Zic2 mutation causes holoprosencephaly via disruption of NODAL signalling

Abstract

The ZIC2 transcription factor is one of the genes most commonly mutated in Holoprosencephaly (HPE) probands. Studies in cultured cell lines and mice have shown a loss of ZIC2 function is the pathogenic mechanism but the molecular details of this ZIC2 requirement remain elusive. HPE arises when signals that direct morphological and fate changes in the developing brain and facial primordia are not sent or received. One critical signal is sent from the prechordal plate (PrCP) which develops beneath the ventral forebrain. An intact NODAL signal transduction pathway and functional ZIC2 are both required for PrCP establishment. We now show that ZIC2 acts downstream of the NODAL signal during PrCP development. ZIC2 physically interacts with SMAD2 and SMAD3, the receptor activated proteins that control transcription in a NODAL dependent manner. Together SMAD3 and ZIC2 regulate FOXA2 transcription in cultured cells and Zic2 also controls the foxA2 expression during Xenopus development. Variant forms of the ZIC2 protein, associated with HPE in man or mouse, are deficient in their ability to influence SMAD-dependent transcription. These findings reveal a new mechanism of NODAL signal transduction in the mammalian node and provide the first molecular explanation of how ZIC2 loss-of-function precipitates HPE.

Introduction

The Zic proteins are a family of multifunctional gene expression regulators which act both as classical transcription factors and as co-factors to regulate gene expression (1). Together the genes are critical for a variety of developmental processes including gastrulation, axis formation, the development of the peripheral and central nervous systems and the axial and appendicular skeleton (2). Germ-line mutation of the Zic genes lead to a variety of congenital defects, with loss-of-function mutation of ZIC2/Zic2 associated with Holoprosencephaly (HPE; MIM 236100) in man and mouse, respectively (3,4). HPE is a common defect of forebrain development, with a frequency as high as 1:250 during human embryogenesis, in which the cerebral hemispheres and deeper cortical structures along the midline of the central nervous system fail to separate (5). Animal and human genetic studies indicate that, in order to prevent HPE, the ventral forebrain neurectoderm must transduce the SHH signal produced by the underlying region of the axial mesendoderm (AME) known as the pre-chordal plate (PrCP). This process can fail if the neurectoderm itself cannot transduce the SHH signal, if the SHH signal produced by the PrCP is faulty, or if the PrCP does not form and no SHH signal is produced. SHH secretion by the PrCP arises via a multi-step process (6). The first step of PrCP development occurs early in gastrulation, at 6.5 days post coitum (dpc), when cells delaminate from the epiblast, transit through the anterior primitive streak (APS), differentiate into AME and migrate to the anterior midline. These are the precursor cells of two tissues; the definitive endoderm (DE) and the PrCP (7). In the second step, the APS induces the formation of the patent node. Third, cells of the epiblast transit through the node of the mid gastrula (7.0 dpc), differentiate into AME and migrate to the anterior midline, taking up a position adjacent to, but posterior of, the cells of the PrCP. These cells form the anterior notochord (ANC) (7). Finally, in step four, the ANC produces a signal that is crucial for both the maintenance of the PrCP and for the onset of SHH secretion from the PrCP (4,8). Previous analysis of the embryonic phenotype caused by a severe loss-of-function allele of murine Zic2 (kumba; Ku) demonstrates that Zic2 is critical for the third step of this process, since in these embryos PrCP precursor cells arise and migrate normally (4) and the node is induced (9), but insufficient cells differentiate into ANC tissues and PrCP development arrests (4). The expression of all genes so far examined that are normally expressed in the mid-gastrula node is depleted in the Zic2 kumba mutants (4,9) and Zic2 may function at the top of the ANC differentiation cascade.

This functional analysis is consistent with the node of the 7.0 dpc embryo (the structure that produces the ANC) being the only unique site of Zic2 gene expression at this stage of development. Other closely related Zic genes (Zic3 and Zic5) are co-expressed with Zic2 in all other areas of the gastrula and likely compensate for Zic2 loss-of-function in these cells (10,11). The molecular mechanism by which Zic2 expression in the node controls ANC differentiation and thus HPE aetiology is unknown. It is, however, striking that Foxa2 expression in the patent node (but not APS) is downregulated in the Zic2 kumba mutant (4), since the FOXA2 transcription factor is known to control Shh expression (12) Foxa2 expression, in turn, can be stimulated by SHH signalling (13). It is therefore possible that ZIC2 indirectly controls Shh expression in the AME by controlling Foxa2 transcription in the node and initiating the Foxa2/Shh auto-induction loop.

The analysis of mouse mutant phenotypes suggests the early steps of AME development, including Foxa2 expression, are under the control of the NODAL signal that emanates from the posterior epiblast during gastrulation (14). An extensive series of murine alleles in Nodal itself or components of the signal transduction pathway demonstrate that NODAL signalling is required for for the formation and function of the anterior primitive streak (APS). A small decrease in NODAL signalling activity prevents differentiation of the DE and PrCP precursors via transit through the APS (15,16) and manifests as moderate anterior truncation. Further loss also interferes with node induction by the APS and results in severe anterior truncation, absent AME and somite fusion across the midline (17). Likewise, loss of Foxa2 activity results in embryos that lack APS function and show severe anterior truncation, absent AME and somite fusion (18,19). This indicates Foxa2 expression is downstream of the NODAL signal, but the mechanism by which Foxa2 expression responds to NODAL signalling remains unclear.

The propagation of the NODAL signal utilizes SMAD molecules which act as both transducers and transcription factors. Ligand binding and receptor activation causes phosphorylation of some members of the SMAD family; the receptor-associated SMADs (R-smads). SMAD2 and SMAD3 are R-smads phosphorylated in response to TGF-β, Activin or Nodal signals. This enables interaction with the common mediator smad (Co-smad), SMAD4, resulting in nuclear localization and formation of higher order transcriptional complexes. SMAD proteins interact only weakly with DNA either by themselves or in combination with other SMADs. For robust transcriptional control heteromeric Smad complexes (i.e. SMAD2/4 or SMAD3/4) must associate with additional nuclear co-factors or tissue specific transcription factors (for review see (20)). The analysis of murine conditional alleles or compound mutants demonstrates that each of SMAD2, 3 and 4 is required once gastrulation has begun to pattern primitive streak derivatives. Sequentially lowering the SMAD2/3 intracellular pool leads to loss of APS derivatives (DE, PrCP and the node) followed by loss of more posterior primitive streak derivatives (heart and pre-somitic mesoderm) (21,22). Ablation of epiblast SMAD4 similarly causes a failure of all APS functions with accompanying loss of DE, PrCP and node (23). Thus SMADs 2, 3 and 4 are required for steps 1 and 2 of PrCP development.

The weak DNA binding ability of the SMAD proteins is circumvented by SMAD interaction with tissue-specific transcription factors. For example, mutation of the Foxh1 transcription factor prevents the formation and function of the APS and results in embryos that phenocopy various NODAL pathway alleles with severe anterior truncation, absent AME and somite fusion. This is because FOXH1 forms complexes with heteromers of SMAD4 and either phosphorylated SMAD2 or SMAD3 (24,25). The binding of this complex to DNA is stabilized by SMAD4 contact with DNA at Smad-binding elements (SBE) that lie adjacent to the FOXH1-binding site. FOXH1 does not contain a transcriptional activation domain and requires SMAD interaction for transcriptional regulation (26). The T-box transcription factor eomes is also able to bind SMAD2 and is thought to regulate gene expression within the APS required for definitive endoderm formation (27,28). It is likely that NODAL dependent gene expression in the node also requires tissue-specific transcription factors and that additional Smad interacting transcription factors remain to be discovered.

As described above, both NODAL signalling and Foxa2 expression in the APS are required for the first two steps of PrCP development. Mouse mutants in which these early steps are defective prevent differentiation of PrCP precursors and do not allow assessment of subsequent steps in PrCP development. The Ku allele of Zic2 therefore represents a unique opportunity to explore these later steps since in these embryos the APS derived PrCP precursors differentiate and the node is induced (4). We hypothesize, given the central role of NODAL signalling in the establishment of the AME, that NODAL signalling will be critical for these later steps too. Additionally, Foxa2 is also expressed at a slightly later stage of development in the patent node and this expression seems likely to be critical for HPE prevention because of the Foxa2/Shh auto-induction loop (12,13). We therefore sought to determine whether ZIC2 intersects the NODAL signalling pathway during node function and whether it directly influences Foxa2 expression.

We now provide evidence that Zic2 and Nodal genetically interact during AME formation and that the ZIC2 transcription factor acts downstream of NODAL signalling. The ZIC2 protein physically interacts with the NODAL stimulated R-Smad proteins (SMAD2 and SMAD3) via the ZIC2 N-terminus and is able to modulate SMAD-dependent transcription in cell-based Luciferase assays. The mutant form of ZIC2 expressed in the kumba mouse Holoprosencephaly mutant retains the SMAD binding ability but not the ability to modulate SMAD-dependent transcription. In addition to the previously demonstrated requirement for ZIC2 function to promote Foxa2 expression in the mid-gastrula murine node (4), ZIC2 modulation alters foxA2 expression in the Xenopus gastrula and in A549 cultured cells. In A549 cells, FOXA2 expression is promoted in a ZIC2/SMAD3 dependent manner and both proteins occupy a known TGF-β-responsive SMAD3 binding site within the FOXA2 promoter. Evidently ZIC2 can interpret the NODAL signal and promote transcription of critical ANC differentiation genes such as FOXA2. The mutant ZIC2 proteins encoded by HPE-associated variant ZIC2 sequences are compromised in their ability to modulate SMAD-dependent transcription. These findings provide the first molecular explanation of how ZIC2 loss-of-function can lead to HPE.

Results

Zic2 genetically interacts with the nodal signalling pathway to promote the formation of the embryonic midline

During murine development, Zic2 function is required in the mid-gastrula node to promote the formation of the ANC (4). Node induction by the APS is dependent upon correct levels of NODAL signal. It is therefore possible that Zic2 interacts with the Nodal pathway at the newly induced node and that a reduction of NODAL signalling activity in the Zic2 mutant background would exacerbate the mild defect in anterior axis formation found in Zic2 mutants. To test this hypothesis a series of embryos with compound mutations in Zic2 and Nodal were analysed (Figure 1) . Embryos heterozygous for the severe loss of function kumba (Ku) allele of Zic2 (MGI: 1862004; Zic2^Ku/+) (29,30) have no discernible phenotype, whereas embryos homozygous for this allele exhibit forebrain dysmorphology that can include a mild loss of telencephalic tissue, but never loss of diencephalon or mesencephalon (also see Table 1 and (4); Figure 1). Embryos with loss-of-function Nodal mutations exhibit anterior truncations, the extent of which is dependent upon the level of remnant NODAL activity, with lowest levels correlating with the most-severe truncation. For example, a hypomorphic mutation of Nodal (Δ600) (Nodal^Tm2Rb; MGI:2181583) (31) has no phenotypic consequence in the heterozygote, but when homozygous can cause moderate or severe anterior truncation at low penetrance (see Figure 1F–H and Table 1). Embryos heterozygous for a null allele of Nodal (Nodal^Tm1Rob; MGI:2180793) have no phenotype (see Figure 1C and Table 1), but homozygosity of this allele is lethal at gastrulation (32). As shown in Figure 1I and Table 1, decreasing Nodal function on the Zic2 Ku background leads to a proportion of Ku embryos exhibiting moderate or severe anterior truncation. For example, 38% of Zic2^Ku/Ku embryos present with moderate or severe anterior truncation when also heterozygous for the Δ600 hypomorphic allele of Nodal, while when homozygous for this allele, 100% of Zic2^Ku/Ku embryos exhibit anterior truncation. Overall, there is a trend towards increasing anterior truncation as the level of NODAL activity decreases (Table 1). The proportion of Zic2^Ku/Ku embryos with truncations is significantly increased (P < 0.01) in all conditions of decreased NODAL activity.

Figure 1.

A genetic interaction between Zic2 and Nodal. (A–I) Lateral view of 9.5 dpc embryos of the genotypes shown (at the Nodal and Zic2 loci respectively) following WMISH with Otx2 and T. Otx2 is expressed throughout the entire forebrain and midbrain territories and T is expressed in the primitive streak and AME-derived notochord. (A-D) Embryos heterozygous for any of the mutations used in this study have normal forebrain and midline development. (E) Embryos homozygous for the Ku allele of Zic2 have mild anterior truncation and disrupted midline development but never exhibit entire loss of forebrain or any loss of midbrain. (F-H) Embryos homozygous for the Δ600 allele of Nodal exhibit a spectrum of defects, ranging from no phenotype (F) through a moderate anterior truncation in which some of the midbrain is retained (G) to a severe anterior truncation in which all forebrain and midbrain tissue is absent and AME does not develop (H). (I) Decreasing Nodal dosage on the Zic2^Ku/Ku background results in embryos with Nodal-like anterior truncations, in this case a severe truncation and loss of AME. (J) RT-qPCR analysis of Foxa2 and Nodal mRNA levels in 7.0 dpc wild-type and Zic2^Ku/Ku embryos (n = 10 per genotype), data is presented as the overall mean value obtained by averaging the results from three repeat PCR experiments from each of the 10 embryos per genotype. Error bars SEM, *P<0.05, Students t-test. f: forebrain, m: midbrain, ps: primitive streak; nc: notochord.

Table 1.

The effect of decreased Nodal dosage in the Zic2 mutant background. All embryos were scored at 9.5 dpc and the following phenotype classes noted. Ku: phenotype corresponds to that of Nodal^+/+;Zic2^Ku/Ku embryos, None: phenotype corresponds to that of Nodal^+/+;Zic2^+/+(i.e. wild-type) embryos, Mod: Moderate anterior truncation in which variable amounts of forebrain and/or midbrain are retained, Sev: Severe anterior truncation associated with complete loss of forebrain and midbrain, loss of AME and somite fusion across the midline. *P<0.01, Binomial test, one tailed. N.B. To generate two outcomes for analysis the number of embryos with no anterior truncation was combined (i.e. Ku + None classes) as was the number of embryos with anterior truncation (i.e. Mod + Sev).

	Nodal+/+				NodalΔ600/+				Nodal+/-				Nodal Δ600/Δ600
	Ku	None	Mod	Sev	Ku	None	Mod	Sev	Ku	None	Mod	Sev	Ku	None	Mod	Sev
Zic2+/+	0	100	0	0	0	100	0	0	0	100	0	0	0	80	12	8
	n = 87				n = 84				n = 10				n = 51
Zic2Ku/+	0	100	0	0	0	100	0	0	0	100	0	0	0	28	44	28
	n = 59				n = 61				n = 11				n = 15 *
Zic2Ku/Ku	100	0	0	0	62	0	20	18	22	0	78	0	0	0	33	67

Zic2 encodes a transcription factor and could promote NODAL epiblast signal either by activating Nodal expression or by acting as a transcriptional mediator of the NODAL signal. To discriminate between these possibilities the level of Nodal mRNA expression at mid-gastrulation was compared between wild-type and Zic2^Ku/Ku embryos (n = 10 per genotype) using RT-qPCR analysis. This analysis showed Nodal expression is unaltered in the absence of functional ZIC2 protein whereas, using the same technique, the expression of Foxa2 is significantly decreased (P < 0.05) in accordance with the previous in whole mount in situ hybridisation studies (4) (Figure 1J). This suggests that the ZIC2 transcriptional regulator does not directly nor indirectly influence Nodal expression, but rather acts downstream of the NODAL signal.

ZIC2 physically interacts with the SMAD transcriptional mediator complexes of the NODAL signal

NODAL signals are transduced by SMAD molecules. Activated Nodal receptors phosphorylate the receptor-regulated Smads (R-Smads), which form homomeric complexes and heteromeric complexes with the common Smad (SMAD4). These activated Smad complexes accumulate within the nucleus, where they often complex with tissue specific transcription factors to directly regulate transcription of target genes (33). If ZIC2 transduces the NODAL signal it should physically interact with one or more of the NODAL-responsive R-Smads (SMAD2 and SMAD3) (34). ZIC2 interaction with each of these SMADS was tested in complex-immunoprecipitation (Co-IP) experiments after overexpression in HEK 293T cells. Following co-transfection of a FLAG-tagged ZIC2 expression construct with one of MYC-tagged SMAD2 or SMAD3, isolated ZIC2 complexes were found to contain SMAD2 or SMAD 3 (Figure 2A). To confirm that ZIC2 can interact with activated SMAD2, HEK 293T cells overexpressing FLAG-ZIC2 were stimulated via treatment with recombinant TGF-β and purified FLAG-ZIC2 complexes interrogated for the presence of phosphorylated SMAD2. ZIC2 was found to complex with endogenous, phosphorylated SMAD2, strengthening the physiological significance of the observed ZIC/SMAD interactions (Figure 2B).

Figure 2.

ZIC2 physically interacts with the receptor activated SMAD2 and SMAD3 proteins. (A–D) Top panels; Western blot analysis of whole cell lysates from HEK 293T cells following transfection of the constructs shown immunoprecipitated with anti-FLAG or anti-V5 as shown, along with an Ig control immunoprecipitate. Bottom panels; Western blot of input lysates using anti-MYC, anti-FLAG or anti-V5 to detect exogenous proteins or anti-P-SMAD2 to detect endogenous phosphorylated SMAD2 protein. The lysates analysed in (B) were obtained from cells stimulated by recombinant TGF-β prior to harvesting. For each condition, n = 3 with typical immunoreactive bands from one experiment shown. (E) Schematic representation of ZIC2 constructs used, the C6 construct is denoted V5-ZIC2-Ku in panel C.

To determine which part of the ZIC2 molecule interacts with SMAD molecules, HEK 293T cells were co-transfected with a full length SMAD3 expression vector and a vector that expresses either a full length or truncated form of FLAG-tagged ZIC2 (bottom WB, Figure 2D). Myc-tagged SMAD proteins co-immunoprecipitated with an anti-FLAG Ab were analyzed by WB and detected by an anti-MYC Ab (Figure 2D). ZIC2 constructs that contained the N-terminus (i.e. full length, 1–415 and 1–255) were able to complex with SMAD3, whereas those with a N-terminal deletion (140-532 and 255–532) were not, indicating that amino acids 1–140 are required to facilitate ZIC2/SMAD3 interaction. As shown in Figure 2E this region of the protein contains the ZOC box; a protein-protein interaction domain found within the ZIC Subclass-A proteins (ZIC1-3) (2).

The murine kumba mutation is caused by an amino acid substitution at the second canonical cysteine within the fourth zinc finger (the C369S variant of the murine ZIC2 protein) (Figure 2E) (29). This mutation prevents DNA binding (35), but the domain mapping studies suggest that it may not interfere with SMAD interaction. To test this, expression vectors designed to produce either V5-tagged wild type ZIC2 (V5-ZIC2) or the equivalent human kumba mutation (V5-ZIC2-Ku), were co-transfected into HEK 293T cells with the MYC-SMAD3 vector and Co-IP experiments performed. Consistent with the domain mapping studies, the C370S (Ku) form of ZIC2 was found to retain its ability to interact with SMAD3 (Figure 2C).

ZIC2 alters transcription at SMAD response elements

To determine the functional consequence of ZIC2/SMAD interaction, trans-activation at SBEs was examined in HEK 293T cells using well-established SMAD response assays. First the FOXH1, SMAD2/4 dependent Activin Response Element (ARE) was assayed using the A3-Luc reporter (which contains 3 AREs; (36)). SMAD2 does not bind DNA, but complexes with SMAD4 which (in combination with FOXH1) binds to each ARE. Luciferase expression from this vector is constrained in this cell line but can be stimulated by co-transfection of plasmids encoding NODAL, CRIPTO and FOXH1 (36). In the absence of NODAL stimulation, ZIC2 did not alter trans-activation of the A3-Luc reporter, but did significantly inhibit NODAL stimulated trans-activation of the A3-Luc reporter (Figure 3A), implying that ZIC2 is able to antagonize SMAD2-dependent transcription. The effect of ZIC2 on the CAGA-Luc reporter which contains 12 repeats of the consensus SMAD3/4 binding site was also examined. High-level transcription at this element can be stimulated by either co-transfection of the ALK4 receptor or co-culture with recombinant TGF-β (37). In the absence of ALK4/TGF-β stimulation, ZIC2 did not alter trans-activation of the CAGA-Luc reporter, but did significantly inhibit NODAL stimulated trans-activation of the CAGA-Luc reporter (P < 0.05, Figure 3B and data not shown). The ability of ZIC2 to interfere with SMAD mediated trans-activation of the CAGA-Luc vector increases in a dose dependent manner and the kumba (Ku) form of human ZIC2 (C370S) exhibits decreased inhibition (Figure 3C). At the lowest dose used V5-ZIC2-Ku did not significantly alter trans-activation of the CAGA-Luc reporter, but both higher doses led to a significant decrease in luciferase activity relative to that observed in the absence of ZIC2 (P < 0.05). This implies that the ZIC2 zinc finger domain is also required for inhibition of NODAL stimulated trans-activation of the CAGA-Luc reporter.

Figure 3.

ZIC2 modifies transcriptional activity at SMAD-binding elements. (A–D) Relative Luciferase activity following transfection of the expression constructs shown into HEK 293T cells. (A) Activity at the A3-Luc reporter in basal culture conditions or following R-smad stimulation via NODAL/CRIPTO/FOXH1 co-transfection. (B and C) Activity at the CAGA-Luc reporter in basal culture conditions or following R-smad stimulation via ALK4 co-transfection. (D) Activity domain mapping experiment using the CAGA-Luc reporter, basal or ALK4 stimulated cells and ZIC2 deletion constructs. Data from one representative experiment is shown with mean values calculated from three internal repeats (error bars SD) and western blots showing the overexpressed ZIC2 protein. Statistical significance was assessed via ANOVA of raw data pooled from three repeat experiments and relevant P values (<0.05) reported in the text.

To further map regions of ZIC2 required for this activity, the ability of the deleted ZIC2 variant proteins to inhibit SMAD dependent trans-activation of the CAGA-Luc reporter was examined. Inhibitory activity was retained only by the C2 construct N terminally deleted beyond the zinc finger domain (Figure 3D; no significant difference between C1 and C2). The C3 construct which is N-terminally deleted prior to the zinc finger domain is unable to inhibit SMAD dependent transcription at the CAGA-Luc reporter. However, physical interaction between C3 and SMAD3 remained (cf- Figure 2D), indicating that both the N-terminal SMAD3 binding domain and the zinc finger domain are required for ZIC2/SMAD3 transcription control.

To determine whether the ZIC2/SMAD complex can influence expression at ZIC response elements, activity at the ZIC2 responsive B:luc2:Z3M2:β-globin luciferase reporter which contains six tandem copies of a ZIC binding site (J. Ahmed, R. Arkell et al.; manuscript in preparation), was assessed. Activation of SMAD complexes (either by co-transfection of ALK4 or NODAL/CRIPTO/FOXH1) did not stimulate transcription at this reporter and activation of SMAD complexes in the presence of ZIC2 did not drive significantly altered expression relative to transfection of ZIC2 alone (Supplementary Material, Fig. S1). Overall, these experiments demonstrate that in HEK 293T cells, NODAL/TGF-β signals do not influence the activity at ZIC response elements but can alter trans-activation at SMAD dependent elements.

ZIC2 regulates transcription at the FOXA2 locus in a SMAD dependent manner

The FOXA2 gene is known to be important for lung development, and the control of FOXA2 gene expression is well studied in lung cancer cell lines. The FOXA2 gene is expressed in the A549 lung adenocarcinoma cell line where its promoter is unmethylated (38) and where a genome-wide screen for SMAD3 targets has identified a TGF-β responsive SMAD3 binding site at the FOXA2 promoter (39). To test whether ZIC2 may directly modify FOXA2 expression the A549 cell line was selected for further analysis. Zhang et al. (39) showed that in response to TGF-β treatment, FOXA2 expression is repressed and SERPINE1 expression is increased in A549 cells and this result was reproduced here (P < 0.05; Figure 4A, Supplementary Materials, Figs S2 and S3). When ZIC2 was overexpressed in this cell line a significant induction (P < 0.05) of FOXA2 mRNA level was observed whereas overexpression of the Ku form of ZIC2 did not elicit significantly increased levels of FOXA2 mRNA (Figure 4D, Supplementary Material, Fig. S3). To determine whether SMAD3 is required for the ZIC2-dependent increase in FOXA2 expression, the experiments were repeated in the presence of a SMAD3 specific inhibitor (SIS3; (40)). As shown in Figure 4B, the significant FOXA2 induction was completely abrogated by treating the cells with SIS3 indicating that the TGF-β repression of FOXA2 can be reversed in a ZIC2/SMAD3 dependent manner. To confirm the involvement of the previously identified SMAD3 binding site in the FOXA2 promoter (39) Chromatin immunoprecipitation (ChIP) analysis was performed on WT cells or cells treated with recombinant TGF-β. SMAD3 occupancy was significantly increased in response to TGF-β treatment as was the occupancy of ZIC2 (P < 0.05, Figure 4C). These data suggest that, in the presence of SMAD3, ZIC2 can be recruited to the FOXA2 promoter and increase transcription at this locus.

Figure 4.

ZIC2 promotes FOXA2 expression in the A549 cell line. (A) Quantification of FOXA2 mRNA expression by RT-qPCR in A549 cells following lentiviral transduction with the constructs shown or TGF-β treatment. The expression level of ZIC2 in each condition is shown in Supplementary Material, Fig. S3 and in all cases is found to be significantly elevated following transfection. (B) Quantification of FOXA2 mRNA expression in A549 cells by RT-qPCR following lentiviral transduction with the constructs shown in the presence or absence of the SIS3 inhibitor. The expression level of ZIC2 in each condition measured is shown in Supplementary Material, Fig. S3. (C) The promoter occupancy of ZIC2 and SMAD3 assayed by ChIP in the presence or absence of TGF-β treatment. Precipitated FOXA2 promoter was quantified by qPCR across the SMAD3 binding site normalized to a non-specific binding site >2 kb upstream. (D) Quantification of FOXA2 mRNA expression by RT-qPCR in A549 cells following nucleofection with the constructs shown. The expression level of ZIC2 in each condition is shown in Supplementary Material, Fig. S3. Data from one representative experiment is shown with mean values calculated from three internal repeat PCR assays (error bars SD). Statistical significance was assessed via ANOVA of raw data pooled from three repeat experiments and relevant P values (<0.05) reported in the text.

Zic2 dependent foxA2 transcription also occurs during xenopus development

The ability of zic2 to control foxA2 expression during Xenopus gastrulation was also investigated. To determine if these two genes are expressed in the same structures during gastrulation, the expression pattern of both genes was examined by WMISH (Figure 5A). At the early and mid-gastrula (stage 10-11), zic2 and foxA2 transcripts are detected in the organizer tissue (the functional equivalent of the mouse node) and are restricted to the notochord at the late gastrula (Stage 12).

Figure 5.

foxA2 expression is modulated by Zic2 during Xenopus embryonic development. (A) The timing and spatial expression of zic2 and foxA2 was monitored by WMISH using DIG-labeled antisense RNA probes on albino embryos at the indicated developmental stages. Shown with a red arrow is the area associated with the Spemann Organizer center which gives rise to the notochord in the later embryonic stage (st12). These two genes are also co-expressed in the surrounding blastopore lip containing the mesoderm progenitor cells. (B) Both zic2 gain-of-function and loss-of-function by means of mRNA or morpholino (zic2MO) injections, respectively, led to foxA2 expression inhibition as marked by erased signal shown with white arrows. The white dashed line indicates the expected invagination cleft that failed to be formed. lacZ mRNA (blue staining) was used as a tracer. (C) Xenopus animal cap assay was performed as illustrated. Total RNA was extracted from pools of 50 caps and analyzed by RT-qPCR. Data is presented as the mean value obtained by averaging the results of three repeat PCR experiments from RNA samples isolated from one experiment (Error bars SD). Statistical significance was assessed via ANOVA of raw data pooled from three repeat experiments and relevant P values (<0.05) reported in the text. a. u.: arbitrary unit, St: stage, cc: control caps.

The animal cap of the Xenopus laevis embryo provides a versatile, quick and reliable tool to evaluate gene regulatory networks in embryonic development. Therefore, to further examine whether foxA2 transcription might be regulated by zic2 in the Xenopus laevis organizer region, zic2 mRNA was injected into the animal pole of each blastomere of 4-8 cell Xenopus embryos. The embryos were grown until the blastula stage and animal caps explanted and cultured to the equivalent of the mid-gastrula stage. RT-qPCR analysis of cap explant mRNA showed that zic2 overexpression, in this model system where active TGF-β/Nodal signalling is absent, led to significant induction of foxA2 expression (P < 0.05), whereas overexpression of mRNA coding for the zic2-Ku variant elicited greatly decreased foxA2 expression (Figure 5C).

To determine whether zic2 can regulate foxA2 expression at the node during Xenopus gastrulation, zic2 mRNA was overexpressed, or zic2 translation blocking morpholinos were injected into the equator region of 4-8 cell stage embryos and the embryos grown till mid-gastrulation. Both zic2 overexpression and morpholino knock-down lead to depletion of foxA2 expression in the Spemann Organizer (Figure 5B). Thirty-one out of forty morphants showed reduced foxA2 expression, suggesting that zic2 is required for foxA2 expression. Strikingly, overexpression of zic2 mRNA in the equatorial region of the whole embryos also inhibited the expression of foxA2 (37 out of 37 injected embryos). It is of note that, both zic2 mRNA and zic2-MO also prevented the formation of the blastopore ring. Taken together, these data indicate that foxA2 expression is zic2 dependent in vivo during organizer formation.

Proband-specific ZIC2 variant proteins have altered activity at SMAD response elements

To assess the likely relevance of ZIC2 antagonism of SMAD3 activity to HPE, the ability of proband-specific ZIC2 variant proteins to inhibit SMAD3 dependent trans-activation of the CAGA-reporter in NODAL stimulated HEK 293T cells was measured. The variant proteins selected for analysis have all been previously expressed in cultured HEK 293T cells and all known to produce stable ZIC2 proteins that predominately localize to the nucleus (as does wildtype ZIC2 protein) (35). When co-expressed with the CAGA-Luc reporter the ZIC2 D152F variant and the construct containing an alanine tract expansion (25) in the ZIC2 C-terminus were significantly less able to inhibit NODAL stimulated SMAD3 dependent transactivation of the CAGA reporter (P < 0.05, Figure 6). In contrast the Q36P variant was not significantly different from wild type with respect to SMAD3 inhibition but the Q36P protein was not expressed at detectable levels in this experiment (Figure 6). A further experiment was therefore conducted using increased amounts of DNA (Supplementary Material, Fig. S4) which showed that even when expressed at higher levels the Q36P variant retained inhibition properties. Overall the data are consistent with a loss of SMAD3 antagonism contributing to HPE aetiology.

Figure 6.

HPE-associated ZIC2 variant proteins have impaired activity at SMAD-binding elements. Relative Luciferase activity following transfection of the expression constructs shown into HEK 293T cells. Activity at the CAGA-Luc reporter in basal culture conditions or following R-smad stimulation via ALK4 co-transfection is shown in the presence of wildtype or variant ZIC2 expression constructs. Data from one representative experiment is shown with mean values calculated from three internal repeats (error bars SD) and western blots showing the overexpressed ZIC2 protein. Statistical significance was assessed via ANOVA of raw data pooled from three repeat experiments and relevant P values (<0.05) reported in the text.

Discussions

The molecular mechanism underlying ZIC2-associated HPE

HPE is a common disorder of forebrain development with a complex genetic and embryological origin. The ZIC2 locus was one of the first shown to be associated with HPE (3) but its molecular mode of action has remained unclear. The recent studies of large series of HPE probands reveal a discrete ZIC2-associated clinical spectrum characterized by severe structural brain anomalies, distinct craniofacial features and defects in organ systems other than the brain (41,42). An increased understanding of both the mechanism of ZIC2-associated HPE and likely accompanying organ defects is required. We have previously used the Ku mouse model of ZIC2-associated HPE to investigate the cellular mechanism of ZIC2 requirement and to demonstrate that Zic2 does not act downstream of the SHH signal to the neurectoderm to contribute to classical HPE (4). We hypothesised instead that ZIC2 intersects with the NODAL signalling pathway. The present study, the first to address this hypothesis, reveals that ZIC2 does act downstream of NODAL signalling during mammalian gastrulation. It is one of a growing list of tissue-specific transcription factors able to complex with SMAD proteins and influence transcription at SBEs. The FOXA2 transcription factor is critical for SHH expression at the murine embryonic midline (13) and we provide evidence in cultured human cells that ZIC2 acts in concert with SMAD proteins to promote FOXA2 expression and that it also controls foxA2 expression during Xenopus development. In combination with our previous work showing ZIC2 dependent expression of Foxa2 during murine gastrulation (4) we postulate that ZIC2 directly controls Foxa2 expression during mammalian gastrulation; an interaction critical for the prevention of HPE.

Human cases of HPE are thought to arise due to the combined effects of multiple genetic and/or genetic and environmental factors (44). Genetic factors include chromosome abnormalities as well as loss-of-function mutations in individual genes, with four principal susceptibility genes identified (SHH, ZIC2, TGIF, SIX3). The molecular basis by which these four major effect HPE genes operate is now known. TGIF functions within the Nodal pathway to fine tune the response to NODAL signals (45). The SIX3 transcription factor directly regulates SHH expression (46) and SHH provides the ‘middleness’ signal from the PrCP to the overlying forebrain neurectoderm (47). The work reported here shows all four of these genes are connected during embryogenesis by the requirement for NODAL signalling to produce the PrCP from which SHH is secreted. ZIC2, the last of the major effect genes to be positioned on this continuum, sits at the junction of these pathways. It interacts with phosphorylated R-Smads and influences the expression of NODAL targets in the mid-gastrula node. Of particular relevance to HPE, it can act in combination with activated SMAD3 to promote transcription at the FOXA2 gene; itself a rare HPE-associated locus (48) and a transcription factor known to be required for Shh expression in the AME (13).

The mouse genetic experiments were conducted using mouse strains congenic on the same inbred background (129Sv/Ev). The embryos analysed were (other than for de novo sequence variants) genetically identical at all but the Zic2 and Nodal loci, each of which is a known HPE-susceptibility gene (3,49,50). This provides the first direct test of the hypothesis that digenic inheritance of severe loss-of-function alleles in HPE-associated genes remains insufficient for full expression of the disease (44). Embryos heterozygous for the Zic2 allele were overtly abnormal only when activity at the Nodal locus was reduced by more than 50%. Evidently on the 129/SvEv background, digenic heterozygous inheritance of null alleles at Zic2 and Nodal are insufficient to produce HPE in the mouse. Although extrapolation of these findings to the human is not without difficulty (because mice are less sensitive to haploinsufficiency than humans) (43) the data are consistent with the idea that additional modifier and/or environmental factors contribute to the ‘multiple hits’ required for HPE manifestation.

ZIC2 is a novel tissue-specific SMAD interacting transcription factor

The work reported here defines ZIC2 as a new tissue-specific SMAD interacting transcription factor. Endogenous p-Smad can complex with ZIC2 following receptor stimulation in cultured cells, suggesting this is a physiologically relevant interaction. This is supported by the demonstration that Zic2 can antagonise FOXH1/SMAD dependent trans-activation at AREs and that Zic2 acts genetically downstream of Nodal during AME formation at mouse gastrulation. The ZIC2/SMAD interaction takes place via the ZIC2 N-terminus, a region that contains the well-characterised ZOC protein interaction domain (51), however, modulation of SMAD-dependent transcription requires a functional ZIC2 zinc finger domain (Figures 3D, 4A, 5C and 6). Like some other SMAD binding proteins the effect of ZIC2 interaction with respect to transcriptional modulation varies with context (52–54), for example in A549 cells ZIC2 expression increases transcription at the FOXA2 locus but decreases expression at the SERPINE1 locus (Figure 4A, Supplementary Material, Fig. S2). Notably, however, ZIC2 co-expression consistently opposes SMAD activity. Further work is required to determine how ZIC2 drives divergent effects at different promoters, but differential recruitment of a co-activator versus co-repressor, or varied interaction with permissive versus repressive chromatin modifiers, would be consistent with this ability (see Figure 7). The complex nature of ZIC regulation of expression control is also highlighted by the Xenopus studies in which zic2 overexpression leads to induction of foxA2 expression in animal caps but repression in whole embryos. Differences in these experimental paradigms include the site of mRNA injection and tissue isolation in the animal cap experiments. The studies may point to a non-cell autonomous component of foxA2 regulation following zic2 overexpression, or that the concentration of Zic2 is critical for SMAD inhibition at the foxA2 locus. The result is also consistent with ectopic expression of Zic2 leading to a fate change that would not normally occur during embryogenesis.

Figure 7.

Models of ZIC2 function at SMAD binding elements. (A and B) ZIC2 drives differential effects at the FOXA2 and SERPINE1 promoters in A549 cells possibly because the ZIC2/SMAD complex is able to interact with a variety of proteins with differing roles in transcription and/or chromatin modification that (A) activate or (B) repress transcription. SBE; SMAD binding element; S3: SMAD3; S4: SMAD4; ACT: Proteins that activate transcription, such as Co-Activators or permissive chromatin modifiers; RNA pol: RNA polymerase; REP: Proteins that repress transcription, such as Co-Repressors or restrictive chromatin modifiers.

In all examined contexts, it was found that the ZIC2 protein encoded by the Ku allele of murine Zic2 was unable to oppose SMAD-dependent transcription. The Ku protein can bind SMAD protein but fails to inhibit SMAD-based transactivation and cannot promote expression at the FOXA2/foxA2 locus in either cell-based or Xenopus assays. Previous work has shown that this protein cannot bind DNA (35) as a consequence of the missense mutation at a canonical cysteine within the fourth zinc finger (29). Our investigations have not directly revealed a requirement for ZIC DNA binding during SMAD inhibition. In A549 cells, both SMAD3 and ZIC2 are found to occupy the fragment containing a known SBE at the FOXA2 promoter. Further work is required to determine whether ZIC may bind DNA within the FOXA2 promoter to enhance FOXA2 transcription. Regardless of the mechanism, the loss of SMAD inhibitory activity and accompanying failure to up-regulate target gene expression is sufficient to explain the classical HPE associated with the murine Ku mutation. Similarly, it was found that some of the ZIC2 variant proteins predicted to be encoded in HPE variants interfere with ZIC2 SMAD inhibitory activity and could explain these HPE cases. Interestingly, this was not the case for one ZIC2 variant protein (Q36P); a variant that also behaved differently to other HPE-associated variant proteins in transactivation assays (35). Whether this variant has a major contribution to the semi-lobar HPE phenotype in this case remains unresolved.

ZIC2 promotes expression of NODAL targets by antagonising transcription inhibition

In some contexts ZIC2 dampens SMAD-dependent transcription but the mouse genetics experiments presented here indicate that during mammalian gastrulation the predominant role of ZIC2 is to promote expression at SBEs. It is of note that in our experiments the Δ600 homozygous phenotype is more severe than previously reported (31) which may be explained by the shift of this allele onto a 129/SvEv congenic background. Lowering the dose of Nodal on the Zic2 mutant background produces phenotypes characteristic of mutants in which APS function is partially affected (moderate anterior truncation) or lost (severe anterior truncation, absent AME and somite fusion across the midline) demonstrating that the normal function of ZIC2 is to promote NODAL target gene expression. In combination with the cell-based assays, which identify ZIC2 as a SMAD antagonist, a model in which expression at node-specific enhancers is initially repressed and subsequently converted to expression activation in the presence of SMAD/ZIC2 complexes is likely. There is precedent for inhibition at NODAL target genes during murine gastrulation. A complete loss of APS function is seen following mutation of the E3 ubiquitin ligase Arkadia (17). Arkadia promotes NODAL signalling by targeting the SNON and Ski transcriptional repressors for degradation (55,56). We postulate that ZIC2 also promotes NODAL target gene expression by preventing transcriptional repression.

In conclusion, ZIC2 is a critical component of NODAL signalling activity that controls AME production at gastrulation. Zic2 loss-of-function leads to HPE due to a failure to activate target genes in the mid-gastrula node including Foxa2. Foxa2 in turn is required to elicit timely activation of Shh expression in the AME. The work supports the notion that multiple genetic and/or environmental factors are required to produce HPE in human embryogenesis and that the overall strength of the relevant signalling pathways is the crucial determinant of phenotypic outcome. Following this premise, that cumulative effects on pathway members determine the penetrance and expressivity of HPE, other molecules that interact with ZIC2 to overcome repression and ensure Foxa2 expression at the mid-gastrula node represent candidate HPE genes.

Materials and Methods

Mouse strains and husbandry

The following pre-existing strains of mice were used in this project: the kumba (Ku) allele of Zic2 (MGI:1862004) (29,30), a targeted null (-) allele of Nodal (NodalTm1Rob; MGI:2180793) and a targeted hypomorphic allele (Δ600) of Nodal (NodalTm2Rb; MGI:2181583). Each strain was made congenic on the 129/SvEv background and subsequently maintained by continuous backcross to 129/SvEv mice. In all cases, mice from backcross 10 or beyond were used in the breeding protocols required to generate embryos for analysis. Mice were maintained in a light cycle of 12 h light: 12 h dark, the midpoint of the dark cycle being 12 A.M. For the production of staged embryos, 12 P.M. on the day of the appearance of the vaginal plug is designated 0.5 dpc. Mice were genotyped by PCR screening of genomic DNA extracted from ear biopsy tissue using the HRMA assays previously described (57). Embryos were genotyped using a fragment of extra embryonic tissue/ectoplacental cone or yolk sac (depending on stage) using HRMA analysis (Nodal alleles) or TaqMan assay (Zic2). Genomic DNA was extracted from embryonic tissue described previously (58).

Whole mount in situ hybridization (WMISH)

Mouse WMISH was performed as previously described (29) using mouse probes Otx2 (59) and T (60). Upon completion of the WMISH procedure, embryos were post-fixed in 4% PFA and transferred via a glycerol series to 100% glycerol. For photography, embryos were flat-mounted under a glass coverslip and photographed in a Nikon SMZ 21500 Stereomicroscope and DS-Ri1 camera (Nikon). Xenopus WMISH was performed as described by Harland et al. (61). The pCMVSport6-zic2 (IMAGE: 6862275; Thermofischer) and pCMVSport6-foxA2 (IMAG E:6862846; Source BioScience LifeSciences), were linearized with EcoRI and KpnI restriction enzymes, respectively. Digoxigenin (DIG)-labelled antisense RNA probes were generated using MEGAscript® T7 Transcription kit (Ambion).

Xenopus embryos and micro-manipulations

Wild-type (pigmented) and albino X. laevis embryos were obtained by hormone-induced egg laying and in vitro fertilization and staged according to Nieuwkoop and Faber (1967). For animal cap assays, pigmented embryos were cultured in 0.5 x MBS containing 1% Ficoll 400 (Sigma) until the 4–8-cell stage, when each blastomere was injected at its animal pole with 500 pg of zic2 or zic2-Ku capped-RNA in ∼ 5 nl volume. These mRNAs were in vitro transcribed using the Sp6 mMessage mMachine kit (Ambion) and Not I-linearized Myc-tagged pCS2+ plasmid containing zic2 or zic2-Ku. As illustrated in Figure 5C, embryos were further cultured until blastula stage and dissected for explant isolation. The animal cap explants, dissected in 1× Steinberg solution supplemented with 0.5% BSA using an adapted eyelash unit, where cultured for 5 h post-dissection at RT (20°C) (equivalent of mid-gastrula stage). Alternatively, albino embryos were injected unilaterally with the same amount of zic2 mRNA or with 20 ng (in ∼ 5 nl volume) of a zic2 translation-blocking morpholino antisense oligonucleotides (Zic2MO: 5’ACTGGG GACCAGC GTC TAGTAGCAT-3’, Gene Tools) in the equatorial region (mesoderm forming area) at 4-8 cell stage for whole embryo functional analysis. In this case, 250 pg of lacZ mRNA were co-injected as tracers and embryos were fixed at mid-gastrula stage (st10.5) in 4% PFA and stained with Xgal (blue) before WMISH.

Cell culture

The human embryonic kidney 293T cell line (HEK 293T) and the Human lung carcinoma cell line A549 were obtained from the American Type Culture Collection and maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (Thermo Fisher Scientific; SH30071.03). R-smad activation was achieved via supplementation of the medium with 2 ng/ml recombinant TGF-β (R&D Systems; 240-B-002) reconstituted in sterile 4 mM HCl containing 1 mg/mL bovine serum albumin, for 2 h before harvesting. Specific SMAD3 inhibition (40) was achieved via supplementation of the medium with 10 µM of SIS3 dissolved in DMSO for 4 h prior to harvesting (SIS3, EMD Chemicals, Inc., San Diego, CA).

Plasmids

The FLAG-tagged human full-length ZIC2 expression construct (FLAG-ZIC2), its deletions (FLAG-ZIC2-1-415, FLAG-ZIC2-1-255, FLAG-ZIC2-140-532 and FLAG-ZIC2-255-432), the Xenopus construct (Xzic2-MYC) and Xenopus Xzic2 Morpholino were previously described (62). The kumba mutant Xzic2 construct (XZic2-Ku) was generated from the wild-type Xzic2 construct using the QuikChange II XL Site-Directed Mutagenesis Kit (Agilent; 200521) according to the manufacturers protocol with the following primers: KU_t1334a_F (GTG CGA ATT TGA AGG CAG CGA CAG GC GAT TTG C) and KU_t1334a_R (GCA AAT CGC CTG TCG CTG CCT TCA AAT TCG CAC). The A3-Luciferase (A3-Luc) reporter, together with the NODAL, CRIPTO and FOXH1 expression plasmids were obtained from Iratni et al. (36). The MYC-SMAD2 and MYC-SMAD3, ALK4 AND CAGA-reporter (CAGA-Luc) plasmids were obtained from Imamura et al. (63). The V5 epitope-tagged, full length Human ZIC2 cDNA expression vector (V5-ZIC2-WT) was previously described (64). To generate a corresponding vector with the C370S variant form of the Human ZIC2 cDNA, the Entry clone containing Human ZIC2 wild-type cDNA (pENTR™3C-ZIC2; (64)) was subjected to mutagenesis via overlap extension PCR, using the following primers: ZIC2 end primers, Ark1150_F (ATC CGG TAC CGA ATT CAG TGT GGT GGA ATT CCT GGC C) and Ark1168_R (GTG CGG CCG CGA ATT CGA GGG TTA GGG ATA GGC TTA C) and ZIC2 mutagenesis primers, Ark1304 _F (CCA GTG TGA GTT TGA GGG CAG CGA CCG GCG CTT CGC) and Ark1305_R (GCG AAG CGC CGG TCG CTG CCC TCA AAC TCA CAC TGG). All PCRs were performed with the Immomix Taq/Buffer reagent (Bioline) in the presence of 80 mM Betaine and an annealing temperature of 65°C. Both the final PCR fragment (1765 bp in length) and the pENTR™3C-ZIC2 plasmid were digested with BstAPI, and the restricted vector dephosphorylated (Antarctic Phosphatase; NEB) before vector/insert ligation. The insert was subsequently transferred to the destination clone pcDNA3.1/nV5-DEST™ (Life Technologies) via a Gateway™ LR Clonase reaction (as per manufacturer’s instructions, Life Technologies) to produce the V5-ZIC2-C370S plasmid. The viral ZIC2 plasmid was obtained as follows: First, the full-length Human ZIC2 cDNA was PCR-amplified from the ZIC2 expression construct (62) using the platinum Taq enzyme (Invitrogen; 11304-029) with the inclusion of 2.5X PCRx enhancer solution (Invitrogen; 11495-017). The primers incorporated a 5T-stretch and a HindIII-site on the Forward- and a 5A-stretch and a XhoI-site on the Reverse-primer (F:TTT TTT AAG CTT TCA CTC CTG GAC GCG GGT CCG and R:AAA AAA CTC GAG TCA CAC GTA CCA TTC ATT GAA GTT GGA G). Next, a 3X-FLAG dsDNA fragment was generated by annealing complementary primers that contained a BamHI and HindIII site respectively (F:AAA AAA GGA TCC ATG GAC TAC AAA GAC CAT GAC GGT G; R:TTT TTT AAG CTT CTT GTC ATC GTC ATC CTT GTA ATC GA). The 3X-FLAG and ZIC2 PCR fragments were HINDIII digested and ligated. The ligated fragment was subsequently restricted with BamHI and XhoI enzymes and ligated into the pENTR1A vector, which was cut with identical enzymes using standard protocols to generate the pENTR™1A-ZIC2-WT-FLAG plasmid. The FLAG-tagged ZIC2 cDNA was transferred to the plx-301 destination vector via a Gateway™ LR Clonase reaction (as per manufacturer’s instructions, Life Technologies) to generate the plx-301-ZIC2-3X-FLAG lentiviral vector. An identical experiment was performed in parallel using the EGFP-pDONR221 as a donor vector to obtain an EGFP control plx-301-EGFP lentiviral vector. All viral plasmids were obtained from Addgene (EGFP-pDONR211-25899; plx301-25895; pENTR1A-11813-011). The ZIC2 shRNA was obtained from Sigma, TRC consortium (TRCN0000062372). The ZIC2 mutant constructs (Q36P, D152F, 25 Alanine tract expansion) were provided by Prof. Brown and were cloned into a pIRES-GFP vector containing a 3X-FLAG tag (35).

Western Blotting (WB)

Cells were lysed in lysis buffer (50 mM Tris HCl (pH 7.4), 150 mM NaCl, 1 mM EDTA, and 1% Triton X-100) supplemented with a proteinase inhibitor mixture (Roche Applied Science; 5892988001) and phosphatase inhibitor mixture (Sigma; P5726-5ML and P0044-5ML). Protein samples were run on a 10% BisTris gel (Invitrogen; NP0301BOX) and transferred to Hybond-C nitrocellulose membranes (Amersham Biosciences; GEHERPN303E). After incubation with antibody and four times 10 min wash steps with TBS-T, the blots were developed using ECL Western blotting detection reagent (Amersham Biosciences; RPN2106). The following antibodies were used: anti-FLAG monoclonal antibody M2 (Sigma; F9291; 1/5000) and anti-Myc monoclonal antibody (Sigma; B7554; 1/5000), anti-V5 antibody (Abcam; ab27672; 1/5000) and P-smad2 antibody (Cell Signalling; 3101S; 1/1000). All antibodies were diluted in a 5% skim milk (Nestlé) 1X TBS-T solution.

Complex Immunoprecipitation (Co-IP)

Cells were collected in lysis buffer containing Proteinase Inhibitor mixture (Roche Applied Science; 5892988001) and Phosphatase Inhibitors (Sigma; P5726-5ML and P0044-5ML) 24 h post-transfection. After sonication using the Branson Sonifier 250 (2 rounds of 30 seconds on output control 5, constant duty cycle), the mixture was centrifuged at 12,000 rpm for 30 min. The cell lysates were precleared on protein G-Sepharose beads (Amersham Biosciences) at 4°C for 3 h on a rotating wheel according to manufacturer’s protocol. Meanwhile, antibody against FLAG, MYC (Sigma), or V5(Abcam) was coupled to protein G-Sepharose beads during incubation on a rotating wheel at 4°C for 3 h using the advised quantities. 200 µg of precleared cell lysates were divided over the antibody-coupled protein G-Sepharose beads and an equal amount of protein G-Sepharose beads coupled with IgG as a control. After four washing steps with lysis buffer, the bound protein complexes were solubilized in 40 µl of 3X water-diluted sample buffer (Invitrogen; SKU# NP0007) and analysed by Western blotting using the appropriate antibody.

Chromatin Immunoprecipitation (ChIP)

ChIP experiments were mainly performed as previously described (62). Activation of Nodal signalling in A549 cells was achieved by treatment with 2 ng/ml TGF-β (R&D Systems, 240-B-002) for 2 h before cross-linking. A total of 5 μg of the indicated anti-SMAD3 antibody (Abcam, ab28379), ZIC2 (Santa-Cruz, sc-28153) or IgG (Abcam, ab18413) were incubated with sheared cross-linked chromatin and BSA-blocked protein G beads (Life Technologies, 10003D). Input and immunoprecipitated DNAs were subjected to RT-qPCR using primers designed against the FOXA2 or unbound region as a negative control as previously described (39).

Luciferase Reporter Assays

HEK 293T cells were transiently transfected in triplicate with a combination of plasmids using FuGENE 6 (Roche Applied Science). The total DNA concentration for each transfection was kept constant via the inclusion of the corresponding empty vector, and cells were transfected according to the manufacturer’s protocol. β-Galactosidase expressing plasmid (5 ng/well) was co-transfected to normalize the transfection efficiency (62). 24 h post-transfection, cells were washed with cold PBS and lysed in Passive Lysis Buffer (Promega). Luciferase reporter enzyme activity (Promega) and β-galactosidase (Applied Biosystems) assays were used according to the manufacturers' protocol. In some cases NODAL stimulation of HEK 293T cells was achieved via co-transfection of NODAL/CRIPTO/FOXH1 (in a 10/10/1 ratio) or ALK4 expression constructs. Three external repeats of all assays were performed.

Real-time quantitative PCR (RT-qPCR)

RNA was isolated from individual (n = 10 per genotype) 7.0 dpc mouse embryos, RNA integrity and absence of genomic DNA contamination confirmed and reverse-transcribed as previously described (64). RNA from Human A549 cells was isolated according to the manufacturer’s protocol (Qiagen, RNeasy kit; 74104) and absence of genomic DNA contamination confirmed. RNA (1 μg) was reverse-transcribed using Superscript transcriptase II (Invitrogen). RNA was extracted from pooled (50 per sample) Xenopus animal cap explants under various treatment conditions using the RNAspin Mini RNA isolation kit (GE Healthcare) and samples confirmed to be free of genomic DNA contamination by 26-cycle PCR amplification using histone H4 primers (65). Complementary DNA was synthesized from 1 µg of total animal cap explant RNA with iScript cDNA synthesis kits (Bio-Rad). In all cases (mouse, human, Xenopus), a Reverse-Transcriptase negative control synthesis reaction was performed in parallel from each RNA sample. The relative amount of gene transcripts was determined by RT-qPCR using the ABI 7500 system (Applied Biosystems) using the SuperScript III Platinum SYBR Green One-Step RT-qPCR reagent for mouse and Taqman Univeral PCR Master Mix (Applied Biosystems; 4364341) for human samples. For Xenopus samples, transcript levels were assayed in the LightCycler® 480 Real-Time PCR System with the LightCycler® 480 SYBR Green I Master, 2× concentration (Roche). Species specific GAPDH (mouse or Xenopus) or RPL13A (human expression) was assayed and used to normalise the data. Standard curves for target and reference genes were generated from the Ct (threshold cycle) values, and the relative concentrations of the standards and samples were calculated from the Ct values and the equation of the curves. The values obtained for the targets were divided by the values of housekeeping genes to normalize for differences in reverse transcription. The following Applied Biosystems assays were used for mouse and Human samples: Mouse GAPDH Endogenous Control FAM, Mouse Foxa2 FAM assay (Mm01976556_s1) and Mouse Nodal FAM assay: Mm00443040_m1, Human FOXA2 assay (Hs00232764_m1), human SERPINE1 assay (Hs01126604_m1) and human RPL13A as control (Hs01926559_g1). For Xenopus assays, the following primers were used: GAPDH (sense 5’- GTG AAG GTC GGT GTG AAC G-3’; antisense 5’-AGG GGT CGT CTG ATG GCA ACA-3’) and foxA2 (Sense 5’- CTT CTG GAC CCT ACA CCC TG; antisense 5’-TCG ATG CCC CTT CTG AAA GT-3’).

Statistics

Mouse phenotype data were analysed using the Binomial test (one tailed) with a P value of less than 0.05 considered statistically significant. For RT-qPCR, data statistical significance was determined by Student’s t-test, with a P value of less than 0.05 considered statistically significant. Raw data from three external repeats of each Luciferase reporter experiment were analysed by ANOVA and a p value of less than 0.05 considered statistically significant.

Supplementary Material

Supplementary Material is available at HMG online.

Acknowledgements

We thank Dominic Norris for the murine Nodal alleles, Steve Brown for ZIC2 plasmids, the ANU Statistical Consulting Unit for assistance and An Zwijsen, Danny Huylebroeck, Przemko Tylzanowski and Vasso Episkopou for helpful discussions and insights.

Conflict of Interest statement. None declared.

Funding

This work was supported by a National Health and Medical Research Council grant [#366746 to R.M.A.]; a Sylvia and Charles Viertel Senior Medical Fellowship [to R.M.A.]; a Ph.D. grant of the Agency for Innovation by Science and Technology [IWT #101485 to R.H.]; a Fonds De La Recherche Scientifique grant [#CRD-14646056] and the fund for Scientific Research Flanders [#1.8.320.07.N.00, #1832017N to S.T.].

Author notes

^†The authors wish it to be known that, in their opinion, the first two authors should be regarded as joint First Authors.

^‡The authors wish it to be known that, in their opinion, the last three authors should be regarded as joint Last Authors.