Bone biology: insights from osteogenesis imperfecta and related rare fragility syndromes

Abstract

The limited accessibility of bone and its mineralized nature have restricted deep investigation of its biology. Recent breakthroughs in identification of mutant proteins affecting bone tissue homeostasis in rare skeletal diseases have revealed novel pathways involved in skeletal development and maintenance. The characterization of new dominant, recessive and X-linked forms of the rare brittle bone disease osteogenesis imperfecta (OI) and other OI-related bone fragility disorders was a key player in this advance. The development of in vitro models for these diseases along with the generation and characterization of murine and zebrafish models contributed to dissecting previously unknown pathways. Here, we describe the most recent advances in the understanding of processes involved in abnormal bone mineralization, collagen processing and osteoblast function, as illustrated by the characterization of new causative genes for OI and OI-related fragility syndromes. The coordinated role of the integral membrane protein BRIL and the secreted protein PEDF in modulating bone mineralization as well as the function and cross–talk of the collagen specific chaperones HSP47 and FKBP65 in collagen processing and secretion are discussed. We address the significance of WNT ligand, the importance of maintaining endoplasmic reticulum membrane potential and of regulating intramembrane proteolysis in osteoblast homeostasis. Moreover, we also examine the relevance of the cytoskeletal protein Plastin-3 and of the nucleotidyltransferase FAM46A. Thanks to these advances, new targets for the development of novel therapies for currently incurable rare bone diseases have been and, likely, will be identified, supporting the important role of basic science for translational approaches.

Graphical Abstract

Recent breakthroughs in the identification of mutant proteins affecting bone tissue homeostasis in rare skeletal diseases have revealed novel pathways involved in skeletal development and maintenance. The characterization of new forms of the rare brittle bone disease osteogenesis imperfecta (OI) and other related bone fragility disorders was a key player in this advance and will likely allow the identification of new targets to develop novel therapies for currently incurable diseases.

Introduction

In the past two decades, the identification of causative mutations for multiple rare diseases, favored by the advance in sequencing techniques, proved to be a very powerful tool to shed light on normal metabolic processes in many organs and tissues [1]. The complexity of bone biology, whose study is made challenging by limited access to samples, and by its mineralized nature which complicates cellular investigation, has been successfully addressed by the discovery of mutations in different proteins in rare skeletal diseases [2]. The generation of animal models reproducing the human disorders and the possibility to investigate mutation consequences at the cell level facilitated recognition of novel fundamental mechanisms responsible for skeletal homeostasis.

The identification of mutant proteins affecting bone tissue in various ways made possible the use of in vitro and in vivo models to better define pathways not yet known to be involved in skeletal development and maintenance. For example, the characterization of new dominant, recessive and X-link forms of the brittle bone disease osteogenesis imperfecta (OI) represented a milestone for understanding bone biology [3].

Independent of their severity, OI individuals are generally characterized by growth delay, bone fragility and frequent fractures often not associated with trauma, whilst extraskeletal features such as blue sclerae, dentinogenesis imperfecta, hearing impairment or joint laxity have variable expression [4]. OI is traditionally associated with dominant mutations in the genes encoding type I collagen; about 85% of individuals carry mutations in COL1A1 and COL1A2 [5]. Beginning in 2006, recessive mutations in cartilage associated protein (CRTAP), prolyl-3-hydroxylase (P3H1) and peptidyl-prolyl cis-trans isomerase B (PPIB) genes, responsible for autosomal recessive OI, unveiled the existence of a then unknown endoplasmic reticulum (ER) resident complex. The three proteins are required for the unusual 3-hydroxylation of specific proline residues in the procollagen α chains, versus standard 4-hydroxylation of multiple helical prolines, whose known relevance for collagen structure has been confirmed by the identification of prolyl-4 hydroxylase-B (P4HB) mutations in Cole Carpenter syndrome [6] and more recently in OI [7, 8]. These findings opened a new research area and shed light on the role of proline hydroxylation both intracellularly and in the bone extracellular matrix (ECM) [9–17]. The expected importance of the procollagen maturation process was supported by the identification of the severe skeletal phenotype caused by homozygous mutations in the C-propeptidase (Bone Morphogenic Protein 1, BMP1) responsible for the removal of the C-propeptide from the molecule [18], as well as of the unique high bone mass form of OI caused by dominant substitutions in the procollagen C-propeptidase cleavage site [19, 20]. The identification of recessive mutations in the osteoblast-specific transcription factor osterix/SP7 in individuals with OI confirmed its essential function in bone formation and homeostasis as a regulator of osteoblast differentiation and maturation [21, 22]. The pathways associated with the above-mentioned OI mutations have been extensively reviewed elsewhere [3, 4, 23–25] and allowed to broaden the OI definition from a dominant heritable disease caused by mutations in collagen type I coding genes, to a collagen type I-related group of heritable disorders characterized by impairment in collagen amount, synthesis, post translational modification, secretion and/or extracellular processing and altered osteoblast differentiation and function, often associated with high bone turnover and hypermineralization.

The identification of the new causative genes involved in specific biochemical pathways suggested to extend the longstanding Sillence OI classification, in which a successive number is attributed to each newly identified gene, resulting in a more functional OI classification system [3]. We will use the classification reported in Table 1 whenever referring to specific types of osteogenesis imperfecta.

Table 1:

Bone molecular mechanisms identified following the characterization of new causative genes in OI forms or related bone fragility diseases.

Protein	Acronym	Disease	Transmission	MIM	Role in bone biology
Compromised bone mineralization
Bone restricted ifitm-like protein	BRIL	OI type V	AD	610967	-To enhance mineralization acting on osteoblast specific gene expression
Pigment epithelium-derived factor	PEDF	OI type VI	AR	613982	-To enhance osteoblast differentiation and mineralization favoring osteogenic gene expression -To inhibit osteoclast maturation -To regulate the expression of VEGF known to promote osteoblasts differentiation and matrix mineralization -To activate in osteocytes a WNT independent β-catenin pathway
Compromised collagen processing and crosslinking
Heat shock protein 47	HSP47	OI type X	AR	613848	-To chaperone procollagen -To act as effector of the Ire1α branch of the Unfolded Protein Response
FK506 binding protein 65	FKBP10	OI type XI	AR	610968	-To favor telopeptide collagen lysyl hydroxylation
Altered osteoblast differentiation and function
Trimeric intracellular cation channel B	TRIC-B	OI type XIV	AR	615066	-To regulate Ca2+-dependent cell signaling and enzyme functions in osteoblasts and osteoclasts
Wingless-type MMTV integration site family 1	WNT1	OI type XV	AR	615220	-To stimulate β-catenin stabilization and to favor its nuclear translocation involved in gene transcription
Old astrocyte specifically induced substance	OASIS	OI type XVI	AR	616215	-To stimulate type I collagen transcription -To regulate ER-Golgi trafficking
SPARC (osteonectin)	SPARC	OI type XVII	AR	616507	-To regulate procollagen processing and assembly in the ECM -To modulate osteoid mineralization
S1 protease	S1P	osteochondrodysplasia	AR	603355	-to regulate skeletal progenitor populations for bone and cartilage -to regulate Golgi-lysosomal trafficking
S2 protease	S2P	OI type XVIII	XR	300294	-To regulate collagen transcription and secretion
OI-related bone fragility disorders
Plastin 3	PLS3	Childhood-onset primary osteoporosis	AR	300910	-To bind and stabilize actin bundles -To regulate osteoclasts activity -Unknown role in matrix vesicles
Terminal nucleotidyltransferase 46, Member A	FAM46A	Stuve-Wiedermann syndrome	AR	601559	-To modulate BMP signaling

In this review, we selected the OI forms and OI-related bone fragility diseases whose causative mutations have been identified in the last decade and whose investigation in recent years provided new insight into molecular mechanisms relevant for bone homeostasis.

Insight from the most rare OI forms

BRIL and PEDF: modulators of bone mineralization

The skeletal role of the bone restricted ifitm-like protein (BRIL), encoded by the interferon-induced transmembrane protein 5 (IFITM5) gene, and of the pigment epithelium-derived factor (PEDF), encoded by serpin family F member 1 (SERPINF1), was clarified following the identification of mutations in these genes in the rare forms of OI, type V and VI, respectively.

Dominantly inherited OI type V and recessively inherited type VI, both having hypermineralized bone, a feature that distinguishes OI from rickets and osteomalacia [26, 27], are characterized by distinctive defects in bone mineralization and lamellar organization in the presence of normal type I collagen post-translational modifications [28, 29]. OI type V, the only dominant form caused by mutations in a non-collagen type I coding gene, is predominantly caused by a single recurrent heterozygous c.−14C˃T nucleotide change in the 5′-untranslated region (5′UTR) of IFITM5 [30, 31]. This mutation creates a novel in-frame translation start site that adds 5 amino acid (Met-Ala-Leu-Glu-Pro, MALEP) at the cytosolic N-terminus of BRIL, a 132 amino acid palmitoylated type II transmembrane protein (N-in/C-out topology) present in the osteoblasts plasma membrane (Figure 1) [32, 33].

Figure 1: BRIL and PEDF regulation of bone mineralization.

The bone restricted ifitm-like protein (BRIL) on osteoblasts membrane regulates the expression of the SERPINF1 gene, encoding for pigment epithelium derived factor (PEDF). In presence of the dominant c.−14C˃T mutation in the 5′-untranslated region of the interferon-induced transmembrane protein 5 (IFITM5) gene, a novel in-frame translation start site is created that adds 5 amino acid (MALEP) at the cytosolic N-terminus of BRIL. MALEP-BRIL increases the expression of PEDF. In presence of the heterozygous p.Ser40Leu substitution in BRIL, palmitoylation is impaired and the mutant protein is retained in the Golgi, being unable to regulate gene expression. Dotted lines refer to the presence of still unknown factors in the pathway.

On osteocyte membrane PEDF binds to PEDF-R inducing phosphorylation of ERK. Active ERK phosphorylates glycogen synthase kinase 3-beta (GSK-3β), targeting it for destruction and leading to the stabilization of β-catenin. β-catenin migrates into the nucleus inducing transcription of the target genes.

Over a hundred OI type V individuals were described worldwide, all carrying the same mutation, but with considerable heterogeneity in their clinical manifestations [34]. In general, they have a moderate form of the disorder, clinically similar to OI type IV in severity. OI type V shares with the other OI types an increased fracture incidence due to low bone mass [35] and impaired bone quality [34]. However, it has distinctive features such as formation of hyperplastic callus during fracture healing, periosteal hyperplastic expansion, calcification of the interosseous membrane in the forearm, subephyseal radiodense line and distinctive mesh-like lamellation of the collagen on bone sections. Neither blue sclerae nor dentinogenesis imperfecta are present [28]. Primary osteoblasts from OI type V bone biopsies show high expression of osteoblastic markers and increased mineralization in the presence of decreased type I collagen expression, secretion, and matrix incorporation [36]. These data are consistent with the hypermineralized bone matrix and high osteocyte lacunar density in immature bone recently described in iliac crest bone samples from affected children [37]. These observations, together with low trabecular bone volume and bone formation rate [28], point to a peculiar characteristic of this OI form, in which exuberant primary bone formation is associated with an osteoporotic phenotype [4].

The normal synthesis, post-translation palmitoylation, and sub-cellular localization of mutant MALEP-BRIL in MC3T3 osteoblasts support a dominant negative effect for the mutation [38], but the molecular mechanism of BRIL in bone still remains unclear. In osteoblast culture BRIL expression level peaks during early mineralization phase [32] and, similarly, in vivo its expression in the skeletal system of murine embryos is predominantly starting from E14.5, when osteoblasts begin to deposit mineralized matrix [39]. In vitro overexpression of Ifitm5 upregulates alkaline phosphatase expression and mineral deposition, whereas its silencing has opposite effects, suggesting a role for BRIL as a positive modulator of mineralization [32].

On the contrary, transgenic mice overexpressing BRIL under Col1a1 promoter do not show any skeletal outcome [40] and surprisingly, skeletal formation is only mildly affected in mice lacking the protein [39, 41], suggesting the existence of compensatory mechanisms in skeletogenesis. Transgenic overexpression of MALEP-BRIL in mice, as well as knock-in of the c.−14C˃T recurrent mutation causing type V OI, result in a lethal phenotype associated with severe bone deformities and abnormal mineralization. The murine outcome is strikingly more severe than in humans, although it confirms the causative effect of the mutation and its dominant transmission [42].

Recently, BRIL has been found to disrupt the interaction between CD9 and FKBP11, important regulators of the expression of interferon-induced genes, revealing a possible connection between type V OI and the immune response [43]. Interestingly, an inflammation response was identified in the knock-in MALEP-BRIL mouse model by whole transcriptome analysis, but further investigations are required [42].

Compromised bone mineralization is also a hallmark among individuals with OI type VI who have homozygous or compound heterozygous mutations in the SERPINF1 gene, which encodes PEDF [44, 45], a secreted collagen binding glycoprotein expressed by a wide range of cells and known to have neurotropic and antiangiogenic properties and a role in cell cycle control, fat metabolism, and tumorigenicity [46, 47]. Although PEDF belongs to the serpin family of serine protease inhibitors, it does not have any protease inhibitory activity.

A variety of recessive mutations in the SERPINF1 gene, including missense, nonsense, splicing, deletions and insertions, resulting in absence or synthesis of non-functional PEDF, have been reported in OI type VI patients with moderate to severe clinical features [44, 45, 48] (osteogenesis imperfecta variant database http://www.le.ac.uk/ge/collagen/). OI type VI individuals show a progressive severe degeneration of bone characterized by reduced bone mineral density, vertebral compressions and long bone fractures, generally beginning after the first year of life. Longitudinal growth stalled after the age of 6 to 8 years in 13 analyzed patients [49]. The distinctive bone characteristic in this OI form is the coexistence of highly mineralized bone matrix with seams of osteoid showing abnormally low mineral content, although overall mineral to matrix content is increased [50]. Notably, atypical collagen fibril organization is detectable in the perilacunar region of young osteocytes, suggesting impairment in early mineralization steps [50]. Bone histology reveals increased osteoid, prolonged mineralization lag time and loss of the normal orientation of the lamellae, also called fish-scale lamellar pattern [44, 45]. The Serpinf1 knock-out mouse shows reduced bone volume and increased un-mineralized osteoid, associated with increased mineral:matrix ratio, as described in human OI type VI [51].

PEDF is absent from the serum of OI VI individuals, suggesting the potential feasibility of protein replacement therapy [48]. However, the systemic restoration of PEDF in the knock-out mice yielded contradictory results. Induction of human PEDF expression in murine liver upon injection of viral particle failed to improve the bone phenotype [52], whereas intraperitoneal delivery of PEDF increased bone mass and partially ameliorated bone parameters [53]. The difference in the ages of the treated mice, as well as amount of exogenous PEDF, could account for the differences, but further studies are needed.

The mechanism by which PEDF regulates bone mineralization is still incompletely understood. PEDF promotes human MSCs differentiation and osteoblast mineralization in vitro, by favoring the expression of osteogenic genes, while it inhibits osteoclast maturation by stimulating osteoprotegerin expression [54], and suppresses the expression of osteocytic proteins with negative role in bone mineralization [55]. In osteocytes, binding of PEDF to its receptor PEDF-R induces phosphorylation of ERK and of glycogen synthase kinase 3-beta (GSK-3β), favoring disassembly of the β-catenin degradation complex and thus promoting its nuclear role in regulating bone-specific gene expression in a WNT-independent mechanism [56] (Figure 1).

Interaction of PEDF with ECM components such as collagen, heparan sulfate proteoglycans, and hyaluronan may account for PEDF specific bone functions [57]. PEDF also regulates expression of vascular endothelial growth factor (VEGF), which can directly promote osteoblast differentiation and matrix mineralization in bone [58]. Several reports indicate a balanced VEGF/PEDF ratio as a key factor in regulating osteoblastic mineralization [59].

The identification of the heterozygous BRIL p.Ser40Leu mutation revealed the intersection of BRIL and PEDF metabolic pathways, although the two proteins have opposite effects on mineralization, with increased ectopic ossification in type V and impaired mineralization in type VI [60]. The BRIL p.Ser40Leu substitution interferes with its palmitoylation and causes Golgi retention of the mutant protein [38]. Interestingly, the individuals carrying this mutation show the typical bone histology pattern described in OI type VI and none of the clinical or radiographic findings peculiar for OI type V. Consistent with the bone histology, fibroblasts and osteoblasts carrying BRIL p.Ser40Leu have impaired PEDF secretion, opposite to the increased SERPINF1 expression and PEDF secretion detected in OI type V individuals with MALEP-BRIL production. These data reveal the role of BRIL in regulating PEDF expression/secretion and further support the gain-of-function mechanism in OI type V c.−14C>T mutation (Figure 1) [60]. Since co-immunoprecipitation failed to demonstrate direct BRIL-PEDF binding, the existence of a more complex regulatory pathway by which BRIL influences SERPINF1 transcripts and PEDF secretion is hypothesized. The connection between BRIL and PEDF functions in bone mineralization requires further elucidation.

HSP47 and FKBP65: chaperones with a role in collagen biosynthesis and secretion

SERPINH1 and FKBP10 are two genes encoding two ER resident chaperones known to have triple helical procollagen molecules as preferred substrates [61, 62]. SERPINH1 encodes the heat shock protein 47 (HSP47), a chaperone modulating its own transcription level coordinately with collagen expression [61, 63]. FKBP10 encodes for FKBP65 a chaperone with prolyl cis-trans isomerase activity (PPIase) crucial for normal collagen synthesis [64].

The involvement of HSP47 in bone health became clear following the identification of SERPINH1 mutations in individuals affected by a severe to lethal form of OI (OI type X) characterized by short stature, osteopenia, long bone deformity, and multiple fractures [65–69]. Extraskeletal features such as blue sclerae, joint hyperlaxity, dentinogenesis imperfecta and hearing loss, although present in some cases, do not represent hallmarks of the phenotype. Due to the limited number of patients reported, a genotype to phenotype relationship is still unclear.

Before the identification of SERPINH1 mutations as causative for human OI, a general Serpinh1 knock-out murine model was generated, supporting an important role for HSP47 in collagen homeostasis and bone development [70]. Impairment in basement membrane and collagen fibril formation results in the embryonic death of mutant mice at E11. Further, Serpinh1^−/− murine embryonic fibroblasts show impaired collagen processing, increased collagen protease sensitivity and reduced collagen secretion, with consequent accumulation of procollagen in the ER and associated enlargement of the cisternae, activation of unfolded protein response (UPR) and autophagy [71–73], features shared with the most severe forms of classical OI [74, 75].

Similar findings are present in canine fibroblasts from dachshunds carrying an HSP47 p.Leu326Pro substitution and with a severe OI phenotype. In mutant dogs, aberrant collagen structure and cross-links were detected in bone tissue [76, 77].

Further support for a role for this chaperone in skeletal homeostasis is found in the impairment of both cartilage and endochondral bone formation associated with conditional ablation of HSP47 in chondrocytes, obtained by mating conditional floxed Serpinf1 mice with Col1a2-Cre expressing mice [78].

Interestingly, patients’ fibroblasts have normal collagen synthesis and post translational modification, although a delay in secretion, increased sensitivity to trypsin digestion and elevated melting temperature suggest an altered procollagen structure [65, 66]. The almost complete instability of HSP47 p.Leu78Pro results in partial procollagen retention in the Golgi apparatus while the HSP47 p.Met237Thr substitution is associated with vesicle-like structures negative for Golgi markers and positive for the ER marker PDI [65, 66].

Procollagen synthesis requires a sequence of events beginning with translation of α chains in the ER, the stoichiometric recognition of two proα1 and one proα2 chains at their C-terminal end and specific post-translational modifications that occur until the trimer is completely folded [79]. The HSP47-procollagen interaction in the ER at the consensus binding sequence (Gly-Xaa-Arg) [80, 81] is essential to stabilize the procollagen triple helical structure, and to prevent local unfolding and lateral aggregation and/or fibril formation [61]. HSP47 in collagen-producing cells chaperones procollagen molecules from the ER to the ER-Golgi intermediate compartment (ERGIC) in a pH dependent manner. HSP47 recycling back to the ER is mediated by the presence of a C-terminal RDEL retention signal [82–84]. Procollagen trafficking from the ER requires larger coat protein II complex (COPII)-vesicles, whose formation at ER exit sites is indeed associated with HSP47 anchorage to the SH3 domain of the ER protein TANGO1 [85] (Figure 2).

Figure 2: HSP47 role in type I collagen triple helix stabilization and secretion.

Heat shock protein 47 (HSP47) is an ER resident chaperone able to specifically bind type I procollagen molecules, preventing local unfolding and lateral aggregation and/or fibril formation. In the ER, HSP47 interacts with the chaperone peptidyl prolyl cis-trans isomerase 65-KDa FK506-binding protein (FKBP65) and the lysyl hydroxylase 2 (LH2). The stability of this complex is regulated by the master effector of the unfolded protein response (UPR) pathway, Bip. Upon binding to HSP47, procollagen trafficking towards Golgi is mediated by the formation of larger COPII vesicles derived by the HSP47 anchorage to the SH3 domain of the ER protein TANGO1. Once reaching the Golgi, HSP47 releases its cargo in a pH dependent manner and it is recycled back to the ER thanks to the presence of a C-terminal RDEL retention signal.

HSP47 interacts also with the Ire1α effector of the UPR, thus modulating the adaptive UPR.

The chaperone properties of HSP47 and the HSP47-TANGO1 interaction provide a possible unifying model for recruitment of other macromolecules for assembly into larger COPII vesicles targeted for Golgi secretion [85]. Notably, fibronectin networks, fibrillin fibril structures, and secretion of small leucine-rich proteins (SLRPs), such as decorin and lumican, were altered in the absence of HSP47 [85, 86]. Indeed, HSP47 interacts with the N-terminal of fibrillin-1, decorin, and lumican. The HSP47-TANGO1 interaction and its role in assembly of larger COPII vesicles in ER-Golgi trafficking are intriguingly in concert with transcriptional regulation of COPII components by the S1P-S2P-CREB3L axis in Regulated Intramembrane Proteolysis (RIP) pathway [87, 88]. Hence, transcriptional regulation and protein-protein interaction provide distinctive layers to control gene expression, cargo selection, and assembly of larger COPII vesicles for Golgi secretion. These overlapping controls allow fine tuning of the formation of a collagen-based extracellular matrix.

Recessive defects in FKBP10 have been identified as causative in OI alone (OI type XI) [89, 90], in OI with contractures (Bruck syndrome) and in Kuskokwim syndrome, a rare congenital contracture disorder with minor skeletal features [91].

In OI type XI, collagen type I structure is normal, although impairment of trimer stability is described [89, 92]. Nevertheless, in absence of FKBP65, procollagen molecules accumulate in aggregates within the cells, causing dilation of the ER and matrix insufficiency [89]. Importantly, extracellular matrix collagen crosslinks are dramatically reduced, similar to reports in individuals with mutations in lysyl hydroxylase 2 (LH2) causing Bruck syndrome, which shares severe joint contractures with OI type XI [93]. This observation supports cross-talk between FKBP65 and LH2, which accords with the recent reports of the FKBP65 role in assisting LH2 dimerization [89].

The Fkbp10 knock out mouse model shared the human findings showing generalized connective tissue defects and a reduction in telopeptide lysine hydroxylation, although perinatal lethality limited the investigation [94]. More recently, the bone specific Fkbp10 conditional knock out provided further insight on the function of FKPB65 in skeletal tissue, demonstrating its role in modulating bone quality, rather than bone quantity. Conditional Fkbp10^−/− have minimal alterations in the amount of bone and no differences in bone matrix mineralization, indicating that osteoblast derived Fkbp10 is dispensable for postnatal bone mass. A decrease in hydroxylysine-aldehyde crosslinking, a reduction in mineral-to-matrix ratio and crystal size abnormalities associated with reduced bone biomechanical strength in Fkbp10^−/− underlie prominent qualitative bone defects [95].

The recent identification of direct binding of HSP47 with FKBP65 suggests a potential role for HSP47 as a central hub during collagen folding and secretion, with the ability to direct other molecules to their target site on collagen. Indeed, FKBP65 binding to HSP47 does not compete with HSP47-collagen interaction; when both proteins are present the collagen folding rate is increased, suggesting a role for HSP47 in favoring FKBP65-collagen interactions [96]. Mutations in either HSP47 or FKBP65 result in distinct forms of recessive osteogenesis imperfecta with overlapping phenotypes, characterized by increased or reduced telopeptide lysine hydroxylation, respectively [3, 77, 92]. Recently, an HSP47-FKBP65-LH2 complex has been identified, the stability of which is regulated by the chaperone Bip, which is well known to favor collagen trimerization and acts as an UPR effector (Figure 2) [97]. Importantly, HSP47 is also an effector of the Ire1α branch of the UPR, able to replace Bip binding upon ER stress thus modulating UPR response (Figure 2) [98].

In summary, HSP47 interacts with collagen, TANGO1, SLRPs, fibrillin-1, FKBP65, LH2, and plausibly, additional macromolecules of the extracellular matrix. Many of the HSP47 interacting proteins contribute to key steps in collagen regulation and function, including helix stability, telopeptide hydroxylation, collagen secretion, and matrix assembly. It is not surprising that mutations in HSP47 lead to skeletal defects like OI, with reduced collagen function and dysregulated matrix organization.

TRIC-B: an ER resident K⁺ channel influencing bone cells homeostasis by modulating Ca²⁺ flux

Mutations in TMEM38B cause the recessive OI type XIV [99, 100] (OMIM 611236). TMEM38B encodes the ubiquitous voltage-dependent cation channel TRIC-B (trimeric intracellular cation channel subtype B) [101]. TRIC-B maintains membrane potential in the ER by importing K⁺ to offset Ca²⁺ release by the inositol-3-phosphate receptor (IP₃R). Thus, TRIC-B deficiency compromises Ca²⁺ release and leads to dysregulated Ca²⁺ kinetics (flux). Since Ca²⁺ is an important second messenger in cell signaling and a key co-factor for many proteins, including enzymes important for multiple collagen post-translational modifications, dysregulated Ca²⁺ kinetics would severely impair an array of cellular processes and lead to skeletal dysfunction. Indeed, Tmem38b null mice die shortly after birth due to respiratory failure with insufficient surfactant secretion, a Ca²⁺-regulated process [102].

TMEM38B was initially delineated as an OI-causative gene among Bedouins with bone fragility [100] and soon after identified in several other pedegrees. Homozygous genomic deletions, splice site mutations and nonsense variants were described [99, 103, 104].

Clinical features of OI type XIV probands had especially wide variation in severity, even in siblings with identical mutations [105]. Their phenotypes ranged from moderate severity, with bowed femora, non-vertebral fractures, cardiac/skeletal muscle defects, blue or white sclera, to clinically normal. Most OI type XIV probands showed reduced bone density.

At the cellular level, both osteoblasts and osteoclasts of OI type XIV individuals were affected by the loss of TRIC-B [105]. Defects in cells with opposing functions may account for the normal bone mineral density of some patients. The decreased osteoblast number found in bone biopsy of probands was corroborated by impaired in vitro osteoblast differentiation and reduced expression of osteoblast differentiation markers. Proband bone tissue also exhibited a low osteoclast number, consistent with expression of TMEM38B, rather than TMEM38A in osteoclasts [105]. Furthermore, Ca²⁺ signaling plays an important role in maturation of osteoclasts [106]. Thus, decreased osteoclast number in the patients is likely an intrinsic defect.

Mechanistically, the loss of TRIC-B lowered cytoplasmic resting state Ca²⁺ level in osteoblasts [99]. Ca²⁺ release from intracellular stores upon activation was also reduced. In addition, PERK-dependent, but not ATF6- or IRE1- mediated, ER stress response was potentiated in proband fibroblasts, by the Ca²⁺ imbalance. Interestingly, however, the steady state Ca²⁺ level in the ER lumen was not altered by the TRIC-B deficiency. Hence, Ca²⁺ flux is impaired in the absence of TRIC-B, resulting in changes in duration and amplitude of Ca²⁺ rise and altered activation kinetic of various downstream Ca²⁺ effectors.

The intracellular Ca²⁺ dynamics, which are mediated, in part, by TRIC-B, also regulate expression of collagen modifying enzymes, such as lysyl hydroxylses, and affect collagen biochemistry (Figure 3) [99]. Faster collagen migration in SDS electrophoresis was the result of reduced collagen hydroxylysine content in OI type XIV proband fibroblasts and osteoblasts. However, expression of lysyl hydroxylase 1 (LH1) protein was increased, possibly due to an attempted cellular compensation in proband cells. Elevated expression of LH2 protein in the OI type XIV fibroblasts, but not osteoblasts, was associated with increased hydroxylation at the telopeptide. Together, dysregulated hydroxylation and misfolded collagen likely contributed to the intracellular accumulation, reduced secretion, and decreased thermal stability of collagen in OI type XIV.

Figure 3: TRIC-B, SP2, OASIS and SPARC function in bone homeostasis.

Collagen is synthesized within the endoplasmic reticulum (ER), the major intracellular Ca²⁺ store. In response to extracellular stimuli, Ca²⁺ is released from the ER lumen into the cytoplasm mainly via inositol-3-phosphate receptor (IP₃R) and it is transported back into the ER via SERCA pumps. The trimeric intracellular cation channel subtype B (TRIC-B) regulates Ca²⁺ fluctuations indirectly mediating transmembrane K⁺ flux to maintain electroneutrality across the ER membrane. In the absence of TRIC-mediated K⁺ flux, ER-resident Ca²⁺-binding chaperones, including BiP, Cyclophilin B (CyPB) and protein disulfide isomerase (PDI), that directly interact with collagen chains for assembly and folding, are dysregulated. Thus, absence of TRIC-B affects synthesis and secretion of collagen.

S2P is a protease in the Golgi membrane involved in the Regulated Intramembrane Proteolysis (RIP) of transcription factors, such as old astrocyte specifically induced substance (OASIS). Following ER stress or sterol deficiency, OASIS is transported from the ER to Golgi membrane for sequential processing by S2P.

SPARC is a calcium-binding protein, one of the most abundant non-collagenous protein expressed in mineralized tissues and relevant for cellular-matrix interaction. In osteoid, SPARC has been proposed to bind collagen and hydroxyapatite crystals and release calcium ions, perhaps enhancing mineralization of the bone collagen matrix. SPARC seems to have also an intracellular chaperone role since dermal fibroblasts isolated from OI individuals with SPARC mutations exhibited delay in collagen folding.

In summary, TRIC-B plays a key role in balancing ER membrane potential upon Ca²⁺ release from intracellular stores. Ca²⁺ kinetics are altered in the absence of TRIC-B and dysregulated Ca²⁺ kinetics severely impair cell signaling and Ca²⁺-dependent enzymes in osteoblasts and osteoclasts, resulting in skeletal dysfunction.

WNT1: a key player in bone homeostasis

Wingless-type MMTV integration site family 1 (WNT1) belongs to a family of 19 secreted glycoproteins, highly conserved throughout species, and involved in the activation of signal transduction pathways modulating in autocrine or paracrine mode cell functions in various tissues, including bone [107, 108]. Upon WNT1 binding to a membrane dual-receptor complex, consisting of low density lipoprotein receptor-related proteins (LRP5 or LRP6) and the seven-transmembrane protein Frizzled, the second messenger β-catenin is released from a cytosolic degradation complex and translocates to the nucleus where it regulates the expression of genes involved in osteoblast differentiation and activity [109].

The role of WNT1 in bone biology was initially uncovered by the identification of WNT1 heterozygous and homozygous mutations in dominant early-onset osteoporosis with a variable phenotype and in moderate to severe cases of recessive OI type XV, respectively [110–113]. This unusual semidominant transmission points to a critical dose-dependent requirement for this ligand in bone homeostasis. Monoallelic missense mutations are associated with low bone mineral density and multiple peripheral and vertebral compression fractures without extraskeletal manifestations. Heterozygous subjects have normal growth without dysmorphic features. In contrast, more detrimental outcomes are associated with homozygous WNT1 missense mutations, splice site substitutions, deletions and duplications, which either change highly conserved residues or compromise WNT1 synthesis or stability. Individuals with homozygous WNT1 mutations manifest severe skeletal abnormalities, reduced bone density and multiple fractures, growth delay and typical dominant OI extraskeletal findings such as blue sclerae and dentinogenesis imperfecta [114]. Developmental defects in the central nervous system (CNS) associated with intellectual disability are also occasionally described and may be linked to WNT1 expression in brain tissue [115, 116]. WNT1 mutants impair the β-catenin transport into the nucleus and the expression of downstream genes, compromising bone cells homeostasis (Figure 4) [110, 111]. Interestingly, normal collagen biochemistry, bone collagen lamellation pattern and bone matrix mineralization are described in individuals with WNT1 mutations, which distinguish them from classical OI caused by COL1A1 and COL1A2 mutations [112, 117]. Furthermore, low bone turnover, pointing to an imbalance between bone forming and resorbing cell activity, is also distinctive in patients carrying WNT1 mutations and opposite to the high bone turnover described in dominant OI [118]. Osteoclast number and/or activity are unchanged in OI type XV individuals, which may explain the limited effect of bisphosphonate treatment compared with results reported in dominant OI [117, 119].

Figure 4: WNT1 and FAM46A regulate osteoblast gene transcription.

Wingless-type MMTV integration site family 1 (WNT1) regulates bone formation in several ways. In the canonical pathway, WNT1 binding to a membrane dual receptor complex LRP5/6-Frizzle favors β-catenin release from a degradation complex and its migration to the nucleus where it regulates the expression of genes involved in osteoblasts differentiation and activity [109].

Recently, a WNT1 stimulatory effect on bone formation, independent from LRP5 and mediated by WNT1 regulation on the mammalian target of rapamycin (mTOR) receptor, has been hypothesized (gray shade). The WNT1/mTORC1 complex activates protein synthesis through the phosphorylation of the ribosomal protein S6. Question mark refer to the presence of still unknown factors in the pathway.

Following bone morphogenetic protein (BMP) association with the BMP receptor (BMPR) on osteoblast membrane, FAM46A binds to the receptor regulated SMAD (R-SMAD), stabilizing it and avoiding its degradation. Then, FAM46A, R-SMAD, and Co-SMAD form a complex that migrates to the nucleus and regulates gene expression activating, among others, the transcription of the BMP target genes [183].

There are two general knock-out mouse models for WNT1. The first was generated by gene targeting a neo cassette within exon 2. Homozygous Wnt1^−/− are delivered at normal Mendelian ratio, but die in the first day of life due to absence of a large portion of the brain. No bone phenotype was noted [120]. A spontaneous recessive mutation in Wnt1 was later identified in the swaying mouse model (Wnt1^sw/sw), originally reported for ataxia [121]. This mouse carries a single nucleotide deletion (c.565delG, p.Glu189Argfs*10) in the third exon of Wnt1, similar to the genomic location of the human nonsense mutation (c.565G>T, p.Glu189*) responsible for OI type XV [111, 121]. Wnt1^sw/sw mice also display a brain phenotype, similar to that described in Wnt1^−/− animals, as well as typical OI features such as severe osteopenia, spontaneous fractures, reduced bone strength and impaired matrix mineralization, thus representing a good model for the human condition. Osteoblast function is impaired in Wnt1^sw/sw, whereas no changes are found in osteoclasts [122].

The site and mechanism of WNT1 action in bone has only recently begun to be elucidated, thanks to elegant studies using a conditional Wnt1 mouse model (Wnt1^fl/fl) in which the ablation of the gene in osteoblasts or in late osteoblasts/osteocytes is achieved using specific Cre-deleter mice, Runx2-Cre and Dmp1-Cre, respectively [123, 124]. The Wnt1^fl/fl Dmp1-Cre and the Wnt1^fl/fl Runx2-Cre phenotypes resemble the Wnt1^sw/sw outcome and the clinical features described in individuals with WNT1 mutations. Both models show multiple fractures and low bone mass, and thus provide direct evidence that WNT1 loss-of-function in osteoblasts is responsible for OI type XV.

Transgenic mice overexpressing Wnt1 in bone (Dmp1-Cre Rosa26^Wnt1/+ and Col1a1-tTA-Wnt1) and characterized by increased bone mass, osteoblast number, mineralizing surface, mineral apposition rate and bone formation rate, further support the anabolic role of WNT1 in bone. Recently, WNT1 modulation of bone formation and mineralization has been hypothesized to be associated with non-canonical mTOR pathway activation, independent of β-catenin (Figure 4) [123]. However, this observation was not confirmed in primary osteoblasts in a subsequent study and further investigations are required [124]. Joeng KS et al showed increased phosphorylation of S6, one of the targets of mTORC1, in stromal ST2 cells overexpressing Wnt1 as well as rescue of the bone phenotype of transgenic mice overexpressing Wnt1 in bone (Dmp1-Cre Rosa26^Wnt1/+) after 1 month of treatment with rapamycin, a well-known inhibitor of the mTORC1. Interestingly, the extent of rescue differs in trabecular and cortical compartments, suggesting that mTORC1 may differentially transduce WNT1 signaling in bone forming cells. Furthermore, activation of mTOR in Wnt^sw/sw mice by conditional ablation in bone of Tsc1, a negative regulator of mTORC1, improved bone properties and reduced bone fractures compared to Wnt^sw/sw [123]. Recently, a WNT1 stimulatory effect on bone formation, independent of LRP5 has also been demonstrated (Figure 4). A role for LRP5 in mediating the bone-anabolic activity of WNT1 is supported by the high bone phenotype described in humans and in murine models carrying Lrp5 gain-of-function mutations. Nevertheless, the overexpression of Wnt1 in Lrp5-deficient mice leads to increased osteoblast number and surface area. The hypothesis that WNT1 acts mainly via LRP6 in bone is appealing, but needs further corroboration. Finally, WNT1 expression acts very quickly on bone mass, as elegantly described using a doxycycline dependent Col1a1-Wnt1 transgenic mouse. The mode of action of WNT1 is proposed to be juxtacrine, rather than autocrine, based on in vitro studies of cells overexpressing the protein ligand [124].

S2P/S1P and OASIS: insight into the role of Regulated Intramembrane Proteolysis in osteoblast homeostasis

Mutations in MBTPS2 cause the recessive X-linked OI type XVIII [125] (OMIM 300294). MBTPS2 is located on the X-chromosome and encodes the Golgi membrane protease S2P, which acts in concert with the protease S1P (encoded by MBTPS1) for regulated intramembrane proteolysis (RIP) [126, 127]. RIP plays a key role in sterol synthesis and ER stress homeostasis via proteolytic cleavage and activation of ER bound transcription factors. Current models indicate that transcription factors, such as members of SREBP, ATF6 and CREB3 families, traffic to the Golgi upon activation by S1P and S2P, which proteolytically cleave off the N-terminal domain that is then translocated to the nucleus for gene regulation (Figure 3).

MBTPS2 mutations c.1376A>G and c.1515G>C have been identified in males from two OI type XVIII pedigrees [125]. Both probands exhibited moderate/severe OI with numerous fractures in childhood. In addition, short stature, white sclerae, vertebral compressions and scoliosis, barrel chest, and reduced DXA bone density were observed in both probands. The c.1376A>G and c.1515G>C mutations lead to p.Asn459Ser and p.Leu505Phe substitutions, respectively, in S2P. These substitutions are located in or near to the conserved metal ion binding domain of the enzyme. Both mutations impaired S2P function and resulted in reduced cleavage of RIP substrates. Indeed, the reduced activation of OASIS, an S2P target, may underlie the decreased secretion of collagen from type XVIII OI osteoblasts. Furthermore, lysyl hydroxylase 1 expression and hydroxylation of its substrate residue K87 in type I collagen chains were also reduced in OI type XVIII proband osteoblasts. The elevated lysylpyridinolyne/hydroxylysylpyridinolyne (LP/HP) cross-link profile in proband urine and bone tissue supported altered collagen crosslinking in the matrix. These data suggest that reduced cross-linking, together with decreased collagen secretion, affect the quality and quantity of ECM and undermine bone strength in type XVIII OI individuals.

Mutations in MBTPS2 have also been identified in individuals with ichthyosis follicularis, atrichia and photophobia syndrome (IFAP) or keratosis follicularis spinulosa decalvans (KFSD) [128–130]. Substitutions found in IFAP or KFSD individuals map to different locations in S2P than p.Asn459Ser or p.Leu505Phe. OI probands do not exhibit skin lesions, atrichia or photophobia typical of these diseases. Some IFAP/KFSD individuals have mild skeletal abnormalities, such as vertebral defects, but do not have a generalized skeletal dysplasia. The diverse pathological effects due to different MBTPS2 mutations suggest that enzyme kinetics of intramembrane cleavage may be differentially affected upon S2P substitutions. Alternatively, substrate specificity and affinity to the catalytic site may be altered by S2P substitutions.

Mutations in OASIS, an ER bound transcription factor encoded by CREB3L1 and one of the S1P/S2P substrates, cause recessive OI type XVI [87, 88, 131, 132] (OMIM 616215). The role of Creb3l1 in skeletal development was first uncovered while examining osteopenia in old astrocyte specifically induced substance (OASIS) null mice [133]. Reduced expression of type I collagen (Col1a1), decreased secretion of non-collagenous bone matrix proteins, and enlarged ER were reported. As a transcription factor, OASIS binds to the Col1a1 promoter and stimulates type I collagen gene transcription (Figure 3). Transcription of non-collagenous matrix proteins, such as bone sialoprotein and matrilin, may also be upregulated by OASIS.

Three independent OI type XVI probands have been identified with homozygous CREB3L1 mutations. In one individual, molecular analysis identified the mutation c.1284C>A, which leads to a premature stop codon p.Tyr428* [132]. The second proband was terminated in utero due to severe skeletal dysplasia [87]. Molecular analysis revealed c.934_936delAAG, causing a 3-bp in-frame deletion (p.Lys312del) in OASIS. In vitro biochemical studies revealed that exogenous p.Lys312del OASIS showed defective DNA binding. Transcription assays indicated that OASIS regulates expression of SEC24D, a component of COPII in ER-Golgi trafficking. The third proband carried lethal homozygous c.911C>T mutation (p.Ala304Val) in the conserved nuclear localization sequence and reduced SEC24D expression [88, 131].

Since S2P functions coordinately with S1P to process substrates such as OASIS, deficiency of S1P would reasonably be expected to cause a syndrome with OI types XVIII and XVI phenotype overlapping. Interestingly, the skeletal consequences to MBTPS1 defects are osteochondral, as shown in zebrafish [134], mice and in a recently identified proband [135, 136]. An 8-yr old child was reported with compound defects in MBTPS1, resulting in 1% functional transcripts. A moderately severe skeletal dysplasia became apparent after 6 months, including growth retardation, spondyloepiphyseal dysplasia, facial dysmorphism similar to Russell-Silver and reduced bone density. Mannose-6-phosphate modification was impaired, impeding Golgi-lysosomal trafficking. These factors may underlie the ER retention of collagen by chondrocytes, decreased transcription of COPII components, and elevated secretion of lysosomal enzymes in the proband [136].

Thus, an emerging model suggests that the RIP pathway involving S2P cleavage of members of the CREB3 family is critical to normal skeletal function, two aspects of which concern regulation of collagen transcription and of components of COPII vesicles for collagen secretion [87, 88, 133]. Future identification of novel, and potentially distinct, S1P/S2P substrates may illuminate the RIP pathway in bone regulation and function.

SPARC: a matricellular protein with intracellular chaperone function

The identification of mutations in SPARC, encoding for the secreted protein, acidic and rich in cysteine (SPARC), shed new light on the role of this extracellular matrix protein in mineralized tissue [137]. SPARC, also referred to as osteonectin or basement membrane protein 40 (BM-40), is a calcium-binding protein and one of the most abundant non-collagenous protein expressed in mineralized tissues. It is classified among the matricellular proteins without structural function, but relevant for cellular-matrix interaction [138, 139]. In osteoid, SPARC has been proposed to bind collagen and hydroxyapatite crystals and release calcium ions, perhaps enhancing mineralization of the bone collagen matrix [138]. The two individuals with an OI diagnosis, who were reported with biallelic mutations in SPARC, carry non-synonymous homozygous variants c.497G>A (p.Arg166His) in exon 7 and c.787G>A (p.Glu263Lys) in exon 9, respectively. These individuals presented with progressive and severe bone fragility, multiple vertebral compression fractures and bone hypermineralization [137].

Arg166 and Glu263 are evolutionary conserved amino acids in the Ca²⁺ binding domain of SPARC, which form a salt bridge necessary for collagen binding [138]. The bone phenotype associated with the substitution of these amino acids supports a critical role of SPARC-collagen interaction in skeletal tissue. Interestingly, dermal fibroblasts isolated from OI individuals exhibited delay in collagen secretion and a mild post-translational overmodification, supporting a previously hypothesized intracellular chaperone role for SPARC [137, 140]

A general Sparc knock-out murine model generated before the identification of SPARC mutations as causative for human OI had already suggested a role for SPARC in bone. Sparc-null mice develop progressive osteoporosis, due to a defect in bone formation [141]. SPARC-null bones show alterations in collagen structure, mineral composition, and apatite crystallite morphology [142]. As in human cases, high mineral content of bone was found, associated with increased crystal size in mice. SPARC-null mice exhibit reduced collagen content and smaller collagen fibrils [143]. Osteoblastic precursors are also reduced in SPARC-null mice, and osteoblast differentiation from SPARC-null bone marrow derived cells is impaired [144].

In a recent study, a peptide (ON29) derived from the mineral-binding region of SPARC induced changes to the disordered phase and apatite crystallite morphology in an in vitro assay of matrix mineralization, supporting a role of the protein in bone formation through its capacity to bind mineral [145]. Interestingly, during in vitro differentiation of WT osteoblasts, SPARC mRNA level remains relatively constant [144], but the protein level is higher at the initial stages of differentiation and subsequently decreases as cells mature and begin to express late osteoblastic markers [144]. A post-transcriptional regulation mediated by miR-29 has been proved to play a role. miR29 interacts with the 3′-UTR of SPARC mRNA and increased levels of miR-29 were found during osteoblast differentiation [146]. Since the expression of SPARC by osteoclasts has not been reported to date, SPARC seems to limit osteoclast formation by a still-unknown mechanism [147].

Thus, studies from mice and humans revealed a critical role of SPARC in maintaining bone mass and quality. The multi-level mechanisms include the regulation of procollagen processing and assembly in the bone matrix, mineralization, and/or osteoblast/osteoclast differentiation and activity (Figure 3).

Insight from OI-related bone fragility syndromes

Plastin 3: a cytoskeletal protein involved in bone homeostasis

Plastin 3 (PLS3) belongs to the group of proteins whose role in bone homeostasis was discovered during the investigation of the molecular basis of childhood-onset primary osteoporosis. PLS3 mutations were first detected in the male members of five families with clinical features of OI or early-onset osteoporosis associated with osteoporotic features, without COL1A1 and COL1A2 defects [148]. Seventeen different mutant PLS3 alleles have been described in patients affected by idiopathic juvenile osteoporosis with X-linked inheritance characterized by reduced bone mineral density, vertebral compressions and long bone fractures [149–156].

PLS3 is located on the X chromosome (Xq23); it is composed by 16 exons and encodes plastin 3, also named Plastin-T, a ubiquitous cytoskeletal protein involved in the formation of F-actin bundles and known to be expressed in osteoblasts, osteocytes and osteoclasts [149, 157–159]. Interestingly, altered expression of cytoskeletal proteins has been recently reported as potential modulators of severity in the Brtl OI murine model [160]. PLS3 missense mutations, small and large deletions, and small insertions resulting in premature termination codons or impairment of plastin 3 function were identified. The range of severity in individuals with PLS3 mutations has not currently been associated with a genotype-phenotype relationship. The X-chromosomal location of the gene may partially explain the more severe phenotype in hemizygous male subjects, although females carrying heterozygous mutations can also display severe bone fragility, possibly related to the extent of X-inactivation of the mutant allele or the presence of modifiers [153]. Of note, a rare variant in PLS3 shows the strongest association with osteoporosis in elderly post-menopausal women, suggesting that osteoporosis may also be influenced by PLS3 [148]. In individuals with PLS3 mutations, type I collagen has normal structure and a normal level of synthesis. In the majority of cases, an extraskeletal phenotype is not described, making this osteoporotic disease distinctive from both classical and recessive forms of OI. Nevertheless, several exceptional cases should be mentioned to provide a full picture of the consequences of PLS3 mutations to allow proper differential diagnosis. In two male individuals carrying a large PLS3 genomic deletion encompassing exons 4 to 16, dysmorphic features such as round face, narrow deep-set eyes, bushy eyebrows and large ears have been reported [152]. Features such as blue sclerae, yellow teeth, loss of tooth enamel, soft skin, joint laxity and mild aortic valve regurgitation, described in some male and female subjects carrying missense mutations responsible for premature stop codons and the substitution of an highly conserved amino acid, will require differentiation from an OI diagnosis [153].

Iliac crest bone biopsies obtained from several patients with PLS3 mutations had a normal bone lamellation pattern, low BV/TV and reduced trabecular and cortical thickness [149, 150, 152, 155]. Quantitative Backscattering Electron Imaging (qBEI) was reported on bone from 4 individuals with PLS3 defects. Subjects carrying PLS3 p.Asp332*, a large deletion from exon 4 to 16 and p.Leu432* mutations, respectively, had normal mineralization, hypomineralization and hypermineralization [149, 152, 155], findings which are not concordant with the hypermineralization described in the vast majority of OI forms [161, 162]. Even taking into consideration the variability of bone properties with age and site of biopsy, the data support a complex albeit poorly understood role of PLS3 in bone mineralization. Overall, PLS3 mutations cause low bone formation and resorption, which would be expected to increase matrix mineralization over time. Consistent with a role in bone mineralization, PLS3 has been demonstrated in matrix vesicles (Figure 5) [163, 164].

Figure 5: PLS3 function in osteoclasts.

Plastin 3 (PLS3) is a cytoskeletal protein involved in the formation of F-actin bundles. Osteoclast resorptive activity, migration and adhesion are dependent on formation of large actin filaments containing ring structures known as podosomes, which are impaired when PLS3 levels are decreased. In osteoclasts PLS3 interacts in the cytosol with the NF-kB repressing Factor (NKRF) favoring its translocation into the nucleus. Here NKRF inhibits the transcription of NFATC1, a key regulator for osteoclastogenesis, causing an impairment in osteoclast activity.

The role of PLS3 in bone fragility is still puzzling. A bone-regulatory role for PLS3 was demonstrated in vivo in zebrafish, in which morpholino knock-down of pls3 caused severe craniofacial abnormalities with impaired muscles [148]. Based on the abundance of fimbrin, a chicken PLS3 homologue, in osteocyte dendrites and on the role of dendrites in osteocyte function, a possible effect of PLS3 mutations on osteocytes mechanosensing function can reasonably be hypothesized [148].

Although neuromuscular abnormalities are not reported in most individuals with PLS3 mutations, a few probands exhibit “waddling gait” and mild muscular hypotonia [148, 152]. Indeed, overexpression of PLS3 is known to have a protective effect in spinal muscular atrophy, with a positive role on axonal growth and pre-synaptic F-actin dependent processes at the neuromuscular junctions [165, 166]. While normal intellectual ability is reported for almost all individuals with PLS3 mutations, mild developmental delay was described in two subjects [167], one with a diagnosis of autism spectrum disorder, revealing the necessity of also assessing Central Nervous System function [155].

Investigation of general Pls3 knock-out and Pls3 overexpressing mice shed new light on the role of PLS3 in bone homeostasis, revealing a crucial role of plastin 3 in osteoclast function, further setting defects in this gene apart from classical osteogenesis imperfecta, and its characteristic high-turnover bone histomorphometry [151]. Pls3 knock-out mice have a cortical and trabecular osteoporotic phenotype which is more severe in male than in female mice, resembling humans with loss-of-function PLS3 mutations. Biomechanical investigation demonstrates reduced breaking force, ultimate stress and stiffness in males, but not in females, further supporting bone fragility. Pls3 overexpressing mice have increased femoral cortical thickness. Lack of Pls3 increases osteoclasts activity in knock-out mice, whereas Pls3 overexpression decreases resorption activity in transgenic animals. Osteoclast resorptive activity, migration and adhesion are dependent on formation of large actin filaments containing ring structures known as podosomes [168], which are impaired in the presence of altered PLS3 levels (Figure 5) [151]. Furthermore, PLS3 interacts in the cytosol with the NF-kB repressing Factor (NKRF), favoring its translocation into the nucleus. Nuclear NKRF inhibits transcription of a key regulator of osteoclastogenesis, Nuclear Factor of Activated T-cells Cytoplasmic 1 (NFATC1), which underlies the increased osteoclast activity described in knock-out mice and some PLS3 patients (Figure 5) [151, 169–171]. Finally, an additional possible effect of PLS3 on osteoclasts may be mediated by impaired intracellular trafficking, since PLS3 is known to be involved in vesicular dynamics [151, 172].

The fact that PLS3-inactivating mutations cause a predominantly bone phenotype in mice and humans suggests that other actin binding proteins may compensate in other tissues.

FAM46A: a nucleotidyltransferase with unexpected bone regulatory function

FAM46A is the most recent example of a gene associated with a skeletal dysplasia with some overlapping clinical features with OI. The identification of FAM46A mutations shed new light on its unexpected function in mineralized tissue [173]. FAM46A is one of the four members of the mammalian FAM46 paralogs, which also includes FAM46B, FAM46C and FAM46D, belonging to the nucleotidyltransferases (NTase) fold superfamily [174]. In humans, it is located on chromosome 6 and encodes a soluble protein, whose function is not yet defined. A loss-of-function mutation in FAM46A was first identified by exome sequencing undertaken to investigate the molecular basis of Stuve-Wiedermann syndrome (SWS) in an individual who did not have the Leukemia Inhibitory Factor Receptor (LIFR) mutation, previously identified as causative for this disease [175]. SWS and OI share some skeletal defects such as bowing of the lower limbs and spontaneous fractures, and similar extraskeletal findings including blue sclerae, dentinogenesis imperfecta, hearing loss and joint hyperlaxity [3, 176, 177]. A homozygous c.612_613 duplication in FAM46A, causing a premature stop codon (p.Ser205Tyrfs*13) and the likely absence of the protein, was identified in a subject clinically overlapping with severe OI, including frequent fractures, bowing of long bones, wormian bones, blue sclerae and joint laxity. This finding prompted the search for FAM46A defects in other severe OI patients with unknown mutations. Among 25 individuals, two novel homozygous missense mutations were detected in two families and predicted to be causative for the disease [173].

All the three mutant alleles are located in the DUF1693 domain of the protein, which is highly conserved in the NTase fold superfamily of proteins [178]. FAM46A expression in primary osteoblasts suggests a role in bone homeostasis, that is further supported by the phenotype described in the recently identified ENU-derived mouse model carrying a homozygous missense mutation, FAM46A c.469G>T (p.Glu157*), which likely causes absence of the protein and is located in the same conserved domain altered in humans [179]. The mice reveal alterations in both endochondral and membranous ossification. Mutant mice have growth delay, multiple long bone and rib fractures, severe limb, rib, pelvis and skull deformities, minimal trabecular bone and reduced cortical and calvarial thickness, and elevated plasma alkaline phosphatase. Fam46a is expressed in most adult and embryonic tissues with strong expression in mineralized tissues such as femur and calvaria, suggesting a possible dual role in bone development and postnatal bone homeostasis [179, 180]. The nucleotidyltransferase FAM46A may be involved in the addition of nucleotides to either cytosolic and nuclear RNAs, thus influencing gene expression and possibly cell differentiation. A functional proteomic study revealed FAM46A as a binding partner for the TGFβ effector SMADs [181], major players of a signaling pathway with a critical role in bone development and homeostasis [182]. In this regard, FAM46A has been recently demonstrated to function as a modulator of BMP signaling, via its binding to Smad1 and Smad4 during early differentiation of Xenopus (Figure 4) [183]. A detailed analysis of the bone properties, by microCT and/or qBEI, is not available for either patients or homozygous Fam46a mutant mice, thus, their classification within the OI spectrum awaits further investigations.

Also awaiting functional illumination are the Variable Number of Tandem Repeats (VNTR) in FAM46A exon 2, a domain which has been previously associated with predisposition to tubercolosis and osteoarthritis of hip and knee joints [184, 185], as well as the initial linkage of mutations in FAM46A to retinitis pigmentosa [186].

Conclusion

The last dozen years has been an exciting time for rare bone disorders. The discoveries of novel and often unpredicted molecular mechanisms regulating bone cellular differentiation, activity and cross-talk revealed bone as an extremely dynamic tissue. Mutations identified in rare bone diseases such as osteogenesis imperfecta and related bone fragility disorders proved to be an invaluable tool to shed light on new pathways. The generation of murine and zebrafish models, as well as the development of in vitro models for these diseases, corroborate the molecular findings and help to dissect the role of several previously unknown molecular players in bone biology. The characterization of BRIL and PEDF mutations revealed an unexpected cross-talk between an integral membrane protein highly expressed in osteoblasts and a secreted cytokine known to inhibit angiogenesis, respectively, in modulating mineral deposition in presence of normal collagen matrix. HSP47 and FKBP65 mutations in recessive OI highlighted the role of ER chaperones in collagen post-translational modification and ER to Golgi secretion, allowing the identification of an unknown chaperone-lysyl hydroxylation complex affecting collagen cross-linking and mineral deposition. TMEM38B defects revealed the relevance of intracellular calcium homeostasis in multiple steps of collagen biosynthesis. mTOR identification as possible new transducer of WNT1 was derived from intensive investigation following WNT1 identification as cause of recessive OI form. The unexpected requirement of specific membrane proteases in maintaining bone homeostasis came from the characterization of MBTPS2 and OASIS mutations, whereas PLS3 mutations revealed the role of actin in both osteocyte and osteoclast activity. A possible cross-talk between nucleotidase and signal transduction pathways that play an important role in bone homeostasis was hypothesized following the identification FAM46A mutations in skeletal dysplasia. The new knowledge substantially enhanced our understanding of normal bone physiology. Moreover, it will also provide potential new targets for the development of novel therapies for currently incurable rare bone diseases, as already demonstrated by the preclinical trials with recombinant PEDF for OI type VI [53].

Abbreviations

OI

osteogenesis imperfecta

CRTAP

cartilage associated protein

P3H1

prolyl-3-hydroxylase 1

PPIB

peptidyl-prolyl cis-trans isomerase B

ER

endoplasmic reticulum

P4HB

prolyl-4 hydroxylase-B

ECM

extracellular matrix

BMP1

bone morphogenic protein 1

BRIL

bone restricted ifitm-like protein

IFITM5

interferon-induced transmembrane protein 5

PEDF

pigment epithelium-derived factor

SERPINF1

serpin family F member 1

5′UTR

5′-untranslated region

MALEP

Met-Ala-Leu-Glu-Pro

GSK-3β

glycogen synthase kinase 3-beta

VEGF

vascular endothelial growth factor

HSP47

heat shock protein 47

PPIase

prolyl cis-trans isomerase

UPR

unfolded protein response

ERGIC

ER-Golgi intermediate compartment

COPII

coat protein II complex

SLRPs

small leucine-rich proteins

RIP

regulated intramembrane proteolysis

LH

lysyl hydroxylase

WNT1

Wingless-type MMTV integration site family 1

LRP

low density lipoprotein receptor-related proteins

CNS

central nervous system

LP/HP

lysylpyridinolyne/hydroxylysylpyridinolyne

IFAP

ichthyosis follicularis, atrichia and photophobia syndrome

KFSD

keratosis follicularis spinulosa decalvans

OASIS

old astrocyte specifically induced substance

SPARC

secreted protein, acidic and rich in cysteine

PLS3

plastin 3

qBEI

quantitative backscattering electron imaging

NKRF

NF-kB repressing factor

NFATC1

nuclear factor of activated T-cells cytoplasmic 1

NTase

nucleotidyltransferases

LIFR

leukemia inhibitory factor receptor

VNTR

variable number of tandem repeats

IP₃R

inositol-3-phosphate receptor

CyPB

cyclophilin B

mTOR

mammalian target of rapamycin

R-SMAD

receptor regulated SMAD

PDI

protein disulfide isomerase

MBTPS2

membrane bound transcription factor peptidase site 2

MBTPS1

membrane bound transcription factor peptidase site 1

SREBP

sterol regulatory element binding protein

ATF6

activating transcription factor 6

CREB3

cAMP responsive element binding protein 3

SERCA

sarcoplasmic endoplasmic reticulum calcium ATPase