PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
 
Dev Biol. 2010 December 1; 348(1): 97–106.
PMCID: PMC2997430

Wnt/β-catenin signaling in the dental mesenchyme regulates incisor development by regulating Bmp4

Abstract

Loss- and gain-of function approaches modulating canonical Wnt/β-catenin activity have established a role for the Wnt/β-catenin pathway during tooth development. Here we show that Wnt/β-catenin signaling is required in the dental mesenchyme for normal incisor development, as locally restricted genetic inactivation of β-catenin results in a splitting of the incisor placode, giving rise to two incisors. Molecularly this is first associated with down-regulation of Bmp4 and subsequent splitting of the Shh domain at a subsequent stage. The latter phenotype can be mimicked by ectopic application of the BMP antagonist Noggin. Conditional genetic inactivation of Bmp4 in the mesenchyme reveals that mesenchymal BMP4 activity is required for maintenance of Shh expression in the dental ectoderm. Taken together our results indicate that β-catenin together with Lef1 and Tcf1 are required to activate Bmp4 expression in order to maintain Shh expression in the dental ectoderm. This provides a mechanism whereby the number of incisors arising from one placode can be varied through local alterations of a mesenchymal signaling circuit involving β-catenin, Lef1, Tcf1 and Bmp4.

Keywords: Canonical Wnt-signaling, Tooth development, Mesenchyme, BMP, Msx1, Shh, Incisor, Supernumerary

Introduction

The dentition of mice is highly reduced, with only one incisor separated by a diastema region from three molars in each quadrant. The first sign of mammalian tooth development for both type of teeth, molars and incisors, is a thickening of the oral epithelium in the tooth area around embryonic day (E) 11. Molar and incisor developments are in principle similar, but differ slightly in developmental timing and with respect to gene expression. Tooth development requires sequential and reciprocal interactions between the dental epithelium and dental mesenchyme mediated by different signaling pathways (Caton and Tucker, 2009). The thickened dental epithelium expresses growth factors such as the Fibroblast growth factor 8 (Fgf8) and Sonic hedgehog (Shh). Concurrent with its thickening the proliferating epithelium invaginates into the underlying mesenchyme. The mesenchyme condenses around the invaginated epithelium, resulting in the formation of a dental bud stage around E12.5 (bud stage) in the incisor region. Subsequently, the invaginating epithelium extends farther into the mesenchyme and in response to mesenchymal signals, such as bone morphogenetic protein-4 (BMP4), the enamel knot, a non-proliferating transient epithelial structure, appears at the cap stage (around E13.5 for the incisor). Mesenchymal Bmp4 expression is regulated by transcription factors such as Msx1 and Pax9 (Chen et al., 1996; Ogawa et al., 2006). Amongst other molecules the enamel knot expresses Shh and Fgf4 and serves as a signaling center that regulates dental papilla formation in the adjacent mesenchyme and proliferation of the remaining dental epithelium (Thesleff and Jernvall, 1997). The epithelium extends further by differential proliferation and forms a bell shaped structure enclosing the mesenchyme (bell stage; ≥ E16 for the incisor). During the bell stage, epithelial cells adjacent to the mesenchyme differentiate further into enamel-producing ameloblasts and the mesenchymal cells into dentin-producing odontoblasts.

Previous studies provided evidence for a key role for the Wnt/β-catenin pathway in the dental epithelium during tooth morphogenesis. The pathway is activated by binding of a Wnt-ligand to a receptor complex composed out of a frizzled family member and LDL receptor related protein (Lrp5/6), which leads to the inhibition of a cytoplasmic destruction complex containing the Adenomatosis polyposis coli (APC), Axin and GSK3 kinase proteins. Thus cytosolic β-catenin accumulates and translocates to the nucleus where it assembles into a transcriptional activating complex with members of the Tcf/Lef transcription factor family (Logan and Nusse, 2004). Loss of Lef1, which is expressed in the dental mesenchyme and epithelium causes arrest of tooth development at the bud stage (van Genderen et al., 1994) and tissue recombination experiments showed that Lef1 activity is required in the dental epithelium to regulate Fgf4 expression and cell survival (Kratochwil et al., 2002; Sasaki et al., 2005). Constitutive activation of the Wnt/β-catenin pathway in the oral epithelium results in the formation of ectopic invaginations, giving rise to ectopic teeth (Jarvinen et al., 2006; Kuraguchi et al., 2006; Liu et al., 2008; Wang et al., 2009). A similar but less severe phenotype is observed when Lef1 is overexpressed in the oral epithelium (Zhou et al., 1995). Inhibition of the Wnt/β-catenin pathway by overexpression of the secreted inhibitor Dickkopf 1 (Dkk1) in the oral epithelium results in an early arrest of tooth development (Andl et al., 2002; Liu et al., 2008). Genetic deletion of β-catenin in the oral epithelium results in a similar albeit less severe phenotype (Liu et al., 2008). Until recently, most of the published experimental approaches, with the exception of inhibiting the pathway by Dkk1 overexpression, which may also affect the dental mesenchyme, demonstrated a role for Wnt/β-catenin in the dental epithelium and not in the mesenchyme during tooth development. Direct evidence for a role of the canonical Wnt-pathway in the mesenchyme, was for the first time provided through conditional inactivation of β-catenin in the dental mesenchyme, which lead to an arrest of molar development at the cap stage (Chen et al., 2009).

Here we investigated the requirement for Wnt/β-catenin signaling in the incisor mesenchyme using a conditional approach and showed that locally restricted loss of β-catenin from the mandibular incisor region results in the formation of double incisors. Molecular analysis revealed that β-catenin is required in the mesenchyme to positively regulate Bmp4 expression. This occurs independent of Msx1 and is directly mediated by the activity of Tcf/Lef transcription factors. Furthermore, we provide in vitro and genetic in vivo evidence that mesenchymal BMP4 activity is required to maintain Shh expression in the overlying dental epithelium. Thus our results show that mesenchymal activity of β-catenin is required for incisor development and indicate that incisor numbers can be altered through local modulation of β-catenin activity.

Materials and methods

Animals and preparation of embryonic tissues

The transgenic mouse lines for β-catenin, β-catfl (conditional allele) and β-catlacZ (null allele), Bmp4, Bmp4[increment]LacZ (null allele), Bmp4hf (hypomorphic conditional allele) and Bmp4fl (conditional allele), Lef1 (null allele), Tcf1 (null allele) and Prx1-Cre used in this study have been described earlier and genotyping was performed accordingly (Benazet et al., 2009; Chang et al., 2008; Galceran et al., 1999; Huelsken et al., 2000, 2001; Logan et al., 2002). Intercrosses to generate the mutants were done as described previously (Benazet et al., 2009; Galceran et al., 1999; Hill et al., 2005). Bmp4[increment]Prx1/[increment]GL mutants were generated taking advantage of the germ line activity of Prx1-Cre females through intercrosses of Prx1-Cre females with Bmp4fl/fl males. From the offspring Bmp4[increment]GL/+;Prx1Cre males were crossed with Bmp4fl/fl females to generate the desired mutants. Embryos were staged according to morphological criteria; plug date was considered to be E0.5. Whole heads or lower jaws were dissected in PBS, fixed in 4% paraformaldehyde, dehydrated with methanol and ethanol for whole mount and paraffin section in situ hybridizations, respectively. All mouse experiments were performed in accordance with local and institutional regulations and licenses.

Cre recombinase and Wnt activity detection

For the detection of Cre recombinase Prx1-Cre mice were crossed with Z/AP mice (Lobe et al., 1999) and stained for alkaline phosphatase (Hill et al., 2005). For the detection of Wnt activity in E11.5–E13.5 mandibles, heads from BAT-gal (Maretto et al., 2003), TOP-gal (DasGupta and Fuchs, 1999) and Axin2-lacZ (Lustig et al., 2002) mice were dissected and fixed in 2% paraformaldehyde/0.2% glutaraldehyde in PBS and stained in X-gal solution for 20 h to detect β-galactosidase activity as previously described (Jarvinen et al., 2006).

Skeletal preparations, in situ hybridizations and immunohistochemistry

Heads of newborns (P0) embryos were skinned, fixed in 95% ethanol and stained with alizarin red and Alcian blue (McLeod, 1980). Whole mount (on dissected mandibles) and section (5 μm paraffin-sections) in situ hybridizations were performed as described (Murtaugh et al., 1999; Riddle et al., 1993). Double-color and double fluorescence in situ hybridization on serial paraffin sections were performed using Digoxigenin (DIG) and Fluorescein (Flu) and Biotin and DIG labeled probes, respectively, as described previously (Brent et al., 2003; Tylzanowski et al., 2003). All probes used are available upon request. Immunohistochemistry was performed on paraffin sections using anti-β-catenin mAb (BD Transduction Laboratories, 1:200) as described (Spater et al., 2006). Ki67 (Novocastra, 1:1000) and Caspase-3 (Cell signaling, 1:100) immunohistochemistry was performed using the automated Discovery XT system (Ventana Medical Systems).

Tissue recombination experiments and kidney capsule transplantation

Incisor anlagen from E12.5 mandible of WT and β-catfl/fl mice were dissected in Dulbecco's PBS (pH 7.4) under a stereomicroscope. Dental epithelium and mesenchyme were separated after incubation in a 2.25% Trypsin (Sigma)/0.25% Pancreatin (Sigma)/Tyrode's solution (pH 7.4) for 2 min at room temperature, followed by incubation in DMEM/10% FCS (Autogen Bioclear) to stop Protease digestion. Given that it has been shown that excessive removal of dental mesenchyme can affect incisor number (Munne et al., 2009), we took care not to remove any dental mesenchyme. Separated β-catfl/fl mesenchyme was inoculated with 1 × 1010 pfu/ml Adeno-GFP or Adeno-Cre each (Gene transfer vector core, University of Iowa) in DMEM/10% FCS for 1 h at 37 °C in 5% CO2. β-catfl/fl adenovirus-infected mesenchyme recombined with uninfected wild-type epithelium in the right orientation on Nuclepore filters, and cultured on metal grids at 37 °C in 5% CO2. After 1 day in culture, the recombined samples were transplanted under the kidney capsule of anesthetized immuno-compromised mice. The cut was sutured and after 3 weeks the mouse was killed, and the kidneys were carefully removed.

Bead implantations

Implantation of NOGGIN and BSA coated beads were performed by following published protocol (Vainio et al., 1993). Affi-Gel blue agarose beads (Bio Rad) were incubated in 100 ng/μl recombinant mouse Noggin (R&D) for 45 min at 37 °C. Heparin acrylic beads (Sigma) incubated for 45 min at 37 °C in PBS containing 0.1% BSA were used as controls. Control beads were implanted near the presumptive incisor regions in the left quadrant and NOGGIN beads in the right quadrant of dissected mandibular explants from WT E11.5 embryos. Explants were then placed on Nuclepore filters and cultured supported by metal grids in DMEM / 10% FCS. After 48 h the explants were fixed in 100% methanol for 5 min followed by an overnight fixation in 4% paraformaldehyde, and processed for whole-mount in situ hybridization.

Real-time PCR

For real-time PCR analysis, DNase I-treated total RNA (1 μg) was reverse transcribed using oligo (dT) to produce first-strand cDNA. Real-time PCR was performed using SYBR greenI nucleic acid gel stain (Roche) and TAKARA rTaq. Values were calculated using the relative standard method and normalized to mouse Hprt1 expression. Primers sets were tested by dilution series and products were analyzed by gel electrophoresis and melting curves. cDNA were amplified using the following primer pairs:

  • Bmp4: sense, CGTTACCTCAAGGGAGTGGA and anti-sense ATGCTTGGGACTACGTTTGG;
  • Bmp7: sense, TACGTCAGCTTCCGAGACCT and anti-sense, GGTGGCGTTCATGTAGGAGT;
  • Hprt1: sense, AGCTACTGTAATGATCAGTCAACG, and anti-sense, AGAGGTCCTTTTCACCAGCA.

Luciferase promoter studies

The mouse Bmp4 reporter plasmids were constructed by cloning the −3000/+120 fragment (3 kb) and the −4820/+120 fragment (4.8 kb) of the mouse Bmp4 genomic region into the pGL3-Basic luciferase vector (Promega). HEK293T cells (8 × 105 cells/well) were co-transfected with each reporter plasmid (0.5 μg) and a total of 2 μg DNA containing the respective concentrations of expression plasmids (kindly provided by H. Clevers) and pcDNA3.1 (Invitrogen) together with 0.05 μg of Renilla luciferase reporter plasmid pRL-CMV (Promega) using calcium phosphatase co-precipitation procedure. 24 h after transfection, firefly and Renilla luciferase activities were measured using the Dual-Glo luciferase assay system (Promega) following the manufacturer's protocol on a Synergy 2 Multi-Detection Microplate Reader (BioTek Instruments). Firefly activity was standardized against the Renilla luciferase control activity.

Results

Inactivation of β-catenin in the dental mesenchyme affects incisor number

Previous data have established important roles for Wnt/β-catenin signaling for tooth development in the dental epithelium and recently in the mesenchyme. In various Wnt-reporter mice such as Top-GAL, BAT-GAL and Axin2-lacZ mice reporter activity in the dental mesenchyme can first be detected in the incisor regions from E11.5 onwards (Supplementary Fig. 1A–C and data not shown), suggesting a potential role for Wnt/β-catenin signaling in the dental mesenchyme for incisor development. In order to delete β-catenin activity in the dental mesenchyme we used the Prx1-Cre line (Logan et al., 2002). Using the Z/AP reporter mouse we could show that Prx1-Cre has mosaic activity in the mesenchyme of the mandibular incisor region from E11.5 onwards, while we could not detect any activity in the molar region (Supplementary Fig. 1D–H). Skeletal preparations of P0 β-catenin conditional mutant heads derived from intercrosses of Prx1-Cre;β-cat+/lacZ males with homozygous β-catfl/fl females revealed in 5/12 β-cat[increment]Prx/lacZ mutants the presence of two incisors on one side of the mandible only, while 2/12 mutants had two incisors on both sides. In contrast, wild-type embryos have only one incisor per mandibular half (Fig. 1A, and data not shown). H&E staining of sagittal and frontal sections of E18.5 β-cat[increment]Prx/lacZ heads showed that the two incisors developed on top of each other (Fig. 1B, C). In addition, we noticed that the incisors in the mutants were shorter and showed altered ameloblast distribution (Fig. 1B and Supplementary Fig. 2). The observed shortening of the incisors is probably a secondary event and due to the presence of ectopic cartilage (asterisks in Fig. 1B, C). Formation of ectopic cartilage primarily occurred within the bone region and is probably due to a differentiation of chondrocytes instead of osteoblasts as previously reported in the limbs and skulls of conditional β-catenin mutants (Day et al., 2005; Hill et al., 2005). The first clear morphological differences with regard to incisor development were observed as early as E13.5 (n = 3/6, Fig. 1D), while at E12.5 and E11.5 the incisors in mutants resembled those of wild-types (n = 3, Fig. 1E, and data not shown).

Fig. 1
Deletion of β-catenin in the dental mesenchyme increases the number of incisors. (A) Skeletal preparation of WT and β-cat[increment]Prx/lacZ newborn (P0) showing one half of a mandibular jaw. Note the double incisors at the distal tip (indicated ...

To further support the in vivo observations by an independent experimental approach we performed the following in vitro experiments. Single incisor anlagen were dissected from E12.5 β-catfl/fl mandibles and the dental mesenchyme was infected with adenovirally expressed Cre-recombinase to inactivate the β-catenin gene. Subsequently, the partially infected, partly mutant mesenchyme was recombined with wild type E12.5 dental epithelium, cultured for 24 h, and then implanted into the kidney capsule and grown there for 3 weeks. The frequency in recovering incisors and not only bone in those in vitro experiments was very low, with incisors present in only 8/35 adeno-Cre treated and 5/34 adeno-GFP treated control recombinant explants. In 1/8 adeno-Cre treated cases two incisors instead of one developed from a single incisor anlage (Supplementary Fig. 2M). This was not observed in the control tissue recombination experiments (infected with adeno-GFP; n = 5). Thus, the preliminary result from the in vitro recombination experiment supports the in vivo genetic analysis (Fig. 1A–D), suggesting that local inactivation of β-catenin in the mesenchyme of the incisor anlage can lead to incisor duplication.

The loss of β-catenin alters Shh and Bmp4 expression

In order to identify the underlying molecular alterations we examined the expression of genes critical for tooth development such as Shh, Msx1, Msx2, Pax9, Axin2, Fgf3, Fgf8, Fgf9, Fgf10, and Bmp2, Bmp4, and Bmp7 in the incisor region. At E11.5, the distribution and levels of Fgf8 (n = 6), Shh (n = 5), Fgf9 (n = 5) and Fgf10 (n = 6) were not altered in the dental epithelium (Fig. 2A, B, and data not shown). The expression of mesenchymal genes such as Msx1 (n = 5), Msx2 (n = 6) and Pax9 (n = 6) expression domain were not noticeably altered (Fig. 2C, D, and data not shown). This was confirmed for Msx1 by section in situ analysis (data not shown). In contrast, Bmp4 expression appeared patchy in the mutant mesenchyme in comparison to the homogenous expression in wild-types (n = 5/5, Fig. 2E). A similar patchy down-regulation was observed for the bona fide Wnt target gene Axin2 (n = 3/4; Fig. 2F) and confirmed on sections by in situ hybridization (Supplementary Fig. 3B). The expression of other BMP family members, such as Bmp2 and Bmp7, which were expressed in the mandible at E11.5 in the epithelium or in the epithelium and mesenchyme, respectively, were not altered in β-catenin mutants (Supplementary Fig. 3C–E). Quantification of total transcript levels in the mandibular region by real-time PCR showed a significant reduction of Bmp4, but not Bmp7 expression (Supplementary Fig. 3F). From E12.5 onwards, the Shh expression domain appeared to be split into two small dots either bi- (n = 4/11) or unilaterally (n = 7/11) (Fig. 2G). The split in the Shh expression domain was confirmed by section in situ hybridization using frontal sections (n = 3/4, Fig. 3A, C). Hybridization of alternating sections to Bmp4 or using two-color in situ detection showed that Bmp4 is highly expressed in the mesenchyme directly underneath the Shh expression domain in wild-type controls (Fig. 3B, C, E). In mutants with a split Shh expression domain, two high level Bmp4 expression domains are apparent in conjunction with a slight reduction of the overall Bmp4 levels in the mandibular mesenchyme (Fig. 3B, C). Consistent with the whole mount in situ observations Msx1 was still expressed throughout the dental mandibular mesenchyme, including the region where Bmp4 was down-regulated (n = 2, Fig. 3D). We used double fluorescent section in situ hybridization to establish that Bmp4 expression was specifically lost from regions deficient in β-catenin (Fig. 3E–G). In contrast to what we had observed in the Z/AP reporter staining, recombination for the β-catenin locus was more restricted at E12.5, resulting in a patchy, mosaic loss of β-catenin expression (Fig. 3F). The epithelial and mesenchymal cells surrounding the split Shh expression domains were proliferating normally and no increase in cellular apoptosis was observed (Supplementary Fig. 4). Expression of Fgf10 was not altered at E13.5, while Fgf3 expression was reduced (data not shown). Thus the first change that we are able to detect in mutants lacking β-catenin in a mosaic manner in their mesenchyme was a mosaic down-regulation in expressions of Bmp4 and the bona fide Wnt/β-catenin target gene Axin2 in the dental mesenchyme.

Fig. 2
Alterations in gene expression in β-cat[increment]Prx/lacZ mutant mandibles at early stages of tooth development. (A–F) In situ hybridization showing the expression of Fgf8, Shh, Msx1, Pax9 and Bmp4 at E11.5 and (G) Shh at E12.5 in wild-type ...
Fig. 3
Correlation between alterations in mesenchymal Bmp4 and epithelial Shh expression upon inactivation of mesenchymal β-catenin. (A–G) Frontal section showing incisor tooth buds at E12.5. (A, B) Alternating sections showing a split Shh expression ...

Alteration of BMP4 activity and expression affect Shh expression in the incisor

As down-regulation of Bmp4 expression in the dental mesenchyme of β-cat[increment]Prx/lacZ embryos seems to precede the alterations in Shh expression, we asked whether there is a causal connection. Therefore, BMP activity was inhibited in in vitro explant-cultures using the BMP antagonist NOGGIN. NOGGIN-soaked beads were implanted near the incisor region of wild-type E11.5 mandibles, which were then cultured for 48 h. In about 50% of all NOGGIN treated cultures, the Shh expression domain was split (8/15), associated with a down-regulation of Shh in 4/8 cases showing the split expression (Fig. 4A, and data not shown). Next we asked whether genetic inactivation of Bmp4 in dental mesenchyme would alter Shh expression. This was achieved by crossing females homozygous for the hypomorphic Bmp4hf allele with males double-heterozygous for the Bmp4[increment]lacZ null allele and the Prx1Cre transgene (Benazet et al., 2009). At E12.5 (n = 2) Shh was completely absent from the incisor and molar regions of mandibles of Bmp4 deficient (Bmp4hf[increment]Prx/[increment]lacZ) embryos, while its expression was unaffected in littermate controls (Fig. 4B). In contrast, Shh expression was only reduced and the two domains fused at the midline in the Bmp4 (Bmp4hf[increment]Prx/[increment]lacZ) deficient embryonic mandibles at E11.5 (n = 2; Fig. 4C). When we examined the efficiency of the Bmp4 deletion in Bmp4hf[increment]Prx/[increment]lacZ mandibles at E11.5 using a Bmp4 probe detecting the deleted exon 4, it became apparent that Bmp4 expression was in general severely reduced but not completely lost (n = 2; Fig. 4D). Interestingly, when we looked at the expression of Msx1 in the Bmp4 mutants at E11.5, we did not observe a decrease in the expression levels (n = 2/2; Fig. 4E). As expected, Bmp4 expression was also reduced in the mandibles of Bmp4[increment]lacZ/hf embryos (n = 2/3), but Shh expression was not lost at E12.5 (n = 1; data not shown). At E14.5, either no (n = 1/3) or only weak Shh expression in one or two tiny spots was detected at the tip of Bmp4hf[increment]Prx/[increment]lacZ mandibles (n = 2/3) (Fig. 4E). In contrast to the local reduction of Bmp4 expression in β-cat[increment]Prx/lacZ mandibles (Fig. 3C) the almost complete loss in Bmp4hf[increment]Prx/[increment]lacZ embryos provides a feasible explanation for why Shh expression is lost in Bmp4hf[increment]Prx/[increment]lacZ rather than being split. In addition, we examined Shh expression in Bmp4[increment]Prx/[increment]GL mutants, in which a none-hypomorphic conditional allele of Bmp4 (Chang et al., 2008) had been deleted using the Prx1-Cre in the presence of a germ-line deleted Bmp4[increment]GL/+ allele. However, using this Bmp4 loss-of-function allele, only minimal changes in Bmp4 expression (n = 3/4) and no significant changes in Shh expression (n = 3) were observed in the mandibles (Supplementary Fig. 5). In contrast to the mandibles, mesenchymal Bmp4 expression is completely lost in the limb, but again we did not observe a loss of Msx1 expression (n = 2; Supplementary Fig. 5C, D). Taken together, this genetic analysis indicates that BMP4 activity is likely to be required for maintenance but not initiation of epithelial Shh expression. Furthermore, our genetic data suggest that Msx1 is not a Bmp4-dependent gene.

Fig. 4
Alterations in BMP4 activity affect Shh expression in the dental epithelium. (A) In situ detection of Shh expression in E11.5 mouse mandibular explants cultured for 48 h with BSA- (white) or NOGGIN- (blue) soaked beads. Note the split in the ...

Bmp4 is a direct target of a β-catenin–Tcf1–Lef1 complex

Since Msx1, which has been shown previously to be required for Bmp4 expression in the dental mesenchyme (Chen et al., 1996), was not affected by the loss of β-catenin activity, we determined whether Bmp4 is a transcriptional target of Wnt/β-catenin signaling in the dental mesenchyme. In silico analysis using the transfac database revealed that a 4.8 kb upstream region of the Bmp4 locus encodes approximately 33 putative TCF/LEF binding sites (Fig. 5A). Some of those binding sites are conserved between rat and mouse or rat, mouse and human based on a CLUSTALW analysis (indicated by one or two asterisks respectively, see Fig. 5A). Bmp4-reporter constructs encoding either the 4.8 kb up stream region or a shorter 3 kb region (containing approximately 21 potential TCF/LEF binding sites) were transactivated in luciferase assays upon co-transfection of expression plasmids for both Lef1 and β-catenin, Tcf1 and β-catenin or an expression plasmid encoding a TCF1-β-catenin fusion protein (Fig. 5B, and data not shown). Next, we determined whether Bmp4 expression is altered in the dental mesenchyme of Tcf1 and/or Lef1 mutant mandibles at E11.5. However, no significant difference in Bmp4 expression was observed in the incisor regions of Lef1−/− or Tcf1/ embryos in comparison to controls at E11.5 (n = 3; Fig. 5C). In compound Tcf1;Lef1 mutant embryos (Tcf1/;Lef1+/ n = 5/6 and Tcf1+/;Lef1/ n = 4/4) the overall expression of Bmp4 was slightly reduced (Fig. 5D), while Bmp4 expression was completely lost in Tcf1/:Lef1/ mutant embryos at E11.5 (n = 2/2; Fig. 5D). Given the severity of the embryonic defects and the developmental retardation of the Tcf1/Lef1 double mutant embryos at E11.5 it is possible that the loss of Bmp4 is due to secondary effects. However, when we analyzed E9.5 double mutant embryos, which are fairly similar to wild-type and single mutant embryos, we could already at this stage observe a reduction in the Bmp4 expression in the mandibles of Tcf1/:Lef1/ compared to littermate Tcf1/;Lef1+/+ mutant embryos (Fig. 5E). At this stage a rudimentary forelimb was still present, which also showed reduced Bmp4 expression (data not shown).

Fig. 5
In vitro and in vivo regulation of Bmp4 expression by β-catenin in combination with Tcf/Lef family members. (A) Schematic overview over the −4.8 kb mouse Bmp4 promoter region indicates the location of putative Tcf/Lef binding sites ...

Although all four Tcf/Lef family members are expressed in the dental mesenchyme at E11.5 (Supplementary Fig. 6), the almost complete absence of Bmp4 expression in Tcf1/Lef1 double mutants would suggest that Tcf1 and Lef1 are the primary transcriptional co-factors, which in a complex with β-catenin, regulate Bmp4 expression in the dental and possibly also in the limb mesenchyme.

Discussion

Mesenchymal β-catenin activity regulating Bmp4 is required for incisor development

Various ectodermal appendages including teeth depend on canonical Wnt/β-catenin signaling (Mikkola and Millar, 2006). Until recently it was thought that this pathway is primarily required in the dental epithelium during tooth development. This notion was supported by the finding that TCF/LEF reporters seem not to be active in the molar mesenchyme (Liu et al., 2008). However, here we show that there is activity of various Wnt-reporter mice in the early incisor mesenchyme of the lower jaw. Recently, the analysis of a conditional β-catenin mutant by Chen and colleagues established that β-catenin is required in the dental mesenchyme during tooth development (Chen et al., 2009). Their study showed that following β-catenin inactivation, molar and incisor formation arrests at bud stage but the underlying molecular alterations resulting in this arrest remained unknown. In contrast we show that sporadic deletion of β-catenin in the incisor mesenchyme of the lower jaw using the Prx1-Cre recombinase transgene resulted in a double-incisor phenotype in about 50% of all mutant embryos. Our molecular analysis revealed that in parallel to the onset of Prx1-Cre activity, gene expressions were altered starting around E11.5. In particular, mesenchymal Bmp4 and Axin2 expressions were down-regulated in a patchy manner as a likely first molecular alteration in the mosaic β-catenin deficient mesenchyme. The changes in Bmp4 expression appear to occur in an Msx1-independent manner and cell-autonomously, coinciding with the local loss of β-catenin. These cell-autonomous changes together with the Bmp4 promoter analysis suggest that Bmp4 might be a direct target of β-catenin/TCF/LEF mediated transcription in the dental mesenchyme. β-catenin has been shown to regulate Bmp4 in other tissues such as cancer cells, lung, limb and eye (Hill et al., 2006; Kim et al., 2002; Shu et al., 2005; Zhou et al., 2008). Consistent with our in vitro data, it has been shown previously that β-catenin can activate and bind to a 2.4-kb Bmp4 promoter region in vitro (Shu et al., 2005; Song et al., 2009). Based on the results from transgenic promoter studies this region is, however, not sufficient to drive expression in the primordial tooth anlage (Feng et al., 2002). Thus the more upstream located Tcf/Lef binding sites might be the ones, which regulate Bmp4 expression in the early tooth anlage. This needs to be demonstrated in the future by in vivo ChIP experiments and transgenic promoter studies. Nevertheless given that Bmp4 expression was reduced in mandibles of compound Tcf1/Lef1 mutants and lost in mandibles of double mutants supports our model that β-catenin, together with Tcf1 and Lef1, which are both expressed in the dental mesenchyme at E11.5, directly regulates Bmp4 expression in the dental mesenchyme (Fig. 6). So similar to what has been reported in the limb, Tcf1 and Lef1 seem to act redundantly in the mandibular mesenchyme (Galceran et al., 1999). Furthermore, our data establish that Tcf1 plays a functional, albeit redundant role in tooth development. Interestingly, Bmp4 has been shown to regulate Lef1 expression in the dental mesenchyme (Chen et al., 1996), thus establishing a positive feed-back loop (see model Fig. 6). In addition to the Bmp4 down-regulation, Shh expression in the overlying dental epithelium was altered in the conditional β-catenin mutant mandibles. Bmp4 has been previously implicated in regulating Shh and Bmp2 expression in the dental epithelium of the molars based on the analysis of Msx1 deficient mouse embryos and in vitro explant studies (Zhang et al., 2000). Alteration of BMP activity in mandibular explants by NOGGIN-bead implantation affects Shh expression in a manner similar to loss of mesenchymal β-catenin activity, which supports the notion that BMP activity is required to maintain Shh expression in the incisor region. In addition, it has been recently shown that treatment of incisor explants with NOGGIN resulted in splitting of the Shh expression domain and the development of to up to three incisors, dependent on the NOGGIN concentration (Munne et al., 2010). This further supports the idea that alteration of BMP-signaling influences Shh expression and incisor number. Importantly, genetic inactivation of Bmp4 in the dental mesenchyme establishes that it is the mesenchymal activity of BMP4, which is indeed required for the maintenance of Shh in the dental epithelium of molars and incisors, but not for its induction. Therefore, we hypothesize that the Shh expression domain splits only in those cases where the regions lacking Bmp4 expression in the mesenchyme were substantial and directly abutting the overlying epithelium (Fig. 6). The study by Zhang et al. (2000) reported that Bmp4 might also inhibit Shh expression at least at high concentrations in vitro. Thus, BMP4 in combination with other factors might regulate Shh expression differentially over time in a concentration-dependent manner, which might explain why residual Shh expression was detected in 2 out of 3 Bmp4 deficient embryos at E14.5. Alterations in the Shh expression domain might ultimately affect the invagination of dental epithelium, as loss of Shh at the early bud stage and inhibition of Shh activity at the epithelial thickening stage is known to alter growth and morphogenesis of teeth (Cobourne et al., 2001; Dassule et al., 2000). Furthermore, ectopic application of SHH protein to oral epithelium induces invagination of non-dental epithelium (Hardcastle et al., 1998). Thus the generation of a second Shh expressing center within the dental epithelium might be one of the key steps triggering the second invagination that causes the formation of a second incisor in β-cat[increment]Prx/lacZ mutant embryos.

Fig. 6
Model of the β-catenin/BMP4 circuit maintaining Shh expression in wild-type and β-catenin mutant incisor regions. (A) β-catenin in a transcriptional complex with Tcf1 and Lef1 activates the expression of Bmp4 (indicated by the ...

These phenotypic and molecular alterations were not observed in the previous genetic analysis by Chen et al. (2009). This might be due to differences in the onset of Cre-recombinase activity in the mandibular region as at least for the Osr2-IresCre no recombinase activity was reported prior to E14.5 in incisor regions (Lan et al., 2007). On the other hand the deletion efficiency might vary between the different Cre-lines. In particular, we have observed that deletion of β-catenin in different β-cat[increment]Prx/lacZ mutant embryos was patchy and variable. This observation probably also explains why the penetrance of the double-incisor phenotype was not 100%. Similarly, only incomplete and patchy inactivation of β-catenin activity was achieved in the adenovirally infected mesenchymal explants, explaining probably why a double-incisor was only observed in one case.

Local alterations in the β-catenin/BMP4 circuits affect incisor number

Although adult mice have only one upper and lower incisor, mouse embryos still show a more ancestral pattern with about six vestigial incisor placodes in the upper and three in the lower half of the developing jaw (Peterkova et al., 2006). The entire epithelial area, comprising all placodes, folds normally into the mesenchyme, but is thought to form only one incisor through an incorporation process potentially due to the failure to form inhibitory zones (Peterkova et al., 2006). These inhibitory zones might be regulated by reaction–diffusion interactions of growth inhibitors and activators similar to Turing-type mechanisms, but may have been lost or became suppressed during rodent evolution (Peterkova et al., 2002; Turing, 1952). Recently it has been shown that the mesenchyme surrounding the incisor tooth germ has the capability to restrict tooth induction (Munne et al., 2009). This repressive potential of the mesenchyme is in part mediated by ectodin, which is encoded by the Sostdc1 gene (Munne et al., 2009). Ectodin, is related to Sclerostin and is known to modulate the Wnt-pathway and to potentially inhibit BMP-signaling (Itasaki et al., 2003; Laurikkala et al., 2003; Lintern et al., 2009). However, the timing and mode of supernumerary incisor formation in Sostdc1 mutants is distinct from the duplications observed following β-catenin inactivation (Munne et al., 2009; Murashima-Suginami et al., 2007). Thus, we propose that in addition to the inhibitory potential of the dental mesenchyme there is also an activatory potential, which β-catenin and BMP4 are part of. This mesenchymal activatory pathway maintains the expression of critical factors in the dental epithelium, such as Shh and the Bmp4 dependent cell cycle inhibitor p21 (Jernvall et al., 1998). A recent finding showed that treatment of mandibular explants with defined concentrations of the BMP antagonist Noggin leads to a destabilization of the large placode and the formation of two to three incisors (Munne et al., 2010). In agreement with these findings, our data suggest that local or non-uniform reduction of mesenchymal BMP activity due to the loss of β-catenin contributes to the formation of supernumerary incisors by splitting the signaling center, which results in the formation of two invaginations. Thus, a potential mechanism to alter incisor number would be to locally modulate β-catenin activity in the dental mesenchymal, thereby altering Bmp4 expression and subsequently epithelial Shh expression. An alternative explanation, for the observed phenotype might be that reduced Wnt/BMP4 signaling somehow affects survival/revitalization of the posterior incisor placode, which normally regresses around E13.5. This explanation is in agreement with earlier findings by Zhang and colleagues, which showed that high levels of BMP4 inhibit Shh expression in the molar and incisor region (Zhang et al., 2000). However, based on this alternative model one would have expected increased Shh expression upon reduction of BMP4 activity in the mesenchyme, which we did not observe.

Acknowledgments

We thank Rudolf Grosschedl and Hans Clevers for providing the Lef1 and Tcf1 mutant mice and expression plasmids, Elina Järvinen and Riikka Santalahti for technical advice, Vukoslav Komnenovic and Mihaela Ziba for help with the immunohistochemistry. This research was supported by Boehringer Ingelheim. HN was supported by the Austrian Science Found (FWF Grant No. P19281-B16). Research by RZ is supported by the Swiss National Science foundation (SNF Grant. No. 31003A-113866) and AG is supported by "The Fundação para a Ciência e a Tecnologia” (FCT; SFRH/BD/24301/2005).

Appendix A. Supplementary data

Supplementary materials

References

Andl T., Reddy S.T., Gaddapara T., Millar S.E. WNT signals are required for the initiation of hair follicle development. Dev. Cell. 2002;2:643–653. [PubMed]
Benazet J.D., Bischofberger M., Tiecke E., Goncalves A., Martin J.F., Zuniga A., Naef F., Zeller R. A self-regulatory system of interlinked signaling feedback loops controls mouse limb patterning. Science. 2009;323:1050–1053. [PubMed]
Brent A.E., Schweitzer R., Tabin C.J. A somitic compartment of tendon progenitors. Cell. 2003;113:235–248. [PubMed]
Caton J., Tucker A.S. Current knowledge of tooth development: patterning and mineralization of the murine dentition. J. Anat. 2009;214:502–515. [PubMed]
Chang W., Lin Z., Kulessa H., Hebert J., Hogan B.L., Wu D.K. Bmp4 is essential for the formation of the vestibular apparatus that detects angular head movements. PLoS Genet. 2008;4:e1000050. [PMC free article] [PubMed]
Chen Y., Bei M., Woo I., Satokata I., Maas R. Msx1 controls inductive signaling in mammalian tooth morphogenesis. Development. 1996;122:3035–3044. [PubMed]
Chen J., Lan Y., Baek J.A., Gao Y., Jiang R. Wnt/beta-catenin signaling plays an essential role in activation of odontogenic mesenchyme during early tooth development. Dev. Biol. 2009;334:174–185. [PMC free article] [PubMed]
Cobourne M.T., Hardcastle Z., Sharpe P.T. Sonic hedgehog regulates epithelial proliferation and cell survival in the developing tooth germ. J. Dent. Res. 2001;80:1974–1979. [PubMed]
DasGupta R., Fuchs E. Multiple roles for activated LEF/TCF transcription complexes during hair follicle development and differentiation. Development. 1999;126:4557–4568. [PubMed]
Dassule H.R., Lewis P., Bei M., Maas R., McMahon A.P. Sonic hedgehog regulates growth and morphogenesis of the tooth. Development. 2000;127:4775–4785. [PubMed]
Day T.F., Guo X., Garrett-Beal L., Yang Y. Wnt/beta-catenin signaling in mesenchymal progenitors controls osteoblast and chondrocyte differentiation during vertebrate skeletogenesis. Dev. Cell. 2005;8:739–750. [PubMed]
Feng J.Q., Zhang J., Tan X., Lu Y., Guo D., Harris S.E. Identification of cis-DNA regions controlling Bmp4 expression during tooth morphogenesis in vivo. J. Dent. Res. 2002;81:6–10. [PubMed]
Galceran J., Farinas I., Depew M.J., Clevers H., Grosschedl R. Wnt3a−/−like phenotype and limb deficiency in Lef1(−/−)Tcf1(−/−) mice. Genes Dev. 1999;13:709–717. [PubMed]
Hardcastle Z., Mo R., Hui C.C., Sharpe P.T. The Shh signalling pathway in tooth development: defects in Gli2 and Gli3 mutants. Development. 1998;125:2803–2811. [PubMed]
Hill T.P., Spater D., Taketo M.M., Birchmeier W., Hartmann C. Canonical Wnt/beta-catenin signaling prevents osteoblasts from differentiating into chondrocytes. Dev. Cell. 2005;8:727–738. [PubMed]
Hill T.P., Taketo M.M., Birchmeier W., Hartmann C. Multiple roles of mesenchymal beta-catenin during murine limb patterning. Development. 2006;133:1219–1229. [PubMed]
Huelsken J., Vogel R., Brinkmann V., Erdmann B., Birchmeier C., Birchmeier W. Requirement for beta-catenin in anterior–posterior axis formation in mice. J. Cell Biol. 2000;148:567–578. [PMC free article] [PubMed]
Huelsken J., Vogel R., Erdmann B., Cotsarelis G., Birchmeier W. beta-Catenin controls hair follicle morphogenesis and stem cell differentiation in the skin. Cell. 2001;105:533–545. [PubMed]
Itasaki N., Jones C.M., Mercurio S., Rowe A., Domingos P.M., Smith J.C., Krumlauf R. Wise, a context-dependent activator and inhibitor of Wnt signalling. Development. 2003;130:4295–4305. [PubMed]
Jarvinen E., Salazar-Ciudad I., Birchmeier W., Taketo M.M., Jernvall J., Thesleff I. Continuous tooth generation in mouse is induced by activated epithelial Wnt/beta-catenin signaling. Proc. Natl. Acad. Sci. USA. 2006;103:18627–18632. [PubMed]
Jernvall J., Aberg T., Kettunen P., Keranen S., Thesleff I. The life history of an embryonic signaling center: BMP-4 induces p21 and is associated with apoptosis in the mouse tooth enamel knot. Development. 1998;125:161–169. [PubMed]
Kim J.S., Crooks H., Dracheva T., Nishanian T.G., Singh B., Jen J., Waldman T. Oncogenic beta-catenin is required for bone morphogenetic protein 4 expression in human cancer cells. Cancer Res. 2002;62:2744–2748. [PubMed]
Kratochwil K., Galceran J., Tontsch S., Roth W., Grosschedl R. FGF4, a direct target of LEF1 and Wnt signaling, can rescue the arrest of tooth organogenesis in Lef1(−/−) mice. Genes Dev. 2002;16:3173–3185. [PubMed]
Kuraguchi M., Wang X.P., Bronson R.T., Rothenberg R., Ohene-Baah N.Y., Lund J.J., Kucherlapati M., Maas R.L., Kucherlapati R. Adenomatous polyposis coli (APC) is required for normal development of skin and thymus. PLoS Genet. 2006;2:e146. [PubMed]
Lan Y., Wang Q., Ovitt C.E., Jiang R. A unique mouse strain expressing Cre recombinase for tissue-specific analysis of gene function in palate and kidney development. Genesis. 2007;45:618–624. [PMC free article] [PubMed]
Laurikkala J., Kassai Y., Pakkasjarvi L., Thesleff I., Itoh N. Identification of a secreted BMP antagonist, ectodin, integrating BMP, FGF, and SHH signals from the tooth enamel knot. Dev. Biol. 2003;264:91–105. [PubMed]
Lintern K.B., Guidato S., Rowe A., Saldanha J.W., Itasaki N. Characterization of wise protein and its molecular mechanism to interact with both Wnt and BMP signals. J. Biol. Chem. 2009;284:23159–23168. [PubMed]
Liu F., Chu E.Y., Watt B., Zhang Y., Gallant N.M., Andl T., Yang S.H., Lu M.M., Piccolo S., Schmidt-Ullrich R., Taketo M.M., Morrisey E.E., Atit R., Dlugosz A.A., Millar S.E. Wnt/beta-catenin signaling directs multiple stages of tooth morphogenesis. Dev. Biol. 2008;313:210–224. [PMC free article] [PubMed]
Lobe C.G., Koop K.E., Kreppner W., Lomeli H., Gertsenstein M., Nagy A. Z/AP, a double reporter for cre-mediated recombination. Dev. Biol. 1999;208:281–292. [PubMed]
Logan C.Y., Nusse R. The Wnt signaling pathway in development and disease. Annu. Rev. Cell Dev. Biol. 2004;20:781–810. [PubMed]
Logan M., Martin J.F., Nagy A., Lobe C., Olson E.N., Tabin C.J. Expression of Cre Recombinase in the developing mouse limb bud driven by a Prxl enhancer. Genesis. 2002;33:77–80. [PubMed]
Lustig B., Jerchow B., Sachs M., Weiler S., Pietsch T., Karsten U., van de Wetering M., Clevers H., Schlag P.M., Birchmeier W., Behrens J. Negative feedback loop of Wnt signaling through upregulation of conductin/axin2 in colorectal and liver tumors. Mol. Cell. Biol. 2002;22:1184–1193. [PMC free article] [PubMed]
Maretto S., Cordenonsi M., Dupont S., Braghetta P., Broccoli V., Hassan A.B., Volpin D., Bressan G.M., Piccolo S. Mapping Wnt/beta-catenin signaling during mouse development and in colorectal tumors. Proc. Natl. Acad. Sci. USA. 2003;100:3299–3304. [PubMed]
McLeod M.J. Differential staining of cartilage and bone in whole mouse fetuses by alcian blue and alizarin red S. Teratology. 1980;22:299–301. [PubMed]
Mikkola M.L., Millar S.E. The mammary bud as a skin appendage: unique and shared aspects of development. J. Mammary Gland Biol. Neoplasia. 2006;11:187–203. [PubMed]
Munne P.M., Tummers M., Jarvinen E., Thesleff I., Jernvall J. Tinkering with the inductive mesenchyme: Sostdc1 uncovers the role of dental mesenchyme in limiting tooth induction. Development. 2009;136:393–402. [PubMed]
Munne P.M., Felszeghy S., Jussila M., Suomalainen M., Thesleff I., Jernvall J. Spitting placodes: effects of BMP and Activin on the patterning and identity of mouse incisors. Evol. Dev. 2010;12:383–392. [PubMed]
Murashima-Suginami A., Takahashi K., Kawabata T., Sakata T., Tsukamoto H., Sugai M., Yanagita M., Shimizu A., Sakurai T., Slavkin H.C., Bessho K. Rudiment incisors survive and erupt as supernumerary teeth as a result of USAG-1 abrogation. Biochem. Biophys. Res. Commun. 2007;359:549–555. [PubMed]
Murtaugh L.C., Chyung J.H., Lassar A.B. Sonic hedgehog promotes somitic chondrogenesis by altering the cellular response to BMP signaling. Genes Dev. 1999;13:225–237. [PubMed]
Ogawa T., Kapadia H., Feng J.Q., Raghow R., Peters H., D'Souza R.N. Functional consequences of interactions between Pax9 and Msx1 genes in normal and abnormal tooth development. J. Biol. Chem. 2006;281:18363–18369. [PubMed]
Peterkova R., Peterka M., Viriot L., Lesot H. Development of the vestigial tooth primordia as part of mouse odontogenesis. Connect. Tissue Res. 2002;43:120–128. [PubMed]
Peterkova R., Lesot H., Peterka M. Phylogenetic memory of developing mammalian dentition. J. Exp. Zool. B Mol. Dev. Evol. 2006;306:234–250. [PubMed]
Riddle R.D., Johnson R.L., Laufer E., Tabin C. Sonic hedgehog mediates the polarizing activity of the ZPA. Cell. 1993;75:1401–1416. [PubMed]
Sasaki T., Ito Y., Xu X., Han J., Bringas P., Jr., Maeda T., Slavkin H.C., Grosschedl R., Chai Y. LEF1 is a critical epithelial survival factor during tooth morphogenesis. Dev. Biol. 2005;278:130–143. [PubMed]
Shu W., Guttentag S., Wang Z., Andl T., Ballard P., Lu M.M., Piccolo S., Birchmeier W., Whitsett J.A., Millar S.E., Morrisey E.E. Wnt/beta-catenin signaling acts upstream of N-myc, BMP4, and FGF signaling to regulate proximal-distal patterning in the lung. Dev. Biol. 2005;283:226–239. [PubMed]
Song L., Li Y., Wang K., Wang Y.Z., Molotkov A., Gao L., Zhao T., Yamagami T., Wang Y., Gan Q., Pleasure D.E., Zhou C.J. Lrp6-mediated canonical Wnt signaling is required for lip formation and fusion. Development. 2009;136:3161–3171. [PubMed]
Spater D., Hill T.P., Gruber M., Hartmann C. Role of canonical Wnt-signalling in joint formation. Eur. Cell Mater. 2006;12:71–80. [PubMed]
Thesleff I., Jernvall J. The enamel knot: a putative signaling center regulating tooth development. Cold Spring Harb. Symp. Quant. Biol. 1997;62:257–267. [PubMed]
Turing A.M. The chemical basis of morphogenesis. Philos. Trans. R. Soc. (B) 1952;236:37–72.
Tylzanowski P., De Valck D., Maes V., Peeters J., Luyten F.P. Zfhx1a and Zfhx1b mRNAs have non-overlapping expression domains during chick and mouse midgestation limb development. Gene Expr. Patterns. 2003;3:39–42. [PubMed]
Vainio S., Karavanova I., Jowett A., Thesleff I. Identification of BMP-4 as a signal mediating secondary induction between epithelial and mesenchymal tissues during early tooth development. Cell. 1993;75:45–58. [PubMed]
van Genderen C., Okamura R.M., Farinas I., Quo R.G., Parslow T.G., Bruhn L., Grosschedl R. Development of several organs that require inductive epithelial–mesenchymal interactions is impaired in LEF-1-deficient mice. Genes Dev. 1994;8:2691–2703. [PubMed]
Wang X.P., O'Connell D.J., Lund J.J., Saadi I., Kuraguchi M., Turbe-Doan A., Cavallesco R., Kim H., Park P.J., Harada H., Kucherlapati R., Maas R.L. Apc inhibition of Wnt signaling regulates supernumerary tooth formation during embryogenesis and throughout adulthood. Development. 2009;136:1939–1949. [PubMed]
Zhang Y., Zhang Z., Zhao X., Yu X., Hu Y., Geronimo B., Fromm S.H., Chen Y.P. A new function of BMP4: dual role for BMP4 in regulation of Sonic hedgehog expression in the mouse tooth germ. Development. 2000;127:1431–1443. [PubMed]
Zhou P., Byrne C., Jacobs J., Fuchs E. Lymphoid enhancer factor 1 directs hair follicle patterning and epithelial cell fate. Genes Dev. 1995;9:700–713. [PubMed]
Zhou C.J., Molotkov A., Song L., Li Y., Pleasure D.E., Pleasure S.J., Wang Y.Z. Ocular coloboma and dorsoventral neuroretinal patterning defects in Lrp6 mutant eyes. Dev. Dyn. 2008;237:3681–3689. [PMC free article] [PubMed]