Regulation of ocular lens development by Smad-interacting protein 1 involving Foxe3 activation

ZFHX1 family transcription factors, comprising δEF1 and Sip1 in higher vertebrates, contain bipartite zinc-finger clusters and a homeodomain-like motif, named after the Drosophila counterpart Zfh-1(Fortini et al., 1991).δEF1 was identified by its binding to a CACCT sequence of aδ -crystallin enhancer(Funahashi et al., 1991; Funahashi et al., 1993; Kamachi and Kondoh, 1993),while Sip1 (Smad-interacting protein 1) was identified by its binding to Smad proteins mediated by the Smad-binding domain (SBD) of Sip1(Verschueren et al.,1999).

Sip1 and δEF1 are very similar in structure and share DNA-binding sequence specificity to the E2-box-like motif CACCT(G) in vitro; their activity as a transcriptional repressor has been demonstrated using several reporter constructs (Comijn et al.,2001; Kamachi et al.,1995; Postigo and Dean,1997; Remacle et al.,1999; Sekido et al.,1994; Sekido et al.,1997; van Grunsven et al.,2001), suggesting a shared regulatory function. However,δEF1 lacks the SBD sequence and does not bind Smad proteins in vitro(van Grunsven et al., 2003),suggesting a unique Smad-dependent regulation exerted by Sip1. Expression patterns in the mouse embryo partly overlap but are generally diversified between two protein genes, e.g. only Sip1 is expressed in the lens and null mutant mouse phenotypes produced by targeted gene inactivation are distinct(Takagi et al., 1998; Van de Putte et al.,2003).

Although the augmentation of the Sip1 binding of Smads by Alk receptor-mediated phosphorylation through their MH2 domain has been demonstrated (Verschueren et al.,1999), whether or how the binding of a Smad protein affects transcriptional regulation by Sip1 has not been elucidated. Although all experiments using full-length Sip1 and CACCT(G)-containing target sequences indicate a repression activity, supported by its binding to the co-repressor CtBP (van Grunsven et al.,2003) as for δEF1(Furusawa et al., 1999), it may still be only one of the functions of this multi-faceted protein.

During the embryonic development of mice, early Sip1 expression in gastrula (e.g. E8.0) is observed primarily in the neural plate, neural crest and paraxial mesoderm (Van de Putte et al., 2003); however, late Sip1 expression occurs in various tissues (T. Miyoshi, M. Maruhashi, T. Van de Putte, H.K., D. Huylebroeck and Y.H., unpublished). The Sip1-null knockout mouse embryos die around E9.5 after heart dysfunction and embryo turning failure(Van de Putte et al., 2003). Thus, the floxed (flanked by loxP sites) Sip1 allele was generated(Higashi et al., 2002), which enables cell-lineage-specific Sip1 inactivation.

As demonstrated in this study, in the lens lineage, Sip1 is expressed after lens placode induction. The lens is a simple tissue and is one of the best characterized in terms of transcriptional regulation(Kondoh, 1999; Kondoh, 2002). The lens-lineage-specific ablation of the floxed Sip1 gene can be achieved by using the lens enhancer of Pax6(Kammandel et al., 1999; Williams et al., 1998) in controlling Cre recombinase. Therefore, the consequence of lens-specific Sip1 inactivation was investigated, and two steps in lens development dependent on Sip1 activity were characterized: (1) the separation of the lens epithelium and surface ectoderm by the removal of the connecting lens stalk;and (2) the progression of lens fiber precursors in the bow region intoγ-crystallin-expressing mature fiber cells. In this study Foxe3activation, which is involved in the first step, was demonstrated to be dependent on Sip1 activity. The Foxe3 promoter was activated by Sip1 and this activation was augmented by the specific interaction of Sip1 with Smad8 in a transfection assay. This study is the first clear demonstration that Smad-Sip1 interactions are significant in transcriptional regulation.

Given evidence of the involvement of Sip1 in many important processes of embryogenesis, not limited to the lens(Eisaki et al., 2000; Papin et al., 2002; Sheng et al., 2003; Van de Putte et al., 2003; van Grunsven et al., 2000),this study provides new insight into the regulatory functions of this interesting transcription factor.

The mouse line carrying the floxed Sip1 allele has been described(Higashi et al., 2002). The Pax6(LP)-Cre transgenic line [Lens-Cre (see Ashery-Padan et al., 2000)] was provided by Drs A Mansouri and P. Gruss through Dr S. Watanabe of the University of Tokyo. The Pax6(Lens)-Cre transgene was constructed by ligating the 340 bp lens-specific enhancer of Pax6(Kammandel et al., 1999; Williams et al., 1998) to the Pax6 P0 promoter and Cre-encoding sequence. From seven founder lines,the one showing the highest Cre activity, as determined by crossing with the R26R mouse (Soriano, 1999)(obtained from the Jackson Laboratory) was employed. The dyl mutant mouse (Blixt et al., 2000; Brownell et al., 2000) was also from the Jackson Laboratory. The Pax6(LP)- or Pax6(Lens)-Cre transgene in a male mouse was introduced into homozygous floxed Sip1 background by mating for two generations with floxed Sip1 homozygotes, and embryos derived from the mating of a floxed Sip1-homozygous female and a Cre-carrying homozygous male were used in this study. In some cases, the R26R transgene was also introduced in experimental embryos to monitor Cre recombinase activity. X-gal staining for β-galactosidase activity in the embryos was carried out as previously described(Muta et al., 2002). The presence of the Cre transgene in mice was determined by detecting a 235 bp PCR product of DNA extracted from an ear punch, using the primers 5′AGGTTCGTTCACTCATGGA3′ (forward) and 5′TCGACCAGTTTAGTTACCC3′ (reverse), and the lacZ sequence by detecting a 478 bp product, using the primers 5′TTGCCGTCTGAATTTGACCTG3′ (forward) and 5′TCTGCTTCAATCAGCGTGCC3′ (reverse). Normal and floxed Sip1 alleles were distinguished analogously(Higashi et al., 2002). Mice were maintained in C57B6/C3H mixed background.

Anti-crystallin antibodies were raised in rabbit by injecting the following peptides ligated to keyhole limpet hemocyanin. Anti-α-crystallins(recognizing both αA and αB), CVSREEKPSSAPSS;anti-βA-crystallins without cross-reaction to the βB class,CHAQTSQIQSIRRIQQ; and anti-γ-crystallins, GKITFYEDRSFQGRC. The embryos were fixed with 4% paraformaldehyde in phosphate-buffered saline at 4°C overnight, dehydrated, embedded in paraffin and cut into 6 μm serial sections. The sections were treated with anti-crystallin primary antibodies and Alexafluor568-conjugated anti-rabbit IgG (Molecular Probe) antibodies,stained for nuclei with DAPI and mounted in Permafluor anti-fade reagent(Immunotech).

Embryo sections were hybridized with specific probes as described previously (Uchikawa et al.,1999). The full-length Sip1 cDNA probe(Van de Putte et al., 2003), Pax6 3′ UTR probe (Xu et al., 1999), Sox1 3′ UTR probe(XhoI-StuI fragment), and probes for Foxe3, Maf,γ-crystallins and Pdgfra(Yamada et al., 2003) were used.

Apoptotic cells in histological sections were detected by the TUNEL technique using an Apo-Alert DNA Fragmentation Assay kit (Clontech). TUNEL-positive nuclei among DAPI (4′,6-diamidino-2-phenylindole)-stained nuclei in the anterior and posterior lens halves were counted in meridian sections through lenses, and data of individual embryo specimens were combined.

Lens epithelial cells were prepared from E14 chicken embryos and cultured for transfection as previously described(Muta et al., 2002). Collagen-coated 24-well plates were inoculated with one-fifth of the epithelial cells derived from one lens per well. Similarly, E10 gizzard cells were inoculated at 4×10⁴ cells per well. A 1.5 μg plasmid-DNA mixture for transfection, typically containing 100-200 ng of a Foxe3 promoter-ligated GL3 firefly luciferase gene (Promega), 1-500 ng of effector plasmids and 10 ng of phRG-TK Renilla luciferase expression plasmid, was transfected into cells in a well using 3 μg of Fugene6(Roche). Luciferase activity was measured after 48 hours using a Dual Luciferase Reporter Assay kit (Promega) and an LB940 Mithras Multilabel Reader(Berthold Technologies), normalizing firefly luciferase activity using Renilla luciferase. Transfection was carried out at least in triplicate. The activity of Smad expression vectors (provided by Drs M. Kawabata and K. Miyazono)driven by the elongation factor I enhancer/promoter was confirmed by the activation of p3GC2-Lux (Smad1, 5 and 8) or p3TP-Lux (Smad 2 and 3)(Ishida et al., 2000) in lens epithelial cells.

Sip1 expression in early eye development was examined by the in situ hybridization of transcripts. Sip1 is expressed widely in embryonic tissues, and in both lens and retinal components of the eye, but its expression pattern in the lens dynamically changes with developmental stage(Fig. 1). At E9.0, before the occurrence of lens induction, Sip1 is not expressed in the surface ectoderm (Fig. 1A), but in parallel with lens placode induction at E9.5, Sip1 expression is initiated (Fig. 1B). As the lens vesicle is formed (E10.5-E11.5), Sip1 is expressed in the entire vesicle (Fig. 1C,D). After E12,primary lens fiber cells differentiate in the posterior lens, where the Sip1 expression level is very low, while a high level of Sip1 expression continues in the lens epithelium and immature lens fibers in the equatorial bow region (Fig. 1E).

To clarify the intrinsic functions of Sip1 in lens cells, the Sip1gene was ablated in a lens-lineage-specific fashion, using embryos homozygous for the floxed Sip1 allele, in which the action of Cre recombinase causes the loss of detectable Sip1 protein(Higashi et al., 2002)(Fig. 2A).

A lens-lineage-specific Cre-expressing mouse line developed by Ashery-Padan et al. (Ashery-Padan et al.,2000) and another line established in this study were used; both lines use the Pax6-lens enhancer. The Pax6 lens enhancer of 340 bp located 3.7 kb upstream of the P0 promoter gains its activity in the surface ectoderm of the prospective eye area, starting at E9.0 and its activity is maintained in its derivatives, lens epithelium and cornea in later development (Kammandel et al.,1999; Williams et al.,1998). The mouse line Pax6(LP)-Cre(Ashery-Padan et al., 2000)carries a large Pax6 upstream region of 6.5 kb, including a pancreas enhancer (Fig. 2B). By crossing the mouse with an R26R reporter mouse(Soriano, 1999), Cre recombinase activity in the surface ectoderm, not confined to the eye area but extending to the olfactory epithelium area, was demonstrated(Fig. 2C, parts a,c). By contrast, the second mouse line developed in this study, carrying only the minimal lens enhancer, Pax6(Lens)-Cre (Fig. 2B), showed a strong activity in the lens and surface ectoderm confined to the eye-proximal region (Fig. 2C, parts b,d). Using histological sections, the specificity of Cre recombinase action in the lens and ectoderm was confirmed(Fig. 2C, parts e-h).

When these Cre transgenes were introduced into homozygous floxed Sip1 embryos, they developed defective eyes; otherwise, they exhibited normal growth and fertility. In the homozygous floxed Sip1mouse population, the Cre transgenes were transmitted according to the Mendelian ratio. The lens defect was identical between the two Cre transgenic lines, Pax6(LP)-Cre and Pax6(Lens)-Cre; therefore, data using these two Cre lines were combined and used in the following analysis.

The consequence of the loss of Sip1 activity in the lens lineage was investigated at the histological level using various markers. Expression of Cre did not interfere with lens placode development (E9.5) (data not shown) or invagination (E10.5) (Fig. 3A,D); however, the first marked defect was observed at E11.5(Fig. 3B,E) when the lens vesicle was normally separated from the surface ectoderm by tissue reorganization and local apoptosis involving the connecting lens stalk(van Raamsdonk and Tilghman,2000). The Cre recombinase action in the lens lineage was confirmed using R26R mouse background. In the Sip1-defective lens, a thick stalk connecting the vesicle and ectoderm was persistent, and the vesicle was smaller (Fig. 3E). The lens vesicle that developed in the floxed Sip1 embryo lacked Sip1 expression in the surface ectoderm, stalk and anterior region of the vesicle, but had some residual Sip1 expression in the posterior half, as determined by in situ hybridization(Fig. 3F). This is in contrast to the uniform expression of Sip1 in the lens vesicle in the normal embryo (Fig. 3C). Thus, in the absence of Sip1 activity, the cells positioned in the lens stalk persist.

At E14.5 and even at newborn (P0) stages the lens stalk still remains when Sip1 is inactivated in the lens lineage(Fig. 3I,J), and a defect in lens mass development is evident. A stalk-persistent lens produced in dyl/dyl (Foxe3 mutant) lens is shown in Fig. 3K for comparison.

When apoptotic cell distribution was measured using the TUNEL method from E10.5 to E12.5, apoptotic cells were mainly distributed in the anterior half of both normal and Sip1-defective lenses. In Sip1-defective lenses, the apoptotic cell population increased significantly, but the increment in apoptosis rate relative to that observed in the normal lens was comparable between the anterior and posterior halves at E10.5 and E11.5 without any specific suppression of apoptosis in the stalk region, and higher in the anterior half at E12.5 (Fig. 3L,M). Therefore, the increased apoptosis rate in Sip1-defective lenses accounts for the smaller lens size, but does not appear to contribute to either the persistence of lens stalk, or the specific loss of mature lens fibers in the posterior lens to be described below.

To clarify the cellular and molecular bases of lens development defects,various lens markers were examined at the histological level. At E12.5,anti-αA/B-crystallin antibodies stained all cells of normal and Sip1-defective lenses (Fig. 4A,F). βA-crystallins are expressed at a low level in immature fiber cells in the bow region where the Sip1 expression level is high (Fig. 1E), and at a high level in mature lens cells (Fig. 4B). In Sip1-defective lenses, only a lowβA-crystallin expression level was observed in the posterior-most side of the lens vesicle, which may correspond to the bow region of the normal lens. Mature lens fiber cells are marked by γ-crystallins in normal lenses(Fig. 4C). The most distinct characteristic of a Sip1-defective lens is the total absence ofγ-crystallin expression (Fig. 4H). This absence of γ-crystallin expression was confirmed using in situ hybridization (data not shown).

The above observations indicate that the Sip1-defective lens lacks the region of a mature lens with γ-crystallin expression, located between the arcs indicated in Fig. 4C. As the Sip1-defective lens containsβ-crystallin-positive regions (Fig. 4G), which presumably correspond to those outside arcs in normal lenses (Fig. 4B), the data are consistent with the model of a strong Sip1 expression in the bow region promoting the maturation of fiber cells; in the absence of Sip1 expression in the bow region, cells arrest at the premature stage.

γ-Crystallin genes are regulated by Sox1 and Maf, which synergistically function in the activation of their lens-specific promoters(Kamachi et al., 1995; Kawauchi et al., 1999; Kim et al., 1999; Nishiguchi et al., 1998; Ring et al., 2000), but the inactivation of Sip1 does not affect expression of these transcription factor genes, as indicated by in situ hybridization(Fig. 4D,E,I,J).

Thus, two major defects were observed in the Sip1-defective lens:(1) persistent lens stalk; and (2) the arrest of lens fiber cell maturation at the bow-region stage.

A persistent lens stalk is called Peter's anomaly in human congenital diseases, and accompanies the inactivation of the transcription factor Foxe3(Blixt et al., 2000; Brownell et al., 2000). Homozygous dyl (dysgenetic lens) mice with a mutation in the Foxe3 gene (Blixt et al.,2000; Brownell et al.,2000) or a low Pax6 activity affecting Foxe3 expression(Brownell et al., 2000; Dimanlig et al., 2001) are documented examples. This prompted us to examine the possibility that Foxe3 activity is affected in Sip1-defective lenses(Fig. 5). In addition to Foxe3 (Fig. 5A,E),Pdgfra gene under Foxe3 regulation(Blixt et al., 2000)(Fig. 5B,F), and an upstream gene Pax6 required for Foxe3 expression(Brownell et al., 2000)(Fig. 5C,G) were analyzed for their expression using in situ hybridization.

Foxe3 expression in the normal lens was divided into two domains:a strong expression in the anterior domain, including the lens epithelium and bow region; and a weak expression in the posterior domain(Fig. 5A). The Sip1-defective lens lacked the strong Foxe3 expression in the anterior domain (Fig. 5E),accounting for the persistence of the stalk analogous to the Foxe3mutant. However, a weak Foxe3 expression in the posterior domain of lens vesicle remained. This may reflect a low-level Sip1 transcript remaining in the posterior half of the Sip1-ablated lens(Fig. 3F) or, alternatively, Foxe3 expression in posterior lens cells may be regulated by a different mechanism. Pdgfra is regulated by Foxe3 and its expression pattern in the normal lens mirrors that of Foxe3(Blixt et al., 2000)(Fig. 5B). In the Sip1-defective lens, Pdgfra expression was downregulated in the anterior domain, paralleling the loss of Foxe3 expression(Fig. 5F). These observations place Sip1 upstream of Foxe3. Consistently, Sip1 is expressed throughout the lens at a normal level in Foxe3-mutant dyl mice (Fig. 5D,H).

Pax6 expression was basically unaffected in the Sip1-defective lens (Fig. 5C,G). In contrast to Pax6 expression in the normal lens,which is downregulated in the posterior region, Pax6 expression prevails throughout the lens in Sip1-defective lens, but this difference is accounted for by the lack of a mature fiber compartment in the latter. Thus, Sip1 and Pax6 are both assigned as upstream regulator of Foxe3. As Pax6-null embryos develop no lens structure(Hill et al., 1991; Hogan et al., 1986), it was not determined whether Sip1 itself is regulated by Pax6.

The 6.2 kb 5′ flanking region upstream of the SmaI site has promoter activity sufficient for controlling the expression of a lacZtransgene in the mouse lens (Brownell et al., 2000). Whether the Foxe3 promoter is regulated by Sip1 and whether this regulation depends on interaction with Smads were investigated. The 6.2-kb promoter sequence was ligated to a luciferase reporter gene (Fig. 6A), then transfected with a Sip1 expression vector into primary-cultured lens epithelial and gizzard (smooth muscle) cells of chicken embryos. Results using these cells were similar and data for the lens epithelial cells are shown in Fig. 6.

The activity of the 6.2 kb promoter was progressively augmented by the exogenous expression of Sip1, sevenfold activation by 100 ng of expression vector and 20-fold activation by 500 ng of expression vector(Fig. 6B). The same increase was observed using a mutant Sip1 lacking the Smad-binding domain (SBD)(Fig. 6B), indicating that this activation occurs without interaction with a Smad protein.

Under the same transfection conditions, various Smad proteins were expressed by co-transfection (Fig. 6C). Smad1, Smad5 and Smad8, mediating BMP signals, and Smad2 and Smad3, mediating TGFβ signals, bind Sip1 in vitro through the SBD(Verschueren et al., 1999). Of these Smads, only Smad8 caused a significant Sip1-dependent activation of the Foxe3 promoter, threefold more activation than Sip1 alone, while Smad8 by itself had no effect on the Foxe3 promoter(Fig. 6C). However, using SBD-deleted Sip1, exogenous Smad8 did not augment Sip1-dependent activation(Fig. 6C). Under the same transfection condition, Smad1, Smad5 and Smad8 exhibited comparable levels of transactivation of the 3GC2-luciferase reporter gene(Fig. 6D). It was rather unexpected that Smad1 or Smad5 had no significant effect under the transfection condition, although they are generally considered to act analogously to Smad8 (ten Dijke and Hill,2004). When an inhibitory Smad, Smad6 or Smad7, was co-expressed with Smad8, the effect of Smad8 was cancelled and only the activation level attainable by Sip1 alone remains (Fig. 6E). The effect of Smad8 was saturated at approximately threefold the activation level attained by Sip1 alone, regardless of whether the original activation level by Sip1 alone was sevenfold(Fig. 6F) or nearly 20-fold(Fig. 6F).

These observations indicate that augmentation by Smad8 of Sip1-dependent activation of the Foxe3 promoter is through SBD-mediated direct interaction between Sip1 and Smad8. They also indicate a two-step mechanism for the activation of the Foxe3 promoter by Sip1, with moderate activation without the assistance of Smad8, and high-level activation with the Sip1-Smad8 complex.

The 6.2 kb promoter region was divided into four blocks, A to D, from the proximal side, and the block responsible for activation by Sip1 and Smad8 was then investigated. The removal of the upstream blocks B to D did not greatly affect the activation of the Foxe3 promoter by Sip1 or by Sip1 plus Smad8, and the 1.26 kb block A promoter region is sufficient(Fig. 6G). As shown below using transgenic mouse embryos, block A is not involved in the lens-specific regulation of the Foxe3 promoter. The Sip1-dependent activation of the Foxe3 promoter and its augmentation by Smad8 is also observed in gizzard cells (see Fig. S1 in the supplementary material), confirming their tissue non-specific effect.

The shortening of the promoter sequence to 287 bp maintains the capacity for Sip1-dependent activation and further augmentation by Smad8(Fig. 6G). Further shortening of the promoter to 127 bp resulted in the loss of response to Sip1. The unrelated δ-crystallin promoter was not affected by either Sip1 or Sip1 plus Smad8 (Fig. 6G). This observation suggests that the activation of the Foxe3 promoter by Sip1 and Smad8 involves the proximal region.

When the 6.2 kb Foxe3 promoter was ligated to a lacZtransgene construct and primary transgenic mouse embryos were produced, lens-and brain-restricted lacZ expression was observed at E12.5(Fig. 7A), confirming a previous report (Brownell et al.,2000). To investigate tissue-specific regulation, the effect of deleting various blocks was examined (Fig. 7A). The removal of the most distal block D, resulting in the promoter blocks A+B+C, did not have any appreciable effect, but further deletion of block C leaving promoter blocks A+B caused a large decrease in the expression level in the lens and the loss of expression in the brain. When block B was removed from A+B+C blocks, leaving A+C blocks of the promoter region, transgene expression in the lens and brain was indistinguishable from that using the full 6.2 kb sequence. By contrast, with only the most proximal block A, transgene expression was not observed. These results indicate that block C includes the major lens-specific element and a brain element, and block B contains a minor lens element, and that the combination of the activity of these blocks with the Sip1-dependent, cell type nonspecific activity of block A elicits Foxe3 expression in embryonic lenses, as summarized in Fig. 7B.

This study clarified the roles of the Sip1 transcription factor in ocular lens development, using a lens-lineage-specific gene ablation strategy(Fig. 2A). Two lens-specific Cre transgenic lines were used, both taking advantage of the lens-specific enhancer of the Pax6 gene. Pax6(LP)-Cre(Ashery-Padan et al., 2000)carries a large Pax6 upstream region driving a gene in the head ectoderm and pancreas, in addition to the lens lineage, and Pax6(Lens)-Cre carries only the lens/cornea enhancer developed in this study(Fig. 2B). These two Pax6-enhancer-dependent Cre lines gave identical results when crossed with the floxed Sip1 mouse. Given the narrower tissue restriction of Cre recombinase action (Fig. 2C), the Pax6(Lens)-Cre transgenic mouse is useful for investigating a gene function in the lens-restricted lineage.

In lens development, Sip1 is first activated in the lens placode,then after the lens vesicle is formed, Sip1 expression is confined to the vesicle without detectable expression in the surface ectoderm. After mature lens fibers develop, strong Sip1 expression is confined to the lens epithelium and bow region, and the expression is very low in the mature lens fibers (Fig. 1). The consequence of the lens-lineage-specific ablation of Sip1 revealed two major Sip1-dependent steps in lens development(Fig. 3), consistent with the Sip1 expression pattern.

The first significant defect is a persistent stalk connecting the surface ectoderm and lens epithelium (Fig. 3). This defect, called Peter's anomaly as a congenital disease in humans, is shared by defects in the transcription factors Pax6 or Foxe3, and a common denominator is the loss of functional Foxe3(Blixt et al., 2000; Brownell et al., 2000; Dimanlig et al., 2001). A Sip1 defect also causes the downregulation of Foxe3 in the lens epithelium(Fig. 5). This observation indicates that the Foxe3 gene is downstream of Sip1 in the regulatory pathway.

The second defect of the Sip1-defective lens is lack of mature lens fibers expressing γ-crystallins (Figs 3, 4). Sip1 is strongly expressed in the bow region in the normal lens where βA-crystallins are already expressed, signifying the initiation step of fiber differentiation. The bow region has been thought of as merely a zone of transition between the epithelial fiber precursor and mature lens fibers. However, the arrest of lens fiber differentiation in the βA-crystallin-positiveγ-crystallin-negative bow region state strongly suggests that Sip1 expression in immature fiber cells promotes lens fiber maturation.

The possible involvement of the Foxe3 promoter in the Sip1-dependent activation of Foxe3 was examined, using cell transfection. Up to 20-fold activation of the 6.2 kb Foxe3 promoter was observed from exogenous Sip1 (Fig. 6A,B). For this activation, a 1.2 kb promoter sequence was sufficient (Fig. 6F).

The activation of the Foxe3 promoter by exogenous Sip1 allowed examination of the effect of exogenous Smads on its regulation. Of the Smads that bind Sip1 in vitro (Verschueren et al., 1999), only Smad8 further augmented Sip1-mediated Foxe3 promoter activation by threefold(Fig. 6). However, this Smad8 effect was not observed using SBD (Smad-binding domain)-deleted Sip1,demonstrating the involvement of a direct Sip1-Smad8 interaction. Amino acid sequence comparison of Smad8, Smad1 and Smad5 indicates that Smad8 has a considerably shorter and diversified linker sequence between MH1 and MH2 domains than the other two (see Fig. S2 in the supplementary material). However, given the demonstration of similar activities of Smad1, Smad5 and Smad8 in various assays (Moustakas et al.,2001; ten Dijke and Hill,2004), it is possible that Smad1 and Smad5 also contribute to Sip1-dependent gene regulation in different contexts. In any case, this is the first clear demonstration that Smad interaction affects the regulatory potential of the Sip1 protein.

As previously demonstrated, the 6.2 kb Foxe3 promoter controls the expression of a lacZ reporter gene in the lens and mid-forebrain region of transgenic mouse embryo (Brownell et al., 2000) (Fig. 7A). By the deletion of the promoter region using blocks A to D from the proximal side, a major lens element in block C and a minor lens element in block B were identified (Fig. 7A). Thus, lens-specific regulatory elements are separable from those involved in Sip1-dependent activation assigned to block A(Fig. 7B). Indeed, the Sip1-and Smad8-dependent activation of the Foxe3 promoter was observed in non-lens cells (see Fig. S1 in the supplementary material), showing that this activation is not specific to lens cells. Block A of the Foxe3promoter is sufficient for this activation to occur in transfection assay(Fig. 6G), but by itself does not allow transgene expression in mouse embryos(Fig. 7A). Therefore, the combined action of lens element (blocks B and C) and Sip1-dependent (block A)promoter element appears to be required for Foxe3 expression in embryonic lenses.

The block C sequence has a region strongly conserved between mouse Foxe3 and human FOXE3, and with the aid of this sequence conservation, Grainger's group has independently identified the corresponding region in Xenopus as the lens element of the Foxe3 promoter(H. Ogino and R. Grainger, personal communication).

Gene activation by the action of Sip1 shown in this study expands the horizon of gene regulation involving ZFHX1 family transcription factors. Sip1 and δEF1 bind almost identical sets of sequences, owing to their highly conserved Krueppel-type zinc finger sequences(Funahashi et al., 1993; Verschueren et al., 1999). In addition, the bipartite zinc-finger clusters each bind to a similar set of sequences with a consensus of CACCT(G)(Remacle et al., 1999; Sekido et al., 1997), it has been postulated that ZFHX1 proteins bind a pair of CACCT(G) sequences in a two-footed fashion (Remacle et al.,1999). Under such conditions full-length Sip1 or δEF1 clearly exhibited the repression of gene transcription(Comijn et al., 2001; Funahashi et al., 1993; Kamachi and Kondoh, 1993; Papin et al., 2002; Remacle et al., 1999; Sekido et al., 1994; Sekido et al., 1997).

However, several lines of evidence support the view that the gene repression thorough a CACCT(G) pair is just one of many modes of regulatory function associated with ZFHX1 proteins.

(1) With a knockout allele of δEF1 lacking C-terminal zinc fingers,only a minor nonlethal phenotype develops in homozygous mouse(Higashi et al., 1997), in contrast to more severe lethal defects with a null allele(Takagi et al., 1998). This indicates that N-terminal zinc fingers are sufficient for DNA binding and exerting a regulatory function.

(2) The binding consensus CACCT(G) of N- and C-terminal zinc fingers was determined using oligonucleotide sequence pools preferentially binding to respective zinc finger clusters (Sekido et al., 1997). Re-examination of these sequences indicated that N-terminal zinc fingers bind DNA with a more relaxed specificity, including,for example, CACANNT.

(3) The 6.2 kb promoter sequence of Foxe3 contains frequent recurrent CACCT sequences, many located in blocks B, C and D, but the removal of these upstream blocks did not significantly affect the response to exogenous Sip1 (Fig. 6G).

It has not been determined whether the Sip1 protein directly binds to the Foxe3 promoter DNA, but further analysis of Foxe3 promoter activation will reveal how Sip1 is involved in gene activation and how interaction with Smad affects its regulatory potential.

Since the discovery of Sip1 as a Smad-binding protein, Smads have been implicated in Sip1-mediated transcriptional regulation(Verschueren et al., 1999),but this study provides the first definitive evidence that Smad-Sip1 interaction has an impact on Sip1-dependent gene regulation.

The mechanism of augmenting the Sip1-dependent activation of the Foxe3 promoter from interaction with Smad8 is not clear, but this interaction is not required for basal activation by Sip1 (up to 20-fold activation of the Foxe3 promoter), as the same activation level is achieved using SBD-deleted Sip1 (Fig. 6B). A possible model would be that Sip1 itself possesses an activation domain that is exposed upon binding to a proper sequence, and Smad8 bound to Sip1 provides an additional activation domain.

Many BMP signals are implicated in lens development(Faber et al., 2002). It has been demonstrated that BMP4 is required for the activation of Sox2 in the prelens ectoderm (Furuta and Hogan,1998), while BMP7 deficiency causes variable defects in later lens development from the absence of a lens to a slightly smaller lens(Jena et al., 1997; Wawersik et al., 1999). Developmental lens defect caused by lens lineage-specific Sip1inactivation and the involvement of Smad-Sip1 interaction in Foxe3promoter regulation underscore the importance of the BMP-Smad-Sip1 signaling pathway in lens development.

Sip1 gene activity is implicated in many steps of embryogenesis:gastrulation in chickens (Sheng et al.,2003), mesodermal development in Xenopus(Papin et al., 2002) and neural crest development in mice (Van de Putte et al., 2003). This study revealed important aspects of gene regulation mediated by Sip1 leading to gene activation. Most importantly,interaction with Smad proteins, at least with Smad8, modifies the transcriptional regulation mediated by Sip1. These new features of Sip1-mediated transcriptional regulation should help understanding of processes involving Sip1.

We appreciate the stimulating discussions with Drs Y. Kamachi, D. Huylebroeck and members of the Kondoh laboratory. We thank Drs P. Gruss, A. Mansouri, K. Miyazono, S. Watanabe, P. Xu and R. Yamada for providing the mouse lines and for sharing genetic reagents. This work was supported by grants (168207 to A.Y., 15570718 and 16027230 to Y.H. and 12CE2007 to H.K.)from the Ministry of Education, Culture, Sports, Science and Technology of Japan, and from the Human Frontier Science Program (RGP0040/2001-M) to H.K. A.Y. is a Research Fellow of the Japan Society for the Promotion of Science.