Multiple modes of Lrp4 function in modulation of Wnt/β-catenin signaling during tooth development

Wnt/β-catenin signaling plays a pivotal role in the patterning, morphogenesis and growth of a variety of tissues and organs during development and in homeostasis in the adult. Aberrant Wnt signaling activity is causally linked to congenital defects, degenerative diseases and cancers (Clevers and Nusse, 2012). Therefore, understanding the molecular mechanisms that regulate the outputs of this signaling pathway in different in vivo contexts and expanding our knowledge of how this is governed through dynamic crosstalk among different tissues and cell types are fundamentally important.

In the Wnt/β-catenin signaling pathway, initiation of signaling requires interaction between Wnt ligands, their frizzled (Fz) receptors and Wnt co-receptors low-density lipoprotein receptor-related proteins 5 and 6 (Lrp5/6) (MacDonald and He, 2012). These interactions on the cell membrane trigger a cascade of intracellular events leading to stabilization and nuclear localization of β-catenin, which together with TCF/LEF transcription factors activates the expression of target genes (MacDonald and He, 2012; MacDonald et al., 2009).

A variety of secreted Wnt antagonists have been shown to inhibit Wnt/β-catenin signaling at the earliest step, presumably by altering or blocking the formation of Wnt/Fz/co-receptor complexes (Cruciat and Niehrs, 2013). In vitro binding studies have suggested that, among the Wnt antagonists, sclerostin (Sost) and Wise (also known as Sostdc1) can inhibit Wnt/β-catenin signaling via their ability to bind to the extracellular domains of Lrp5/6 (Ellies and Krumlauf, 2006; Itasaki et al., 2003; Li et al., 2005; Semenov et al., 2005). Sost and Wise are closely related, as they emerged through genome-wide duplication and divergence, but they display mostly non-overlapping expression patterns (Collette et al., 2013). The function of Sost and Wise in Wnt regulation via direct binding to Lrp5/6 has been further supported by genetic interaction studies in multiples tissues where they play a crucial role in development and homeostasis (Ahn et al., 2010, 2013; Chang et al., 2014b).

Lrp4 has emerged as an important component of the Wnt/β-catenin signaling pathway. The sequence and structure of its extracellular domain are similar to those of Lrp5 and Lrp6. Since the Lrp4 intracellular domain lacks some of the motifs in Lrp5 and Lrp6 known to be essential for Wnt co-receptor function, Lrp4 was proposed to be a negative regulator of Wnt signaling (Herz and Bock, 2002; Johnson et al., 2005; Weatherbee et al., 2006; Willnow et al., 2012). Supporting this idea, overexpression of Lrp4 results in decreased Wnt/β-catenin signaling activity in cultured cells (Johnson et al., 2005; Li et al., 2010; Ohazama et al., 2008). In in vitro binding assays, the extracellular domain of Lrp4 can directly interact with Sost and Wise, suggesting that the Wnt inhibitory function of Lrp4 may depend on its interaction with the Wnt antagonists (Choi et al., 2009; Karner et al., 2010; Ohazama et al., 2008).

In support of interaction between Lrp4 and Wise, mice deficient for Lrp4 or Wise share similar developmental defects in the ectodermal tissues, e.g. teeth, hair and mammary glands (Ahn et al., 2013; Narhi et al., 2012; Ohazama et al., 2008). Early development of these tissues requires reciprocal interactions between the epithelium and underlying mesenchyme, and Wnt signaling along with other major signaling pathways has diverse roles in the control of patterning and morphogenesis at different stages (Ahn, 2015; Balic and Thesleff, 2015; Biggs and Mikkola, 2014). In the tooth germ, Lrp4 is expressed in the epithelial signaling centers, while Wise is expressed in the surrounding epithelial and mesenchymal cells (Ahn et al., 2010; Laurikkala et al., 2003; Ohazama et al., 2008). Mice homozygous for a hypomorphic Lrp4 allele phenocopy Wise-null mice and display various tooth defects, such as supernumerary teeth and molar fusion (Ahn et al., 2010; Ohazama et al., 2008). Since the tooth defects in Wise-null mice are caused by elevated Wnt/β-catenin signaling (Ahn et al., 2010), it is possible that Lrp4 functions through its interplay with Wise as part of an important molecular mechanism for modulating Wnt/β-catenin signaling in teeth and other contexts.

To address this question, we utilized gain- and loss-of-function mouse models and in vitro reporter assays to investigate how Lrp4 interacts with Lrp5/6 and Wise. Our genetic interaction analyses focused on tooth development indicate that Lrp4 negatively regulates Wnt/β-catenin signaling to control tooth number, morphology and growth through potentiation of the Wnt inhibitory function of Wise. In addition, our study provides evidence suggesting a Wise-independent role for Lrp4 through its interaction with Wnt ligands and Fz receptors. This work has uncovered novel and diverse mechanisms by which Lrp4 contributes to the modulation of Wnt/β-catenin signaling during development.

We investigated the spatiotemporal expression pattern of Lrp4 in the diastema and molar region of the mandible during early tooth development. In mice, two tooth vestigial buds, namely MS and R2, develop sequentially in the toothless diastema region, but they undergo degeneration without advancing to the cap stage of tooth development (Ahn, 2015; Peterkova et al., 2006) (Fig. 1D). Consistent with a previous report (Ohazama et al., 2008), Lrp4 transcripts were detected in MS and R2 at E12.5 and E13.5, respectively, similar to the expression pattern of the TopGal Wnt activity reporter (Fig. 1A,B). At E14.5, Lrp4 expression is diminished in degenerating R2, while strong expression is observed in the more proximal region of the dental epithelium where the first molar (M₁) develops (Fig. 1A). Compared with TopGal, the Lrp4 expression domain at E13.5 and E14.5 is broader and extends more proximally to reach the proximal end of the tooth germ (Fig. 1A,B). The Lrp4 expression pattern was further examined in a reporter knock-in model, Lrp4^lacZ, and found to mimic the in situ hybridization pattern of Lrp4 in the developing incisor and molar regions (Fig. 1A,C). At E14.5, residual β-galactosidase activity marks degenerating R2, which is likely to be due to the stability of the lacZ transcript and/or β-galactosidase protein itself. These expression patterns are consistent with roles for Lrp4 in determining the fate of the diastema buds and M₁.

Supernumerary cheek teeth have been reported with a relatively low frequency in the mandible of mice homozygous for a hypomorphic allele, Lrp4^ECD (Ohazama et al., 2008). To clarify whether the supernumerary teeth arise from R2, we analyzed multiple stages of tooth development in mice homozygous for the Lrp4-null alleles Lrp4^mte or Lrp4^mitt (Weatherbee et al., 2006). First, Lrp4^mitt mice were crossed with Lrp4^lacZ mice and the presence of two domains of reporter expression at E14.5, as compared with controls, suggests that R2 continues to develop in Lrp4-null mice (Fig. 1E). Furthermore, an additional Shh expression domain is observed distal to M₁ at E15.5, suggesting continuous development of R2 in Lrp4-null mice (Fig. 1F). Lastly, narrower and weaker expression domains of enamel knot markers and their downstream target genes indicate abnormal tooth development in Lrp4-null mice (Fig. S1).

The progressive development of R2 and M₁ facilitated by loss of Lrp4 was further monitored utilizing the TopGal reporter line and histological sections (Fig. 1G-I). Consistent with the idea of survival of R2 in Lrp4-null mice, TopGal expression is sustained in R2 at E14.5, whereas M₁ development is delayed as it arises in the more proximal region (Fig. 1G). Survival and continued development of R2 has also been observed in Wise-null mice (Ahn et al., 2010). However, there are significant differences in the temporal and spatial expression patterns of TopGal in the Wise and Lrp4 mutants (Fig. 1G,H). In Wise-null mice, the continued development of R2 is greatly enhanced, as evidenced by a strong and enlarged TopGal expression domain at E14.5. Conversely, M₁ development is greatly delayed in Wise-null mice, as the TopGal expression domain marking M₁ is not detected until E15.5. R2 continues to develop to form a prominent supernumerary tooth, T₁ at E18.5 (Fig. 1I). By contrast, in Lrp4-null mice R2 is maintained with a relatively modest level of TopGal expression at E14.5-15.5 and the delay in M₁ development in Lrp4-null mice is not as significant as in Wise-null mice (Fig. 1G,H). The distance between R2 and M₁ is variable and generally shorter in Lrp4-null mice than in Wise-null mice at E15.5. Consistent with these differences, T₁, which originates from R2, is smaller in Lrp4-null mice at E18.5 (Fig. 1I). Together with reduced expression of the enamel knot markers and their targets, this suggests that R2 is not undergoing accelerated development in Lrp4-null mice and this is associated with a weak inhibition of M₁.

The comparative analyses above highlight differences as well as similarities in tooth phenotypes of Lrp4 or Wise mutant mice, suggesting the possibility of overlapping and independent mechanisms by which Lrp4 and Wise control tooth development. With respect to overlapping mechanisms, it is possible that similar effects of deficiency of Lrp4 or Wise on tooth development arise through their involvement in the regulation of a common signaling pathway. Our primary focus was the Wnt/β-catenin signaling pathway, as mutants of these two genes show altered Wnt signaling activity (Fig. 1G). We performed genetic interaction studies in order to explore this possibility and gain a deeper mechanistic understanding of their roles in tooth germs. We first investigated how Lrp4 and Wise impact each other's function during tooth development by crossing Lrp4^mitt mice with Wise-null mice and monitoring TopGal expression patterns in double mutants. This enabled us to determine whether inactivation of both genes results in significant changes in Wnt signaling and tooth defects compared with the individual mutants (Fig. 2A,B). Lrp4^mitt/mitt;Wise^−/− mice display no sign of exacerbated tooth defects and instead show variable distance between R2 and M₁, similar to Lrp4^mitt/mitt mice. This indicates that tooth defects of Lrp4^mitt/mitt;Wise^−/− mice are milder than those of Wise-null mice, in which a greater distance between R2 and M₁ is invariably maintained.

Since Lrp4^mitt/mitt mice die immediately after birth (Weatherbee et al., 2006), we also used Lrp4^ECD/ECD mice to evaluate how these changes during embryogenesis are translated into the number, size and shape of adult teeth. We observed that embryonic tooth development is similarly disrupted in Lrp4^ECD/ECD mice as in Lrp4-null mice (Fig. S2). After obtaining Lrp4^ECD;Wise compound mutants, we categorized mandibular and maxillary tooth patterns into distinct groups based on the severity of fusion and presence of the supernumerary cheek teeth (T₁) to aid our comparative analyses of these tooth phenotypes (Fig. 2C, Fig. S3). In the mandible, Lrp4^ECD/ECD mice display tooth defects similar to, but generally milder than, those of Wise-null mice (Fig. 2C). In contrast to the full penetrance of the T₁ phenotype in Wise-null mice, only 32% of Lrp4^ECD/ECD;Wise^+/− mice display T₁. This indicates that in the Lrp4-deficient mice, even though R2 initially escapes degeneration, the majority of R2 buds fail to develop into T₁ and instead merge into M₁ during later development. In addition to T₁, Wise-null mice frequently develop lateral supernumerary teeth and fusions between neighboring cheek teeth due to overgrowth, and these defects are rarely observed in Lrp4^ECD/ECD or Lrp4^ECD/ECD;Wise^+/− mice (Fig. 2C; data not shown). In the maxilla, both mutants display fusions of distal molars, with a higher penetrance observed in Wise-null mice (Fig. S3).

Comparison of mandibular tooth phenotypes among Lrp4^ECD;Wise compound mutants indicated that Lrp4^+/+;Wise^−/− mice display the most severe defects. Removing one allele of Lrp4 in a Wise^−/− background (Lrp4^+/ECD;Wise^−/−) and then the second Lrp4 allele (Lrp4^ECD/ECD;Wise^−/−) progressively reduces the severity of the phenotypes (Fig. 2C). For example, fully or partially separated T₁ was observed with 100%, 66% and 46% penetrance in Lrp4^+/+;Wise^−/−, Lrp4^ECD/ECD;Wise^−/− and Lrp4^ECD/ECD;Wise^+/− mice, respectively. A similar trend was observed in the maxilla of the compound mutants (Fig. S3). In order to rule out the possibility that the released extracellular domain of Lrp4 expressed from the Lrp4^ECD allele (Dietrich et al., 2010) has an effect on tooth development and complicates our analyses, we performed the same analyses with another hypomorphic mutant, Lrp4^mdig, combined with Lrp4^mitt mice and obtained similar results (Fig. S4). These genetic analyses indicate that Lrp4 makes an independent contribution to the generation of the severe tooth defects in Wise-null mice.

The more mild phenotypes observed upon loss of Lrp4 in Wise mutants suggests that Lrp4 might exert a stimulatory effect on Wnt/β-catenin signaling in a Wise-independent manner during tooth development. To explore whether elevated Wnt/β-catenin signaling is causally associated with the tooth defects in Lrp4 mutants, we crossed Lrp4^ECD mice with Lrp5- and Lrp6-null mice and examined the effect of reduced Lrp5/6 gene dosages on the tooth phenotypes among littermates. A supernumerary incisor is observed with ∼85% penetrance in the maxilla of Lrp4^ECD/ECD and this phenotype is rescued by reduced dosages of Lrp5/6 (Fig. 3A). In the mandible, a supernumerary cheek tooth (T₁) is observed with high frequency among Lrp4^ECD/ECD mice (90%, n=26) in this strain background, and reduced dosages of Lrp5/6 lead to a lower frequency of T₁ (Fig. 3B, top). In the maxilla, molar fusions common in Lrp4^ECD/ECD mice are rescued by reduced dosages of Lrp5/6 (Fig. 3B, bottom). This dosage-dependent rescue of Lrp4 tooth defects by deficiency in Lrp5/6 implies that elevated Wnt/β-catenin signaling is primarily responsible for the abnormalities in incisor and molar development in Lrp4 mutant mice. Similar to the results from our previous study with Wise-null mice (Ahn et al., 2010), we observed differences between Lrp5 and Lrp6 in their ability to rescue different aspects of the tooth defects with reduced gene dosages.

Although genetic analyses in teeth (Fig. 3A,B) and other tissues (Ahn et al., 2013) indicate that Lrp4 and Lrp5/6 generally have opposite roles in Wnt signaling, our data lead us to hypothesize that this relationship might be altered in the absence of Wise, revealing a potential Wise-independent role for Lrp4 in positively modulating Wnt/β-catenin signaling. To genetically test this idea, we investigated how the Lrp4-null allele Lrp4^mitt interacts with Lrp5 and Lrp6 in Wise-null mice. Interestingly, inactivating a copy of Lrp4 enhances the effect of reduced Lrp5 and Lrp6 gene dosage on Wise-null tooth defects and results in further rescue in the mandible (Fig. 3C). These data suggest that, in the absence of Wise, Lrp4 can positively regulate Wnt/β-catenin signaling, mimicking the normal roles of Lrp5/6. Together, these genetic analyses have uncovered dual roles for Lrp4. In the presence of Wise, Lrp4 negatively regulates Wnt/β-catenin signaling as its major mechanism of action to control tooth number and growth. In the absence of Wise, Lrp4 can play a stimulatory role in Wnt/β-catenin signaling through a different mechanism. This illustrates that Wise can alter the impact of Lrp4 on Wnt signaling in a context-dependent manner.

Overexpression of Wise in the dental epithelium leads to delay and hypoplasia in molar development (Ahn et al., 2010). We investigated whether this Wise gain-of-function phenotype requires Lrp4. Since K14-Wise mice were not able to breed, we instead utilized the Tet-off system, in which the keratin 14 (K14) promoter drives expression of a transactivator (tTA) in the first transgene (K14-tTA), and in the absence of doxycycline tTA in turn activates the expression of Wise and eGFP in the second transgene (tetO-Wise) (Ahn et al., 2013) (Fig. 4A). In K14-tTA;tetO-Wise mice, eGFP, which serves as an indicator of Wise expression, is detected specifically in the dental epithelium (Fig. 4B). After birth, molars appear to be smaller and the third maxillary molar is frequently missing, reminiscent of the phenotypes of the K14-Wise mice (Fig. S5).

These two transgenic lines were then crossed with Lrp4 mutant lines to generate Lrp4-deficient mice overexpressing Wise. In Lrp4^+/mdig mice, overexpression of Wise results in a reduced domain of the dental epithelial cells expressing TopGal and in hypoplasia of tooth germs at E15.5 (Fig. 4C). In Lrp4^mitt/mdig mice, R2 continues to develop, similar to our observation in other Lrp4 mutants, and this phenotype is not altered by overexpression of Wise (Fig. 4C). The lack of a Wise gain-of-function phenotype in Lrp4-deficient tooth germs suggests that Wise depends on Lrp4 to exert its Wnt inhibitory activity in tooth development.

With its highly restricted expression pattern in the dental epithelium, Lrp4 might provide a spatial cue for the action of Wise, which is broadly expressed in the tooth germ (Ahn et al., 2010; Laurikkala et al., 2003; Ohazama et al., 2008). To test this idea, we performed gain-of-function analyses of Lrp4 by generating a tetO-Lrp4 expression line (Fig. 5A). The transgenic line was first tested for its ability to rescue the limb defects of Lrp4 mutants. For this rescue experiment, we also generated Lrp4BAC-tTA driver lines, which express tTA in the Lrp4 expression domains, including the apical ectodermal ridge (AER) of limb buds (Fig. S6A). Severe patterning defects in the distal limb of Lrp4 mutants are fully rescued in Lrp4BAC-tTA;tetO-Lrp4 mice, indicating that functional Lrp4 protein is expressed from the transgene (Fig. S6B).

The tetO-Lrp4 line was then crossed with the K14-tTA line to test the effect of ectopic Lrp4 expression in the dental epithelium (Fig. 5B). Intriguingly, in K14-tTA;tetO-Lrp4 mice, overexpression of Lrp4 results in temporary survival of R2 (Fig. 5G,H), which fails to form a supernumerary tooth and instead merges into M₁ at later stages, leading to abnormal cusp patterning in the distal region of M₁ (Fig. 5C,D). We hypothesized that ectopic expression of Lrp4 disrupts the normal distribution and/or function of Wise and hence causes elevation in Wnt/β-catenin signaling in the epithelial signaling center of R2. This idea was supported by the observation that a half dose of Wise exacerbates the Lrp4 gain-of-function tooth phenotypes. Wise^+/−;K14-tTA;tetO-Lrp4 mice display more sustained survival of R2 (Fig. 5I,J), which gives rise to a supernumerary cheek tooth in the mandible (Fig. 5E,F). In the maxilla, overgrowth/fusion of distal molars was frequently observed in Wise^+/−;K14-tTA;tetO-Lrp4 mice reminiscent of Wise-null tooth phenotypes (Fig. 5C′-J′). The idea was also tested by simultaneously overexpressing Wise and Lrp4 in the dental epithelium. Tooth development is further delayed in K14-tTA;tetO-Wise;tetO-Lrp4 compared with K14-tTA;tetO-Wise mice, as evidenced by much smaller domains of TopGal expression at E15.5-E16.5 (Fig. 5K-R). Together, these data support a model whereby Lrp4 directs Wise function in the epithelial signaling center of developing teeth via direct interactions.

To gain insights into the molecular interactions between Lrp4, Wise and Lrp5/6 uncovered from our in vivo studies, we utilized an in vitro reporter system in which the activity of Wnt/β-catenin signaling is measured in a human cell line expressing different combinations of the proteins (Fig. S7). As expected, expression of Wnt1 or human WNT3A leads to a dramatic increase in Wnt reporter activity, indicating that the amount of Wnt ligand is a limiting factor for signaling activation in the cultured cells. Lrp4 and Wise antagonize the activity of WNT3A as well as Wnt1, consistent with the Wnt inhibitory potential of Lrp4 and Wise over these two classes of Wnt ligands. Co-expression of Lrp4 and Wise leads to further reduction in Wnt reporter activity.

The extracellular domain (ECD) of Lrp4 is similar to those of Lrp5 and Lrp6, whereas its intracellular domain (ICD) is distinct. To investigate whether this difference in the ICD underlies the Wnt inhibitory role of Lrp4, we generated chimeric proteins of Lrp4 and Lrp6 and tested their activity in the cultured cells (Fig. 6A-C). In general, the level of the Wnt co-receptor appeared to be another limiting factor, as co-expression of Lrp6 with the Wnt ligands results in further elevation of reporter activity. When Wnt ligands and Lrp6 are co-expressed, Wise is ineffective in suppressing the reporter activity. Lrp4ECD-Lrp6ICD (L4L6) mimics Lrp6, facilitating activation of Wnt/β-catenin signaling by Wnt1 or WNT3A, although not as efficiently as Lrp6 itself. Conversely, Lrp6ECD-Lrp4ICD (L6L4) mimics Lrp4, leading to a reduction in Wnt reporter activity. These results suggest that the respective ICD determines whether these two Lrp receptors play a stimulatory or inhibitory role in Wnt/β-catenin signaling. Furthermore, the Wnt stimulatory activity of L4L6 implies that Lrp4ECD can act like Lrp6ECD and interact with the Wnt ligands. L4L6 is more responsive to Wise co-expression, resulting in a more dramatic decrease in reporter activity compared with Lrp6. By contrast, L6L4 alone is more potent than Lrp4 in Wnt inhibition and this inhibitory activity is not enhanced by Wise co-expression. This suggests that Wise interacts with Lrp4ECD more efficiently than with Lrp6ECD to impact receptor activity.

We tested the in vivo relevance of the above chimeric receptor findings by expressing Lrp6 and L4L6 in transgenic mice utilizing the Tet-off system. As predicted, expression of Lrp6 in the dental epithelium results in the ectopic formation of tooth bud-like structures accompanied by TopGal expression, indicating increased activation of Wnt/β-catenin signaling (Fig. 6D). A similar phenotype was observed when L4L6 is expressed in the dental epithelium, consistent with its in vitro activity of stimulating Wnt signaling (Fig. 6D). Together, our in vitro and in vivo data indicate that the ECDs of Lrp4 and Lrp6 share an ability to interact with both Wnt ligands and Wise.

It has been shown that the ICD is dispensable for Lrp4 function in the neuromuscular junction as a receptor for agrin (Choi et al., 2013; Gomez and Burden, 2011). In this regard, it is unclear whether the tooth defects of Lrp4^ECD/ECD mice result from lack of anchorage to the cell membrane or from lack of ICD. Therefore, we utilized CRISPR/Cas9 technology to generate a new Lrp4 allele, Lrp4^ΔICD, which produces an Lrp4 protein lacking the ICD. Mice homozygous for this allele are viable and display tooth defects comparable to those of Lrp4^ECD/ECD mice, indicating that the ICD is essential for the normal function of Lrp4 in tooth development (Fig. 6E). The majority (77.8%, n=36) of Lrp4^ΔICD/ΔICD mice show a typical pattern of T₁, T₂-T₃ and T₄, with rare occurrence of later supernumerary teeth (6.7%) in the mandible. Some (16.7%) show signs of incomplete separation of T₁ and T₂ teeth, resulting in the pattern T₁-T₂-T₃ and T₄. Overall, the tooth defects of Lrp4^ΔICD/ΔICD mice are milder than those of Wise-null mice based on criteria such as the relative size of T₁ and the frequency of lateral supernumerary teeth. This highlights the important role of the Lrp4 ICD in tooth development.

The relatively mild tooth phenotypes of various Lrp4 mutants and the genetic interaction of Lrp4 with Wise and Lrp5/6 together suggest the existence of Wise-independent roles for Lrp4 in tooth development. Validation of this hypothesis requires a context in which the Wise-dependent roles are blocked while other aspects of Lrp4 function are preserved. Since no such mouse models were available, we created mice carrying the human R1170W mutation (G to T), which has been shown to abolish SOST binding to LRP4 (Leupin et al., 2011). This is likely to abolish Wise binding to Lrp4, without affecting the normal cell membrane localization of Lrp4 (Leupin et al., 2011). Mice homozygous for the Lrp4^R1170W allele are viable, indicating that the Lrp4 function in the neuromuscular junction is not significantly affected. To our surprise, Lrp4^{R1170W/R1170W} mice display tooth phenotypes more severe than those of Lrp4^ECD/ECD and Lrp4^ΔICD/ΔICD mice and comparable to those of Wise-null mice (Fig. 7A). For example, T₁ is typically longer with multiple cusps and lateral supernumerary teeth are frequently observed in Lrp4^{R1170W/R1170W} mice. This suggests that the single amino acid substitution is sufficient to prevent Lrp4 from inhibiting Wnt/β-catenin signaling during tooth development. It also implies that the substitution efficiently blocks the Lrp4-Wise interaction.

Severe syndactyly and/or oligodactyly are common to all known Lrp4 mutant models and are attributed to disruption in AER patterning (Li et al., 2010; Pohlkamp et al., 2015; Simon-Chazottes et al., 2006; Weatherbee et al., 2006). Similarly, Lrp4^ΔICD/ΔICD mice display severe defects in distal limbs (Fig. S8). Surprisingly, Lrp4^{R1170W/R1170W} mice display no apparent limb defects, indicating that the arginine residue is not essential for the Lrp4 function in patterning of the AER (Fig. S8). Since the limb defects are causally associated with elevated Wnt/β-catenin signaling (Ahn et al., 2013), our data suggest that Lrp4 negatively regulates Wnt/β-catenin signaling largely independently of Wise/Sost during early limb patterning. Supporting this idea, single and double mutants for Wise and Sost display no, or very subtle, limb defects (Collette et al., 2013).

Lrp4^R1170W mice were then crossed with Wise-null mice to test whether the Lrp4^R1170W allele interacts with the Wise-null allele differently, as compared with other Lrp4 alleles. Whereas no tooth defects are present in Lrp4^+/R1170W and Wise^+/− mice, a small portion (8.1%, n=37) of transheterozygotes display T₁, which has presumably developed from R2 (Fig. 7B,C). In contrast to Lrp4^ECD/ECD;Wise^−/− and Lrp4^mitt/mdig;Wise^−/− mice, in which the Wise-null tooth phenotypes are ameliorated, there is no significant difference in the severity of tooth defects between Lrp4^+/R1170W;Wise^−/− and Lrp4^{R1170W/R1170W};Wise^−/− mice (Fig. 7B-D). Together, these genetic data support a model whereby Lrp4 mediates the Wnt inhibitory function of Wise in tooth development, and the R1170 residue is essential for the Lrp4-Wise interaction. Since Lrp4^{R1170W/R1170W} mice phenocopy Wise-null mice with severe tooth defects, we conclude that Lrp4 has a Wise-independent Wnt stimulatory activity that is not disrupted by the missense mutation.

In this study, an extensive series of genetic analyses in teeth provide strong experimental evidence for a hypothesis that Lrp4 and Wise physically interact with each other to negatively regulate Wnt/β-catenin signaling. The findings also provide new insights into molecular mechanisms associated with interactions between Lrp4, Wise and the Wnt co-receptors Lrp5/6. We demonstrate that Lrp4 acts as both inhibitor and activator of the Wnt/β-catenin pathway dependent upon its interactions with Wise. In vitro and in vivo chimeric receptor analyses unveil roles played by the extracellular and intracellular domains of Lrp4 in its interplay with Wnt ligands, antagonists and Lrp5/6 co-receptors. By generating a mouse model of a human missense mutation, we demonstrate that Wise-dependent and Wise-independent roles for Lrp4 can be separated. These findings broaden our understanding of mechanisms that incorporate different inputs to achieve precise spatiotemporal control of signaling activity during development.

In this study, we uncovered not only similarities, but also significant qualitative and quantitative differences in tooth phenotypes between Wise-deficient and Lrp4-deficient mice. We utilized a Wise-null allele and a series of Lrp4-null and hypomorphic alleles and carefully analyzed tooth phenotypes to rule out the possibility that these differences result from partial inactivation of the gene or differences in strain backgrounds. In mice deficient for either of the genes, the diastema bud R2 escapes from degeneration and gives rise to a supernumerary cheek tooth distal to M₁ in the mandible. Whereas this phenotype is ∼100% penetrant in Wise-null mice regardless of strain background, Lrp4 mutants display a wide-range of penetrance (30-90%) in different strain backgrounds. One possible explanation for this difference is that the degree of alteration in signaling activity caused by Lrp4 deficiency is relatively low so that penetrance of the phenotype is affected by genetic modifiers. We discovered that, unlike in Wise-null mice, R2 fails to maintain distance from M₁ and often becomes merged into M₁ in Lrp4-null mice. In the lateral inhibition model, an existing tooth delays development of the next tooth (Kavanagh et al., 2007). This suggests that a lack of accelerated development of R2, and hence reduced lateral inhibition on M₁, results in relatively early development of M₁ in Lrp4 mutants. As R2 loses its inhibitory advantage over M₁, it is more likely to be merged into M₁, mimicking the outcome in normal tooth development. We also discovered that other aspects of tooth defects are milder in Lrp4 mutants. For example, lateral supernumerary teeth are rarely observed and overgrowth is not common in the mandible of Lrp4 mutants compared with Wise-null mice.

We observed that molar defects are more severe in the maxilla than in the mandible in both Wise-deficient and Lrp4-deficient mice leading to overgrowth and fusion of the distal teeth, consistent with published observations (Ahn et al., 2010; Ohazama et al., 2008). This less variable fusion phenotype is attributed to elevation of Wnt/β-catenin signaling to a much higher level in the maxilla of the mutants, as significant rescue of the defect requires removal of at least two copies of Lrp5 and Lrp6 (Ahn et al., 2010) (this study).

That there are significant differences in the severity of tooth defects caused by deficiency of Lrp4 or Wise is in line with the two proteins also having independent roles. In this regard, our previous study in embryonic mammary gland development revealed stage-specific roles for Lrp4 and Wise (Ahn et al., 2013). The earlier and more severe patterning defects in Lrp4 mutant mice compared with Wise-null mice may be attributed to a Wise-independent role for Lrp4 in the mammary placodes. Unexpectedly, in Lrp4;Wise double-homozygous mutants the tooth defects are ameliorated compared with those of Wise-null mice. Furthermore, removing a copy of Lrp4 on top of reduced dosages of Lrp5/6 results in further rescue of Wise-null tooth phenotypes. Since both Lrp4 and Wise single mutants display tooth defects associated with elevated Wnt signaling and these defects are rescued by reduced dosages of Lrp5/6, inhibition of Wnt/β-catenin signaling is likely to be the major mechanism of action of Lrp4 in tooth development. However, our data point to the presence of an additional Wnt stimulatory role for Lrp4 that is independent of Wise. In this scenario, the net effect of Lrp4 deficiency would still be an increase in Wnt/β-catenin signaling, as the inhibitory activity of Lrp4 in the presence of Wise is greater than its stimulatory activity. Consequently, overall tooth phenotypes of Lrp4 mutants would be milder than those of Wise-null mice, in which Lrp4 maintains its stimulatory activity. Lack of the stimulatory activity would explain the amelioration of Wise-null phenotypes in Lrp4^ECD/ECD;Wise^−/− and Lrp4^mitt/mdig;Wise^−/− mice. It is possible that this stimulatory activity is enhanced in the absence of Wise, contributing to the severe tooth defects of Wise-null mice.

Our gain-of-function analyses provided further insights into the interplay between Wise and Lrp4 in tooth development. Whereas overexpression of Wise in the dental epithelium suppresses tooth development, excess Wise fails to exert any significant effect on the dynamics of R2 and M₁ development in Lrp4-deficient mice. This dependence of Wise on Lrp4 suggests that Lrp4 is required for most, if not all, aspects of Wise function in the inhibition of Wnt/β-catenin signaling. Lrp4 gain-of-function tooth phenotypes are more intriguing as they mimic, to a certain degree, Lrp4 loss-of-function phenotypes. Since Lrp4 expression is temporally dynamic and spatially restricted, it is likely that ectopic overexpression disrupts Lrp4 function in its normal expression domain, which overlaps with the epithelial signaling centers. It is possible that this represents the Wise-independent Wnt stimulatory activity of Lrp4 mentioned above. We speculated that secreted Wise proteins become a limiting factor when Lrp4 is in excess, which can alter the distribution and function of Wise. This idea was supported by the observation that a reduced dosage of Wise exacerbates Lrp4 gain-of-function phenotypes and that simultaneous overexpression of Wise and Lrp4 results in stronger suppression of tooth development.

Our chimeric receptor analyses indicate that the ICD determines whether Lrp4 and Lrp5/6 play a stimulatory or inhibitory role in Wnt/β-catenin signaling. The observation that the ECDs of Lrp4 and Lrp6 are somewhat interchangeable suggests that the ECDs of Lrp4 and Lrp5/6 can interact with a similar set of signaling molecules, such as Wnt ligands and antagonists. This might be linked to Wise-independent roles for Lrp4 in certain contexts. Interestingly, Lrp4ECD appeared to be more responsive to co-expression of Wise, and Lrp6ECD more responsive to co-expression of Wnt ligands. This difference might reflect a higher affinity of Lrp4 to Wise and a lower affinity to Wnt ligands.

Our analyses of Lrp4^ΔICD/ΔICD mice indicate that the ICD is essential for Lrp4 function in limb and tooth development, but dispensable in the neuromuscular junction. Since the mutant mice display tooth defects comparable to those observed in mice with other Lrp4 mutations that presumably disrupt most aspects of Lrp4 function, it is likely that removal of the ICD leads to loss of both Wise-dependent and -independent functions of Lrp4. Anchorage to the cell membrane alone appears insufficient for Lrp4 function in the modulation of Wnt/β-catenin signaling. It remains to be investigated whether the ICD is required for proper trafficking to the cell membrane or is directly involved in the modulation of Wnt/β-catenin signaling.

In postnatal bone, loss of Lrp4 in osteoblasts leads to increased bone mass reminiscent of bone phenotypes observed in mice deficient for Sost (Chang et al., 2014a; Collette et al., 2012; Li et al., 2008; Xiong et al., 2015). Human patients with the LRP4 mutation R1170W display bone overgrowth and the mutation results in reduced binding to SOST and abolishes the LRP4 function as a facilitator of SOST in cultured cells (Leupin et al., 2011). The severe tooth phenotypes of Lrp4^R1170W mice and the lack of amelioration of Wise-null tooth defects in Lrp4^{R1170W/R1170W};Wise^−/− mice together suggest that Wise and Sost bind to the same domain of Lrp4 to inhibit Wnt/β-catenin signaling. Sost and Wise originate from a common ancestral gene (Collette et al., 2013), but have since diverged from each other. This study provides evidence that despite significant differences (less than 40% amino acid identity) between the two antagonists, Wise interacts similarly with Lrp4 to regulate Wnt/β-catenin signaling during tooth development. Lrp4^R1170W mice provide a valuable in vivo model in which the antagonist-dependent Wnt inhibitory role of Lrp4 is abolished while its antagonist-independent roles in the modulation of Wnt and other signaling pathways are retained.

With a large ECD possessing multiple protein-protein interaction motifs, Lrp4 can potentially interact with an array of signaling molecules to control Wnt/β-catenin signaling. Based on findings from the current and earlier studies, we propose a model involving multiple modes of action deployed by Lrp4 to modulate Wnt/β-catenin signaling (Fig. 8).

Modes I and II describe two mechanisms for Wise-dependent Wnt inhibitory activity of Lrp4. Lrp4 might act as an anchor/presenter molecule for Wise. During tooth development, Lrp4 recruits Wise to the epithelial signaling centers to inhibit Wnt/β-catenin signaling. Wise might be presented to Lrp5/6 via Lrp4, resulting in displacement of Wnt ligands and inhibition of signaling (mode I, Fig. 8A). Alternatively, Lrp4 together with Wise might compete with Lrp5/6 for binding to Fz receptors (mode II, Fig. 8B).

In some developmental contexts, Lrp4 might inhibit Wnt/β-catenin signaling independently of Wise. In this scenario, Lrp4 might compete with Lrp5/6, interfering with the formation of Wnt/Fz/co-receptor complexes (mode III, Fig. 8C).

The current study also discovered a Wise-independent Wnt stimulatory role for Lrp4. It remains to be investigated whether this role involves a direct interaction between Lrp4 and Wnt/β-catenin signaling or is mediated via another signaling pathway. Lrp4 might bind to Wnt ligands and present them to Lrp5/6 for activation of Wnt/β-catenin signaling (mode IV, Fig. 8D). This mode might be advantageous when the signaling needs to be activated rapidly in a defined domain and then turned off by expression of the antagonists.

Our findings suggest that Lrp4 acts as a modulator of Wnt/β-catenin signaling, integrating multiple inputs, and that its mode of action is determined by the presence and relative concentration of signaling molecules such as Wnt ligands, antagonists and receptors. Although it is unknown how Lrp4 regulates different signaling pathways, our study demonstrates that different domains of Lrp4 can be linked to pathway-specific roles, opening a door to more effective and safer therapeutics to treat disease conditions caused by abnormal LRP4 function (Shen et al., 2015).

TopGal, Lrp4^mitt, Lrp4^mte, Lrp4^mdig, Lrp4^ECD, Wise-null, Lrp5-null, Lrp6-null, K14-tTA and tetO-Wise mice were described previously (Ahn et al., 2013; DasGupta and Fuchs, 1999; Johnson et al., 2005; Kato et al., 2002; Pinson et al., 2000; Simon-Chazottes et al., 2006; Weatherbee et al., 2006) (Table S1). All experiments involving mice were performed under approved protocols issued to R.K. as the principal investigator by the Institutional Animal Care and Use Committee of the Stowers Institute for Medical Research (Protocol ID: 2016-0164).

The Lrp4BAC-tTA was constructed by inserting VP22-tTA-SV40pA (Gossen and Bujard, 1992) in-frame into the first coding exon of Lrp4 in a 134 kb Lrp4 BAC clone (Ahn et al., 2013) using bacterial recombination technology (Lee et al., 2001). The tetO constructs were generated by replacing the Wise ORF in the tetO-Wise construct (Ahn et al., 2010) with coding sequences of Lrp4, Lrp6 and Lrp4ECD-Lrp6ICD. Transgenic founders were generated by pronuclear injection of linearized constructs into (C57BL/10JxCBA)F2 embryos. F0 founder or N1 mice carrying individual tetO transgenes were crossed with K14-tTA mice to identify expression lines that drive eGFP expression in the presence of tTA.

Gene editing was achieved by pronuclear injection of pX330 plasmids expressing Cas9 and single guide RNA (sgRNA) (Cong et al., 2013; Mashiko et al., 2013; Wang et al., 2013). Twenty base pair seed sequences preceding the PAM sequence (NGG) at the target loci were cloned into the pX330 BbsI site. Lrp4^lacZ mice carrying an in-frame insertion of lacZ into the first exon were generated in (C57BL/10JxCBA)F2 embryos by co-injecting a donor plasmid (15 ng/µl) and pX330 plasmid (3 ng/µl) with 5′-GGCGCCCTGCTCTGCGCACA-3′ as a seed sequence (supplementary Materials and Methods).

Lrp4^ΔICD mice were generated by introducing two stop codons after Lys1751 using an oligo donor (5′-ATACCTATAAAGTTCTCAACTGATTTCAGCCCGATTTTTCCTCTTGAAGACACAGAAAATGATAACAGACGGATCCTGGAATGGGAAACCTGACCTATAGCAACCCCTCCTACCGAACTTCCACTCAGGA-3′; mismatched bases against the wild-type allele are underlined) and pX330 with 5′-ACACAGAAAATCCAAGTTCA-3′ as a seed sequence. The R1170W (C to T) mutation was introduced into Lrp4 by co-injecting an oligo donor (5′-TTGGCAACCTGGATGGGTCTATGCGGAAAGTGTTGGTGTGGCAGAACCTTGACAGTCCCTGGGCCATTGTATTATACCATGAAATGGGGTGAGAGCTGGCTTTATCACTCTGAGTGGAC-3′) and pX330 with 5′-TGGTATAATACAATGGCCCG-3′ as a seed sequence. Both Lrp4^ΔICD and Lrp4^R1170W mice were generated and maintained on the FVB/N strain (supplementary Materials and Methods).

To detect β-galactosidase activity from lacZ reporters, embryos were dissected and fixed in either 0.1% paraformaldehyde (PFA)/0.2% glutaraldehyde (E11.5-E13.5) or 4% PFA (E14.0 or older) for 30-60 min on ice. After washes in PBS, fixed samples were stained in X-Gal for 4-20 h at 4°C or at room temperature. Whole-mount in situ hybridization was performed with dissected jaws fixed in 4% PFA overnight according to standard protocols using DIG-labeled antisense riboprobes (Roche) against Lrp4 (Weatherbee et al., 2006) and Shh (Dassule and McMahon, 1998). For histological sections, stained samples were paraffin embedded after post-fixation in 4% PFA, sectioned at 8 µm and counterstained with Nuclear Fast Red.

HEK 293T cells were grown in DMEM (Dulbecco's Modified Eagle Medium with 4 mM glutamine, Gibco) supplemented with 10% FBS. Cells were plated on 24-well plates at ∼50% confluence, transfected with Lipofectamine 2000 (Thermo Fisher Scientific) the following day according to the manufacturer's protocol and harvested 24 h after transfection. Expression constructs were generated by inserting a full-length cDNA of each gene into the multicloning sites of pCS2+ (Turner and Weintraub, 1994). The amounts of individual DNA constructs used per well were 100 ng TOPflash (EMD Millipore), 5 ng Renilla luciferase, 100 ng Wnt1 or human WNT3A, 100 ng Lrp4 or Lrp6 or chimeric receptors, and 200 ng Wise or Sost. The empty vector pCS2+ was added to the DNA mix to keep the total amount of DNA at ∼700 ng/well. Dual-luciferase reporter assays (Promega) were performed according to the manufacturer's protocol. Each experiment was performed in duplicate and representative results are from one of three independent experiments.

We thank Jenny Reynolds, Heidi Monnin and the Stowers Institute Histology Facility for technical assistance and members of the R.K. laboratory for valuable discussion.