Introduction

Paget’s disease of bone is a common metabolic bone disease that is characterized by increased and disorganized bone remodeling [1•]. It preferentially targets the axial skeleton and may affect one or multiple skeletal sites, most commonly the pelvis, spine, skull, femur and tibia. Clinically the raised bone turnover may cause bone pain and this is the most common presenting symptom of the disease. The raised bone turnover may leads to bone expansion, which can manifest clinically as visible deformity, particularly when weight bearing long bones such as the tibia and femur are affected. When the spine and skull are affected, the bone expansion can cause nerve compression syndromes resulting in cranial nerve defects, spinal stenosis or cord compression. Deafness is a recognized complication of PDB affecting the base of skull although this is usually due to a conductive problem as the result of osteosclerosis affecting the temporal bone, rather than compression of the 8th cranial nerve [2]. Bone formation is increased in PDB but the new bone is laid down in a disorganized fashion and is mechanically weak. This can predispose to the development of pathologic fractures, in response to minimal or no trauma. Finally patients with PDB have an increased risk of osteosarcoma. Although osteosarcoma occurs in 0.5 % of patients or less, virtually all osteosarcomas that occur in adulthood do so in patients with PDB [3].

Epidemiology

Paget’s disease has a very striking geographic distribution. The highest incidence in the world is in the UK and within the UK the highest incidence is in the North West of England. Paget’s disease also occurs quite commonly in North West France, Spain and Italy. It also occurs in most other European countries with the exception of Scandinavia where it is quite rare. Paget’s disease is also common in other areas of the world which have been colonized by British and European immigrants such as Australia, New Zealand, the USA and some regions of Canada, most notably Quebec. Paget’s is rare in the Indian subcontinent, Malaysia, Indonesia, China, and Japan. These geographic differences, combined with other evidence from archaeological studies are consistent with a genetic model of predisposition whereby predisposing genetic variants arose in people from the UK many centuries ago and then spread to Europe and the rest of the world through emigration [4, 5]. Although there is little doubt that PDB has a genetic basis, the incidence and severity of the disease has decreased over the past few decades [6, 7], even in patients who are genetically predisposed to develop the disease [8]. Although Paget’s can affect both genders, men are affected more commonly than women. In the UK, the ratio of males to females affected is approximately 1.4 to 1.0 [3]. The occurrence of Paget’s disease is also strongly age-dependent; the disease is unusual before the age of 50 but increases in frequency exponentially thereafter. In the UK, the incidence doubles each decade from the age of 50 onward to reach a peak incidence about 5.4/10,000 population per year among women and 7.6/10,000 population per year among men in those aged 85 and above 85 [3].

Cellular Pathology

At a cellular level the cardinal feature of PDB is increase osteoclast activity. Osteoclast numbers are increased at affected sites and the osteoclasts themselves are much larger and containing many more nuclei than normal. They contain characteristic nuclear inclusion bodies consisting of microcylindrical structures. These were originally thought to represent viral nucleocapsids [9] but it has more recently been suggested that they might represent abnormal protein aggregates due to defects of the autophagy pathway [10]. Osteoclast precursors from PDB patients show increased sensitivity to RANKL and calcitriol compared with those of control subjects [11, 12]. Bone formation is also increased. This is thought to be secondary to the increased osteoclast activity but there is evidence that cells of the osteoblast lineage are also abnormal in PDB [13]. It has been suggested that the osteoclast abnormalities in PDB are caused by paramyxovirus infection. Although over-expression of paramyxovirus proteins has been shown to increase osteoclast activity in vitro and in animal models [14], attempts to isolate paramyxovirus proteins and nucleic acids from patient samples have yielded inconclusive results [15–17].

Inheritance of Paget’s Disease

Familial Paget’s Disease

Patients with Paget’s disease often have a positive family history of the condition and it has been estimated that the risk of PDB occurring in a first degree relative of an affected subject is about a 7-fold increase compared with an unrelated individual. In many families the disease is inherited in an autosomal dominant manner with high but incomplete penetrance with increasing age. Linkage analysis has identified potential candidate loci on chromosomes 2p36, 5q31, 5q35, 10p13, and 18q21. The gene that predisposes to PDB on 5q35 is SQSTM1 [18, 19] and mutations in this gene are the most common cause of familial PDB. Cross-sectional studies indicate that more than 80 % carriers of SQSTM1 mutations develop Paget’s disease by the eighth decade [20, 21]. However, there is evidence that age of onset in familial PDB is delayed in the current generation of people with SQSTM1 mutations as opposed to their parents’ generation [8]. This emphasizes the importance of environmental factors as triggers for the disease.

The causal gene on 10p13 has not yet been identified but families which show linkage to this region show autosomal dominant inheritance with incomplete penetrance as does 5q35-linked Paget’s disease. The strongest candidate gene within the 10p13 region is OPTN [22, 23••].

Other loci identified by linkage in familial PDB lie on chromosome 2q36, 5q31 and 18q21 [24–26]. The 2q36 linkage peak in UK families seems likely to have been a false positive since the signal disappeared when data were reanalyzed after excluding families with SQSTM1 mutations [22]. The status of the 5q31 locus, which was identified in the French Canadian population [24], remains unclear. No gene has yet been identified. The linkage signal previously reported on 18q21 for classical PBD [26] has not been substantiated by other linkage studies [24, 25, 27] although this locus is involved in the PDB-like disorder familial expansile osteolysis [28].

Familial PDB-Like Syndromes

Several rare syndromes with clinical and radiological similarities to PDB have also been described which are clearly inherited. Details are summarized in Table 1. The syndromes of familial expansile osteolysis [29], expansile skeletal hyperphosphatasia [30], and early onset familial Paget’s disease [31] are allelic disorders caused by mutations in TNFRSF11A which encodes RANK. These are severe disorders with high penetrance, which present with deafness and tooth loss in childhood and adolescence followed by the development of osteolytic lesions in early adulthood. The syndrome of juvenile Paget’s disease is a recessive disorder cause by mutations affecting TNFRSF11B which encodes osteoprotegerin [32, 33]. The presentation is during infancy with bone deformity and multiple fractures. The syndrome of inclusion body myopathy, Paget’s disease and frontotemporal dementia (IBMPFD) is inherited in an autosomal dominant fashion and caused by mutations in the VCP gene [34]. It usually presents in the fourth decade with myopathy but about 40 % of patients also go onto develop classical Paget’s disease and 40 % go on to develop dementia.

Table 1 Inherited syndromes with features similar to Paget’s disease of bone
Full size table

Sporadic Paget’s Disease

The majority of patients with Paget’s disease are unaware of a family history of the disorder. Since Paget’s disease is often clinically silent [3] it is unclear to what extent these patients truly have sporadic disease or if they may have had an affected relative in whom the diagnosis was never made. Recent studies have shown that the genetic variants which predispose to PDB have additive effects on susceptibility to the disease and on disease severity [35••, 36•]. In view of this it seems likely that patients who do not carry high penetrance susceptibility alleles such as SQSTM1 mutations, might develop PDB when they have inherited a sufficient number of predisposing smaller effect size variants and have been exposed to the relevant environmental triggers [35••].

Genes and Loci That Predispose to PDB

The genes and loci that predispose to PDB have been identified by a combination of linkage analysis in families and genome wide association studies in unrelated individuals. The following section focuses only on genes and loci with robust evidence for association with PDB or related syndromes.

CSF1

The CSF1 gene located on 1p13 encodes macrophage colony stimulating factor has been implicated as a candidate gene for Paget’s disease by a genome wide association study [23••]. The associated most strongly SNP lay about 87 kb upstream of CSF1. Although the EPS83L gene also lies within the region it is separated from the associated SNP by 2 recombination hot spots. Macrophage colony stimulating factor is a strong candidate gene for PDB since it plays an essential role in promoting differentiation of stem cells to cells of the monocyte-macrophage and osteoclast lineage. The causal genetic variants in CSF1 have not yet been defined. There is evidence that patients with PDB have higher circulating M-CSF levels than controls which would be consistent with the hypothesis that the variants which predispose to PDB at this locus do so by increasing M-CSF expression or altering its stability or secretion from cells.

SQSTM1

The SQSTM1 gene located on chromosome 5q35 was identified as a candidate gene for PDB in families by genome wide linkage analysis and positional cloning. Two groups independently identified 5q35 as predisposing locus for PDB in UK and French-Canadian families [24, 25]. Mutation screening of genes in the region resulted in the identification of a proline to leucine mutation at codon 392 (P392L) of SQSTM1 in the French Canadian population, which segregated with the disease. Subsequently this and other mutations affecting the ubiquitin-associated domain of the protein were identified in the UK population [19] and in other populations worldwide (reviewed by Ralston and Layfield [10]). It has been estimated that between 40 %–50 % of patients with a family history of PDB carry a SQSTM1 mutation compared with 5 %–10 % of those with “sporadic” disease [20, 37]. At the time of writing, 23 different mutations in SQSTM1 have been identified in patients with PDB and most of these affect the ubiquitin associated (UBA) domain.

Mutations of SQSTM1 affect severity and extent of PDB as well as predisposing to the occurrence of PDB. This was first demonstrated in a study of patients participating in the PRISM trial where patients with SQSTM1 mutations had more extensive disease and more complications than those without mutations [38]. These results have recently been confirmed in a multinational study, which showed that the variants identified by genome wide association studies as predisposing to PDB interacted with SQSTM1 mutations to influence disease severity and extent [36•].

Extensive research has been conducted to investigate the mechanisms by which SQSTM1 mutations regulate osteoclast activity. Functional analysis has shown that most mutations cause loss-of-function for ubiquitin binding suggesting that this may be a unifying mechanism by which the disease occurs [39]. Having said that, a missense mutation which results in substitution of serine to threonine at position 349 (S349T) has been described in PDB which does not impair ubiquitin binding or affect NFκB signaling [40]. It has been shown that this mutation impairs the ability of the SQSTM1 gene product p62 to interact with the protein Keap1 and it has been speculated that this might contribute to PDB by regulating expression of genes involved in oxidative stress [40].

Additional studies have shown that several expression constructs with PDB-associated mutations of SQSTM1 activate NFκB signaling compared with wild type SQSTM1 and that osteoclast precursors transfected with the P392L SQSTM1 variant formed more osteoclasts in response to RANKL and M-CSF compared with wild type SQSTM1. The extent to which different mutants can activate NFκB in vitro has recently been shown to correlate with disease extent [41].

It is likely that the UBA domain mutations of SQSTM1 cause osteoclast activation by a number of pathways including activation of atypical protein kinase Cξ/λ, and phosphoinoside dependent protein kinase 1 [42] but an important mechanism seems to involve the deubiquitinating enzyme CYLD [43]. Under normal circumstances, the CYLD protein is recruited to the intracellular domain of the RANK receptor by interacting with the UBA domain of SQSTM1. This, in turn causes deubiquitation of TNF-receptor associated factor 6 (TRAF6), thereby inhibiting signal transduction downstream of RANK. Deletion of the SQSTM1 UBA domain [43] and the P392L mutation [44] has both been shown to prevent CYLD being recruited to the RANK receptor complex, causing activation of NFκB signaling.

Animal models have also been generated in which the P392L human mutation has been knocked into the germ-line of mice. Mice carrying this mutation have been shown to exhibit increased osteoclastogenesis in 1 study [45] and in another to develop bone lesions with remarkable similarity to those found in human PDB [46].

Chromosome 7q33

A genomic region of approximately 350 kb at this locus was identified by an extended genome wide association study. The region is bounded by 2 recombination hotspots and contains 3 genes, NUP205, SLC13A,4, and CNOT4. Although none of these are known to play a role in bone metabolism CNOT4 has E3 ubiquitin ligase activity. Given the importance of the ubiquitin pathway in osteoclast function it is, therefore, plausible that CNOT4 might be the candidate gene in this locus. However, further work will be required to confirm or refute this hypothesis, to identify the causal variants and examine the molecular mechanisms responsible for the association observed.

TM7SF4

The TM7SF4 gene (also known as DCSTAMP) located on 8q22 encodes dendritic cell specific transmembrane protein (DC-STAMP). Common variants at this locus were identified as risk factors for Paget’s disease following a genome wide association study. The region of association extended for about 220Kb but the region of strongest association was within a 18Kb region encompassing the TM7SF4 gene. This gene is a strongest candidate for susceptibility to PDB since its product plays an essential role in promoting fusion of osteoclast precursors to form mature osteoclasts [47, 48]. The causal genetic variants remain to be identified but the associated SNP from the GWAS was strongly associated with TM7SF4 expression in monocytes suggesting that the causal variants most probably influence gene expression. Recently a rare coding variant causing an amino acid change from Leucine to Phenylalanine at position 397 (L397F) was reported in the French Canadian population that was associated with PDB [49]. Using exome sequencing, we have also found the L397F variant and another 2 novel coding variants in a small number of PDB cases (unpublished data). The functional effects of these variants remain to be investigated but these observations could be consistent with a model whereby predisposition to PDB is mediated not only by changes in levels of gene expression but also by rare coding variants.

TNFRSF11B

The TNFRSF11B gene located on 8q24 encodes osteoprotegerin, which is an endogenously produced inhibitor of osteoclast differentiation and activity. Mutations in TNFRSF11B have been identified as the cause of juvenile Paget’s disease (JPD) by family based studies. The first report was that of Whyte and colleagues who described 2 individuals with JPD with an identical deletion in chromosome 8q24 encompassing the whole of the TNFRSF11B gene [32]. Circulating levels of OPG were undetectable and markers of bone turnover greatly elevated. Around the same time linkage analysis in families with JPD had gained suggestive evidence of linkage to the 8q24 region and subsequent mutation screening in patients with JPD identified several homozygous loss of function mutations as the cause of the disease [33, 50]. Candidate gene association studies have suggested that common polymorphism at the TNFRSF11B locus might also predispose to PDB [51, 52] but this locus has not emerged as a candidate for predisposition to PDB in genome wide association studies [23••, 35••]. In view of the role of TNFRSF11B variants as a predisposing factor for classical PDB remains uncertain.

VCP

Mutations affecting the VCP gene located on chromosome 9p13 have been identified as the cause of the autosomal dominantly inherited syndrome of inclusion body myopathy, Paget’s disease and frontotemporal dementia [53]. The VCP gene encodes valosin-containing protein, which has been shown to regulate a variety of cellular processed including cell division, membrane fusion, stress responses, and protein degradation. The most common mutation is an arginine to histidine amino acid change at position 155 (R155H), although several other mutations have been described (reviewed by Metha [54]). Most of the mutations affect the CDC48 domain 2, which is located in the N-terminus of the protein. This is a substrate recognition domain, which binds polypeptides, prevents protein aggregation, and catalyzes refolding of permissive substrates. Although 9p13 did not emerge as a candidate locus for classical PDB either in linkage studies or GWAS studies, 2 groups of investigators have sought to determine if common variants in VCP were associated with PDB. In 1 study [55] mutation screening of VCP was conducted in 44 families with a history of PDB and an association study conducted in 197 sporadic PDB cases and a similar number of controls. No coding mutations were found in the families and there was no association between common VCP variants and PDB in the association study [56]. In another study of Belgian patients, a common SNP within VCP was associated with susceptibility to the disease [57]. In view of this, further studies to investigate the role of VCP as a predisposing gene for classical PDB would be of interest.

OPTN

The OPTN gene, which is located on 10p13 was implicated as a candidate gene for PDB by linkage analysis in families [22] and genome wide association studies [23••, 35••]. Genome wide linkage in UK families identified a candidate locus on 10p13 with the highest LOD score close to the position of OPTN and subsequently GWAS studies narrowed the region to a ~200Kb interval encompassing OPTN. The optineurin protein is involved in regulating multiple cellular processes relevant to the pathogenesis of PDB including autophagy and NFκB signaling although it has no known role in bone metabolism. Our preliminary data showed that knock down of optineurin in mouse bone marrow derived macrophages enhances osteoclast formation suggesting that optineurin acts as a negative regulator of osteoclast differentiation [58]. The most strongly associated SNP from the GWAS was associated with OPTN expression [23••] implying that the causal variant most probably affects regulatory elements involved in controlling transcription of OPTN. At the present time, the identity of the causal variant(s) remains to be discovered.

TNFRSF11A

The TNFRSF11A gene located on Chromosome 18q21 encodes receptor activator of nuclear factor kappa B (RANK), which plays a critical role in osteoclast differentiation and activity. Variants in the TNFRSF11A gene have been identified as genetic risk factors for PDB by studies in familial expansile osteolysis and associated disorders as well as by genome wide association studies in classical PDB. Mutations of TNFRSF11A were identified as a cause of familial expansile osteolysis (FEO) and early onset familial PDB by linkage analysis and positional cloning studies [29]. The causal mutations are duplications, which result in the addition of 6 or 9 amino acids to the signal peptide. This in turn prevents signal peptide cleavage and causes activation of RANK signaling by mechanisms that are still incompletely understood [59•]. Similar mutations were subsequently identified in expansile skeletal hyperplasia a related disorder [30]. Furthermore a distinct 12 bp duplication in the signal peptide region of the RANK gene was recently reported in a patient with a severe panostotic expansile bone disease associated with multiple bone deformities and a massive mandibular tumor, which was comprised of woven bone and fibrous tissue [60••]. The TNFRSF11A locus was also identified as predisposing to classical Paget’s by genome wide association studies. Although protein coding mutations similar to those implicated in FEO have not been identified in classical PDB, two relatively common coding variants have been identified (H141Y and V192A), which have been shown to activate NFκB signaling in vitro, but only when co-transfected with SQSTM1 mutations [61•]. Other researchers who investigated the effects of the H141Y and V192A coding variants in reporter assays could not find evidence of an effect on NFκB signaling [62].

RIN3

The RIN3 gene, which is located on 14q33, was implicated as a candidate gene for PDB by an extended genome wide association study [35••]. The strongest association lay within an interval of 62Kb flanked by recombination hot spots containing only the RIN3 gene. The RIN3 gene encodes Rab and Ras interactor protein 3, which acts as a guanine nucleotide exchange (GEF) for the Rab5 family of proteins including Rab5 itself and RAb31. The RIN family of proteins comprises RIN1, RIN2, and RIN3 [63]. They share in common several functional domains including SH2 and proline rich domains at the amino terminus and RH, VPS9, and RA domains at the C-terminus. Through these domains, members of the RIN family interact with various proteins and play a role in endocytosis, vesicular trafficking and signal transduction. The role of RIN3 in bone metabolism has not specifically been studied, although it is known that Rab proteins play a role in regulating osteoclast function through their effects on vesicular trafficking [64]. The causal variants and mechanism underlying the association remain to be discovered but unlike OPTN and TM7SF4, the associated variants were not associated with expression of RIN3, possibly implying an effect on the coding region.

Chromosome 15q24

A genomic region of approximately 200 kb at the 15q24 locus was identified by an extended genome wide association study. The region contains 2 genes (GOLGA6 and PML) but the strongest association lay within PML at a SNP which actually codes for a phenylalanine to leucine amino acid substitution at codon 645 (F645L). Neither PML nor GOLGA6 are known to play a role in bone metabolism so further work will need to be done to identify the causal variants and examine the molecular mechanisms responsible for the association observed.

Somatic Mutations

The focal nature of PDB has led to speculation that somatic mutations might be responsible for localization of the disease to specific bones and indeed this was first suggested as a mechanism of disease by one of the authors more than 20 years ago [65]. The issue of whether PDB is associated with somatic mutations has been investigated in 2 studies. In one study, 2 of 5 patients were reported to have a somatic P392L mutation in SQSTM1 in affected bone but not blood samples [66]. Similar analysis in 5 osteosarcomas from PDB patients showed that 3 tumors were positive for the P392L mutation whereas blood samples were negative. In both the bone samples and the tumor samples the mutant allele was present at much lower abundance than found in peripheral blood samples from patients heterozygous for the P392L mutation. This led the authors to conclude that there may be mosaicism for the mutation in affected tissue [66]. Another study which looked for the P392L mutation in osteoblasts and bone marrow cells cultured from affected bone showed no evidence of the P392L mutation in 28 samples from 23 patients [67]. Accordingly, the contribution of somatic mutations to the pathogenesis of PDB requires further study and needs to be investigated not only for SQSTM1 but also for other candidate genes and genomic regions.

Genetic Testing for PDB

In view of the fact that many of the genetic variants that predispose to PDB are of large effect size there is a prospect of using genetic testing to identify patients at risk of developing the disease or of developing complications [36•]. In the rare PDB-like syndromes of familial expansile osteolysis, early onset familial PDB, expansile skeletal hyperphosphatasia, and inclusion body myopathy, Paget’s disease and frontotemporal dementia predictive genetic testing is already employed in clinical practice. So far as the authors are aware, this has not yet been implemented for Juvenile PDB, but certainly could be given the high penetrance of the mutations. It has also been suggested that genetic testing should be considered in classical PDB with the aim of identifying carriers at an early stage before complications had developed [68]. This would most easily be implemented for SQSTM1 mutations which have a high penetrance and indeed a clinical trial is currently in progress (the ZiPP study, ISRCTN 11616770) in which children of patients with PDB are offered testing for SQSTM1 mutations and entered into a randomized trial of prophylactic zoledronic acid or placebo. It would also be possible to test for other susceptibility alleles with the aim of developing a genetic risk score for the disease [36•]. Further work would have to be carried out to evaluate how best to do this in clinical practice, but a possible starting point would be in children of patients who are diagnosis with PDB. This would allow carriers to be entered into a program of enhanced surveillance for early signs of disease and treatment where appropriate.

Conclusions

Huge advances have been made in understanding the genetic determinants of PDB over the past fifteen years. Current evidence suggests that genetic susceptibility to PDB is mediated by a combination of rare variants of large effect size and common variants of moderate effect size. These act in an additive manner to influence susceptibility to the disease and disease severity. Several questions remain unanswered however. One of the key issues is to gain better understanding of why the disease has such a curious focal distribution. A possible explanation for this would be somatic mutations and there is some evidence that this might occur in some patients, but the data are inconclusive. Another unanswered question is with regard to the changes in incidence and severity that have occurred over recent years. This suggests that key environmental trigger factors have altered in frequency over recent years but the nature of these factors and mechanisms by which they interact with the genes that predispose to PDB remain to be determined. From a clinical perspective, there is a real prospect that genetic testing for susceptibility could be employed to identify people at risk of the disease for enhanced surveillance and intervention although much work needs to be done to evaluate how such as program would best be implemented.