Novel UBA Domain Mutations of SQSTM1 in Paget's Disease of Bone: Genotype Phenotype Correlation, Functional Analysis, and Structural Consequences

Authors


Abstract

Three novel missense mutations of SQSTM1 were identified in familial PDB, all affecting the UBA domain. Functional and structural analysis showed that disease severity was related to the type of mutation but was unrelated to the polyubiquitin-binding properties of the mutant UBA domain peptides.

Introduction: Mutations affecting the ubiquitin-associated (UBA) domain of Sequestosome 1 (SQSTM1) gene have recently been identified as a common cause of familial Paget's disease of bone (PDB), but the mechanisms responsible are unclear. We identified three novel SQSTM1 mutations in PDB, conducted functional and structural analyses of all PDB-causing mutations, and studied the relationship between genotype and phenotype.

Materials and Methods: Mutation screening of the SQSTM1 gene was conducted in 70 kindreds with familial PDB. We characterized the effect of the mutations on structure of the UBA domain by protein NMR, studied the effects of the mutant UBA domains on ubiquitin binding, and looked at genotype-phenotype correlations.

Results and Conclusions: Three novel missense mutations affecting the SQSTM1 UBA domain were identified, including a missense mutation at codon 411 (G411S), a missense mutation at codon 404 (M404V), and a missense mutation at codon 425 (G425R). We also identified a deletion leading to a premature stop codon at 394 (L394X). None of the mutations were found in controls. Structural analysis showed that M404V and G425R involved residues on the hydrophobic surface patch implicated in ubiquitin binding, and consistent with this, the G425R and M404V mutants abolished the ability of mutant UBA domains to bind polyubiquitin chains. In contrast, the G411S and P392L mutants bound polyubiquitin chains normally. Genotype-phenotype analysis showed that patients with truncating mutations had more extensive PDB than those with missense mutations (bones involved = 6.05 ± 2.71 versus 3.45 ± 2.46; p < 0.0001). This work confirms the importance of UBA domain mutations of SQSTM1 as a cause of PDB but shows that there is no correlation between the ubiquitin-binding properties of the different mutant UBA domains and disease occurrence or extent. This indicates that the mechanism of action most probably involves an interaction between SQSTM1 and a hitherto unidentified protein that modulates bone turnover.

INTRODUCTION

PAGET'S DISEASE OF bone (PDB; MIM 167250, 602080) is a common disorder, affecting ∼3.1% individuals over 55 years of age in the United Kingdom.(1) The cause of PDB is incompletely understood, but genetic factors play an important role, reflected by the fact that between 15% and 40% of affected individuals have at least one affected first-degree relative.(2,3) Numerous extended families have been described where the disease is inherited in an autosomal dominant manner, and several susceptibility loci have been identified by genome-wide searches, including the PDB2 locus on chromosome 18q21(4–6); the PDB3 locus on chromosomes 5q35(7,8); the PDB4 locus on chromosome 5q31(7); the PDB5 locus on chromosome 2p36(8); the PDB6 locus on chromosome 10p13(8); and the PDB7 locus on chromosome 18q23.(9) A recurrent mutation affecting codon 392 (P392L) in the ubiquitin-associated (UBA) domain in the Sequestosome 1 (SQSTM1) gene was recently identified as a common cause of familial and sporadic PDB in the French-Canadian population,(10) and subsequently, we identified this and two other truncating mutations affecting the same region of SQSTM1 as an important cause of PDB in patients of British descent.(11) Recently, Johnson-Pais et al.(12) identified three other mutations in the UBA domain of SQSTM1 in patients with familial PDB from the United States. We now report the identification of further UBA domain mutations of SQSTM1 in familial PDB and the relationship between genotype and phenotype. We have also been able to map the mutations onto the tertiary structure of the UBA domain and directly investigate the effects of the mutations on SQSTM1 UBA domain function.

MATERIALS AND METHODS

Family recruitment, disease ascertainment, and disease extent score

We studied 70 families with a history of PDB who were recruited and ascertained as previously described.(8,13) The non-Pagetic controls and patients with sporadic PDB were characterized and recruited as described before.(11) The pedigree diagrams of families that were included in this study are shown on the bone research group website at the University of Aberdeen (www.abdn.ac.uk/medicine_therapeutics/bone/). Patients with sporadic PDB did not have a family history of the disease, but their relatives did not undergo any type of screening, so we cannot exclude the possibility that sporadic patients may have had relatives with asymptomatic disease. All subjects had given informed consent to being included in the studies, which were approved by the Grampian Research Ethics Committee and the Research Ethics Committees of the other contributing centers. For the phenotype-genotype comparison, we counted the number of bones involved as assessed by radionuclide bone scans. For the purpose of this assessment, the spine was considered as a single bone, because in many cases, spine involvement was recorded, but the number of vertebrae affected were not, and we were unable to obtain the original films for review. We also recorded age at diagnosis, but excluded patients whose diagnosis was made as the result of being included in the study. Comparison of disease extent and age at diagnosis between genotype groups was analyzed by ANOVA and Dunnet's test.

Mutation analysis of SQSTM1

We conducted mutation screening of the coding exons and intron-exon boundaries of SQSTM1 using PCR, followed by automated DNA sequencing as previously described.(11)

Functional and structural analysis of SQSTM1 UBA domain polypeptides

We obtained the wildtype SQSTM1 full-length cDNA as an IMAGE clone (2906264), amplified the region corresponding to the UBA domain by PCR, and cloned the resulting product in to the BamH1 and XhoI sites of plasmid pGEX-4T-1 (Amersham Biosciences). PDB-associated mutations were introduced by site-directed mutagenesis using the Quick Change kit (Stratagene). All constructs were verified by DNA sequencing. Wildtype and mutant recombinant proteins (GST-fusions) were prepared as described previously, using thrombin to release polypeptides equivalent to residues 387-436 of human SQSTM1 with an additional Gly-Ser dipeptide at the N terminus.(14) The cleaved proteins were subjected to a final purification step with a Superdex-200 gel-filtration column, dialyzed to remove excess salt, and verified by ESI-MS. For ubiquitin chain-binding assays, UBA polypeptides (1 mg/ml) were immobilized on CNBr-activated thiol-Sepharose 4B, and 20 μl of beads were incubated with Lys48-linked ubiquitin chain mixture (1 μg in 30 μl of 50 mM Tris, 0.1% [wt/vol] bovine serum albumin [BSA], pH 7.5; Affiniti Research Products) for 30 minutes at room temperature. After washing, bound (Bo) or unbound (Un) proteins were detected by Western blotting (rabbit anti-ubiquitin, 1:1000). Control beads contained no immobilized UBA polypeptide. NMR data were collected on 1- to 2-mM samples of UBA domain polypeptides as previously described.(14)

RESULTS

Mutation screening of affected individuals with familial PDB resulted in the identification of three novel mutations in the SQSTM1 gene. All were found in exon 8 of SQSTM1 and affect the UBA domain. In one family, we identified a T deletion at position +1210 of the cDNA sequence (NM_003900), which causes a frameshift, introducing a stop codon in place of leucine at codon 394 (394X). This mutation was not found in sporadic PDB cases or controls. The family affected had previously been reported in error to have a T insertion mutation at +1225, causing the E396X mutation.(11) Another A/G mutation was identified in 4/70 families (5.7%) at position 1250 (A1250G), which predicts a methionine to valine substitution at codon 404 (M404V). This mutation was found in 1/175 subject (0.6%) with sporadic PDB but was absent in 155 age- and sex-matched non-Pagetic controls. Another G/A mutation was identified in 3/70 (4.2%) of families at position +1271 (G1271A), which predicts a glycine to serine substitution at codon 411 (G411S). This mutation was not found in subjects with sporadic PDB or in controls. Another G/A mutation was identified in 1/70 family (1.4%) at position 1313 (G1313A), which predicts a glycine to arginine substitution at codon 425 (G425R). This mutation was found in 1/175 subjects (0.6%) with sporadic PDB but was absent in 155 age- and sex-matched non-Pagetic controls. All four of these mutations segregated with the disease in affected individuals from each of the families studied. These observations, when combined with our previous work,(11) brings the total number of families in which PDB is associated with SQSTM1 UBA domain mutations to 26 of 70 (37.1%; Table 1).

Table Table 1.. Details of Families With SQSTM1 Mutations
original image

We were able to map the different mutations onto the ribbon structure of the SQSTM1 UBA domain (Fig. 1A), which we recently determined using 2D protein NMR.(14) The UBA domain of SQSTM1 consists of three anti-parallel α helices, beginning at residue 392, which are connected by short interhelical loops. The truncating mutations of SQSTM1 found in PDB (A390X, L394X, and E396X) therefore delete all or most of the UBA domain. P392 is the first residue of helix 1 and acts as an N-terminal capping residue.(14) M404 is found in the loop between helices 1 and 2 and forms part of the hydrophobic patch previously implicated in ubiquitin-binding by other UBA domains.(15) G411 is in the same loop at the N terminus of helix 2. G425 is located in helix 3 and also resides in the hydrophobic patch (Fig. 1B).

Figure FIG. 1..

Paget's disease-causing mutations of SQSTM1 in relation to UBA domain structure. (A) Sequencing traces for novel PDB-causing mutations of SQSTM1 identified in this study are shown in relation to the ribbon structure of the UBA domain. In each panel, the mutant trace is shown above the wildtype. For the missense mutations, the affected codons are underlined. The premature stop codon resulting from the delT-1210 mutation is indicated. (B) Surface representation of the SQSTM1 UBA domain showing the hydrophobic patch (white) with M404 and G425 residues within the patch, and G411 and P392 mutations not within the patch.

We analyzed the effect of the different mutations on polyubiquitin chain-binding by the isolated UBA domain (Fig. 2). Because the E396X mutation deletes most of the UBA domain and ablates ubiquitin chain-binding in the SQSTM1 holoprotein,(14) a similar effect is predicted for the A390X and L394X mutations. We have already shown that the P392L missense mutation modifies the secondary structure of the UBA domain, but that ubiquitin chain-binding is not affected.(14) Analysis of the new mutations showed that the M404V and G425R substitutions virtually abolish the ability of the UBA domain to bind ubiquitin chains, whereas the G411S substitution binds ubiquitin chains comparable with the wildtype and P392L sequence.

Figure FIG. 2..

Effects of PDB substitution mutations on ubiquitin-binding properties of the SQSTM1 UBA domain. UBA domain polypeptides (1 mg/ml) representing the wildtype (WT) and mutant SQSTM1 sequences were immobilized on CNBr-activated thiol-Sepharose 4B and incubated with a Lys48-linked multi-ubiquitin chain mixture (1 μg ubiquitin chains mixed with 20 μl beads). Bound proteins retained on beads (Bo) or in the unbound fraction (Un) were detected by Western blotting (anti-ubiquitin). The number of ubiquitins in each chain is indicated.

We investigated the effects of these mutations on SQSTM1 UBA domain structure using NMR. Analysis of 1D proton NMR spectra of the UBA domain polypeptides showed that the M404V, G411S, and G425R mutants all formed native-like folded structures (Fig. 3). This was evident from the aliphatic methyl region of the spectra, where resonances were widely dispersed as a consequence of core hydrophobic packing interactions with aromatic residue side chains, indicative of a compact globular structure. The spectral dispersion was similar for the wildtype and each of the mutants, indicating that the substitutions are not significantly affecting the overall tertiary structure of the UBA domain. The M404V mutation, while maintaining surface hydrophobicity, seems to modify the putative ubiquitin-binding van der Waals surface. Substitution of the extended side chain of methionine for the shorter β-branched side chain of valine produces a small surface cavity that may be sufficient to reduce ubiquitin-binding affinity through a subtle perturbation to the surface complementarity necessary for protein-protein recognition. The loss of ubiquitin chain-binding in the G425R mutant can be attributed to the less subtle substitution of a glycine on the same hydrophobic ubiquitin-binding patch of the UBA domain for a highly polar arginine side chain.

Figure FIG. 3..

NMR spectra showing folding of UBA domains. 1D proton NMR spectra (600 MHz) at 298K of UBA-wt, UBA-G425R, UBA-M404V, and UBA-G411S showing the 0.0-1.0 ppm aliphatic methyl region. Resonances are characteristic of a compact folded structure involving packing of aliphatic and aromatic side chains and show that all mutants have a native-like structure. The peak at 0.0 ppm corresponds to the trimethylsilylpropionate (TSP) reference peak.

We went on to determine whether the type of mutation influences phenotype, by comparing age at diagnosis and number of involved sites in affected subjects from all 70 families who were categorized according to the type of mutation. The results of this analysis are shown in Table 2. Of the 70 kindreds with familial PDB studied, 26 (37.1%) carried SQSTM1 mutations, all of which cluster in the UBA domain of the gene product. In 5 families (7.1%), truncating mutations were observed that are predicted to delete virtually the entire UBA domain, whereas 21 families (30%) had missense mutations affecting the UBA domain. We did not detect SQSTM1 mutations in 44 families (62%). Overall, we found that familial PDB associated with SQSTM1 mutations was more extensive and had a significantly earlier age at diagnosis than PDB in which SQSTM1 mutations were not detected (p < 0.01; Table 2). Although we had limited statistical power to compare the phenotypic effects of individual mutations, there was a clear trend for increased extent of disease in subjects with truncating as opposed to missense mutations, and when subjects were grouped on this basis, those with truncating mutation had significantly more extensive disease than those with missense mutations (bones involved = 6.0 ± 2.7 versus 3.4 ± 2.4; p < 0.0001). Subjects with truncating mutations had a slightly earlier age at diagnosis, but the difference was not significant. There was also a significant difference in extent of disease when mutations were grouped according to whether or not they affected ubiquitin binding, although the difference was not so great. Thus, the disease extent in subjects with mutations which inhibited ubiquitin binding (i.e., truncating mutations, M404V, G425R) was 4.9 ± 2.8 compared with 3.75 ± 2.67 in the group where the mutations did not inhibit ubiquitin binding (i.e., G411S and G4256R; p = 0.03).

Table Table 2.. Phenotype-Genotype Analysis in Familial PDB
original image

DISCUSSION

PDB is a common condition with a strong genetic component. While PDB is genetically heterogeneous,(5,7,8,16) recent studies have shown that mutations affecting the UBA domain of SQSTM1 are an important cause of the disease.(10,11) A recurrent proline-leucine mutation affecting codon 392 in the UBA domain of SQSTM1 (P392L) was first identified as a cause of PDB by Laurin et al.(10) in a French Canadian population, and we subsequently identified the P392L mutations and two other truncating mutations of the SQSTM1 UBA domain (390X and E396X) in Pagetic patients of predominantly British descent.(11) More recently, three further mutations have been described in the UBA domain of SQSTM1 by Johnson-Pais et al.(12) Two of these resulted in truncation of the UBA domain at codon 394, and one resulted in a proline-leucine amino acid change at codon 387 (P387L). In this study, we identified three novel missense mutations affecting the UBA domain of SQSTM1 in familial PDB (M404V, G411S, and G425R) and identified another truncating mutation (394X) caused by a T-insertion at position 1210, identical to that reported by Johnson-Pais et al.(12) In all cases, the mutations segregated with the disease in affected family members and were absent from a large number of control chromosomes. Two of the missense mutations were also found in patients with “sporadic” PDB, although we cannot exclude the possibility that relatives of these individuals may have had undiagnosed, asymptomatic PDB.

The availability of the NMR structure of the SQSTM1 UBA domain(14) allowed us to precisely map the novel mutations onto the 3D structure of the protein and to determine if individual residues affected specific functional motifs within the UBA domain. We have previously reported that the P392L mutant modifies the secondary structure of the UBA domain by extending the N terminus of helix 1 and that the truncating mutations result in deletion of most of the UBA domain.(14) None of the missense mutations were found to affect the folding of the UBA domain. The M404V and G425R mutations are predicted to affect the hydrophobic patch of the UBA domain that binds ubiquitin in subtly different ways by either modifying the van der Waals surface contours (M404V) or by placing a highly polar side chain in the middle of the hydrophobic patch (G425R) consistent with the loss of ubiquitin chain-binding we observed in our functional assays. The G411S mutation is more remote from the hydrophobic patch and does not perturb ubiquitin chain-binding. Clustering of the PDB-causing mutations in the UBA domain of SQSTM1 raises the possibility that the mutant proteins cause PDB by affecting the ability of SQSTM1 to bind ubiquitin. Although we observed a more severe phenotype in subjects with truncating mutations (that resulted in loss of ubiquitin chain-binding), there was no clear correlation between the ability of the mutant UBA domains to bind polyubiquitin and presence or extent of PDB. For example, polyubiquitin chain-binding was virtually abolished by the G425R and M404V mutants, yet disease extent and age at diagnosis in patients with these mutations was similar to that in those who carried the P392L and G411S mutations, which cause PDB in the absence of an effect on polyubiquitin chain-binding.

While the precise mechanisms by which SQSTM1 UBA domain mutations cause PDB remain to be determined, the data presented here show that the propensity of SQSTM1 mutations to cause PDB is not simply attributable to the polyubiquitin binding properties of the mutant UBA domain. It is possible that the mutations cause selective loss of binding to a specific ubiquitylated substrate (as opposed to generalized loss of ubiquitin-binding), because the ubiquitin-binding experiments were conducted with unanchored ubiquitin chains. Another possibility is that the mutations affect the overall structure of the SQSTM1 holoprotein, thereby affecting its half-life or other protein-protein interactions. A third possibility is that the SQSTM1 UBA domain interacts with non-ubiquitylated substrates that affect bone cell activity and that the mutations in the UBA domain cause a loss of this interaction. It is important to emphasize that environmental factors have also been implicated in the pathogenesis of PDB, and most of the attention has focused on the possibility that PDB arises as the result of a chronic infection of osteoclast precursors with paramyxoviruses.(17) While the role of viruses in PDB remains controversial,(18) it is possible that viral proteins might interact with the mutant forms of SQSTM1 to stimulate osteoclast formation. Even if this were to be the case, the functional studies that we have performed clearly show that SQSTM1 mutations can predispose to the disease without affecting the polyubiquitin-binding properties of the UBA domain.

Whatever the underlying mechanism, this work confirms the importance of UBA domain-specific mutations of SQSTM1 as a cause of PDB. Our studies also indicate that the mechanism of action is not attributable to loss of polyubiquitin binding by the UBA domain, but rather, may involve an interaction between SQSTM1 and a hitherto unidentified protein(s) that modulates bone turnover.

Acknowledgements

This study was supported by grants from the Arthritis Research Campaign (SHR, LJH); the Wellcome Trust (RL); the BBSRC (MSS); the National Association for Relief of Paget's Disease, and the Paget's Disease Charitable Trust (Auckland, New Zealand). We also thank the Arthritis Research Campaign and Medical Research Council for infrastructure support (arc ICAC grant and MRC cooperative group grant). We acknowledge the technical contribution of Angela Duthie, Sheena Main, and Grace Taylor to the DNA extraction and DNA sequencing. We are grateful to the School of Chemistry at the University of Nottingham and EPSRC for supporting and funding NMR facilities.

Ancillary