Genetic approaches to metabolic bone diseases

Metabolic bone diseases comprise a diverse group of disorders characterized by alterations in skeletal homeostasis, and are often associated with abnormal circulating concentrations of calcium, phosphate or vitamin D metabolites. These diseases commonly have a genetic basis and represent either a monogenic disorder due to a germline or somatic single gene mutation, or an oligogenic or polygenic disorder that involves variants in more than one gene. Germline single gene mutations causing Mendelian diseases typically have a high penetrance, whereas the genetic variations causing oligogenic or polygenic disorders are each associated with smaller effects with additional contributions from environmental factors. Recognition of familial monogenic disorders is of clinical importance to facilitate timely investigations and management of the patient and any affected relatives. The diagnosis of monogenic metabolic bone disease requires careful clinical evaluation of the large diversity of symptoms and signs associated with these disorders. Thus, the clinician must pursue a systematic approach beginning with a detailed history and physical examination, followed by appropriate laboratory and skeletal imaging evaluations. Finally, the clinician must understand the increasing number and complexity of molecular genetic tests available to ensure their appropriate use and interpretation.


Introduction
Metabolic bone diseases represent a diverse group of skeletal conditions characterized by alterations in bone cell activity, bone matrix proteins or systemic mineral homeostasis (Table 1) [1,2]. Many metabolic bone diseases have a genetic basis, which may be a germline single gene abnormality (i.e. a monogenic or Mendelian disorder), a somatic single gene defect (i.e. a post-zygotic mosaic disorder) or involve several genetic variants (i.e. oligogenic or polygenic disorders) [3]. Genetic mutations causing Mendelian diseases usually have a large effect (i.e. penetrance), whereas oligogenic or polygenic disorders are associated with several genetic variations, each of which may have smaller effects with greater or smaller contributions from environmental factors (i.e. multifactorial disorders) [3]. Whilst many monogenic disorders result from rare mutations affecting the coding sequence of the responsible gene, the majority of common genetic variants identified in association with polygenic traits are located in non-coding regions, usually in proximity to candidate genes implicated in the respective disorders [4]. Furthermore, there is substantial overlap Table 1 Examples of monogenic metabolic bone disorders, modes of inheritance and genetic aetiology

Autosomal dominant
Osteogenesis imperfecta (OI), types I-IV COL1A1, COL1A2 17q21. 33 Familial expansile osteolysis TNFRSF11A 18q21.33 [34,35] Hypophosphatasia TNSALP/ALPL 1p36.12 [36] Vitamin D-dependent rickets, type 3 CYP3A4 7q22.1 [88] Pseudohypoparathyroidism, type 1a (PHP1a) a GNAS 20q13.3 [39] Pseudopseudohypoparathyroidism (PPHP) a GNAS 20q13.3 [39] Pseudohypoparathyroidism, type 1b (PHP1b) a GNAS, NESP55, STX16 20q13.3 [39] Autosomal recessive (continues) between the genes responsible for monogenic skeletal diseases and those contributing to polygenic bone phenotypes. The elucidation of these loci has provided insights into the molecular pathogenesis of skeletal disease, and highlighted novel therapeutic targets [5][6][7]. This review discusses the genetics of metabolic bone diseases, and outlines the clinical and genetic approach to evaluating these disorders. Mutations within a single gene may give rise to seemingly distinctive skeletal phenotypes [(e.g. familial expansile osteolysis (FEO), expansile skeletal hyperphosphatasia (ESH) and early-onset familial Paget's disease of bone (PDB)], which are rapid remodelling skeletal disorders arising from mutations in the signal peptide of receptor activator of NF-κB (RANK) [34,35]. In some metabolic bone diseases, the severity may be determined by mutant allele dosage and whether a mutation is carried in the heterozygous or homozygous state. For example, the severe perinatal and infantile forms of hypophosphatasia, an inborn-error-of-metabolism characterized by alkaline phosphatase (ALP) deficiency, are inherited Genetic approaches to metabolic bone diseases in an autosomal recessive manner, whilst later-onset and milder forms are typically inherited in an autosomal dominant fashion (Table 1) [36]. Moreover, some disorders of mineral metabolism are caused by loss-or gain-of-function mutations affecting the same gene. Thus, loss-of-function CaS receptor mutations cause FHH or neonatal severe hyperparathyroidism (NSHPT), whereas gain-of-function CaS receptor mutations cause autosomal dominant hypocalcaemia (ADH) or Bartter syndrome type V [31, 37,38]. Furthermore, parental imprinting, which results in non-Mendelian inheritance of a monogenic disorder, may influence the phenotypic consequences of a specific mutation. For example, maternally-inherited inactivating coding-region mutations of G-protein subunit αs (Gαs), which is encoded by the GNAS gene, cause pseudohypoparathyroidism type 1a (PHP1a), which is characterized by PTH resistance together with Albright's hereditary osteodystrophy (AHO) [39]; whereas, paternally-inherited inactivating coding-region GNAS mutations cause pseudopseudohypoparathyroidism (PPHP), which is characterized by AHO without PTH resistance (Table 1) [39]. The phenotype of MAS, which is caused by somatic activating Gαs mutations, is also dependent on parental imprinting, with acromegaly occurring in MAS patients who harbour mutations affecting the maternal Gαs allele [40]. Given this apparent genetic/phenotypic complexity despite genetic 'homogeneity', establishing the genetic cause can be challenging for the evaluation of patients and family members with bone and mineral disorders.

Molecular insights from monogenic and polygenic diseases
Classical gene discovery approaches for monogenic disorders have involved studying affected kindreds for cosegregation with polymorphic genetic markers to define the chromosomal location, followed by DNA sequence analysis of genes located within the candidate region [3]. This approach has been superseded by whole-exome and whole-genome sequence analysis of affected patients or kindreds [41,42]. In contrast, the genetic investigation of complex polygenic disorders such as osteoporosis has utilized genome-wide association studies (GWAS), which involve large populations of cases and controls [5, 6, 16, 17]. Such studies typically involve direct or imputed genotyping of large numbers of common (e.g. minor allele frequency >5%) and infrequent (e.g. minor allele frequency 1-5%) single nucleotide polymorphisms/variants (SNPs/SNVs) to identify genetic loci enriched for the trait [3,43]. The genetic investigation of monogenic diseases has provided a fundamental understanding of the molecular regulation of bone mass and maintenance of skeletal microarchitecture. For example, studies of mutations affecting several Wnt pathway components have demonstrated that Wnt signalling plays a key anabolic role in the skeleton ( Figure 1) [44,45]. Thus, autosomal-recessive loss-of-function mutations of the LRP5 gene, which encodes a key Wnt co-receptor (Figure 1), result in osteoporosis-pseudoglioma syndrome, which is characterized by severe juvenile osteoporosis and congenital or childhood-onset blindness [46]. In contrast, heterozygous activating mutations in LRP5 [47] and LRP6 [48], which encode the cognate co-receptors LRP5 and LRP6, respectively, both lead to autosomal dominant high bone mass. Additionally, individuals with autosomal recessive loss-of-function mutations of the Wnt-β-catenin inhibitor sclerostin (SOST) manifest sclerosteosis, type 1, which is characterized by progressive bone overgrowth throughout life [49,50]

Application of genetic discoveries to the development of targeted therapies
A key aim of the genetic characterization of metabolic bone disorders has been to identify genes, molecules and pathways that may be targeted therapeutically. Thus, the identification of the bone cell OPG/RANKL/RANK/NF-κB signalling pathway led to the development of the monoclonal antibody denosumab, which blocks RANK ligand (RANKL), thereby inhibiting osteoclast-mediated bone resorption [5].
Denosumab is now widely used for the treatment of osteoporosis as it significantly reduces fracture risk in women with postmenopausal osteoporosis [57]. The multinational approval in 2015 of the bone-targeted enzyme-replacement biologic asfotase alfa to treat hypophosphatasia has emphasized the importance of determining the genetic and molecular basis for a metabolic bone disease [36]. The identification that PHEX mutations cause FGF-23 excess, which in turn is responsible for the phosphate wasting in XLH [58,59], has led to the approval in 2018 of burosumab, which is an anti-FGF-23 monoclonal antibody, for the treatment of XLH rickets. Burosumab has been shown to improve serum phosphate concentrations and decrease the severity of rickets in children with XLH [60]. Assessing treatment response according to the genetic aetiology has been investigated in patients with early-onset low-turnover osteoporosis due to WNT1 or PLS3 mutations who were shown to respond to teriparatide therapy [61]. Now, several drugs in development are directed at the Wnt pathway. This includes anti-sclerostin antibodies (e.g. romosozumab), which increase bone formation whilst inhibiting bone resorption [62]. An evaluation of romosozumab in phase 3 clinical trials has shown that it is a potent bone anabolic agent for postmenopausal osteoporosis [63,64].

Clinical approach to the patient with a metabolic bone disease
Medical history and physical examination  Schematic representation of Wnt signalling pathway components reported to be mutated in disorders of bone development and skeletal homeostasis. Activation of the canonical Wnt pathway increases bone mass, and this is mediated by the binding of extracellular Wnt ligands (dark green) to a transmembrane receptor complex comprising the Wnt co-receptor LRP5 or LRP6 (LRP5/6, light blue) and a member of the frizzled (FZD) family (dark blue). In contrast, inhibition of the canonical Wnt pathway decreases bone mass [44,45]. This inhibition is mediated by extracellular factors such as sclerostin (SOST, orange) and Dickkopf-related protein 1 (DKK1, yellow), which bind to the LRP5/6 co-receptor thereby preventing activation by Wnt ligands, as well as recruiting inhibitory transmembrane proteins such as LRP4, which is a SOST-interacting protein (light green), and the Kremen proteins (pink), which are high-affinity DKK1 receptors that functionally cooperate with DKK1 to decrease Wnt signalling [109]. Secreted-frizzled-related proteins (SFRPs, purple) also inhibit the canonical Wnt pathway by sequestering Wnt ligands. The importance of the canonical Wnt pathway for the regulation of bone mass has been highlighted by loss-of-function mutations affecting SOST and LRP4, and by gain-of-function mutations of LRP5 and LRP6, which lead to the disorder called high bone mass [47,49,51,110]; and also by loss-of-function mutations of LRP5 and the Wnt1 ligand, which lead to monogenic osteoporosis disorders [19,46] syndrome; angiofibromas or collagenomas that may be associated with MEN type 1; premature loss of deciduous teeth that occurs in hypophosphatasia; hallux valgus which is found in fibrodysplasia ossificans progressiva; alopecia that occurs in vitamin D-dependent rickets, type 2; brachydactyly which is found in PHP1a and PPHP; syndactyly that occurs in sclerosteosis types 1 and 2; torus palatinus which is found in disorders of high bone mass due to LRP5

Family history
Assessment of the family history is essential for establishing the mode of inheritance of monogenic metabolic bone diseases, and medical records from living or deceased affected family members may establish the diagnosis, guide prognostication, and indicate a safe and effective treatment [3]. In autosomal dominant disease, the affected person often has one affected parent, and the disease occurs in both sexes and is transmitted by either the father or mother. In autosomal recessive diseases, which can affect both sexes, the proband is born to parents who are usually asymptomatic 'carriers' and sometimes related (i.e. consanguineous). In X-linked recessive diseases, usually only males are affected, parents are unaffected yet the mother is an asymptomatic carrier, and there is no male-to-male transmission. In X-linked dominant diseases, both males and females can be affected, although the females are often more mildly and variably affected than males, and 50% of offspring (girls and boys) from an affected woman will have the disease, and 100% of the daughters but 0% of sons of an affected man will have the disease. In Y-linked diseases, only males are affected and, unless representing a sporadic case, they have an affected father (patrilineal inheritance) and all sons of an affected male will have the disease. Mitochondrial inherited disorders (non-Mendelian) can affect both sexes. However, these disorders are only transmitted by an affected mother (matrilineal inheritance) in her egg mitochondrial DNA, and not through the paternal line in the sperm, as the small volume of sperm precludes them from contributing mitochondria to the zygote [3]. These patterns of inheritance may be complicated by: non-penetrance or variable expression in autosomal dominant disorders (e.g. in MEN1) [65]; imprinting whereby expression of an autosomal dominant disorder is conditioned by whether it is maternally or paternally transmitted (e.g. PHP1a versus PPHP) [39]; anticipation, whereby some dominant disorders become more severe (or have earlier onset) in successive generations; pseudo-dominant inheritance of autosomal recessive disorders reflecting repeated consanguineous marriages in successive generations; and mosaicism in which an individual has two or more populations of cells with different genotypes because of postzygotic mutations during their development from a single fertilized egg (e.g. MAS). In the special circumstance of germline mosaicism within eggs or sperm arising from somatic mutation during gametogenesis, there may be confusion about the diagnosis and recurrence risk because of seemingly unaffected parents having multiple affected offspring that would be consistent with autosomal recessive inheritance, but actually reflects an autosomal dominant disorder (e.g. OI type II) [70]. Hence, these inheritance patterns, which can help to diagnose a genetic disorder and identify individuals at risk, can come from a detailed family history [3].

Clinical utility of genetic investigations
Establishing the genetic basis of a metabolic bone disease may aid diagnosis, treatment and prognostication; identify the need for screening of associated clinical features not initially apparent; enable appropriate genetic counselling and testing of first-degree asymptomatic relatives; and facilitate preconception and/or prenatal genetic evaluation (Figure 2). Genetic testing may also aid risk profiling. For example, osteoporosis-associated SNPs have been reported to predict fracture risk in patients taking bisphosphonates [71], and other studies have identified potential genetic markers of bisphosphonate-induced osteonecrosis of the jaw [72]. For patients presenting with a likely genetic metabolic bone disease, several factors require consideration before organizing genetic testing (Figure 2). These include the phenotype of the patient, the likely mode of inheritance, the potential genetic aetiology [e.g., aneuploidy, copy number variation (CNV), or single gene defect], and availability of additional pedigree members (Figure 2). For example, DNA sequencing of 'trios' (i.e. both parents and the affected proband) may facilitate the identification of compound heterozygous or de novo mutations [73]. Selecting the most appropriate genetic test will increase the likelihood of achieving a genetic diagnosis. For example, direct DNA sequencing methods which detect nucleotide abnormalities (e.g. substitutions, micro-deletions and micro-insertions) that cause most monogenic metabolic bone disorders frequently do not detect whole or partial gene deletions that are associated with some monogenic syndromes, and are also not optimal for identifying large chromosomal abnormalities (e.g. 22q11.2 microdeletion in DiGeorge syndrome), whose detection requires alternative approaches (Figure 2 and Table 2) [74]. For other monogenic disorders, it is also important to consider analysis of a panel of genes if genetic heterogeneity is likely (e.g. in FHH or OI) [9, 20]. Thus, it is important to emphasize that genetic testing which fails to identify an abnormality does not exclude a genetic disease, but rather may reflect: an alternative genetic aetiology to the one being tested; limitations of the employed genetic methodology (i.e. inadequate resolution or coverage);

Figure 2
Flowchart outlining considerations for genetic testing in patients with metabolic bone disease Genetic approaches to metabolic bone diseases

Types of genetic tests available to the clinician
Cytogenetic and molecular cytogenetic analyses. Karyotyping represents the initial test for major chromosomal abnormalities including aneuploidy or large insertions, deletions, duplications, inversions or reciprocal translocations, but has a resolution limited to~5-10 Mb of DNA (Table 2) [74,75]. It retains an important place in the diagnosis of Turner and Klinefelter syndrome, each of which may manifest a form of osteoporosis [76,77]. Fluorescence in-situ hybridization (FISH) employs DNA probes that hybridize to specific target regions, which allow the detection of specific chromosomal deletions, duplications, translocations or inversions ( Table 2). The utility of FISH is limited to detecting abnormalities involving pre-determined genomic regions (e.g. detection of 22q11.2 deletion in DiGeorge syndrome). Multiplex-ligation dependent probe amplification (MLPA) detects complete or partial gene deletions by using a pool of custom-designed probes to amplify specific genomic regions of interest (Table 2). MLPA is used in the diagnostic evaluation of monogenic disorders associated with such genetic alterations (e.g. MEN1) [78]. Modifications of the MLPA technique may also be used. For example, in establishing the diagnosis of pseudohypoparathyroidism type 1b (PHP1b), methylationspecific MLPA (MS-MLPA) may be employed to detect genetic (e.g. deletions) or epigenetic (e.g. altered patterns of methylation) abnormalities within the differentially methylated regions (DMRs) of the GNAS locus, although alternate methods such as CpG bisulphite pyrosequencing are frequently used to confirm the presence of specific methylation defects [79]. Microarray-comparative genomic hybridization (aCGH) is undertaken for the genome-wide detection of small chromosomal abnormalities (e.g. CNVs) ( Table 2) and is increasingly used as a first-line investigation for patients with multiple congenital abnormalities, which include skeletal manifestations and/or neurodevelopmental delay [80,81]. However, it is important to note that all individuals harbour many small CNVs without discernible adverse impact on health, whilst several potentially pathogenic CNVs do not cause disease in all individuals (i.e. reduced penetrance). Finally, SNP arrays may detect CNVs as well as facilitating genome-wide genotyping ( Table 2). For example, deletions spanning several adjacent SNPs included on the array may reveal loss of heterozygosity (LOH), whilst copy number gains (e.g. duplication) may be indicated by increased numbers of different genotypes [74]. SNP arrays may also help localize recessive disorders in the offspring of consanguineous parents by facilitating homozygosity mapping [82], whilst regions of LOH can also indicate uniparental isodisomy, which may be relevant to the diagnosis of imprinting disorders such as PHP1b [83,84].

DNA sequence analysis
Sanger sequencing remains the gold standard for detecting DNA sequence variants due to the high accuracy of the DNA polymerase (i.e. base accuracy of >99.99%) employed during DNA amplification [41,85]. However, it remains labour intensive and is typically reserved for disorders with low genetic heterogeneity (e.g. single-or pauci-gene disorders), an example being hypophosphatasia caused only by TNSALP/ALPL mutations [86]. Single-gene testing by Sanger sequencing is increasingly being replaced by next-generation sequencing (NGS) approaches, which facilitates the simultaneous sequencing of large amounts of genetic material. Such NGS methodology has provided a paradigm shift in the investigation and diagnosis of genetic disease. Currently, the three most widely employed uses of NGS are whole genome sequencing (WGS), whole exome sequencing (WES), and disease-targeted gene panel sequencing (Table 2). WGS determines the DNA sequence of the entire genome including coding and non-coding regions, and can identify SNVs, small insertions or deletions ('indels') and CNVs [3]. In contrast, WES analyses the 1-2% of the genome that encodes thẽ 20 000 protein-coding genes (i.e. the 'exome'), which are expected to harbour most disease-associated mutations [3]. WES has been the mainstay of highly successful disease-gene  [87]; and CYP3A4 mutations in vitamin Ddependent rickets, type 3 [88]). Disease-targeted sequencing represents the most widely utilized NGS method in clinical practice, as it can be designed to simultaneously analyse large collections of genes (e.g. <10 to >150 genes) associated with a specific disorder [41,84,89]. Such NGS disease-targeted panels have been established for genetically heterogeneous disorders including OI and other skeletal disorders, as well as for hypophosphataemic rickets and calcium-sensing disorders [90][91][92].

Genetic tests to detect mosaicism
Some metabolic bone disorders only manifest as somatic mosaicism (e.g. GNAS mutations in MAS) [67]. However, other conditions (e.g. OI type II) may also rarely occur as germline mosaicism, arising from somatic mutation during gametogenesis, and may cause diagnostic confusion. In this setting, apparently unaffected parents (with one carrying the mutation limited to their gametes) may give rise to more than one affected child, suggesting possible autosomal recessive inheritance, in contrast to the underlying autosomal dominant inheritance pattern [93]. Detection of mosaicism has been enhanced by improved genome-wide testing strategies (e.g. aCGH, SNP arrays, droplet digital PCR and NGS approaches), which can provide sensitive methods for the detection of low-level mosaicism (e.g. 5% for SNP array) [70,94,95]. However, choosing the optimal test depends on the clinical phenotype, the type of mutation suspected (e.g. SNV, CNV, aneuploidy), the likely extent of mosaicism, and its tissue distribution. Typically, circulating lymphocyte DNA will suffice, but analysis of other affected tissues may be required (e.g. fibroblasts or bone) [96,97].

Genetic tests for prenatal diagnosis
Prenatal genetic testing may be undertaken at preimplantation or prenatal stages, and has been used to detect severe skeletal disorders such as perinatal lethal OI [98]. Preimplantation genetic diagnosis (PGD) uses a single cell taken from the developing embryo several days after in vitro fertilization (IVF) to detect chromosomal abnormalities or single gene defects, thereby allowing selection of the unaffected embryos for implantation [99]. In contrast, prenatal genetic testing is used once pregnancy is established to identify fetuses at risk of genetic disease [99]. Typically, this involves invasive methods such as chorionic villous sampling (CVS) or amniocentesis to obtain cells for genetic evaluation [99]. This may include karyotyping for the detection of aneuploidy, FISH or aCGH to identify smaller chromosomal abnormalities or DNA sequencing to identify single gene defects associated with monogenic disease. Recent progress in the detection of cell-free circulating fetal DNA in the maternal circulation (e.g. after~10 weeks gestation) now offers the potential for non-invasive prenatal genetic diagnosis (NIPD) and/or testing (NIPT) [100]. Thus, a maternal blood sample may allow screening for aneuploidy and fetal sex determination, which is important for X-linked disorders, and may also be used to detect monogenic disorders; however, this is limited to paternally inherited mutations or those arising de novo, as the sample may contain maternal cell-free DNA, and hence the detected abnormality cannot be reliably assigned to the fetus as the methodology cannot distinguish between fetal and 'contaminating' maternal DNA in the sample [100].

Data interpretation and incidental findings
The advent of high-content genetic testing employing NGS approaches has revolutionized the investigation and diagnosis of genetic disease. However, such approaches may also present clinical and ethical challenges [101]. For example, the simultaneous sequencing of large numbers of genes (e.g. disease-targeted gene panels, WES and WGS) inevitably identifies variants of uncertain significance (VUS), whose relevance to the clinical phenotype is ambiguous [102,103]. Indeed, the methods employed to assess variant effects are frequently imprecise leading to inaccurate interpretation, although the provision of recent large-scale population level sequence databases facilitates improved estimates of variant pathogenicity and penetrance [104,105]. In addition, high-content genetic testing may identify clinically relevant genetic abnormalities unrelated to the phenotype under investigation [i.e. incidental findings (IFs)] and these may have important health implications for the patient and their family. Hence, the possibility of identifying ambiguous or incidental results should be part of the informed consent prior to genetic testing ( Figure 2).

Conclusion
Many metabolic bone diseases have a genetic basis, which may be a germline single gene abnormality (i.e. a monogenic or Mendelian disorder), a somatic single gene defect (i.e. a post-zygotic mosaic disorder), or involve several genetic variants (i.e. oligogenic or polygenic disorders). Recognition of these heritable disorders is clinically important, as it can facilitate relevant and timely investigation and treatment for the patients and families. Recent advances in genetics and DNA sequencing methods have resulted in new ways to detect genetic abnormalities. Therefore, it is increasingly important for the clinician to gain an appreciation of these complex genetic tests and to combine this with the fundamental skills of history taking and physical examination to ensure they are used for the benefit of patients.