Structure of the Ubiquitin-associated Domain of p62 (SQSTM1) and Implications for Mutations That Cause Paget's Disease of Bone*

The p62 protein (also known as SQSTM1) mediates diverse cellular functions including control of NFκB signaling and transcriptional activation. p62 binds non-covalently to ubiquitin and co-localizes with ubiquitylated inclusions in a number of human protein aggregation diseases. Mutations in the gene encoding p62 cause Paget's disease of bone (PDB), a common disorder of the elderly characterized by excessive bone resorption and formation. All of the p62 PDB mutations identified to date cluster within the C-terminal region of the protein, which shows low sequence identity to previously characterized ubiquitin-associated (UBA) domains. We report the first NMR structure of a recombinant polypeptide that contains the C-terminal UBA domain of the human p62 protein (residues 387–436). This sequence, which confers multiubiquitin chain binding, forms a compact three-helix bundle with a structure analogous to the UBA domains of HHR23A but with differences in the loop regions connecting helices that may be involved in binding accessory proteins. We show that the Pro392 → Leu PDB substitution mutation modifies the structure of the UBA domain by extending the N terminus of helix 1. In contrast to the p62 PDB deletion mutations that remove the UBA domain and ablate multiubiquitin chain binding, the Pro392 → Leu substitution does not affect interaction of the UBA domain with multiubiquitin chains. Thus, phenotypically identical substitution and deletion mutations do not appear to predispose to PDB through a mechanism dependent on a common loss of ubiquitin chain binding by p62.

The ubiquitin-binding protein p62 (SQSTM1) functions as a scaffold in a range of signaling pathways associated with cell stress, survival, and inflammation and also controls transcriptional activation and protein recruitment to endosomes (1). p62 co-localizes with ubiquitin-containing inclusions in the brain tissue of Alzheimer's, Pick's, and Parkinson's disease patients, suggesting a possible role in the formation of these lesions (2). Recently, mutations in the gene encoding p62 have been found to cause Paget's disease of bone (PDB) 1 (3,4), a common disorder of the elderly characterized by excessive bone resorption and formation (5). All of the p62 mutations identified to date (Pro 392 3 Leu, Glu 396 3 Stop, and an intron 7 splice donor site mutation predicted to generate a frameshift 390Stop protein) cluster within the C-terminal region of the protein (Fig. 1a). This region shows low sequence identity to previously characterized ubiquitin-associated (UBA) domains (6,7) which occur in enzymes of the ubiquitin conjugation/deconjugation pathway (6), as well as in proteins that are downstream regulators of ubiquitin-dependent proteolysis (8). Consistent with the ability of other UBA domains to bind non-covalently to ubiquitin and multiubiquitin chains (9 -12), the C-terminal region of the p62 protein is also reported to have these properties (13,14). We report the first structure of a recombinant polypeptide containing the UBA domain of the human p62 protein (residues 387-436; Fig. 1b) and NMR investigations of the Pro 392 3 Leu mutant (the most common p62 PDB mutation). This structural model allows us to rationalize the differential effects on ubiquitin binding of p62 PDB deletion and substitution mutations detected in parallel in vitro ubiquitin chain binding assays.

EXPERIMENTAL PROCEDURES
Expression and Isolation of p62 UBA and Mutants-The cDNA containing the UBA domain of human p62 was amplified from IMAGE clone 2906264 by PCR and cloned into the BamHI and XhoI sites of plasmid pGEX-4T-1 (Amersham Biosciences). The Pro 392 3 Leu PDB mutation was introduced by site-directed mutagenesis (Quick-Change Mutagenesis kit, Stratagene), and constructs were verified by sequencing. Glutathione S-transferase fusion proteins were expressed in XL-10 Gold Escherichia coli (Stratagene). Protein purification involved growing transformed cells at 37°C until the absorbance at 600 nm reached 0.6 to 0.7, and induction of overexpression was achieved by adding isopropyl-1-thio-␤-D-galactopyranoside to a final concentration of 200 M. Cleavage of purified glutathione S-transferase fusion proteins retained on glutathione-Sepharose beads with thrombin released polypeptides equivalent to residues 387-436 of human p62 with an additional Gly-Ser dipeptide at the N terminus. The cleaved proteins were subjected to a final purification step with a Superdex-200 gel filtration column and dialyzed to remove excess salt.
NMR Spectroscopy and Structure Calculation-NMR data were collected on 2-mM samples of the wild-type p62-UBA domain and the Pro 392 3 Leu mutant at pH 5.8. Two-dimensional TOCSY, DQF-COSY, and 30-, 50-, 100-, 250-, and 300-ms two-dimensional NOESY experiments were performed at 25 and 35°C on a Bruker Avance-600 instrument equipped with a triple-resonance probe with solvent suppression achieved using the WATERGATE sequence (15). Data were processed using Bruker XWINNMR (version 2.5) and ANSIG software (16). NOE restraints from NOESY spectra at 25 and 35°C were classified as weak * This work was supported by grants from the Wellcome Trust (to R. L.) and the Biotechnology and Biological Sciences Research Council (to M. S. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The (1.8 -5.0 Å), medium (1.8 -4.0 Å), or strong (1.8 -3.0 Å) on the basis of cross-peak intensities. A family of 100 structures was calculated using the simulated annealing protocol starting from the template coordinates as implemented in XPLOR version 3.1 (17) using 732 NOE restraints and 56 dihedral angle restraints derived from H␣ chemical shift data and from H␣-H␤ splittings in TOCSY and DQF-COSY for residues Asp 408 , Asn 421 , and Asp 423 . The NOE-derived distances were treated as r Ϫ6 summation averages (sum) using a soft potential function with a force constant of 50 kcal mol Ϫ1 Å Ϫ2 . The simulated annealing procedure started from extended coordinates and consisted of 35 ps of high temperature dynamics at 2000 K followed by a slow cooling to 100 K in 35 ps (18). For the final minimization, the structures were subjected to a 500 steps of restrained Powell energy minimization, and 27 final structures were selected, which exhibit good geometry with no NOE violations Ͼ0.5 Å and no dihedral violations Ͼ5°. Analysis of the , distribution within the structured region of the ensemble (residues 392-431) showed 97% of residues to lie in favored and additionally allowed regions and none in disallowed regions of Ramachandran space. The RMSD from the mean structure (residues 392-431) is 0.51 Å (backbone heavy atoms) and 0.95 Å over all heavy atoms. The coordinates have been submitted to the Protein Data Bank (code 1Q02). The quality of the structures was assessed using the programs PROCHECK and AQUA (19).
Ubiquitin Chain Binding Assay-UBA polypeptides (wild-type and Pro 392 3 Leu, 1 mg/ml) were immobilized on CNBr-activated thiol-Sepharose 4B and incubated with a Lys 48 -linked multiubiquitin chain mixture (Affiniti Research Products Ltd.). Ubiquitin chains (1 g in 30 l of 50 mM Tris, 0.1% (w/v) bovine serum albumin, pH 7.5) were mixed with 20 l of beads for 30 min at room temperature. After washing, bound proteins retained on beads or in the unbound fraction were detected by Western blotting (rabbit anti-ubiquitin, 1:1000). Control beads contained no immobilized UBA polypeptide.

RESULTS
The one-dimensional 1 H NMR spectrum of the p62-UBA recombinant polypeptide (residues 387-436) has a widely dispersed amide and methyl region consistent with a compact folded structure which CD spectra indicate has a high helical content (ϳ55%). NMR resonance assignments and secondary structure were determined from homonuclear 1 H NMR TOCSY and NOESY experiments following well established procedures (20). The pattern and intensity of sequential NOEs and TOCSY cross-peak intensities establish that the p62-UBA domain is helix-rich. Structure calculations based on 735 NOE and 56 dihedral angle restraints show that the structure is well defined between residues Pro 392 and Ile 431 (see Table I); however, the short N-and C-terminal regions are much more dynamic and show no signs of stable secondary structure (Fig. 2).
The polypeptide forms a compact three-helix bundle (Fig.  3a), with a helical alignment analogous to that found for the UBA domains of HHR23A (21,22) (Fig. 3b), a highly conserved protein that functions in nucleotide excision repair. The RMSD between the ␣-helical backbone C␣ atoms of the UBA (2) domain of HHR23A and that of p62 is 1.7 Å indicating a common helical alignment. Although the residues forming the hydro-phobic core are mostly conserved, the major differences between the p62-UBA and HHR23A UBA (2) domain are in the loops (L1 and L2) connecting the helices (Fig. 3b). In the p62-UBA, Glu 409 and Gly 410 extend loop 1 by 2 residues forming a turn partially stabilized by interactions between the side chains of Asp 408 /Glu 409 and residues in helix 2, as clearly evident from the NOE data. The conserved Gly-Phe-Xaa se-   quence in loop 1 (Gly-Phe-Pro in HHR23A-UBA (2) and Gly-Phe-Ser in the p62-UBA), which binds HIV-1 Vpr in HHR23A (23), is still present in the p62-UBA domain suggesting a possible role for this region in binding accessory proteins. The side chain of Phe 406 within loop 1 is, however, partially buried in the hydrophobic core suggesting that it may have a conserved role in structure stabilization and presentation of the GFX motif. The UBA domain of p62, like the UBAs of HHR23A, is stabilized by a well defined hydrophobic core. Despite the low sequence homology between the two UBA domains (Fig. 1b), both have a conserved hydrophobic surface patch previously implicated in ubiquitin binding (21).
Based on our structural model, it is evident that the Glu 396 3 Stop and 390Stop PDB mutations cause deletion of most, or loss of all, of the UBA domain, respectively. The structure also unequivocally places Pro 392 at a key position within the UBA domain acting as the N-terminal capping residue of helix 1 (Fig.  3a). To investigate the structural consequences of the Pro 392 3 Leu mutation, we have characterized by NMR the p62-UBA domain containing this substitution. Analysis of deviations of H␣ chemical shifts (Fig. 4a) and intensities of NH-NH sequential NOEs shows that the Pro 392 3 Leu mutation suggests an extension of the N terminus of helix 1 moving the N-cap backward toward Pro 388 .
By using an in vitro ubiquitin chain binding assay with immobilized recombinant polypeptides, we have confirmed that residues 387-436 of the p62 protein (identical to those used in our structural analysis) confer multiubiquitin chain binding (Fig. 4b). In these experiments we found that the p62-UBA domain binds Lys 48 -linked multiubiquitin chains containing a minimum of two ubiquitins, but we did not detect an interaction with monomeric ubiquitin. Additionally, we did not detect binding of monomeric ubiquitin to the p62-UBA domain in the absence of multiubiquitin chains. We attribute our failure to detect monomeric ubiquitin binding, despite previous reports that p62 binds to (monomeric) ubiquitin-Sepharose (14), to either the relatively low affinity of UBA domains for the monomer (10) or the presence of a second uncharacterized monomeric ubiquitin-binding site in the p62 holoprotein (14). Be-cause the Glu 396 3 Stop PDB mutation that deletes most of the UBA domain ablates multiubiquitin chain binding in the holoprotein (not shown), a similar effect is predicted for the 390Stop mutation. To determine whether the Pro 392 3 Leu mutation similarly causes loss of multiubiquitin chain binding, we assessed the functional consequences of this substitution in the p62-UBA domain. In contrast to the PDB deletion mutations, the Pro 392 3 Leu substitution did not affect either multiubiquitin chain binding in the isolated UBA (Fig. 4) or p62 holoprotein (not shown), suggesting that neither Pro 392 nor residues immediately preceding are critical for multiubiquitin chain binding per se.
Consistent with the above findings, analysis of our structural model shows that Pro 392 does not form part of the hydrophobic surface patch previously implicated in ubiquitin binding in other UBA domains (21). Additionally, residues that are involved in the modification of secondary structure induced by the Pro 392 3 Leu substitution (Glu 389 , Ala 390 , and Asp 391 ) are not included in this proposed ubiquitin-interacting surface. DISCUSSION We have determined the first NMR structure of the p62-UBA domain and shown that the Pro 392 3 Leu substitution mutation, which causes PDB, modifies secondary structure by extending the N terminus of helix 1. The functional significance of the UBA domain in the p62 protein is currently not known; however, it is able to bind unanchored Lys 48 -linked multiubiquitin chains (13), and as evident in Fig. 4b, which when attached to target proteins, are proteasomal degradation signals (24). Additionally, p62 binds to multiubiquitinated proteins that accumulate when proteasome degradative function is impaired (13,25) further suggesting a role for p62 in the regulation of ubiquitin-dependent proteolysis as is the case for other UBA domain proteins (26). We find that in contrast to the loss of multiubiquitin chain binding that results from PDB deletion mutations, modification of secondary structure by the Pro 392 3 Leu substitution does not affect interaction of the UBA domain with Lys 48 -linked multiubiquitin chains. The observation that deletion and substitution mutations are phenotypically identi- cal is therefore strongly suggestive that UBA domain mutations do not predispose to PDB through a mechanism related to loss of regulation of ubiquitin-mediated proteolysis dependent on a common loss of ubiquitin chain binding by p62.
We cannot however discount the possibility that the modification of secondary structure in the N-terminal portion of helix 1 induced by the Pro 392 3 Leu mutation causes a selective loss of binding of a key ubiquitylated p62 substrate (as opposed to a generalized loss of ubiquitylated p62 substrate binding by the deletion mutations), because our binding analyses were conducted using only unanchored multiubiquitin chains. It will also be of interest to determine the ubiquitin chain specificity of p62 (ubiquitin chains may also be linked via Lys 29 and Lys 63 (27)) and investigate the effects of the Pro 392 3 Leu substitution on p62-UBA binding to these species. Additionally, it is also possible that like the UBA (2) of HHR23A which binds directly to HIV-1 Vpr (23), the p62-UBA domain also interacts with non-ubiquitylated substrates, and PDB deletion and substitution mutations in the UBA domain cause a common loss of interaction with such proteins leading to altered osteoclast activity. Finally, the lack of well defined secondary structure in the N-terminal sequence preceding helix 1 (residues 387-391; Fig. 1b) might suggest that this portion of the polypeptide chain constitutes a flexible linker between the PEST and UBA domains in p62. In this context, it is possible that PDB mutations affect the half-life of p62, for example by exposing PEST sequences that render the protein more susceptible to ubiquitindependent proteolysis (28).
Regardless of the precise molecular mechanism by which p62-UBA domain mutations contribute to PDB pathogenesis, it is tempting to speculate that functionally these mutations result in activation of NFB signaling. Such a proposal is based on the observation that signal peptide mutations in receptor activator of NFB, which lead to an increase in NFB signaling, cause the PDB-like condition familial expansile osteolysis (29). Establishing the precise in vivo function of the p62 protein in bone metabolism, and in particular the identification of ubiquitylated and non-ubiquitylated UBA domain-interacting proteins, should provide further insights into the pathogenesis of both familial and sporadic forms of PDB. FIG. 4. a, deviation of H␣ chemical shifts from random coil values in the N-terminal sequence of wild-type p62-UBA domain and the Pro 392 3 Leu mutant (data at 35°C, pH 5.8). b, ubiquitin chain binding assay for the p62-UBA domain. UBA polypeptides (wild-type and Pro 392 3 Leu, 1 mg/ml) were immobilized on CNBr-activated thiol-Sepharose 4B and incubated with a Lys 48 -linked multiubiquitin chain mixture (1 g of ubiquitin chains mixed with 20 l of beads). Bound (Bo) proteins retained on beads or in the unbound fraction (Un) were detected by Western blotting (anti-ubiquitin). Control beads contained no immobilized UBA polypeptide. The number of ubiquitins in each chain is indicated.