Analysis of mouse models carrying the I26T and R160C substitutions in the transcriptional repressor HESX1 as models for septo-optic dysplasia and hypopituitarism

Developmental defects in pituitary gland formation lead to hypopituitarism, which can range from mild phenotypes involving deficiency of a single hormone, through more severe phenotypes affecting multiple pituitary hormone axes, to panhypopituitarism. Isolated growth hormone deficiency (IGHD) is the most frequent form of human hypopituitarism, affecting 1 in 4000–10,000 live births (Vimpani et al., 1977; Lindsay et al., 1994; Procter et al., 1998). Combined pituitary hormone deficiency (CPHD), in which there is a deficiency of more than one pituitary hormone, is less common, but is associated with considerable morbidity and, if not treated promptly and adequately, occasional mortality. Septo-optic dysplasia (SOD; also referred to as de Morsier syndrome) is a phenotypically and aetiologically heterogeneous disorder in humans, characterised by hypopituitarism occurring in conjunction with midline forebrain defects and optic nerve hypoplasia. This congenital disorder (1 in 10,000 live births) is characterised by a highly variable phenotype with varying degrees of abnormalities in the corpus callosum, septum pellucidum, eyes and pituitary gland (Patel et al., 2006; Kelberman and Dattani, 2007).

Phenotypic analyses of mouse mutants have implicated a number of genes in pituitary development, some of which have also been associated with hypopituitarism in human patients with mutations in orthologous genes (Cushman and Camper, 2001). Several homeobox genes in particular have been shown to play a crucial role in both mouse and human pituitary organogenesis (Kelberman and Dattani, 2007; Cushman and Camper, 2001). One such gene encodes the paired-like homeodomain protein HESX1, a highly conserved transcriptional repressor, which is expressed in the early forebrain primordium and Rathke’s pouch during vertebrate development (Thomas and Beddington, 1996; Hermesz et al., 1996). Hesx1-deficient embryos show a significant reduction in anterior forebrain structures, such as the telencephalic and optic vesicles, which is caused by a transformation of anterior to posterior forebrain (Andoniadou et al., 2007). Hesx1^–/– mutants also show severe pituitary gland dysplasia and enhanced cellular proliferation, but terminal differentiation of the hormone-producing cell types is not affected at later stages of development (Dasen et al., 2001). Hesx1-deficient mutants also manifest fully penetrant eye defects, ranging from microphthalmia to anophthalmia, disturbances in midline telencephalic commissural tracts (corpus callosum and anterior commissure) and abnormalities in the olfactory bulbs (Dattani et al., 1998; Andoniadou et al., 2007).

In humans, mutations in HESX1 have been associated with phenotypes affecting the midline forebrain structures, the eyes and, most commonly, the pituitary gland. So far, a total of 13 HESX1 mutations have been identified in association with SOD and/or hypopituitarism (Dattani et al., 1998; Thomas et al., 2001; Brickman et al., 2001; Carvalho et al., 2003; Cohen et al., 2003; Tajima et al., 2003; Sobrier et al., 2005; Sobrier et al., 2006; Coya et al., 2007). Five of them are recessive and the remaining eight are dominant. They vary from missense to frameshift mutations and result in considerable variability in the penetrance and severity of the phenotype in affected patients. At present, the reasons underlying this variability are not clear.

Two previously reported, recessive missense mutations involve the substitution of highly conserved residues at position 26 (isoleucine) and 160 (arginine) by threonine and cysteine (I26T and R160C), respectively (Dattani et al., 1998; Carvalho et al., 2003). I26 maps within the engrailed homology (eh-1) domain, an octapeptide sequence shown to be involved in the interaction of HESX1 with the co-repressor TLE1 (transducin-like enhancer of split 1) that is able to recruit histone deacetylases required for transcriptional repression (Dasen et al., 2001; Carvalho et al., 2003). In vitro, the HESX1-I26T mutant protein can bind to DNA, but its ability to repress transcription is reduced in comparison to wild-type HESX1. Carvalho and colleagues reported on a patient with evolving CPHD who was homozygous for the I26T mutation, but who had normal optic nerves and no telencephalic defects (Carvalho et al., 2003). The parents of the affected individual were heterozygous with respect to the mutation and appeared to be clinically unaffected. The reasons for the lack of ocular and telencephalic defects are not fully understood.

R160 is localised within the recognition alpha helix of the homeodomain, which establishes direct contact with target DNA through the major groove (Wilson et al., 1995; Dattani et al., 1998). In vitro, the ability of the HESX1-R160C protein to bind to DNA is abolished by this mutation, but its repressor activity is retained when fused to the Gal4 DNA-binding domain in a mammalian one-hybrid system (Brickman et al., 2001). It has been postulated that HESX1-R160C may have a dominant negative effect, as it is able to inhibit the DNA binding and repressor activities of wild-type HESX1 in vitro (Brickman et al., 2001). However, Dattani and colleagues observed that although two siblings manifesting panhypopituitarism in association with midline telencephalic commissural defects and severe optic nerve hypoplasia were homozygous for the R160C mutation, their heterozygous parents were phenotypically normal (Dattani et al., 1998). This in vivo observation is in direct contrast to the in vitro dominant negative effect of this mutation. The forebrain defects of these two siblings were mild when compared with the severe abnormalities frequently observed in Hesx1^–/– mice, suggesting that part of the function of HESX1 may be performed independently of its ability to bind to DNA (Dattani et al., 1998; Martinez-Barbera et al., 2000). A more direct analysis of the phenotypic consequences of the R160C mutation in a mouse model can be used to clarify this issue.

Here, we have used an in vivo approach to provide insights into the aetiology, pathogenesis and variable nature of hypopituitarism and SOD associated with the human HESX1 mutations I26T and R160C.

Homozygous loss of Hesx1 in the mouse has been shown previously to cause perinatal lethality, possibly because of the severe forebrain and craniofacial defects, and/or pituitary dysfunction (Dattani et al., 1998; Andoniadou et al., 2007). However, the phenotypic consequences of replacing the wild-type allele with specific HESX1 mutations originally identified in human patients with hypopituitarism and SOD have not been reported previously. To gain further insight into the in vivo consequences of the I26T and R160C mutant proteins, we have generated mouse models carrying these two substitutions (Fig. 1).

Hesx1^I26T/+ mice were phenotypically normal, viable and fertile (n=124). Genotypic analysis of pups and weaners from Hesx1^I26T/+ intercrosses showed a slight deviation from the expected Mendelian ratios. Although this did not reach statistical significance, it may be indicative of sporadic postnatal lethality (Table 1). Of the surviving Hesx1^I26T/I26T pups, 73.7% exhibited eye defects, usually unilateral or bilateral microphthalmia (n=19). This phenotype was not observed in heterozygous or wild-type littermates. Histological analysis of Hesx1^I26T/I26T adult brains revealed normal telencephalic commissural tracts with no apparent abnormalities in the corpus callosum or anterior commissure (100%, n=8).

A similar genotypic analysis of embryos from 8.5–17.5 days post coitum (dpc) showed the expected Mendelian proportions (Table 1). The majority of the Hesx1^I26T/I26T embryos analysed from 12.5–17.5 dpc (n=42) displayed eye abnormalities (76.2%), but telencephalon development appeared normal in most of these embryos (97.6%) when compared with wild-type or heterozygous littermates (Fig. 2C-E; Table 2). The most common forebrain phenotype observed in Hesx1^I26T/I26T homozygous mutants was the presence of eye defects in the absence of any telencephalic abnormalities (n=42) (Table 2). Eye defects ranged from unilateral or bilateral microphthalmia (59.6%) to anophthalmia (16.7%). The right side was more severely affected, as described previously (Andoniadou et al., 2007). Only one Hesx1^I26T/I26T embryo showed severe forebrain defects affecting both the eyes and the telencephalon (2.4%). The remaining embryos displayed no evident forebrain abnormalities (21.4%). When Hesx1 dosage was reduced by 50% in Hesx1^I26T/– hemizygous embryos, the severity and penetrance of the ocular and telencephalic defects were increased (Fig. 2F-H; Table 2). All of the embryos analysed from 12.5–14.5 dpc had developed severe bilateral microphthalmia or anophthalmia (100%, n=19), some of them in conjunction with reduced telencephalic tissue (26.3%). Although it is likely that the enhanced severity is caused by the reduction of Hesx1 gene dosage, we cannot exclude a contribution from a genetic background effect.

Hesx1^R160C/+ mice were normal and fertile. Phenotypic analysis of 146 pups from Hesx1^R160C/+ heterozygous intercrosses indicated a gross deviation from the expected ratio of genotypes, with a significant loss of Hesx1^R160C/R160C animals (Table 1). Only five Hesx1^R160C/R160C mice were obtained at weaning, which represents approximately 3% of pups instead of the expected 25% (Table 1). All of the homozygous animals showed dramatic eye defects, typically bilateral microphthalmia or anophthalmia. Furthermore, histological analysis of 17.5 dpc Hesx1^R160C/R160C embryos and surviving adults indicated abnormal development of telencephalic commissural tracts with agenesis or hypoplasia of the corpus callosum and anterior commissure (77.8%, n=9) (data not shown).

Deviation from the expected Mendelian ratio was not observed in Hesx1^R160C/R160C embryos from 8.5–17.5 dpc (Table 1). The most common phenotype in these mutants was anophthalmia, which was associated with a significant reduction of telencephalic tissue (67.4%; n=43). This was followed by unilateral or bilateral anophthalmia in the absence of telencephalic defects (32.6%) (Fig. 2I-K; Table 2). Halving the Hesx1 dosage in Hesx1^R160C/–hemizygous embryos did not increase the severity of the phenotype; these embryos displayed the same variable expressivity of telencephalic and eye defects that were observed in Hesx1^R160C/R160C mutants (n=15) (Fig. 2L-N; Table 2). Severe craniofacial defects, i.e. defective or absent frontonasal mass, were observed in 27.9% of Hesx1^R160C/R160C embryos from 12.5–17.5 dpc (supplementary material Fig. S1D). Similar defects have been observed in Hesx1^–/–animals (Dattani et al., 1998; Andoniadou et al., 2007) (supplementary material Fig. S1B).

From these analyses, we conclude that the forebrain defects of Hesx1^I26T/I26T and Hesx1^R160C/R160C mutants are variable, but that eye tissue has increased sensitivity to impaired HESX1 function compared with the telencephalon. The Hesx1-I26T mutation leads to a less severe forebrain phenotype when compared with both the Hesx1-R160C and the previously characterised Hesx1 null alleles (P<0.001, on both severity and penetrance of defects between the Hesx1^I26T/I26T and Hesx1^R160C/R160C genotypes). There is a dosage effect for the Hesx1-I26T mutation, which is evidenced by the increased frequency and severity of forebrain abnormalities in embryos carrying one copy of the mutated allele in compound heterozygosity with a null allele (Hesx1^I26T/–), compared with those bearing two copies (Hesx1^I26T/I26T) (P<0.01 for severity and P<0.05 for penetrance). However, there are no significant phenotypic differences between Hesx1^R160C/R160C and Hesx1^R160C/– embryos, and the forebrain defects are identical to those observed in the Hesx1^–/– mutants (Dattani et al., 1998; Andoniadou et al., 2007).

Hesx1^–/– embryos show a significant reduction of anterior forebrain structures, including the telencephalon, ventral diencephalon, hypothalamus and eyes, which is caused by a posterior transformation of the anterior forebrain (Andoniadou et al., 2007). In these mutants, anterior forebrain descendants ectopically populate posterior forebrain regions and give rise to neural crest cells that colonise the frontonasal mass and first branchial arch. This fate transformation is probably the consequence of the ectopic activation of the Wnt–β-catenin signalling pathway within the anterior forebrain at 8.0 dpc. We analysed whether phenotypic differences in forebrain development between the Hesx1^I26T/I26T and Hesx1^R160C/R160C embryos could be traced back to early developmental stages. To determine this, mRNA in situ hybridisation with several diagnostic markers of brain development was performed on mutant and normal littermates between 8.0 and 9.0 dpc.

Hesx1^I26T/I26T embryos showed a variety of neural patterning defects, with some showing mild abnormalities in the expression patterns of diagnostic markers, whereas others exhibited no apparent defects. Hesx1 is normally expressed in the anterior forebrain at 8.0 dpc (Fig. 3A), and by 8.5 dpc transcripts become restricted to the ventral area of the forebrain extending into the proximal regions of the optic stalks (Fig. 3D). At 8.0 dpc, the Hesx1 expression domain was slightly reduced in the medial part of the anterior neural plate in a proportion of Hesx1^I26T/I26T homozygotes (three out of five embryos) (Fig. 3B). This was more apparent at 8.5 dpc, when Hesx1 expression was normal in the prospective ventral forebrain but severely reduced in the developing optic stalks (Fig. 3E) (four out of six embryos). However, four of the analysed Hesx1^I26T/I26T homozygous embryos (n=11) showed a very similar expression pattern to wild-type littermates (data not shown), possibly reflecting embryos that would subsequently develop normal eyes and a normal telencephalon. The expression of Wnt1, a marker of the midbrain and posterior forebrain (Fig. 3G,J) and an important contributor to the fate transformation of anterior forebrain precursors, was clearly expanded in dorsal regions of the anterior forebrain in several Hesx1^I26T/I26T embryos (three out of five embryos) (Fig. 3H,K) (McMahon and Bradley, 1990; Lagutin et al., 2003; Andoniadou et al., 2007). The reduction of presumptive optic tissue was also evident in some Hesx1^I26T/I26T embryos hybridised with riboprobes against Pax6, a marker of forebrain and eyes at these stages (two out of five embryos) (Fig. 3M,N) (Walther and Gruss, 1991). This reduction was further confirmed by analysing the expression pattern of Pax2, whose transcripts are confined to the ventral optic vesicle at early developmental stages (Fig. 3P,Q) (Torres et al., 1996). Bf1 (Foxg1b), a gene essential for proper telencephalic development, was normally expressed in the prospective telencephalon in mutant embryos when compared with wild-type littermates (n=2) (supplementary material Fig. S2I,J) (Xuan et al., 1995). From this marker analysis, we conclude that the I26T substitution leads to defective patterning of the anterior neural plate, with the eye region being more sensitive to impaired HESX1 function. This is consistent with previous data showing that the majority of these homozygous embryos have eye abnormalities but develop a normal telencephalon (Fig. 2C-E; Table 2).

Hesx1^R160C/R160C embryos showed more dramatic and consistent abnormalities in forebrain patterning at early developmental stages, because anterior forebrain tissue was typically reduced in size. The Hesx1 expression domain was significantly smaller at 8.0–8.5 dpc in homozygous mutants compared with wild-type and heterozygous littermates (n=9) (Fig. 3C,F). The degree of reduction of the Hesx1 expression domain was variable and correlated with the amount of anterior forebrain tissue present in the homozygous mutant, but all embryos analysed showed an evident decrease of anterior tissue and Hesx1 expression. This reduction in Hesx1 expression is likely to reflect the ongoing transformation of the anterior forebrain into a posterior fate in the Hesx1^R160C/R160C embryos (Andoniadou et al., 2007). In the Hesx1^R160C/R160C embryos, the expression domain of Wnt1 was anteriorised (Fig. 3I,L) to a similar extent (n=4) to that observed in Hesx1^I26T/I26T mutants (Fig. 3H,K), despite the fact that the Hesx1^R160C/R160C embryos show more severe forebrain defects. This supports the hypothesis that Wnt1 anteriorisation alone cannot account for the anterior defects of Hesx1-deficient embryos (Andoniadou et al., 2007). The Bf1 expression domain was significantly diminished in homozygous mutants when compared with wild-type or Hesx1^I26T/I26T embryos (n=2) (supplementary material Fig. S2I-K). Reduction of telencephalic and optic tissue was further confirmed by an apparent decrease in the Pax6 expression domain across the entire forebrain of Hesx1^R160C/R160C embryos (n=5) (Fig. 3M,O). Expression of Pax2 was either reduced (n=2) or absent (n=3) in optic vesicles, but normal at the mid-hindbrain boundary of these embryos (Fig. 3P,R). As previously suggested (Fig. 2I-K) (Table 2), this mutation leads to a more severe lack of anterior forebrain tissue, including the telencephalon and the eyes, which is a phenotype identical to that observed in Hesx1^–/– mutants (Dattani et al., 1998; Andoniadou et al., 2007).

Finally, we performed cell fate analysis of Hesx1-expressing cells in mutant embryos carrying either the I26T or the R160C alleles. We used two previously characterised mouse models: (1) the Hesx1-Cre mouse line, in which the Cre recombinase gene replaces the Hesx1 coding region creating a Hesx1 null allele. In Hesx1-Cre embryos, Cre expression recapitulates the endogenous Hesx1 expression pattern (Andoniadou et al., 2007). (2) ROSA26-floxstoplacZ reporter (ROSA26-Cond-lacZ), in which lacZ expression is permanently activated upon Cre-mediated excision of a loxP-flanked stop sequence (Soriano, 1999).

Compound embryos of specific genotypes were generated by genetic crosses of existing models and were analysed by X-Gal staining at 10.0 dpc. In Hesx1^Cre/+;ROSA26^Cond-lacZ/+ embryos, which are phenotypically normal, the majority of lacZ-positive cells localised within anterior forebrain structures, including the telencephalon, eyes and ventral diencephalon (n=4) (Fig. 4A,D). In contrast, Hesx1^Cre/I26T;ROSA26^Cond-lacZ/+ and Hesx1^Cre/R160C;ROSA26^Cond-lacZ/+ compound embryos, both of which have impaired HESX1 function owing to the presence of either the I26T or R160C mutations, contained a higher number of lacZ-expressing cells in both the posterior forebrain region and the first branchial arch (n=9) (Fig. 4B,C,E,F). In Hesx1^Cre/R160C;ROSA26^Cond-lacZ/+ embryos, we observed a greater contribution of lacZ-positive cells to the frontonasal mass and first branchial arch than in Hesx1^Cre/I26T;ROSA26^Cond-lacZ/+ embryos. This correlates with a higher degree of cell fate transformation, leading to a more severe phenotype in the former, as also observed in the Hesx1 null mutants (Andoniadou et al., 2007).

Taken together, the marker and cell fate analyses indicate that the anterior forebrain area of the early embryo is abnormally specified in both Hesx1^I26T/I26T and Hesx1^R160C/R160C embryos. However, the R160C substitution has a profound effect on anterior neural patterning and affects both the telencephalic and eye primordia, whereas the I26T mutation allows for better anterior specification of the neural plate and primarily affects the eye precursors.

To assess the development of the pituitary gland in Hesx1^I26T/I26T and Hesx1^R160C/R160C embryos, we performed hematoxylin-eosin (H&E) staining and in situ hybridisation analysis with diagnostic markers at different stages of embryogenesis (I26T, n=34; R160C, n=29).

Pituitary defects were clearly evident at 12.5 dpc onwards in all the embryos analysed, but some phenotypic differences were observed between Hesx1^I26T/I26T and Hesx1^R160C/R160C genotypes. At 12.5 dpc, Hesx1^I26T/I26T embryos exhibiting forebrain phenotypes ranging from normal eyes to bilateral anophthalmia (but normal telencephalon) displayed a single pituitary phenotype, which was typically characterised by the presence of an enlarged and bifurcated anterior pituitary, often connected to the oral cavity (type I phenotype) (Fig. 5D) (n=8). This phenotype was also observed in five out of eight Hesx1^R160C/R160C embryos analysed at this stage (Fig. 5G). In the other three embryos, Rathke’s pouch development was clearly delayed, appearing equivalent to that of an 11.5 dpc wild-type littermate, and remained embedded within the oral ectoderm and appeared rostrally expanded (type II phenotype) (Fig. 5J). Invariably, this latter phenotype was observed in embryos with a significant lack of anterior forebrain tissue, affecting the telencephalon and eyes (Fig. 2I). The variable phenotype observed in Hesx1^R160C/R160C embryos was reminiscent of that previously described in the Hesx1^–/– mutants (Dasen et al., 2001), further suggesting that the R160C mutation yields a null allele. At 15.5 and 17.5 dpc, the majority of the Hesx1^I26T/I26T and Hesx1^R160C/R160C embryos showed a common phenotype in which the anterior pituitary developed in its normal location but appeared enlarged, often interfering with the normal development of the basisphenoid cartilage, which was fragmented, allowing some pituitary tissue to invade the nasopharynx (Fig. 5E,F,H). In histological sections from one Hesx1^I26T/I26T (n=15) and three Hesx1^R160C/R160C (n=13) mutants from 15.5–17.5 dpc, all of which exhibited severe craniofacial defects (supplementary material Fig. S1D), the pituitary gland could not be recognised as a defined structure in its normal location (Fig. 5K,L). It is likely that these abnormalities represent the progression of the type II pituitary phenotype described in a proportion of Hesx1^R160C/R160C mutants at 12.5 dpc. Absent pituitary has been reported previously in a low proportion of Hesx1^–/– null mutants (Dasen et al., 2001).

Rathke’s pouch induction appeared to occur normally in Hesx1^I26T/I26T and Hesx1^R160C/R160C embryos. In both homozygous mutants, the expression domains of both Fgf8, a marker of the ventral diencephalon, and Lhx3, a marker of the developing Rathke’s pouch, were very similar to those observed in wild-type and heterozygous littermates at 10.5 dpc (Crossley and Martin, 1995; Sheng et al., 1996; Treier et al., 1998) (n=11) (supplementary Fig. S2A-H). These data conflict with a previous report suggesting that the Fgf8 expression domain was rostrally expanded in the ventral diencephalon in 10.5 dpc Hesx1-deficient embryos (Dasen et al., 2001). The possibility exists that Fgf8 deregulation might occur in a very small proportion of these embryos. Alternatively, the use of radioactive (Dasen et al., 2001) versus non-radioactive (this manuscript) in situ hybridisation may have contributed to this discrepancy, as radioactive in situ hybridisation is a more sensitive technique. Taken together, these data suggest that recruitment of additional oral ectoderm into Rathke’s pouch at 10.5 dpc is not a major contributor to the enlargement of the anterior pituitary at subsequent developmental stages. In fact, there is a significant increase in cellular proliferation in a proportion of Hesx1-deficient embryos from 12.5 dpc, which may account for the pituitary hyperplasia (Gaston-Massuet et al., 2008).

Despite the abnormal morphology, the expression of several diagnostic markers such as Lhx3 (Fig. 6A-I), Pit1 (Fig. 6P-R), Prop1 (Fig. 6S-U) and Hesx1 (Fig. 6V-X) was normal in the anterior pituitary of both homozygous mutants from 12.5 to 15.5 dpc, although pituitary hyperplasia was evident (Cohen et al., 1996; Treier et al., 1998; Nasonkin et al., 2004). The only significant difference was a reduction in Pomc1 expression in the developing hypothalamus of both mutant embryos at 12.5 dpc (Fig. 6J-O) and in the developing Rathke’s pouch of Hesx1^R160C/R160C embryos showing the type II phenotype (supplementary material Fig. S3). Likewise, no differences were observed in the expression domains of Sox2, Sox3, Wnt4, Wnt5a and Axin2 between Hesx1^I26T/I26T, Hesx1^R160C/R160C and wild-type littermates (Treier et al., 1998; Rizzoti et al., 2004; Cha et al., 2004; Kelberman et al., 2006; Olson et al., 2006) (supplementary material Fig. S4). Differentiation of hormone-producing cells in the anterior pituitary was unaffected, as shown by the normal temporal and spatial expression of several diagnostic markers for specific cell types, such as Gh (somatotrophs) (Fig. 7D-F,P-R), Prl (lactotrophs) (Fig. 7G-I), Tshb (thyrotrophs) (Fig. 7J-L), Cga (encoding glycoprotein hormones, alpha subunit; gonadotrophs and thyrotrophs) (Fig. 7A-C), Lhb (gonadotrophs) (Fig. 7S-U) and Pomc1 (corticotrophs) (Fig. 7M-O). However, in general, the number of hormone-producing cells appeared to be increased in the mutant pituitaries, possibly reflecting the initial enlargement of Rathke’s pouch observed in the Hesx1^I26T/I26T and Hesx1^R160C/R160C embryos at earlier stages. There appeared to be a decrease in expressing cells in only some embryos (mostly Hesx1^R160C/R160C mutants), possibly because part of the anterior pituitary tissue is ectopically located in the pharynx and is lost during processing for histological analysis. Remarkably, in those Hesx1^R160C/R160C mutants where the pituitary was not morphologically recognisable (Fig. 5L), in situ hybridisation analysis with terminal differentiation markers revealed the presence of hormone-producing cells, mostly embedded in the pharyngeal epithelium (Fig. 8). Therefore, rather than being absent, the anterior pituitary tissue was ectopically located in the roof of the nasopharyngeal cavity. A less severe manifestation of this phenotype was found in some Hesx1^I26T/I26T and Hesx1^R160C/R160C embryos, where only part of the anterior pituitary invaded the nasopharyngeal mucosa (Fig. 5E,F,H; data not shown).

Expression of human HESX1 has not been reported previously. This information is relevant to advance our understanding of the relationship between the phenotypic consequences of particular mutations in mice and humans. At Carnegie stage (CS) 11 (∼8.5–9.0 dpc equivalent in mouse), HESX1 expression was detected by in situ hybridisation in the ventral forebrain and in the invaginating oral epithelium of the developing Rathke’s pouch (Fig. 9A,B). By CS 12 (∼9.5–10.25 dpc), neural expression could not be detected but HESX1 transcripts were abundant in Rathke’s pouch, mainly in the dorsal region (Fig. 9C,D). Rathke’s pouch expression persisted until CS 15 (∼11.0–11.5 dpc) (Fig. 9E-H) but was not detected at CS17 (∼12.0–12.5 dpc) (Fig. 9I-J). Two conclusions can be drawn from this analysis: (1) HESX1 expression is transient in the developing human embryo and is restricted to the anterior region, including ventral forebrain and Rathke’s pouch; (2) it establishes the mouse as an ideal model to study the phenotypic consequences of HESX1 mutations associated with forebrain and pituitary defects in humans.

The results presented here have revealed some interesting insights into the function of Hesx1/HESX1 in the aetiology and pathogenesis of SOD and hypopituitarism. We have shown that the expression domain of human HESX1 is comparable to that in the mouse, and includes the ventral forebrain and Rathke’s pouch. Based on cell fate studies in the mouse and chick, which have shown that Hesx1-expressing cells colonise the anterior forebrain (including ventral forebrain) at early somite stages and eyes at later stages (Fernandez-Garre, 2002; Andoniadou et al., 2007), it is likely that HESX1 is also expressed in the anterior neural plate (presumptive anterior forebrain) at earlier stages of human embryogenesis. Unfortunately, owing to the difficulty of obtaining very early-stage human embryos, we could not confirm this hypothesis. The human expression pattern may also provide an explanation for the eye defects seen in patients carrying HESX1 mutations (Kelberman and Dattani, 2007).

We have also generated knock-in mouse mutants harbouring the I26T and R160C substitutions to analyse the phenotypic consequences of impaired HESX1 function in mice. Our data indicate that, within the anterior forebrain, eye defects are more common than telencephalic abnormalities, including commissural tract defects, in Hesx1^I26T/I26T and Hesx1^R160C/R160C mutant embryos. This suggests that the eye precursors are more sensitive to impaired HESX1 function than the antecedents of the telencephalon. This conclusion is supported by further evidence: (1) Hesx1^–/– null embryos show fully penetrant eye defects owing to ectopic activation of the Wnt–β-catenin signalling pathway within the anterior forebrain (Dattani et al., 1998; Andoniadou et al., 2007); (2) transgenic overexpression of Hesx1 within the anterior forebrain of Hesx1^–/– mouse embryos rescues the telencephalic defects at a lower dosage than for eye abnormalities, which require higher levels of HESX1 (Andoniadou et al., 2007); (3) in Xenopus and zebrafish, activation of the Wnt–β-catenin signalling pathway within the anterior forebrain primarily affects eye development (Fredieu et al., 1997; van de Water et al., 2001). The HESX1-I26T mutant protein shows diminished repressing activity when compared with wild-type HESX1, probably because of reduced binding to the co-repressor TLE1, which requires the highly conserved amino acid I26 within the eh-1 domain for proper binding (Galliot et al., 1999; Dasen et al., 2001; Carvalho et al., 2003). Since TLE1 is expressed across the entire forebrain (Allen and Lobe, 1999; Lopez-Rios et al., 2003), telencephalic precursors are probably less sensitive to the lack of HESX1-repressing activity mediated by TLE1. Studies in zebrafish have shown that TLE1 plays a fundamental role in eye development, as TLE1 overexpression leads to enlargement of the eyes with little or no effect on the telencephalon (Lopez-Rios et al., 2003). Therefore, HESX1-I26T is likely to allow interactions with other co-repressors within telencephalic precursors, which may compensate for the impaired HESX1-I26T–TLE-1 interaction. Among them, nuclear co-repressor (N-CoR) and some of the recently identified HESX1 interactors may contribute to the differences in forebrain phenotype between the I26T and R160C mutations (Dasen et al., 2001; Sajedi et al., 2008) (data not shown).

The pituitary defects of Hesx1^I26T/I26T and Hesx1^R160C/R160C are fully penetrant and usually very similar, although the morphogenesis of the pituitary gland is more affected in Hesx1^R160C/R160C embryos that show severe forebrain defects. Our data suggest that the signals controlling terminal differentiation of anterior pituitary cells are acting at the right place and time, but, because of an initial enlargement of Rathke’s pouch at 12.5 dpc, there are more cells that can follow the differentiation pathway. However, a minority of embryos with ectopic pituitary tissue showed fewer numbers of hormone-producing cells compared with wild-type embryos. This is reminiscent of the ectopic pharyngeal pituitary gland that has been reported in humans in association with craniofacial defects (Kjaer and Hansen, 2000; Osman et al., 2006).

The fact that Hesx1^I26T/I26T embryos display pituitary defects that are comparable to Hesx1^R160C/R160C and Hesx1^–/– embryos suggests a fundamental role for the HESX1-TLE1 interaction during normal pituitary development that cannot be compensated for by other HESX1-interacting proteins, as was previously suggested by overexpression experiments in mice (Dasen et al., 2001). Given the complete penetrance of pituitary defects in animals harbouring both the hypomorphic (I26T) and null (R160C) alleles, it is likely that the pituitary gland is highly sensitive to HESX1 dosage, and that the high incidence of hypopituitarism in humans with HESX1 mutations is not a consequence of selection bias. Our studies indicate that the R160C substitution yields a null allele, whereas I26T is a hypomorphic allele. Additionally, they provide in vivo evidence that the R160C substitution does not have a dominant-negative effect as suggested previously by in vitro data (Brickman et al., 2001) as Hesx1^R160C/+ mice are normal, viable and fertile.

We propose that, in the mouse, Hesx1 function is required primarily in Rathke’s pouch, secondarily in eye precursors and finally in the antecedents of the telencephalon. It is possible that this differential sensitivity to HESX1 function is conserved in the human embryo as the Hesx1/HESX1 expression domains are very similar between the two species. Furthermore, phenotypic analysis indicates that anterior pituitary dysfunction, ranging from IGHD to CPHD, is the most common clinical finding in patients with HESX1 mutations, followed by optic nerve hypoplasia and midline commissural defects (Dattani et al., 1998; Thomas et al., 2001; Brickman et al., 2001; Carvalho et al., 2003; Cohen et al., 2003; Tajima et al., 2003; Sobrier et al., 2005; Sobrier et al., 2006; Coya et al., 2007). It is interesting that the human patient carrying the I26T substitution does not have eye defects (Carvalho et al., 2003), although more than 75% of the Hesx1^I26T/I26T mutant mice exhibited an eye phenotype. This may be the result of species-specific differences between mice and humans, or, alternatively, the human patient may just have a milder phenotype, as approximately 21% of Hesx1^I26T/I26T mouse mutants do not have eye defects.

The reasons underlying perinatal lethality in the majority of Hesx1^R160C/R160C animals and perhaps in a minority of Hesx1^I26T/I26T mutant pups are not clear. As pituitary defects are fully penetrant and similar in both mutants, it seems unlikely that these morphological abnormalities alone can cause mortality. The severe craniofacial defects in the Hesx1^R160C/R160C pups are also likely to contribute to some perinatal death; however, pups without gross morphological defects in the brain and craniofacial structures also die soon after birth. Abnormal development of the hypothalamic region and/or its connection to the pituitary gland might underlie the observed perinatal death. Indeed, hypothalamic Pomc1 expression is reduced in mice carrying the I26T or the R160C substitution, with a more severe loss of signal in Hesx1^R160C/R160C embryos. As pituitary function is controlled by the neuroendocrine hypothalamus, the Hesx1^R160C/R160C pups might be expected to show more severe hypopituitarism than the Hesx1^I26T/I26T mice, as described for human patients carrying these mutations. Unfortunately, the Hesx1^I26T/I26T mice do not show any sign of hypopituitarism and were viable and fertile, and the Hesx1^R160C/R160C animals die too early to assess any sign of hypopituitarism in the form of delayed growth. Further investigation will be required to assess, in more detail, the development of the neuroendocrine hypothalamic nuclei and their connections to the pituitary gland. Analysis of a systematic mutational screen has shown that mutations in HESX1 are a rare cause of SOD and hypopituitarism (McNay et al., 2007). It is possible that those surviving patients represent the milder end of the phenotypic spectrum. As with the low numbers of surviving mice, it is likely that a significant proportion of HESX1 mutations in humans are not viable and therefore remain undetected. Although this hypothesis is difficult to test, the mouse data presented here support this idea.

The Hesx1-I26T and Hesx1-R160C targeting vectors were generated using homologous regions obtained from plasmids containing the mouse Hesx1 gene, which have been previously used successfully (Fig. 1A,B). A PGK-Neo cassette flanked by loxP sequences was inserted in the EcoRI site located in the first intron of both targeting constructs. In the Hesx1-I26T targeting vector, the PGK-Neo cassette was cloned in the same orientation of transcription as the Hesx1 locus. However, the orientation was inverted in the Hesx1-R160C targeting vector in an attempt to reduce the expression of the mutated allele, which may have had a dominant negative effect in embryonic stem (ES) cells since the Hesx1 locus is transcriptionally active in these cells. The codons for amino acids I26 and R160 of murine wild-type Hesx1 were mutated by PCR, as indicated in Fig. 1C,D (primer sequences available on request). The mutated codons introduced novel restriction sites that are not present in wild-type Hesx1, and which were very useful as a quick method to identify the presence of the mutated alleles during generation of the mouse lines. The linearised targeting vectors were electroporated in CCE ES cells (129/SvEv) (kindly provided by E. Robertson) and 500 colonies were picked, expanded and screened by PCR and Southern blot, as described previously (Andoniadou et al., 2007). For each construct, two correctly targeted clones were isolated and injected into blastocysts from C57BL/6J (Harlan) mice. Male chimeras were backcrossed to C57BL/6J females to establish the F1 generation of heterozygous mice. F1 animals were crossed with beta-actin-Cre mice, which were kept on a C57BL/6J background, to excise the PGK-Neo cassette (Meyers et al., 1998). After backcrossing with C57BL/6J animals to remove the beta-actin-Cre transgene, Hesx1^I26T/+ and Hesx1^R160C/+ heterozygotes were kept on a C57BL/6J background. The analysis described here used animals and embryos after three backcrosses to C57BL/6J. The Hesx1^+/– mice used in this study had been maintained on a C57BL/6J background for more than 20 generations (Andoniadou et al., 2007). With the exception of the opposite orientation of the remaining loxP sequence after Cre-mediated excision of the PGK-Neo cassette and the specific point mutation, both Hesx1-I26T and Hesx1-R160C alleles were identical. This was relevant in order to eliminate any adverse effect of the remaining loxP site on transcription and splicing efficiency, which might impact the phenotype.

Embryos and neonates were genotyped by PCR on DNA samples prepared from tail tips, yolk sacs or whole embryos (Andoniadou et al., 2007). Primer sequences and PCR protocols are available on request.

H&E and X-Gal staining, and whole-mount in situ hybridisation were performed as described previously (Andoniadou et al., 2007). For in situ hybridisation on paraffin sections, the following protocol was used: mouse embryos at 10.5, 12.5, 15.5 and 17.5 days post coitum (dpc) were dissected and fixed overnight in 4% paraformaldehyde (PFA) (Sigma), followed by dehydration in a graded ethanol series and embedding in paraffin. Sagittal and coronal sections (6–8 mm) were cut using a standard microtome. Sections were de-waxed, hydrated and fixed in 4% PFA (20 minutes). After proteinase K treatment (8 minutes at room temperature) and fixation (4% PFA, 5 minutes), they were treated with 0.1 M triethanolamine plus 0.25% acetic anhydride (10 minutes). Hybridisation was carried out overnight at 55–65°C in 50% formamide, 0.3 M sodium chloride, 20 mM Tris HCl, 5 mM EDTA, 10% dextran sulphate, 1×Denhardt’s reagent (Sigma) supplemented with tRNA (Sigma), RNAase inhibitor (Roche) and the relevant digoxigenin-labelled riboprobe. Stringency washes consisting of 2×SSC (twice, 30 minutes), 50% formamide:50% 2×SSC (twice, 30 minutes) and then 2×SSC (twice, 30 minutes) were carried out the following day at 65°C. Slides were then blocked for 1 hour in 10% foetal calf serum (FCS) and placed in a humid chamber in a buffer containing: 0.1 M Tris pH 7.6, 0.15 M sodium chloride, anti-digoxigenin antibody conjugated with alkaline phosphatase (1:1000 from Roche), and 2% FCS. Staining was carried out the following day using the NBT/BCIP system (Roche). Sections were mounted in Vectamount (Vector Laboratories). Human embryonic/foetal material was obtained from the Human Developmental Biology Resource (HDBR) and from the M. Vekemans Research Group at Necker Hospital, Paris with full ethical approval under both British and French bioethics regulations. In situ hybridisation was performed on histological sections of embryos from Carnegie stages (CS) 11 to 18. Embryos at stages earlier than CS 11 were unavailable.

Deviations from the Mendelian ratios in offspring were analysed for statistical significance using a chi-square test. The severity and penetrance of the forebrain defects between: (1) Hesx1^I26T/I26T and Hesx1^R160C/R160C; (2) Hesx1^I26T/I26T and Hesx1^I26T/–; and (3) Hesx1^R160C/R160C and Hesx1^R160C/– were analysed using a Fisher’s exact test. To detect possible differences in severity, only the proportion of embryos showing the most severe forebrain defects were compared, and for penetrance we used the proportions of embryos with/without forebrain defects (Table 2).

We are grateful to A. Copp for comments on the manuscript; A. McMahon, G. Martin, P. Gruss, M. Rosenfeld, R. Lovell-Badge, B. Hogan and the MRC Geneservice for probes; the MRC/Wellcome-funded Human Developmental Biology Resource; and M. Vekemans and M. Teboul for access to human tissues and in situ hybridisation. This work was supported by grants 068630 and 078432 from The Wellcome Trust, grant 1ZTG from UCL Central Funds, a grant from the Association Francaise des Myopathies (H.C.E.) and by the Medial Research Council (M.T.D. and D.K.). E.S. is the recipient of a PhD studentship funded by the Child Health Research Appeal Trust. Deposited in PMC for immediate release.