The canonical Wnt pathway (or Wnt/β-catenin) mediates signaling through the stabilization of β-catenin in the cytoplasm upon activation. In the presence of Wnt glycoproteins, occur the binding to the complex frizzled receptor to one of the co-receptors low density lipoprotein receptor-related protein (LRP), LRP5 or LRP6, causing the phosphorylation of the Dishevelled (Dsh) protein, which inhibits glycogen synthase kinase-3beta (GSK3β) from phosphorylating β-catenin in the cytoplasm. This causes the accumulation of dephosphorylated β-catenin, which, upon reaching a certain concentration level, translocates into the nucleus (Figure 1). In the nucleus, β-catenin acts as a transcriptional coactivator with lymphoid-enhancer binding factor (LEF)/T-cell specific transcription factor (TCF) to activate the expression of Wnt target genes, such as CYCLIN D1, AXIS inhibition protein 2 (Axin2), c-Myc and peroxisome proliferator-activated receptor [68–72]. In osteoblasts, the activation of the canonical Wnt also activates the expression of osteoprotegerin (OPG, encoded by TNFRSF11B gene), an inhibitor of receptor activator of nuclear factor kappa beta (RANK) ligand (RANKL) [73]. OPG inhibits osteoclastogenesis by blocking the interaction between RANK-RANKL, thus preventing bone resorption [74, 75]. On the other side, in the absence of a Wnt ligand, β-catenin is constantly degraded by the action of the tetramer degradation complex composed of GSK3β, adenomatous poliposis coli (APC), and Axin, which tags β-catenin for ubiquitination and degradation by the proteasomal machinery (Figure 1) [68, 69, 72].

Given the importance of the Wnt pathway, in different cellular processes, several regulatory systems are involved in its control. Based on their sites of action, the regulators may be divided into extracellular or intracellular antagonists. Most extracellular antagonists are secreted proteins that can interact with either Wnt ligands or Wnt receptors. Examples of extracellular antagonists are DKK1, SOST, WIF-1/2, and SFRPs. DKK1 binds to LRP and the transmembrane protein Kremen 1/2, leading to the internalization and degradation of LRP, thus decreasing the amount of LRP for the binding of the Wnt ligand [76, 77]. Similarly, SOST binds to the co-receptor LRP competing with the binding Wnt protein [78, 79]. However, the WIF-1/2 and SFRPs directly bind and neutralize Wnt proteins, preventing them from binding to the frizzled and LRP complex (Figure 1) [80, 81]. At the intracellular level, the most predominant inhibitors of the Wnt pathway are the β-catenin phosphorylation complex composed of GSK-3β, Axin, and APC. It is believed that APC membrane recruitment protein 1 (AMER1), also called WTX is also involved in the β-catenin destruction complex, however, the exact molecular function is under debate [82]. Finally, at the nucleus, the β-catenin/TCF complex could also be a subject of inhibition [83].

In the “non-canonical” Wnt pathway, the signaling response to Wnt proteins is independent of β-catenin and LRPs. The Wnt PCP pathway uses the receptor tyrosine kinase-like orphan receptor 2 (ROR2), neurotrophin receptor homolog 1 (NRH1), receptor like tyrosine kinase (RYK), or tyrosine-protein kinase-like 7 (PTK7) as co-receptor, instead of LRP5/6 (Figure 2). The binding of the non-canonical Wnt protein to their frizzled receptor acts directly on Dsh to form a complex with Dsh associated activator of morphogenesis 1 (DAAM1), which activates the rhodopsin (Rho, the small G protein) and Rac family small GTPase 1 (RAC1). Rho activates the Rho associated protein kinase (ROCK) involved in the modulation of the cytoskeleton and RAC1 stimulates c-Jun N-terminal kinase (JNK) to activate target gene expression and remodel cytoskeletal actin (Figure 2) [68, 84, 85].

The non-canonical Wnt/Ca²⁺ pathway is the most complex of the three Wnt pathways. This pathway regulates intracellular Ca levels, by releasing intracellular Ca by the endoplasmic reticulum and consequent activation of Ca-sensitive enzymes, which causes the activation of transcription factors. More specifically, the non-canonical Wnt protein binding to their frizzled receptor, and acting directly on Dsh and trimeric Gα/Gβ/GƳ, leads to the activation of PLC and PDE. PLC, by PIP2 cleavage, generates IP3 and DAG. IP3 binds to its receptor in the endoplasmic reticulum and releases Ca from intracellular stores. Ca release activates several Ca-sensitive enzymes, such as CamKII, CaCN, and PKC, and subsequently the activation of cAMP response element binding protein, NF-κB, and NAFT (Figure 2). Activated NAFT translocates into the nucleus to regulate the expression of target genes involved in cell adhesion, migration, and other functional aspects. On the other hand, DAG activates PKC and subsequently MAPKs involved in a variety of physiological processes [68, 72, 84, 86, 87].

The association of the Wnt pathway with bone biology was found through the identification of variants in genes related to the pathway, causing bone diseases (Table 1) [88–98]. Also, several studies using mouse models confirm the importance of the Wnt pathway in bone, in which its activation or inhibition leads to increased or decreased bone formation, respectively [99]. The first evidence of the association of the Wnt pathway with osteoblastogenesis came from the identification of gain or loss of function variants in the LRP5 gene causing high or low bone density, respectively [88, 100]. The gain of function variants in the LRP5 gene affects the ligation of the inhibitors DKK1 and SOST to LRP5 decreasing the inhibition of the Wnt signaling pathway, leading to bone formation [101]. Contrariwise, loss of function variants in the LRP5 gene leads to osteoporosis Pseudoglioma syndrome [online mendelian inheritance in man (OMIM)#259770] with an excessively low bone mass and an excessive risk of developing skeletal fractures [88].

Another player of the Wnt pathway is LRP4 which, unlike LRP5/6, is involved in the inhibition of Wnt through interaction with SOST protein. Variants in the LRP4 gene may cause sclerosteosis type 2 (OMIM#614305), a severe disease characterized by progressive bone overgrowth (Table 1). Due to the compromised binding of LRP4 and SOST in this disease, serum levels of SOST are elevated as well as the activity of the Wnt pathway in osteoblasts and consequently increased bone formation [97].

Several evidences show that Wnt signaling is involved in various diseases characterized by ectopic ossification. Niu et al. [104] found that higher levels of OCN and SOST and lower levels of DKK1 are correlated to the ectopic ossification observed in OPLL (OMIM%602475), AS (OMIM#106300), DISH (OMIM#106400), and OYL diseases. According to the authors, the high levels of SOST are counterbalanced by underproduction of DKK1. An interesting study using a mice model shows that DKK1 blockade stimulates the expression of collagen type X, induces the formation of hypertrophic chondrocytes and promotes ankyloses of sacroiliac joints. In these animals was a subsequent enhancement of the Wnt pathway with new bone formation and ankyloses, suggesting that Wnt signaling has a role in new ectopic bone formation in the axial skeleton [109]. To support this hypothesis, SOST and DKK1 gene expression were found to be downregulated in the spine of a mouse model of AS [110]. Another study investigated a positive relationship between DKK1 inhibitor and excessive ossification in DISH and the authors showed that lower levels of DKK1 were associated with more severe spinal involvement [111]. Diarra et al. [112] reported that low expression of DKK1 leads to bone formation and high levels of bone loss. According to the authors, DKK1 inhibitor plays an important function in bone remodeling. According to previous studies, DKK1 inhibitors play important roles in the control of the Wnt pathway; however, variants in the DKK1 gene are not described in the literature as associated with monogenetic bone disorders [101]. Several evidences point to the fact that the alterations in DKK1 levels, found in some skeletal diseases, are not genetically determined, but probably caused by posttranscriptional changes [113]. In fact, one study performed by Daoussis et al. [114], found that functional levels of DKK1 are decreased in AS patients, suggesting that DKK1 in these patients is dysfunctional and may contribute to ectopic bone formation. The SOST gene is a crucial inhibitor of the Wnt pathway, and its production in bone is performed by mature osteocytes. Loss of function variants in the SOST gene are found in patients with sclerosteosis type 1 (OMIM#269500), a severe disease with progressive skeletal overgrowth [94]. In addition, deletion of 52 kilobase in the same gene results in reduced transcription of SOST [115], responsible for causing a very similar and related disease, the van Buchem disease (OMIM#239100) [96]. Both diseases, sclerosteosis, and van Buchem disease, result from osteoblasts hyperactivity and both show autosomal recessive inheritance [116]. In both diseases, in order to compensate for the loss of SOST, an upregulated of DKK1 expression is verified, which is not sufficient to prevent enhanced bone formation, but, it may prevent ectopic calcification [117–119]. Another disease caused by SOST variants is craniodiaphyseal dysplasia (OMIM#122860), a rare autosomal dominant inherited disease, marked by generalized hyperostosis and sclerosis. Kim et al. [105] found in two children with craniodiaphyseal dysplasia two severe variants affecting the secretion signal of the SOST gene. Variants in the SFRP4 gene, another inhibitor of the Wnt pathway, have also been shown to be disease-causing in Pyle’s disease (OMIM#265900), a metaphyseal dysplasia with autosomal recessive inheritance pattern [106–108]. As already mentioned, SFRP4 binds to Wnt ligands and inhibits Wnt signaling activation. Kiper et al. [108], showed that the knockout mice of SFRP4 manifest the same phenotype observed in Pyle’s disease. Osteopathia striata (OMIM#300373) is another disease caused by inactivating mutations in another mediator of the Wnt pathway, the AMER1 or WTX gene, an intracellular inhibitor of Wnt. The disease is X-linked and is virtually lethal in male patients [91].

Several studies have investigated the role of epigenetics in modulating Wnt signaling, suggesting that miRNAs have a role in osteoblast differentiation and bone formation. miRNA are non-coding small RNA molecules responsible for regulating post-transcriptionally the expression of several genes. miR-29a seems to activate bone formation by enhancing Wnt signaling [120], maybe by downregulating the expression of the DKK1 gene [121]. Similarly, Zhang et al. [122] found that the DKK1 gene was downregulated via a 3’ untranslated region (3’ UTR) by the miR-335-5p. As a result of decreased DKK1 protein level, by miR-335-5p, the Wnt pathway is enhanced and promotes osteogenic differentiation and bone formation. In the AS mice model, the miR-96 is increased and seems to promote osteoblast differentiation and bone formation, via Wnt signaling activation, through downregulation of the SOST gene [123]. Another miRNA that interacts with Wnt signaling is miR-27, which inhibits the expression of APC, a player in the degradation complex of β-catenin. This inhibition favors the activation of Wnt signaling and bone formation [124]. In addition, the non-canonical Wnt through Rho could be able to mediate osteogenic differentiation induced by mechanical loading in tendon-derived stem cells, which might contribute to tendon ossification [125].

DNA methylation is also an important epigenetic mechanism that plays a role in the regulation of bone formation. This mechanism is characterized by the introduction of a heritable epigenetic mark involving a reversible modification of cytosines in the context of CpG dinucleotides by DNA methyltransferases [126]. It was shown that DNA methylation modulates SOST gene expression, a potent inhibitor of the Wnt pathway. A decreased methylation of the SOST gene promoter is verified during the differentiation of osteoblasts to osteocytes, thus allowing osteocyte-specific expression of SOST [127]. Another evidence of the involvement of DNA methylation in regulating bone formation comes from a study with osteoporosis patients which showed increased methylation of the SOST gene promoter accompanied by a decreased level of serum SOST [128].

The HH signaling pathway includes HH ligands, the patched (PTCH) receptor, the signal sensor smoothened (SMO) receptor, the suppressor of fused (SUFU), and transcription factors glioma-associated oncogenes homologs (GLIs). The HH signaling pathway can be in active or inactive form, and its control is mediated by the presence or absence of HH ligands, respectively. Normally, the responsive transcription in HH signaling is performed downstream by GLI-mediated transcription factors (GLI1, GLI2, and GLI3), which act as terminal transcription factors of the HH signaling pathway. GLI1 and GLI2 are transcriptional activators of the HH pathway and GLI3 acts as a repressor [89]. GLI1 is a critical transcriptional activator that acts on the HH pathway and is one of the HH target genes [89, 90]. Specifically, in the presence of HH ligands, such as SHH, DHH, or IHH, occurs the binding to the PTCH transmembrane receptor and consequently, the SMO inhibition by PTCH is abolished [132]. Active SMO is released, and in the primary cilium, recruits the GLI/SUFU/Kinesin family member 7 (KIF7) complex resulting in the detachment of GLI from SUFU and KIF7. The detachment of active GLI permits its entrance to the nucleus to promote the transcription of HH target genes, crucial to cellular differentiation, migration, and proliferation (Figure 3) [133–135]. One of the main target genes in HH signaling is PTH-related protein, which is essential to promote the proliferation of chondrocytes during endochondral ossification [136]. On the other hand, in the absence of HH ligands, the PTCH represses the activity of SMO, which is located in the intracellular endosome. The GLI/SUFU/KIF7 complex recruits kinases, such as casein kinase I (CKI), PKA, and GSK3, leading to GLI phosphorylation and consequently promoting the processing of the repressor form of GLI (GLIR). GLIR translocates to the nucleus to inhibit the transcription of HH target genes, and in this case, the pathway remains in its inactive form (Figure 3) [134, 137, 138].

HH pathway is essential in cartilage and bone growth, and its disturbance has been confirmed in several bone disorders, such as osteoarthritis, heterotopic ossification, bone tumors, and others [139]. Specifically, the local aberrant activation of HH signaling participates in the pathophysiology of heterotopic ossification [140]. In a study using a mice model of osteoarthritis, the authors demonstrated that the increased HH signaling is positively related to the severity of osteoarthritis. Importantly, pharmacological inhibition of HH signaling in this model led to attenuation of osteophyte formation and osteoarthritis severity. In human osteoarthritic samples, the authors also showed an enhanced activation of HH signaling [141]. Another evidence of the importance of the HH signaling pathway in pathological bone formation comes from a mouse model of inflammatory arthritis. In these mice, the authors used a blockage of SMO, a key component of HH signaling, and found a reduction of osteophyte formation in arthritis compared with mice without blockage of SMO. In addition, the authors found that the SMO blockage did not affect inflammation, showing that HH signaling is responsible for new bone formation in the context of inflammatory arthritis [142]. One more evidence, showing the importance of the HH signaling pathway in new bone formation, comes from a mouse model genetically modified to have increased HH signaling pathway activity at spinal chondrocytes. These mice developed a fusion of their spine (ankylosing spine) without any signs of inflammation, showing once again that new bone formation and inflammation are two independent processes [143]. Daoussis et al. [144], reported that patients with AS have higher serum levels of IHH secreted factor compared with healthy controls, suggesting that IHH may be involved in new bone formation in these patients. Also, in human cervical OPLL, the overexpression of HH signaling (IHH) promotes atypical chondrocyte differentiation and enhances new bone formation [145].

In progressive osseous heteroplasia (POH; OMIM#166350), a human disease characterized by extensive heterotopic ossification in soft tissues (Table 2), the HH pathway is also dysregulated by loss of function variants in the guanine nucleotide binding protein (G Protein) gene (GNAS). GNAS encodes the alpha-subunit of the stimulatory G protein (Gα_s), an activator of kinase PKA, which is an inhibitor of HH signaling [146–148]. Regard et al. [148] in a study using a GNAS^-/- mice model showed that HH signaling was up-regulated in ectopic osteoblasts and progenitor cells, while Wnt-β-catenin was decreased [148]. However, gain of function variants in GNAS upregulated Wnt-β-catenin signaling in fibrous dysplasia (OMIM#174800), showing that Gα_s is a critical regulator of osteoblasts differentiation by maintaining the balance between the Wnt-β-catenin and HH signaling pathways [149]. In normal soft tissues, Gα_s is the main responsible for maintaining HH signaling suppressed to prevent bone formation [150].

Another bone disease associated with the HH signaling pathway is Gorlin syndrome (OMIM#109400), characterized by multiple tumor formation and skeletal abnormalities, including ectopic ossification, polydactyly, and rib anomalies. The disease is caused by variants in the PTCH1 gene, which encodes the transmembrane protein PTCH, directly involved in the inhibition of the HH signaling pathway (Table 2) [152–154]. Ohba et al. [155], demonstrated that deficient PTCH^+/- mice showed an accelerated osteoblast differentiation with consequently high bone mass due to an increased bone formation. Additionally, the authors verified a reduction of the repressor transcription factor GLI3 in osteoblast precursor cells.

As verified in the Wnt signaling pathway, a variety of miRNAs could also be involved in the osteogenesis process mediated by HH signaling. In one study using human umbilical mesenchymal stem cells, the miR-342-3 was reported to down-regulate SUFU to activate the SHH pathway and consequently accelerate osteogenic differentiation [156]. It is also reported that miR-342-3 overexpression, by activating the HH signaling, upregulates the expression levels of ALP, RUNX2, and OPG, all of them important in the osteogenic process [157]. Otherwise, the miR-467g seems to be an inhibitor for osteoblast differentiation that could downregulate the osteogenesis progression by targeting IHH/RUNX2 signaling. The authors also found that miR-467g silencing led to a significant increase in new bone formation [158].

TGFβ secreted proteins are synthesized as a latent form and are composed of mature TGFβ and latency-associated protein (LAP) [162]. The latent TGFβ ligand is stored in the ECM and its activation depends on osteoclast bone resorption [160]. At the cell surface, the secreted active ligand binds to the complex of two signaling receptors (type II and I) and transduce signals to both the canonical mothers against decapentaplegic homolog (SMAD)-dependent signaling pathway and the non-canonical SMAD independent signaling pathway (Figure 4). In human cells, at least five type II and seven type I receptors can be expressed and generate receptor complexes that function with all the ligands involved in the TGFβ family [163]. Specifically, in the canonical SMAD-dependent signaling pathway, following a TGFβ binding (or other family members such as activins and nodal), the type II receptor phosphorylates and activates the type I receptor, which in turn activates both SMAD2 and SMAD3 proteins by phosphorylation events. In the case of BMPs, three other SMAD proteins are activated, SMAD1, SMAD5, and SMAD8 [164]. The group of SMAD proteins, which are activated by a TGFβ receptor, are usually designated by receptor activated SMADs (R-SMADs). Activated R-SMADs dissociates from the receptor type I and form a trimeric complex with the mediator SMAD4 in cytoplasm. The trimeric complex enters into the nucleus, where it associates with transcription factors and chromatin modifiers proteins, to regulate the transcription of TGFβ/BMP target genes (Figure 4). BMP signaling mediated by SMADS regulates the expression of transcription factors (RUNX2, osterix, Msh homeobox 2, distal-less homeobox5/6, and SOX9) [165], which are essential for chondrogenesis and osteoblastogenesis [166]. Several mediators are implicated in the regulation of TGFβ/BMP signaling. The SMAD6 and SMAD7 proteins are intracellular mediators with the ability to prevent R-SMAD activation and consequently the SMAD signaling propagation. SMAD6/7 recruits the specific E3 ubiquitin ligases 1 (SMURF1) and SMURF2 to the type I receptor, promoting its ubiquitin/proteasome degradation [167–170]. Into the nucleus, SMAD7 could also inhibit the transcriptional activity of the SMAD complex, by binding on the regulatory sequences of TGFβ target genes [161]. TGFβ signaling could also be inhibited or modulated by BMP and activin membrane-bound inhibitor (BAMBI) which is a transmembrane pseudo-receptor that resembles the TGFβ type I receptors, but with an absent intracellular kinase domain, which is required for signaling activation. Specifically, BAMBI can associate with TGFβ family receptors blocking its function by preventing phosphorylation of the receptor and consequent inhibition of BMP and TGFβ ligands [171]. BAMBI can also form a complex to the intracellular inhibitor SMAD7 to inhibit the interaction between type I receptor and R-SMADs blocking the activation of the SMAD signal transduction pathway [172].

Since TGFβ and BMP have important roles in cartilage and bone homeostasis, dysregulation in these signaling pathways is frequently associated with multiple diseases. A number of human genetic variants in genes related to the TGFβ/BMP signaling pathway have been identified in several inheritable bone diseases [173–179]. Gain of function variants in the TGFβ1 gene, especially located in the LPA region of TGFβ1, are causative of Camurati-Engelmann disease (CED; OMIM#131300), an autosomal dominant disease characterized by generalized hyperostosis affecting the skull and the long bones of the arms and legs [180, 181]. Normally, the non-covalent binding of LAP to the active TGFβ1 is important to keep TGFβ1 inactive in the bone matrix. However, in this disease, the release of active TGFβ1 from the bone matrix is increased. Importantly, it was found that after bone resorption, the TGFβ1 is essential for the migration and proliferation of osteogenic cells to the bone region. This shows that due to CED variants, the release of active TGFβ1 from the bone matrix results in disturbed bone remodeling with increased bone turnover [2]. Since the importance of the TGFβ pathway in regulating bone metabolism, several players of this pathway have been extensively studied in several other ectopic bone diseases, such as OPLL. It was verified, in vivo, that TGFβ is present in the ossified matrix and chondrogenic areas of the posterior longitudinal ligament, but it is absent in the non-ossified ligament [179, 182]. The single nucleotide polymorphism (SNP) with the reference SNP identification rs1982073, located in exon 1 of the TGFβ1 gene, is considered a risk factor for genetic susceptibility to OPLL [183]. However, one study found no association between this variant and the occurrence of OPLL [184], and another study found that the C allele seems to be related to the area of the ossified lesion [185]. TGFβ3 gene is another gene presenting two intronic variants (rs226862 and rs22847) with significant association with OPLL disease [184]. In the same way, variants in TGFβ receptors seem to be associated with OPLL disease. Three variants in the TGFBR2 gene seem to be associated with OPLL; however, the exact mechanism is unknown [186]. Variants in genes involved in the inhibition of the TGFβ pathway were also found. LEM domain containing 3 (LEMD3, also called MAN1) gene encodes a nuclear membrane protein with the ability to bind to SMAD proteins and consequently promote inhibition of the TGFβ signaling pathway [1, 9]. Loss of function variants in LEMD3 is the cause of osteopoikilosis (OMIM#166700), an autosomal dominant inherited disease, characterized by increased bone density with sclerotic lesions in one or several bones [187]. In some cases, other tissues, such as skin and muscles, are also affected [1, 9]. One study performed, based on linkage analysis, tried to find an association between LEMD3 variants with both DISH and chondrocalcinosis, but none of the variants were segregated within the studied families [188].

BMP growth factors stimulate chondrocyte and osteoblast differentiation, resulting in cartilage and bone formation [189]. Variants in the BMP2 gene could induce thoracic ossification of the ligamentum flavum in Han Chinese populations. Subsequently, the same authors showed, by functional analysis, that the BMP2 variants upregulate the BMP2 gene expression and promote osteogenic differentiation [175]. Also, a significant expression of BMP ligands and their receptors in ossified and chondrogenic regions has been noted in OPLL and thoracic ossification of the ligament flavum [178, 179].

In fibrodysplasia ossificans progressiva (FOP; OMIM#135100), the gain of function variants in the activin receptor A type 1 (ACVR1)/activin receptor-like kinase-2 (ALK2) gene, which encodes a BMP receptor type I, designated as ACVR1, causes dysregulation of BMP signaling with resulting ectopic endochondral ossification (Table 3) [177]. In FOP disease, the BMP signaling is dysregulated, keeping BMP signaling activated when it should be inactivated [176].

Several evidences show that epigenetic modifications could also control TGFβ and BMP signaling pathways. For example, miR-133 targets RUNX2 and SMAD5 genes to inhibit osteogenesis induced by BMP2 [191]. RUNX2 is one of the main early BMP response genes essential for bone formation. Also, miR-30 family members target RUNX2 and SMAD1 genes to negatively regulate osteoblast differentiation induced by BMP2. Interestingly, the overexpression of miR-30 leads to a decrease in ALP activity, and the opposite increases its activity, showing that miR-30 could be an important regulator during the biomineralization process [192]. The binding of miR-542-3p to the BMP7 gene inhibits bone formation by suppressing osteoblast proliferation and differentiation induced by BMP7. However, the inhibition of miR-542-3p seems to promote the expression of osteoblast-specific genes and increase the activity of ALP and matrix mineralization [193]. On the contrary, the miR-20a is involved in promoting osteogenic differentiation by up-regulating BMP and RUNX2. The miR-20a targets negative regulators of BMP signaling, such as BAMBI, peroxisome proliferator activated receptor gamma (PPARγ), and cysteine rich transmembrane BMP regulator 1 (CRIM1) [194].