R-spondin 3 deletion favors Erk phosphorylation to enhance Wnt signaling and bone formation

Activation of Wnt signaling leads to high bone density. The R-spondin family of four secreted glycoproteins (Rspo1-4) amplifies Wnt signaling. In humans, RSPO3 variants are strongly associated with bone density, but how RSPO3 affects skeletal homeostasis is not fully understood. Here we show that in mice Rspo3 haplo-insufficiency or its targeted deletion in osteoprogenitors lead to an increase in bone formation and bone mass. Contrary to expectations, Rspo3 haplo-insufficiency results in canonical Wnt signaling activation. Using mouse embryonic fibroblasts we show that Rspo3 deficiency leads to activation of Erk signaling, stabilizing  -catenin. Furthermore, Rspo3 haploinsufficiency impairs Dkk1 efficacy in blocking canonical Wnt signaling and prevents the in vivo inhibition of bone formation and bone mass induced by osteoblast-targeted expression of Dkk1 . We conclude that Rspo3 haplo-insufficiency/deficiency boosts canonical Wnt signaling by activating Erk signaling and impairing Dkk1 ’s inhibitory activity, which in turn lead to increased bone formation and bone


Introduction
The Wnt signaling pathway controls cell fate decisions and tissue homeostasis during development and in the adult 1,2 . It is also involved in skeletal development and essential for the regulation of bone mass in the adult skeleton 3-5 . Genetic or therapeutic activation of canonical Wnt signaling is associated with increased bone mass 3 and current therapeutic approaches aim at activating this pathway in patients with osteoporosis or osteogenesis imperfecta for instance 6 .
Wnt signaling involves a large number of receptors and co-receptors, and of endogenous agonists and antagonists that, together, tightly regulate its activation (1-4). Due to this complexity, and even though Wnt signaling has been studied extensively in recent years in bone, several aspects of the mechanisms by which it regulates bone mass remain unclear. Similarly, the specific downstream events regulated by these various components of the Wnt signaling machinery and their interaction with other signaling cascades remain puzzling.
In this context, the fact that several studies have demonstrated that the four Roof plate-specific spondin, R-spondins (Rspo1 to 4), synergize with Wnt ligands to activate Wnt signaling 4,7-13 raises the question of their potential role in skeletal development and homeostasis. This enhancement of Wnt activity is attributed to the ability of Rspos to prevent the ubiquitination and degradation of the Lrp5/6/Fzd receptor complex via Lgr4-6, closely related orphans of the leucin-rich repeat containing G protein-coupled receptors, and the transmembrane E3 ubiquitin ligases ring finger 43 (Rnf43) and zinc and ring finger 3 (Znrf3), sensitizing cells to Wnt ligands 7, [14][15][16][17][18] . Although the role of Lgr receptors in the effects of Rspos is well established, recent reports have shown that Rspo2 and Rspo3 can also enhance Wnt signaling independently of Lgr receptors, possibly by acting as direct antagonist ligands to RNF43 and ZNRF3 11,19,20 .
Of particular interest is the fact that, in contrast to the many studies that have reported that Rspos co-activate Wnt signaling, some studies in Zebrafish have indicated that Rspo3 can also function as a negative regulator of canonical Wnt signaling during dorsoventral and anteroposterior patterning 21,22 .
Additionally, it has been shown that Rspos can potentiate non-canonical Wnt cascades, such as the Wnt/PCP signaling 14,15 and can function as antagonists of BMPR1 in Xenopus 23 , two events that could have a negative impact on bone formation and bone mass. Thus, the mechanisms involved in the Rspos/Wnt signaling axis and their influence on skeletal homeostasis appear be more complex in vivo.
The four Rspos belong to a family of cysteine-rich secreted glycoproteins with high structural similarity and 60% sequence homology 8,11 . Although all Rspos are expressed in bone during development, they appear to have specific functions and their deletion in mice and human leads to different phenotypes 10,13, [24][25][26][27][28][29][30][31] . Within the Rspo family, Rspo3 is of particular interest for bone because it is highly expressed in skeletal elements during development 32 and several human GWA studies have shown that RSPO3 SNPs are strongly associated with bone mineral density and risk of fracture [33][34][35][36][37][38] . Not surprisingly, in vitro studies have shown that overexpression of, or treatment with, Rspos can enhance Wnt-ligand mediated osteoblast (OB) differentiation 10,28,39 . However, it has also been reported that Rspo3 may function as a negative regulator of osteoblast differentiation 40 , raising questions about the true net influence of Rspos, and in particular Rspo3, on skeletal homeostasis.
Thus, despite the consensus that Rspos co-activate Wnt signaling, Rspo3 may play a specific role in skeletal biology and disease, exerting a positive and/or a negative influence on bone. Little is known about its specific effects and on the mechanisms by which Rspo3 affects the skeleton in vivo.
Here, we show that, in mice, decreasing the levels of Rspo3 increases Erk phosphorylation, activates Wnt signaling and is anabolic to bone, suggesting that its specific inhibition could constitute a therapeutic mean to increase bone mass.

Rspo3 haplo-insufficiency increases bone formation and trabecular bone mass.
Rspo3 is expressed in bones during skeletal development 32 . We assessed Rspo3 mRNA levels in primary calvarial osteoblasts (cOBs) and found that it is expressed in these cells and that its expression increases significantly during OB differentiation (Fig. S1a). To determine the physiological role of Rspo3 in skeletal homeostasis, we first used mice in which Rspo3 has been germline-deleted 27 . As global deletion of Rspo3 leads to lethality by E9.5 24,27 , before skeletal development, we analyzed mice lacking only one Rspo3 allele. Rspo3 +/mice develop normally (Fig. S1b), are fertile and are born at the expected Mendelian ratio. These mice continue to be healthy and do not develop any particular pathology as they age (up to one year) (data not shown). Most surprisingly however, and contrary to expectations, two-way ANOVA analysis of the skeletal phenotype at 6, 12 and 18week (wk) of age, revealed highly significant (p<0.001) anabolic effects of Rspo3 haplo-insufficiency on structural, cellular and dynamic parameters in both female and male mice (Fig.1a, Table 1, Table   S1 and Fig. S1c). The skeletal phenotype of Rspo3 +/male and female mice is overall characterized by an increase in trabecular bone mass with high bone formation, mineral apposition rate and OB number and surface whereas bone resorption parameters are not affected (Fig.1a, Table 1,Table S1 and Fig. S1c ). As expected, the two-way ANOVA also demonstrated an effect of age, affecting primarily the structural parameters (Fig. 1a, Table 1 and Table S1 ). At 12 wk of age Rspo3 haploinsufficiency led to a significant increase in trabecular bone mass (BV/TV), trabecular thickness (Tb.Th.) and trabecular number (Tb.N) ( Fig. 1b and 1d, Table 1 and Table S1). In both genders bone resorption parameters (Oc.S/B.Pm and N.Oc/B.Pm) were not changed in Rspo3 +/mice. In contrast, Rspo3 +/mice exhibited an increase in bone formation parameters (BFR/BS) ( Fig. 1b- Table 1 and   Table S1). The increase in BFR was associated with an increase in mineral apposition rate (MAR), indicating a marked increase in the activity of individual OBs in Rspo3 +/mice, in addition to the increase in their numbers (N.Ob/B.pm, Fig. 1d, Table 1 and Table S1). Consistent with these results, the osteoid surface (OS/BS) and the OB surface (Ob.S/B.Pm) were also significantly increased in both genders (Table 1 and Table S1). Despite these marked effects on trabecular bone, all cortical bone parameters were unchanged in Rspo3 +/female and male mice at 12 wk of age (Fig. 1d, Table   S2 and S3), indicating that Rspo3 haplo-insufficiency affects preferentially trabecular bone homeostasis.
Thus, in contrast with the expectation that decreasing the expression of a Wnt signaling potentiator might lead to a decrease in bone formation and bone mass, our data clearly indicates that Rspo3 haplo-insufficiency induces an increase in trabecular bone mass due to a significant increase in bone formation, with no changes in bone resorption. In agreement with our in vivo observations, we found that, although Rspo3 is expressed in bone marrow macrophage (BMM)-derived osteoclasts (OCs) (Fig. S2a), there was no significant differences in the formation of TRAP + multinucleated cells and in the expression of OC marker genes (Ctsk, Trap, Nfatc1) between BMM cultures from wt and Rspo3 +/mice in response to M-CSF and RANKL ( Fig. S2b and 2Sc). In addition, mix-and-matched co-cultures of cOBs and BMMs from wt or Rspo3 +/mice confirmed that Rspo3 haplo-insufficiency does not affect osteoclastogenesis, whether directly or indirectly (Fig. S2d).
Rspo3 haplo-insufficiency leads to an increase in bone marrow precursor cells and in their osteoblast potential.
Given that the OB number was significantly increased in mutant mice ( Fig. 1, Table 1 and Table   S1) we hypothesized an increase in the population of precursor cells. Bone marrow flow cytometry showed that while the total number of bone stromal cells (Lin -CD45 -) was not significantly affected by Rspo3 haplo-insufficiency (4047±1245 in wt compared to 5867+2382 in Rspo3 +/-, mean ± SEM n=10), the mesenchymal stromal cells (MSC) population (defined here as Lin -CD45 -CD31 -CD51 + Sca-1 + ) was significantly increased in Rspo3 +/mice compared with wt littermates (Fig. 2a and 2b) 41 . Consistent with these findings and with the observed increase in OB number and bone formation, Rspo3 haploinsufficiency significantly increased CFU-F and CFU-OB formation (Fig. 2c). These data show that the changes induced by Rspo3 haplo-insufficiency affect the bone marrow MSC lineage and induce an increase in the pool of progenitor cells with an osteoblast potential.

Specific deletion of Rspo3 in cells of the osteoblast lineage mirrors the skeletal phenotype seen with global Rspo3 haplo-insufficiency.
To assess whether the effect on bone mass seen in Rspo3 haplo-insufficient mice was cellautonomous to cells of the OB lineage, we generated mice with deletion of Rspo3 in OB precursors  Table   S4). These findings indicate that the increase in bone formation and in bone mass results from a cellautonomous effect of Rspo3 insufficiency in osteoblast progenitors.

Rspo3 haplo-insufficiency and deletion lead to -catenin stabilization.
The above results raised the question of how reduction in Rspo3, classically considered a potentiator of canonical Wnt signaling, leads to increased bone formation and trabecular bone mass.
To address this question, we first looked at the status of the Wnt signaling pathway in Rspo3 haplo-insufficiency. Surprisingly, but consistent with the observations on bone formation, Rspo3 haploinsufficiency led to a remarkable increase in the expression of the canonical Wnt target genes Dkk1, Axin2 and Opg in marrow-flushed long bones and in the protein levels of activated -catenin ( Fig.4a and 4b). Importantly, we noted no difference in the mRNA levels of Rspo1 and Rspo2 between Rspo3 +/and wt marrow-flushed long bones (Fig.4a). We also observed that Sost expression was, in contrast with that of Dkk1, not affected by Rspo3 haplo-insufficiency suggesting that the bone formation and Wnt signaling changes seen in Rspo3 +/mice are independent of sclerostin levels ( Fig.4a). While both Opg and Rankl expression was increased, the Rankl/Opg ratio remained unchanged in Rspo3 +/long bones, consistent with our in vivo and in vitro observation of unchanged number of OCs. Given the observed increase in OB progenitors in the bone marrow of Rspo3 +/mice ( Fig. 2), we also determined the level of Wnt signaling activation in Rspo3 haplo-insufficient bone marrow MSCs (BMSCs). Canonical Wnt signaling was also activated in these cells, as indicated by increased expression of several canonical Wnt target genes and increased active -catenin levels ( Fig. 5a and 5b). Thus, surprisingly but consistent with the bone and OB phenotypes, Rspo3 haploinsufficiency leads to activation of canonical Wnt signaling, which in turn results in increased bone formation.
To exclude any function of the remaining Rspo3 on canonical Wnt signaling in the Rspo3 +/mice and cells, we generated Rspo3 -/mouse embryonic fibroblasts (MEFs) at E9.5, before embryonic lethality 24,27 . As shown by several groups 7,8,12,29 , we confirmed that while Rspo3 does not activate Wnt signaling by itself, it potentiates exogenous Wnt3a action in wt MEFs as indicated by the Tcf-1/Lef luciferase reporter assay (Fig. 6a). Counter-intuitively, Rspo3 deficiency led to a marked increase in the expression of the canonical Wnt target genes Tcf-1 and Axin2 at steady-state ( Fig. 6b) as well as to a significant increase in the levels of pLrp6, activated -catenin and Tcf-1 (Fig 6c).
Accordingly, TOPflash reporter activity was significantly increased in Rspo3 null MEFs compared to wt MEFs and further increased by Wnt3a treatment (Fig. 6d). Similarly, upon Wnt3a treatment, Axin2 and Tcf-1 expression as well as pLrp6 and activated -catenin levels, were increased in the absence of Rspo3 ( Fig. 6c-d). Thus, these findings confirm that, unexpectedly, haplo-insufficiency and absence of Rspo3 in BMSCs and MEFs respectively lead to -catenin stabilization and enhancement of -catenin-dependent signaling. We then explored the mechanisms by which Rspo3 may regulate Wnt signaling.

Rspo3 haplo-insufficiency and deletion impair Dkk1-Wnt inhibitory activity.
Activation of canonical Wnt signaling results from changes in endogenous activators and/or inhibitors levels and/or activity. Interestingly, we found that Wnt3a decreases the expression of Rspo3 in wt MEFs, whereas it is significantly increased by Dkk1 (Fig. 7a). This raised the possibility that Rspo3 participates in a feedback loop that tones down or balances canonical Wnt activity. Our results below suggest that the reduction in Rspo3 levels induced by Wnt3a may enhance Wnt signaling by decreasing Dkk1 Wnt inhibitory activity. Indeed, as shown in Figure 7b, the ability of Dkk1 to block Wnt3a-dependent activation of canonical Wnt signaling was significantly impaired in the absence of Rspo3: whereas in wt MEFs, 50% reduction in the reporter activity was achieved by 50 ng/mL Dkk1, a dose of 400 ng/mL Dkk1 (8x higher) was needed to obtain the same level of inhibition in Rspo3 null MEFs (Fig. 7b). A similar difference was also observed in pLrp6 and -catenin protein levels (Fig.   7c). To determine whether this relationship between Rspo3 levels and Dkk1 efficacy was also happening in vivo, we crossed Rspo3 +/mice with mice expressing high levels of Dkk1 in OBs (Dkk1-Tg mice), which exhibit impaired canonical Wnt signaling and low trabecular bone mass due to decreased bone formation 43 . Histomorphometric analysis showed that the low BV/TV and decreased bone formation parameters (MAR, BFR/BS and N.Ob/B.pm) seen in Dkk1-Tg mice were significantly rescued by Rspo3 haplo-insufficiency (Fig. 7d, 7e and Table S5). These data, together with the findings that Dkk1 expression is higher in Rspo3 +/long bone, despite the fact that Rspo3 +/mice display higher trabecular bone mass, indicate that Rspo3 haplo-insufficiency counteracts the effect of Dkk1 overexpression on the skeleton.
Since both Rspo3 haplo-insufficiency and its deletion lead to -catenin stabilization, we then asked whether, possibly in addition to or explaining the decreased efficacy of Dkk1, Rspo3 might also be involved in the regulation of other signaling pathways which in turn can stabilize -catenin, stimulating osteoblastogenesis and counteracting Dkk1 efficacy. It has been proposed that Rspo3 binding to Lgr4 inhibits Erk phosphorylation (pErk) 23,40,44 . We therefore investigated Erk signaling, known to activate Wnt signaling and to regulate OB differentiation and bone mass [45][46][47][48] , in our model. Rspo3 deficiency led to a clear and significant increase in pErk basal levels (Fig. 8a). Although in wt cells the specific Erk inhibitor U0126 significantly decreased both the basal and the Wnt3a-dependent increase in pErk, it did not significantly affect active -catenin levels. In contrast, inhibition of pErk in Rspo3 -/cells led to a significant decrease in active -catenin levels in both steady state and Wnt3a-stimulated cultures. A similar effect was also seen for the levels of pLrp6 (Fig. 8a). Confirming these findings, the increase in the expression of the canonical Wnt signaling target genes Tcf1 and Axin2 was also partially rescued by blocking Erk signaling in Rspo3 null MEFs (Fig. 8b). Thus, the stabilization of catenin we observe in the absence of Rspo3 is due, at least in part, to activation of the Erk pathway.

Discussion
Wnt signaling is central to skeletal development and homeostasis in health and disease 3 .
Understanding the biological mechanisms by which this signal operates is therefore of both scientific and clinical interest. R-Spondins, classically considered as positive modulators of Wnt signaling, play an important role in normal development of several tissues and organs including bone, and are implicated in human diseases 8,9,[11][12][13] . Investigating further their exact function therefore holds therapeutic potential. Our results clearly demonstrate, through several independent lines of genetic in vivo and in vitro experiments, that, counter-intuitively, decreasing Rspo3 results in canonical Wnt signaling activation, increased bone formation and high bone mass. This response is mainly driven by increased number of OB progenitors and OBs as well as an increase in their bone forming activity, with no effect on OC number and bone resorption. Thus, our studies reveal a novel and unexpected negative function of Rspo3 on the Wnt signaling machinery in bone homeostasis. This might have implications for our understanding of the multi-faceted aspects of Wnt signaling regulation of skeletal homeostasis, and possibly reveal novel ways to increase bone mass in patients.
Despite the fact that the four Rspos display high structural similarity and sequence homology, they display differential expression profiles 32 and have unique and distinct functions 8,11,49 . Studies in mice have shown that during embryonic development Rspo3 is highly expressed in the skeleton 32 . Human GWA studies have shown that RSPO3 might be specifically involved in bone metabolism due to the strong association between RSPO3 common variants and bone mineral density 33,34,[36][37][38] . Whether these variants lead to RSPO3 gain of-or loss of-function is however not known. Rspos are considered to be co-activators of Wnt signaling 3,7,29 . Accordingly, treatment with Rspos has been reported to have a positive effect on osteogenesis 10,28,39 , such that decreasing its expression levels should have a negative impact on skeletal homeostasis. In contrast, we show here that Rspo3 haploinsufficiency has an anabolic effect on bone. Importantly, targeted deletion of Rspo3 in the OB lineage in mice mimicked the skeletal phenotype seen with Rspo3 global haplo-insufficiency, revealing an OB lineage cell-autonomous effect of Rspo3 in the regulation of skeletal homeostasis.
In vitro studies have demonstrated that overexpression of and treatment with Rspo1 or Rspo2 enhance Wnt ligand-mediated OB differentiation 9,10,28 . Nonetheless, our in vitro, ex vivo and in vivo studies demonstrate that Rspo3 haplo-insufficiency and its deletion result in canonical Wnt signaling activation. As discussed later, and although paradoxical, these two observations are not mutually exclusive or discrepant. The literature and our own in vitro studies indeed confirm that Rspo3 is, as expected, a co-activator of canonical Wnt signaling in the Topflash assay, potentiating Wnt3adependent activation of canonical Wnt signaling in cellular assays ( 49 and Fig. 6). Surprisingly however, the expression of canonical Wnt target genes and the levels of pLrp6 and activated catenin were markedly increased in Rspo3 haplo-insufficient bones and BMSCs and in Rspo3-null MEFs, indicating that, counter-intuitively, the decrease or absence of Rspo3 in bone activates mechanisms that favor -catenin-dependent signaling.
The findings that Rspo3 is strongly repressed by Wnt3a and increased by Dkk1, suggest that Rspo3 may in fact provide a negative feedback-loop helping to dampen canonical Wnt activity by repressing Erk phosphorylation (see below). Interestingly, and again confirming activation of the canonical Wnt signaling pathway, the expression of Dkk1 is also strongly increased by Rspo3 haploinsufficiency and depletion, establishing another Rspo3-dependent negative feedback loop. Although the increase in Dkk1 would be expected to block Wnt activation, we found that Dkk1 efficacy in blocking Wnt3a-dependent activation of canonical Wnt signaling is significantly impaired in the absence of Rspo3. Additionally, our finding that Rspo3 haplo-insufficiency antagonizes the inhibition of bone formation induced by OB-targeted expression of Dkk1 confirms in vivo the fact that Rspo3 haplo-insufficiency counteracts Dkk1 function. The question is whether this occurs at the cell surface or is the result of intracellular changes in alternative pathways regulated by Rspo3. In fact, our findings that the Erk signaling pathway is activated and the basal level of pLrp6 enhanced by Rspo3 depletion suggests the possibility that the activation of the Wnt signaling pathway results from intracellular changes. Alternatively, Rspo3 haplo-insufficiency and deletion might induce cell surface changes that favor pLrp6 basal level and/or sensitivity to Dkk1 and/or Wnt ligands. Further studies will be required to investigate these possibilities. Interestingly, Sost levels were not altered by Rspo3 haploinsufficiency suggesting that the effect of Rspo3 on Dkk1 efficacy and expression represents a specific regulatory mechanism.
Although there is an apparent paradox in that both Rspo3 and its deletion increase -catenindependent signaling, these results can in fact be explained. Our data and that of others 23 suggest that there are alternate Rspo3-mediated signaling mechanisms, separate from the Fzd/Lrp/-catenin Wnt pathway, including the Wnt/PCP signaling 14,15 and that these events can in turn regulate Wnt signaling intra-cellularly. Indeed, -catenin can be stabilized independent of the proximal activation of the canonical Wnt signaling machinery through changes in other signaling pathways 3,48,50 .
In fact, we show here that deletion of Rspo3 enhances Erk signaling which, in turn, stabilizes −catenin independent of the canonical Wnt signaling receptor complex as shown by the significant increase in Erk phosphorylation observed in the absence of Rspo3. In turn, this activation of Erk signaling promotes pLrp6 and -catenin stabilization and has a positive effect on OB differentiation 45,47,48 . Our finding that the increase in Wnt signaling activation seen in the absence of Rspo3 is abrogated by blocking Erk signaling strongly suggests that the activation of Erk signaling associated with Rspo3 deficiency is responsible, at least in part, for the observed Wnt signaling activation.
In agreement with our findings, in vitro studies have reported that Rspo3 silencing leads to increased OB differentiation of human adipose-derived stem cells by activating the Erk signaling downstream of Lgr4 40 . Importantly, although Lgrs function as receptors for Rspos, and Rspos/Lgrs interactions enhance Wnt signaling by inducing the clearance of Rnf43 and Znrf3 16,17,51 , there is strong evidence that Rspos/Lgrs interaction can also activate distinct signaling cascades that can affect bone, including the cAMP/PKA/Creb signaling pathway in Lgr4 null mice 52 (Fig. 9). Thus, haplo-insufficiency and deletion of Rspo3 would dampen Wnt signaling at the cell surface by preventing the Rnf43/Znrf3 effects while enhancing pLrp6 and −catenin stabilization intracellularly, via Erk phosphorylation, which has a more potent effect and overcompensates the decrease in Rspo3-dependent proximal Wnt activation in OBs and their progenitors. Because activation of the Lgr/Rnf43/Znrf3 cascade is not exclusively dependent on Rspo3, deletion of Rspo3 would only hinder canonical Wnt signaling partially. In contrast, lack of Rspo3 promotes the Lgr/Erk cascade, which leads to Erk signaling activation, which in turn not only enhances −catenin stabilization (Fig. 9) but also regulates OB differentiation and bone formation. This model can also explain the observed loss of Dkk1 efficacy in inhibiting Wnt signaling: the Erk-dependent stabilization of −catenin being independent of proximal Wnt receptor activation, Dkk1 cannot dampen the activation of downstream events as they are independent of the LRP5/6-Fzd receptor complex.
Supporting the fact that Rspo3 can also regulate Wnt-independent pathways, a recent study has suggested that Rspo3 acts as an antagonist to BMPR1A, inhibiting BMP signaling during development 23 . Thus, our observations may be due, at least in part, to changes (activation) in BMP signaling, which in turn could lead to the observed increase in pErk 44 . Although this remains a possibility, it seems unlikely. First, in contrast to our observations here, BMP activation in the adult skeleton has been linked to activation of non-canonical Wnt signaling, increased Sost expression and bone resorption 56,57 . Second, several studies have shown that activation of BMP signaling in the osteoblast lineage has a negative impact on bone formation and bone mass [58][59][60][61] .
In conclusion, our studies suggest that Rspo3 regulates bone formation through its interaction not only with the Wnt receptor machinery but also with other signaling pathways that affect −catenin stability. Consequently, if its deletion removes a co-activator of Wnt signaling, potentially decreasing bone formation, it also promotes Erk signaling activation, increasing −catenin stability sufficiently to enhance bone formation and increase bone mass. Furthermore, because Rspo3 depletion increases Dkk1 and Dkk1 increases Rspo3 expression, this study reveals also a novel feedback Wnt signaling regulatory loop. These findings have important implications for understanding the pleiotropic functions of Rspos and Wnt signaling in skeletal homeostasis and reveal alternative mechanisms to increase bone mass.

Flow Cytometry
Bone marrow was analyzed by flow cytometry as previously described 41

Bone Marrow Stromal Cells and calvarial osteoblasts
Bone marrow cells were flushed from femurs and tibiae of 6-8 wk-old wt or Rspo3 +/mice and cultured in DMEM supplemented with 10% fetal bovine serum (FBS) (GIBCO) and 1% penicillin (100 U/ml) and streptomycin (100 µg/ml) for 3 days. Adherent MSCs were counted, re-plated onto at a 5,000/cm 2 density and RNA or protein isolated 3 days later. For colony forming unit assays, flushed bone marrow stromal cells were plated (3X10 6 /6 wells) for CFU-(Fibroblast) F and CFU-OB assays. CFU-F was detected by staining with 0.2% crystal violet in 2% ethanol for 1 hour and CFU-OB was detected by alkaline phosphatase activity with Napthol AS-MX, n,n-dimethylformamide and Fast Blue RR salt (Sigma Aldrich, St. Louis, MO, USA). Calvarial OBs were isolated from 1-3-day old pups via serial enzymatic digestions and cultured as previously reported 63 .

Mouse Embryonic Fibroblasts (MEFs) Primary Culture
To obtain wt and Rspo3 -/-MEFs, Rspo3 +/males and females were crossed and the morning of vaginal plug detection was defined as embryonic day (E) 0.5. At E9.5, whole embryos were isolated, washed in PBS, minced in 0.05% trypsin (GIBCO) followed by incubation at 37C for 10 min. After For co-culture experiments, mouse calvarial osteoblasts were isolated from newborn wt and Rspo3 +/as previously reported 63,64 and seeded in 96-well plates (2.000 cells/well) in complete osteogenic α-MEM containing 100 nM Vitamin D3 and 1 µM prostaglandin E2 (Enzo Life Science). After 3 days, 10,000 BMM from wt and Rspo3 +/mice at 6 to 8-week-old mice were added per well and cocultured for 9 days in complete osteogenic α-MEM. Tartrate-resistant acid phosphatase (TRAP) staining was performed to evaluate the number of osteoclasts according to the manufacture's protocol (Sigma-Aldrich).
Immunoreactivities were assessed using ECL plus kit following the manufacture's protocol (Perkin Elmer). Quantification was performed using Image J. Protein levels were normalized to the levels of housekeeping protein or total protein in within the same sample.

Immunohistochemistry
Paraffin embedded decalcified tibiae were fixed in 10% NBF. Five μm thick longitudinal sections were incubated with antibodies specific to active -catenin (CST8814, Cell Signaling) at 4°C overnight, treated with TSA-biotin and streptavidin-HRP (Perkin Elmer), as per manufacturer's instructions. Analyses were performed on bones from n=3-4 mice/genotype.

Quantitative-Real Time PCR
Total RNA was isolated from cells using the RNeasy Mini Kit (Qiagen) according to the manufacturer's protocols. Total RNA from cortical bone of wt and Rspo3 +/mice was extracted using Trizol reagent (Invitrogen) followed by RNeasy Mini Kit (Qiagen) according to the manufacturer's protocols. cDNA was synthesized using iScript cDNA synthesis kit (BIO-RAD) and quantitative real time PCR performed. mRNA levels encoding each gene of interest were normalized for 2M or actin mRNA in the same sample and the relative expression of the genes of interest was determined using the formula of Livak and Schmittgen 65 . Data are presented as fold change relative to wt cells or animals.

Statistical Analysis
Data are expressed as the mean ± SEM or ± SD. Statistical analysis was conducted using unpaired two-tail Student's t-test, or two-way ANOVA followed by post-hoc test. The difference was considered as significant at p<0.05.      vs Wnt3a-treated same genotype by Student-t test.