartz, 2002) was co-injected with MOs into embryos from gsc-GFP transgenic fish at a dose of 10 ng. Injected embryos were incubated to the shield stage, dechorionated, and mounted animal pole upwards in 3% methylcellulose/0.3X Danieau’s buffer. Two small groups of lateral cells, one on each side of the embryo, were labeled green by 405 nm laser-photoactivation of PA-GFP (optimal photoactivation occurs at 413 nm) on the confocal microscope. In both assays, labeled embryos were incubated in a moistened chamber at 28°C between documented time points.
The 33K Zebrafish oligo array consists of three oligo sets: Compugen (with 16,512 60mers), MWG (with 14,240 50mers) and Operon (with 3,479 70mers). The set includes 170 positive- (known house keeping genes) and 244 negative control oligos (Random sequences) to control the homogeneity and specificity of hybridization. The 5’ amine modified oligos were resuspended in 3xSSC (450 mM NaCl, 45 mM sodium citrate, pH 7.5) and printed on Epoxy slides from Corning Life Sciences.
Three pools of embryos were each divided in two and injected with FoxH1 MO1 and control MO to generate six samples. Embryos were incubated until the 40% epiboly stage and dechorionated with pronase (4 mg/ml). RNA was extracted in Trizol (Invitrogen Inc., Carlsbad, CA) and further purified using a Qiagen RNAeasy Mini column (Qiagen Inc., Santa Clarita, CA). 5 µg of total RNA from each FoxH1 morphant sample was reverse transcribed in the presence of normal and AminoAllyl –modified nucleotides (5.3 mM aa-dUTP, 10 mM dATP, dGTP and dCTP, 5 mM dTTP). The resulting AminoAllyl dUTP cDNA was coupled with Cy5 or Cy3 to generate 3 probes: two Cy5-FoxH1 morphant and one Cy3-FoxH1 morphant. Similarly, 3 probes were generated from control MO samples: two Cy3- control MO and one Cy5- control MO. The experimental-Cy5 and control-Cy3 samples and the one dye swap were pooled pairwise and hybridized overnight on three microarrays. More details are available at: http://magic2.nhgri.nih.gov/mcore/maservice.shtml
Microarray slides were washed and scanned with a confocal laser (Agilent Technologies, Palo Alto, CA) to measure fluorescence intensities and assign quality values to each spot, and background intensities were subtracted, using DEARRAY software (www.scanalytics.com).
Using Avadis software, data points with average quality values below 0.8 were discarded. Remaining data points were log2 transformed and Lowess normalized. Log2 ratios for each biological replicate were calculated and subjected to Student’s t-test with Benjamini-Hochberg correction to generate p-values. Average log2 ratios across the biological replicates were also determined. Using R software, a power analysis was performed to ensure the data satisfied β > 80% and α < 5% (p < 0.05) (Wei et al., 2004). The microarray data is available at NCBI GEO via the following link: http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=dbuhxkcecoegari&acc=GSE8076 To identify classes of enriched GeneOntology (GO) terms, we selected the subset of genes with RefSeq IDs, and either mean intensity log2 ratios > 1 (2 fold enrichment – up regulated) and p<0.01 or mean intensity log2 ratios < −1 (2 fold enrichment – down regulated) and p<0.01, then loaded them along with the complete RefSeq set from the microarray into GeneSifter (http://www.genesifter.net/). Z scores for all available GO terms were determined (Giorgi et al., 2005) and GO clusters of more than two and a Z score >5 were considered for our analysis.
To determine the phenotype of zebrafish with disruptions of the entire FoxH1 protein, we used a splice-blocking MO to disrupt splicing from the exon1 donor site of zygotic foxh1 pre-mRNA and translation-blocking MOs to disrupt translation of maternal and zygotic foxh1 transcripts. Injecting 8 ng of the splice-blocking MO did not produce any midline defects, but did cause a predicted loss of zygotic foxH1 mRNA, as judged by RT-PCR (data not shown), and also led to a failure in heart looping (mesocardia) in 93% (N=88) of scored embryos, compared to a background mesocardia rate of 7.1% (N=70) in control MO-injected embryos (data not shown). Mesocardia has previously been reported for surty68b mutants (Bisgrove et al., 2000). We have observed a similar low penetrance of midline defects accompanied by highly penetrant mesocardia in zygotic surm768 and surty68b embryos (data not shown), and therefore conclude that the foxh1 splice-disrupted phenotype is equivalent to recessive sur phenotypes.
Injecting 8 ng per embryo of either of two foxh1 translation-blocking morpholinos (FoxH1 MO1 or FoxH1 MO2), by contrast, caused a severe phenotype characterized by (1) a developmental delay evident by 6 hours post-fertilization (hpf) (Fig. 1B, C), when control embryos are at shield stage (Fig. 1A), (2) a compromised ability for embryonic cells to epibolize past the embryonic equator (Fig. 1B’, C’), and (3) embryonic death around 14–17 hpf (Fig. 1B”, B’”, C”, C’”) when control embryos are at mid somitogenesis stages (Fig. 1A”, A’”). Thus, interference of maternal and zygotic foxh1 translation via MO injection causes earlier and more severe phenotypes in zebrafish than previously reported for sur alleles (Pogoda et al., 2000; Sirotkin et al., 2000).
We used several approaches to ensure that the translation-blocking MOs are FoxH1-specific. First, a control MO, whose sequence does not target any known zebrafish genes, was injected at the same dose (8 ng) as the FoxH1 MOs, and found to produce no phenotype (Fig. 1A–A’”). This indicates that general MO toxicity is not a concern at 8 ng doses.
To test whether the FoxH1 MOs actually target FoxH1, we co-injected FoxH1 MO1 or control MO along with (1) a plasmid (ARE-luc) that requires Smad/FoxH1 complex binding for coupled transcription/translation of luciferase, (2) mRNA encoding the Nodal-related protein, Squint (to stimulate phospho-Smad/FoxH1 complex formation) and (3) a consitutively expressed control plasmid encoding a luciferase variant that can be independently assayed (Huang et al., 1995; Osada et al., 2000). These embryos were raised for four hours at 28°C, then lysed and assayed for induced and control luciferase activity. This revealed that FoxH1 MO1 specifically decreases Smad/FoxH1 transcription/translation of luciferase (Fig. 1D). Thus, FoxH1 MO1 genuinely disrupts FoxH1 function.
Another critical control is to test whether two MOs targeting different sequences on the same gene can induce the same phenotype. This is done to rule out phenotypes that might arise from chance binding of a MO to a second gene. We therefore injected a second non-overlapping FoxH1 MO (FoxH1 MO2) and obtained very similar morphological perturbations and times of death (compare Fig. 1B–B’” with 1C–C’”).
As further control for the specificity of the FoxH1 MOs, we injected them at lower individual doses (4 ng instead of 8 ng). Each of the two MOs causes a milder phenotype at this lower dose, however co-injecting 4 ng each of the two MOs causes the same phenotype as individual 8 ng injections (data not shown). This ability to produce the same phenotype with lower individual FoxH1 MO doses, which are less likely to cause non-specific defects, further argues that the 8 ng MO phenotypes are FoxH1 specific. The remaining data in this study utilizes FoxH1 MO1.
To further explore the morphological defects of gastrulating embryos injected with FoxH1 MO1, we prepared hematoxylin and eosin-stained sections of 8 hpf FoxH1 morphants and controls, when controls were at 70% epiboly. A comparison of lateral sections through the thickest portions of the animal poles of FoxH1 MO1 morphant (Fig. 1F) and control embryos (Fig. 1E) reveals three notable differences. (1) Morphant embryos are thicker and their nuclei (arrows) rounder, although there are a similar number of cellular tiers in control and FoxH1 morphant embryos, indicating that morphant embryos are less compacted along the mediolateral axis. (2) The EVL (dotted line), which is the outermost cell layer, is more ruffled in morphants. (3) There are widespread gaps (arrowheads) between the eosin-rich yolk globules and the embryonic cells of morphant embryos. Thus, at mid-gastrulation FoxH1 morphant embryos display several defects in cellular stratification.
We examined how FoxH1 depletion affects gene expression, using whole-mount in situ hybridization for selected molecular markers. At 10 hpf, when control embryos were at the bud stage of development, FoxH1 morphant expression of the mesodermal markers no tail (ntl; Fig. 2B), protocadherin 8 (pcdh8; Fig. 2D) and snail1 (Fig. 2F) was restricted to the vegetal margin, resembling the late-blastula stage (5 hpf) expression pattern for those genes in WT embryos (see, e.g., Fig. 2c for a comparison of ntl) (Hammerschmidt and Nusslein-Volhard, 1993; Schulte-Merker et al., 1994; Yamamoto et al., 1998). A similar heterochronic expression pattern was seen for the axial mesoderm markers goosecoid (gsc; Fig. 2H) and sonic hedgehog (shh; Fig. 2J), which in bud stage (10 hpf) FoxH1 morphants are expressed in a small domain near the margin, similar to the expression of these genes in the organizer of late blastula-stage control embryos (see, e.g., Fig. 2d for a comparison of gsc)(Stachel et al., 1993; Strahle et al., 1996). The restricted expression of the preceding genes to cells near the margin suggests that the presumptive mesoderm and endoderm cells of FoxH1 morphants may have defects in internalization and/or migration towards the embryo’s animal pole. The absence in mid-gastrula stage FoxH1 morphants of bone morphogenetic protein 4 (bmp4)-positive mesoderm (Fig. 2L, arrowhead) or sox17-positive endoderm (Fig. 2N) within the dorsal axis also suggests an internalization/migration defect (Chen et al., 1997; Kunwar et al., 2003). We noted one exception to this trend: of twenty foxa2-stained FoxH1 morphants scored, seven displayed deep anterior staining as shown in Fig. 2P, suggesting that dorsal internalization can persist in FoxH1 morphants (Strahle et al., 1993). But this foxa2 exception is variable: of the remaining thirteen scored embryos, seven had staining that remained much closer to the margin and six had nearly undetectable staining, indicating that foxa2-positive cells are usually compromised as well (data not shown).
With respect to neurectoderm development, gastrulation-stage FoxH1 morphants exhibit robust expression for markers of broad territories, as seen for the posterior neural marker hoxb1b (Fig. 2R), the anterior neural marker otx2 (Fig. 2T) and the early ectoderm marker, gata2 (Fig. 2V) (Imai et al., 2001; Kramer et al., 2002; Vlachakis et al., 2000). By contrast, more regionalized patterning of the CNS is disrupted, seen in bud-stage FoxH1 morphants as a loss of pax6a transcripts (Fig. 2X), which normally mark the presumptive forebrain and hindbrain, in pax2.1 transcripts (Fig. 2Z), which normally mark the presumptive midbrain-hindbrain boundary and the pronephros, and krox20 transcripts (Fig. 2b), which normally mark the presumptive hindbrain rhombomeres 3 and 5 (Reim and Brand, 2006; Strahle et al., 1993; Strahle et al., 1996).
As noted, FoxH1 perturbations cause gastrulation movement defects in frogs and in a significant fraction of mouse mutants. Although zebrafish sur mutants show no substantial alterations in gastrulation movements, FoxH1 morphants have dramatic epiboly defects (compare Fig. 1A’ with 1B’ and 1C’). We went on to examine other gastrulation movements in FoxH1 morphants. To assess internalization, we co-injected embryos with either FoxH1 MO1 or control MO, together with Kaede protein, a coral-derived green fluorescent protein that fluoresces red after photocleavage by UV and near-UV light (Ando et al., 2002). These embryos were incubated until the 40% epiboly stage (~5 hpf) at which time we photolabeled two small groups of cells situated on or near the vegetal margin (Fig. 3A, 3B). The fate of these labeled cells was documented 95 minutes later. In control embryos, some of the labeled cells had clearly internalized and migrated towards the animal pole (Fig. 3A’, white brackets). Other labeled cells did not internalize and remained at the margin (Fig. 3A’, white arrowheads), which itself was displaced towards the vegetal pole as a result of continuing epiboly. In FoxH1 morphants, there was no movement of labeled cells, indicating an absence of internalization as well as a lack of epiboly within the experimental time frame (Fig. 3B’, white arrowheads). We obtained similar results from eleven independent trials, demonstrating a dramatic overall disruption of internalization in FoxH1 morphants. Because the dorsal-ventral position of labeled cells was random in these trials and because the assay time was limited to 95 minutes, it is possible that subtle instances of partial internalization were missed, for instance on the dorsal side where we frequently detect deep foxa2 staining at a substantial distance from the margin of FoxH1 morphants (Fig. 2P). These considerations notwithstanding, our internalization assay demonstrates a widespread disruption of internalization in FoxH1 morphant embryos.
To assess convergence in FoxH1 morphants, we used gsc-GFP transgenic zebrafish, which have green fluorescent protein (GFP) under the control of the goosecoid (gsc) promoter (Doitsidou et al., 2002). Gsc is first expressed in the dorsal gastrula organizer and GFP recapitulates this pattern in gsc-GFP embryos, allowing us to visualize the dorsal side of these embryos. We injected gsc-GFP embryos with either control MO or FoxH1 MO1 together with photoactivatable-GFP, a GFP variant that has weak fluorescence until photoactivated by violet light (Patterson and Lippincott-Schwartz, 2002). When injected embyos reached the shield stage, we photolabeled two groups of cells close to the margin situated 180 degrees apart and flanking the dorsal gsc-GFP-positive cells (Fig. 3C–C’ and 3D–D’). After 2-hours of further incubation, the labeled cells in control MO injected embryos had clearly converged ~40 degrees towards the dorsal midline (Fig. 3C”). By contrast, labeled cells in FoxH1 morphants showed no detectable convergence (Fig. 3D”). Similar results were obtained in six independent experiments. We did not directly assay extension movements, however this movement relies both on convergence to populate the midline and on epiboly to create space for elongation, so extension is necessarily compromised as well. In summary, our various analyses reveal a critical role for FoxH1 in all four gastrulation movements.
We wished to identify molecular alterations that might account for the defects in FoxH1 morphants. Our initial analysis of marker gene expression elucidated some of the dynamics of FoxH1 morphant differentiation, but the variety of genes examined was limited and we failed to identify genes with altered expression prior to the onset of morphological defects. To undertake a broader search for genes with altered expression in early FoxH1 morphants, we used microarrays to compare the late blastula (5 hpf) gene expression profile of FoxH1 MO1-injected embryos with that of control MO-injected embryos. This approach yielded 75 RefSeq genes with significantly reduced expression in FoxH1 morphants and 100 RefSeq genes with significantly elevated expression (P<0.01 and >two-fold change). To identify categories of co-regulated genes within these two sets, we used GeneSifter software.
We identified five over-represented categories of up-regulated genes: genes whose products are involved in cell division and/or regulation of translational initiation, genes whose products localize to chromosomes and/or the nuclear envelope and genes whose products have oxidoreductase activity. We also found three over-represented categories of down-regulated genes: genes encoding intermediate filament components, genes whose products are involved in cytoskeleton organization and biogenesis, and genes whose products are involved in mesoderm development. We cloned genes from these categories to confirm their regulation by in situ hybridization. Our selection was biased towards cytoskeletal proteins and proteins involved in the cell cycle, because these seemed the most likely candidates for causing the observed gastrulation movement defects and developmental delays in FoxH1 morphants.
In this way we confirmed the indicated down regulation of five cytoskeletal genes in FoxH1 morphants at 5 hpf: cytokeratin1 (cyt1), cytokeratin II (cyt2), keratin (krt) 18, krt8 and krt4 (previously called zf-k8), but not capzb, transgelin2 or tubulinα1 (Fig. 4A–F; I–N and data not shown). Strikingly, all of the downregulated cytoskeletal genes turned out to be keratins with expression patterns restricted to the most superficial layer of cells in the embryo: the EVL (Imboden et al., 1997; Sagerstrom et al., 2005; Thisse et al., 2001; Thisse and Thisse, 2004). EVL expression is seen, for instance, in a lateral view of cyt1 expression in a control embryo (Fig. 4A). The expression of four of these five keratin genes recovers by bud stage in FoxH1 morphants (Fig. 4W–Z), whereas cyt1 expression remains low (Fig. 4V). We also looked at the cyt1 expression levels in surty68b/ty68b embryos, and saw no difference from sibling controls (data not shown).
We further confirmed the indicated up regulation of two cell cycle genes: her5 and cdc14, but not cdc45-like, cyclinA1 or cyclinB2 (Fig. 4G–H, O–P and data not shown). To further investigate the possibility of cell cycle alterations in FoxH1 morphants, we looked at mitosis and cell death at bud stage, by anti phospho-histone 3 staining and TUNEL assays respectively, but saw no difference from controls (data not shown). In addition, we assessed the relative numbers of G0/G1, S and G2/M phase cells at various stages using propidium iodine staining and FACS. No differences between FoxH1 morphants and controls were seen during early and mid-gastrulation, but a characteristic increase in the G0/G1 cellular fraction at the expense of the G2/M cellular fraction seen in controls at the two-somite stage (11 hpf) was substantially depressed in FoxH1 morphants (Zamir et al., 1997). Considering that only two of the cell cycle genes we examined had validated changes in expression, and the only cell-cycle related phenotype we identified was well after the onset of the FoxH1 morphant phenotype, we instead focused our attention on the keratins.
To better understand the requirement for maternal FoxH1, we explored various ways of rescuing, perturbing or phencopying the FoxH1 morphant phenotype. We first tested whether the FoxH1 morphant phenotype could be rescued by co-injecting foxh1 mRNA. To remove the risk of the FoxH1 MO1 interfering with translation of exogenous foxh1 mRNA, or exogenous foxh1 mRNA diluting the effect of the MO, we excluded foxh1’s 5’ UTR and introduced two silent mutations, eliminating 16 of 25 complementary nucleotides. Co-injecting 150 pg of this foxh1 mRNA with 8 ng of FoxH1 MO1 fails to rescue any morphological defects, but it partially rescues the reduced expression of pax2.1 (Fig. 5A–C) and cyt1 (Fig. 5D–F), but not that of pax6.1 or krox20 (data not shown). The pax2.1 partial rescue is seen as an increased fraction of embryos displaying pax2.1-positive midbrain-hindbrain boundary precursor cells, while the cyt1 partial rescue is seen as a general increase in transcript levels. While complete rescue would have provided convincing proof of the specificity of the FoxH1 morphant phenotype, our inability to completely rescue does not necessarily reflect non-specificity; it may rather reflect an inability for non-regulated exogenous FoxH1 to compensate for the loss of regulated endogenous FoxH1. In support of this, all of our other controls (Fig. 1) have indicated specificity, the gastrulation defects we observe do not fall within the spectrum of typical off-target morpholino effects and difficulty rescuing Foxh1 loss of function has been reported in Xenopus (Ekker and Larson, 2001; Howell et al., 2002; Watanabe and Whitman, 1999). Although morphological defects in FoxH1 morphants injected with 8 ng of the FoxH1 MO1 cannot be rescued, a milder set of morphological defects resulting from injection of 4 ng of FoxH1 MO1can be rescued. At 24 hpf, embryos injected with 4 ng of FoxH1 MO1 display a dramatic shortening of the body axis and microcephaly and darkened head tissue, indicating necrosis (Fig. 5H–H’). Co-injection of 150 pg foxh1 mRNA dramatically reduces the number of embryos displaying this phenotype, instead producing embryos with longer body axes, and larger, less opaque heads (Fig. 5I–I’). It is of course possible that this lower-dose rescue represents a qualitative rather than a quantitative rescue.
FoxH1 is believed to drive transcription of the zebrafish Nodal-related genes and their antagonists, the lefties. We observed a characteristic heterochronic expression of the Nodal genes squint and cyclops in FoxH1 morphants (data not shown). To directly test whether the FoxH1 morphant phenotype reflects an imbalance in Nodal signaling, we co-injected either squint mRNA or lefty1 mRNA along with FoxH1 MO1, and found that neither was able to visibly alter the morphological phenotype, but these potent mRNAs did produce the expected gain of Nodal function (squint) and loss of Nodal function (lefty1) morphological alterations in sibling embryos in which control MO was co-injected (data not shown). Although squint, lefty1 and (previously) foxh1 mRNA each failed to alter the FoxH1 morphant phenotype, this does not reflect a general inability of FoxH1 morphants to efficiently translate ectopic genes; in our luciferase assays, for instance, embryos injected with either the FoxH1 MO or the control MO produced similar levels of ectopic renilla luciferase (data not shown). We also injected FoxH1 MO1 into maternal-zygotic squinthi975/hi975 embryos, and the phenotype was indistinguishable from the FoxH1 morphant phenotype seen in WT embryos (data not shown). Taken together, our data point to an early requirement for zebrafish FoxH1 that is epistatic to Nodal signaling.
To address whether the FoxH1 morphant phenotype is attributable to the loss of keratin expression, we used two approaches. We attempted to rescue the high dose FoxH1 morphant phenotype by co-injecting cyt1 mRNA, looking for changes in morphology as well as changes in the expression of three molecular markers, but saw no evidence of rescue (data not shown). This may reflect a need for the additional down-regulated keratins or a need to deliver Cyt1 specifically to the EVL. Intriguingly, injection of 200 pg cyt1 mRNA often causes WT embryos to develop with one eye smaller than the other (16%, N=31), perhaps due to unequal distribution of mRNA, and a similar phenotype is seen in a similar fraction of WT embryos (23%, N=39) injected with 150 pg foxh1 mRNA (Fig. S1 A’, B’, C’).
We also asked whether removal of keratins from WT embryos could phenocopy the FoxH1 morphant phenotype. Injection of MOs targeting cyt1 translation (Fig. S1) or splicing (data not shown) caused epiboly defects and decreases in neural marker expression that were consistent with the FoxH1 morphant phenotype, but significantly milder (compare Fig. S1 E–E”” and F–F””; H and I; K and L).
Considering that multiple keratins are downregulated in FoxH1 morphants, we went on to ask whether simultaneous depletion of several keratins might produce a better phenocopy, and this is indeed the case. We injected a cocktail of MOs targeting cyt1, cyt2, k18, k8 and k4, comprising a total of 16 ng MO. The resulting phenotype was compared with that seen in embryos injected with 8 ng of FoxH1 MO1, 16 ng of control MO or no MO at all. Because 16 ng is a higher dose of MO than typically used, we compared control-injected and uninjected embryos, revealing that this dose had only mild effects on embryonic development (compare Fig. 6A’” and 6B’”). By contrast, the keratin MO cocktail produces a severe phenotype that is indistinguishable from the 8 ng FoxH1 morphant phenotype, with respect to the delay in epiboly (compare Fig. 6C and D; C’ and D’; C” and D”) and the time of death (compare Fig. 6C’” and D’”). A striking match between the keratin- and the FoxH1-depleted phenotypes is also seen in the reduced expression of the EVL marker cyt1 (compare Fig. 6I and M), the reduced expression of the regionalized neural marker pax2.1 (compare Fig. 6J and N), the heterochronic expression of the paraxial mesoderm marker snail1 (compare Fig. 6K and O) and the reduced expression of the endoderm and forerunner cell marker sox17 (compare Fig. 6L and P).
While sur phenotypes are comparable to the mildest classes of mouse FoxH1 perturbations, the FoxH1 morphant phenotype has more resemblance to Xenopus embryos injected with either FoxH1 MOs or anti-FoxH1 antibodies targeting the Smad-interaction domain (Howell et al., 2002; Watanabe and Whitman, 1999). These Xenopus FoxH1 loss-of-function models and our zebrafish model feature delays in epiboly (blastopore closure) as well as embryonic disintegration after gastrulation stages. Howell et al. specifically noted defects in the epidermal layer (homologous to the zebrafish EVL) and persistence of mesodermal marker expression in Xenopus FoxH1 morphants, as we have reported here for zebrafish FoxH1 morphants. These phenotypic similarities reconcile certain perceived differences between the roles of FoxH1 in zebrafish and Xenopus laevis, two organisms with relatively homologous modes of development.
Our FoxH1 splice-disrupting MO yielded phenotypes consistent with previously reported sur mutants, but when we blocked translation of maternal and zygotic foxh1 message with either of two non-overlapping MOs, we observed an earlier and more severe phenotype. The difference between our splicing and translation-blocking phenotypes suggests a critical role for maternal FoxH1, which in theory is uniquely targeted by the translation-blocking MOs. A role for maternal FoxH1 has already been demonstrated through comparisons of zygotic surm768 and maternal-zygotic surm768 (MZsurm768) mutants as well as comparisons between antisense-treated Xenopus embryos and oocytes (Howell et al., 2002; Kofron et al., 2004; Pogoda et al., 2000; Sirotkin et al., 2000), however the FoxH1 morphant phenotype we describe here is much more severe than the MZsurm768 phenotype (Sirotkin et al., 2000) or the MZsurty68b phenotype (Dirk Meyer, personal communication).
Why do FoxH1 MOs and FoxH1 mutations yield such distinct phenotypes in zebrafish? Our specificity controls argue that this difference is not due to non-specific MO toxicity. Another explanation could be that the two FoxH1 MOs share a common second target, such as a FoxH1 paralogue or another gene. This seems unlikely, since BLASTs of FoxH1 MO1 and FoxH1 MO2 against the Ensembl Danio rerio assembly 40 show that the only high affinity target (25 of 25 matching base pairs) for these MOs is the single zebrafish foxh1 gene, and there is no common gene among their lower-affinity targets (alignments down to 16 of 25 matching base pairs considered); but the zebrafish genome is not yet complete, so this remains a possibility.
Another possibility is that both sur alleles are hypomorphic, something already suggested by the relatively mild phenotypes of zebrafish sur mutants compared to mouse and Xenopus FoxH1 loss-of-function models. Each of the sur alleles has a point mutation in the DNA binding domain, leading to an R94H substitution in the surm768 allele and a K97N mutation in the surty68b allele. These appear to be bona fide loss of function mutations, with regard to DNA binding. This conclusion is based on the fact that the R94H substitution in the surm768 allele ablates the rescuing activity of a FoxH1 DNA binding construct (Pogoda et al., 2000). We also found that TGFβ failed to induce coupled transcription/translation of ARE-luciferase plasmids in NIH3T3 cells that were co-transfected with either surty68b or surm768 forms of foxh1, whereas co-transfection of WT foxh1 rendered NIH3T3 cells TGFβ-responsive (data not shown).
If the sur alleles are indeed hypomorphic, yet lack DNA binding activity, this would imply that their mutant protein products retain activity outside of the winged helix DNA-binding domain, which is located near the N terminus. In addition to the DNA binding domain, FoxH1 has two characterized downstream domains: one (the SIM domain) that binds both Smad2 and Smad3 and one (the FM domain) that uniquely binds activated Smad2 (Germain et al., 2000; Randall et al., 2002). With this in mind, perhaps the phenotypic differences between FoxH1 mutants and morphants point to an essential role for these or other more C-terminal domains of FoxH1. Further evidence for a functional role of FoxH1’s C-terminus comes from our examination of human FOXH1 variant alleles associated with cardiac laterality disorders, where we find that most debilitating mutations lie in the C terminus (manuscript in preparation). We are naturally interested in generating and analyzing zebrafish carrying foxh1 truncations or mutations downstream of the DNA binding domain, as this would provide a definitive test of our hypothesis, as well as validating our observations in general (Wienholds et al., 2003).
In zebrafish there is a striking difference between the FoxH1 morphant phenotype and even the most severe Nodal loss of function phenotypes, typified by squint;cyclops double mutants and maternal-zygotic one-eyed pinhead mutants (Feldman et al., 1998; Gritsman et al., 1999). Like FoxH1 morphants, these Nodal signaling mutants have an internalization defect, but unlike FoxH1 morphants, other gastrulation movements continue in the Nodal signaling mutants and they exit gastrulation sufficiently intact to survive for several days, despite severe mesoderm and endoderm deficits. In addition to these phenotypic differences, we have determined that the FoxH1 morphant phenotype is epistatic to alterations in Nodal signaling.
Despite these broad differences between the FoxH1 morphant phenotype and loss of Nodal phenotypes, we did detect certain molecular alterations in FoxH1 morphants that are consistent with disrupted Nodal signaling. The FoxH1 morphants displayed reductions in the expression of the midline mesoderm and endoderm marker foxa2 (Fig. 2P), and loss of expression of the late endoderm marker, sox17 (Fig. 2N). This is consistent with observations that these and other endoderm markers are downregulated in zebrafish Nodal pathway mutants, including sur embryos, and FoxH1 knockout mouse embryos (Feldman et al., 1998; Hoodless et al., 2001; Kunwar et al., 2003; Yamamoto et al., 2001). Our microarray studies also showed a significant reduction in lim1 (2.3-fold reduced, P=0.0008), a dorsal organizer gene that is directly regulated by FoxH1 downstream of Nodal signaling (Watanabe et al., 2002).
We have provided evidence that cyt1 and four other keratin genes act downstream of FoxH1 and that early loss of their expression mediates the FoxH1 morphant phenotype. This raises the question of how and why keratins are required for gastrulation movements. Given that three distinct gastrulation movements are disrupted in FoxH1 morphants, answering this question will likely require extensive experimentation. In the absence of more data, we will limit ourselves to suggesting one model: a proposal for how the loss of keratins could lead to epiboly defects. This model assumes that reduction of these keratin genes in the EVL compromises this tissue’s function. In support of this idea, we note that the outermost cell layer of FoxH1 morphants, where the EVL is situated, is substantially ruffled (Fig. 1F, dotted line).
We propose that a combination of the five keratins we have studied are necessary for EVL cells to provide a necessary scaffold or signal for the thinning and spreading of superficial deep layer (DEL) cells through a process known as radial intercalation that has been previously proposed to drive epiboly (Kane et al., 2005). That previous proposal was based on the coincidence of epiboly and radial intercalation defects in zebrafish half-baked (hab) mutants, which have mutations in e-cadherin, a gene normally expressed in peripheral DEL cells. Although we have not assayed radial intercalation in FoxH1 morphants, we do note a failure in the overall thinning of the DEL (Fig. 1F) and a tendency for FoxH1 morphant nuclei to remain rounded (Fig 1F, arrows). Thus, an essential interaction between Keratin-expressing EVL cells and E-cadherin-expressing DEL cells may be perturbed in FoxH1 morphants.
In conclusion, using FoxH1 translation-blocking MOs we find that maternal FoxH1 is essential for epiboly, most internalization and convergence, and these defects are likely mediated by a loss of multiple keratins. A comparison of this phenotype to other mutant and morphant models indicates that the loss of maternal FoxH1 is epistatic to Nodal-signaling and suggests that molecular interactions outside of FoxH1’s winged-helix domain are critical for early development.
(A–C) Similar effect of cyt1 and foxh1 mRNA over-expression on eye development. Wild type embryos were injected with 150 pg of foxh1 mRNA, 200 pg of cyt1 mRNA or 300 pg of gfp mRNA and scored after one day of development. No obvious defects arose from gfp mRNA injection. A reduced eye phenotype, as in B’, arose in 16% (N=31) of embryos injected with cyt1 mRNA and a similar reduced eye phenotype, as in C’, arose in 23% (N=39) of embryos injected with foxh1 mRNA. Apostrophes (’) indicate the front view for the same embryos. (D–F) Cyt1 MO injection partially phenocopies the FoxH1 morphant phenotype. Wild type embryos were injected with 8 ng of the indicated MOs, embryos were dechorionated and soft-mounted in agar/agarose, then simultaneously time lapse documented. Apostrophes (’) indicate different time points for the same embryos. Injection of both Cyt1 MO and FoxH1 MO caused epiboly delays at earlier stages (i.e. shield, 80% epiboly and 3-somite stages), and embryonic death later, at the 16-somite stage. However, the Cyt1 morphant phenotype was less severe than the FoxH1 morphant phenotype. (G–L) Reduced neurectodermal marker expression in both Cyt1 morphants and FoxH1 morphants. Wild type embryos were injected with 8 ng of the indicated MOs, then fixed at the bud stage (10 hpf) for whole mount in situ hybridization. G–I, lateral views, dorsal to the right. J–L, dorsal views, animal pole to the top. Signature markers were reduced in Cyt1 morphants, but reductions were more pronounced in FoxH1 morphants.
We thank Alexander Schier for the pCS2+/FoxH1 plasmid; Aaron Steiner and Daniel Kessler for the pGL3-3ARE and pRL-CMV plasmids; George Patterson for the PA-GFP protein; Dirk Meyer for the gsc-GFP transgenic fish. Thanks are also due to Jamie Brown for help with confocal microscopy; Martha Kirby for help with FACS analysis; Howard Metger for fabricating time-lapse molds and Blake Carrington for fish care and Igor Dawid, David Bodine, Erich Roessler and Shawn Burgess for comments on the manuscript. This research was supported by the Intramural Research Program of the National Human Genome Research Institute, National Institutes of Health.
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 maybe discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.