Dimerisation of the UBA Domain of P62 Inhibits Ubiquitin Binding and Regulates NF-Κb Signalling

Author's personal copy

doi:10.1016/j.jmb.2009.11.032 J. Mol. Biol. (2010) 396, 178–194

Available online at www.sciencedirect.com

Dimerisation of the UBA Domain of p62 Inhibits Ubiquitin Binding and Regulates NF-κB Signalling

Jed Long1,2, Thomas P. Garner1,2, Maya J. Pandya3, C. Jeremy Craven3, Ping Chen1,2, Barry Shaw4, Michael P. Williamson3, Robert Layfield4 and Mark S. Searle1,2⁎

1Centre for Biomolecular The ubiquitin (Ub)-binding p62 scaffold protein (encoded by the SQSTM1 Sciences, University Park, gene) regulates a diverse range of signalling pathways leading to activation Nottingham NG7 2RD, UK of the nuclear factor kappa B (NF-κB) family of transcription factors and is an important regulator of macroautophagy. Mutations within the gene 2School of Chemistry, encoding p62 are commonly found in patients with Paget's disease of bone University Park, and largely cluster within the C-terminal ubiquitin-associated (UBA) Nottingham NG7 2RD, UK domain, impairing its ability to bind Ub, resulting in dysregulated NF-κB 3Department of Molecular signalling. However, precisely how Ub-binding is regulated at the Biology and Biotechnology, molecular level is unclear. NMR relaxation dispersion experiments, coupled Western Bank, with concentration-dependent NMR, CD, isothermal titration calorimetry University of Sheffield, and fluorescence kinetic measurements, reveal that the p62 UBA domain ∼ − μ Sheffield S10 2TN, UK forms a highly stable dimer (Kdim 4 12 M at 298 K). NMR analysis shows 4 that the dimer interface partially occludes the Ub-binding surface, School of Biomedical Sciences, particularly at the C-terminus of helix 3, making UBA dimerisation and University of Nottingham, Ub-binding mutually exclusive processes. Somewhat unusually, the Queen's Medical Centre, monomeric UBA appears to be the biologically active form and the dimer Nottingham NG7 2UH, UK appears to be the inactive one. Engineered point mutations in loop 1 (E409K Received 7 September 2009; and G410K) are shown to destabilise the dimer interface, lead to a higher received in revised form proportion of the bound monomer and, in NF-κB luciferase reporter assays, 11 November 2009; are associated with reduced NF-κB activity compared with wt-p62. accepted 12 November 2009 © 2009 Elsevier Ltd. All rights reserved. Available online 17 November 2009 Keywords: ubiquitin; ubiquitin-associated domain; Paget's disease of bone; Edited by A. G. Palmer III NMR structural analysis; UBA dimerisation

Introduction More recently, p62 has been shown to be a critical regulator of the degradation of ubiquitinated pro- 3,4 The p62 protein (encoded by the SQSTM1 gene) teins by macroautophagy. The p62 protein has a domain structure consistent with its participation in functions as a ubiquitin (Ub)-binding scaffold, 5 which regulates a diverse range of signalling path- multiple signalling complexes, including a C- terminal Ub-associated (UBA) domain through ways leading to activation of the nuclear factor 6,7 kappa B (NF-κB) family of transcription factors.1,2 which p62 binds non-covalently to Ub. An important role of p62 is in the control of induced osteoclastogenesis,8 a process dependent on p62's *Corresponding author. E-mail address: ability to regulate RANK-mediated NF-κB signal- [email protected]. ling, probably through the K63-linked ubiquitination Abbreviations used: UBA domain, ubiquitin-associated of TRAF6 and other signalling intermediates. Ac- domain; mUBA, monomeric UBA; Ub, ubiquitin; mUb, cordingly, mutations affecting the SQSTM1 gene are monomeric ubiquitin; NF-κB, nuclear factor kappa B; commonly found in patients with the skeletal PDB, Paget's disease of bone; CPMG, Carr–Purcell– disorder Paget's disease of bone (PDB), a chronic Meiboom–Gill; HSQC, heteronuclear single quantum condition characterised by excessive bone turnover coherence; CSP, chemical shift perturbation; NOE, nuclear leading to bone expansion, structural weakness, Overhauser enhancement; NOESY, nuclear Overhauser deformity and pain.9 The vast majority of the PDB- enhancement spectroscopy; GST, glutathione S-transferase. associated mutations identified to date cluster within

p62 UBA Domain Dimerisation Inhibits Ub-Binding 179 the UBA domain, impairing p62's ability to bind Ub p62 UBA domain forms a highly stable symmetrical and resulting in dysregulated NF-κB signalling.10,11 dimer. We show that the Ub-binding surface of the The structure–function relationship between the monomeric UBA and the dimer interface partially sites of PDB-associated mutations in the p62-UBA overlap such that dimerisation and Ub-binding are domain and their effects on protein folding, mutually exclusive events, with the former strongly Ub-recognition and dysregulation of NF-κB signal- modulating the latter. Somewhat unusually, the ling are still largely unclear, as are the mechanisms monomeric UBA is the biologically active form, that regulate normal Ub-recognition. Human p62 is whilst the dimer is inactive with a role in regulating a 440-residue protein, and the isolated C-terminal levels of signal activation. We demonstrate this polypeptide sequence (residues 387–436) forms a using a number of point mutations in loop 1, which compact three-helix bundle,7 with close structural destabilise the dimer interface and lead to a higher – homology to other UBA and CUE domains.12 20 All proportion of Ub-bound UBA domain than ob- have been shown to bind non-covalently to mono- served for wt-UBA. The effects of these mutations on meric Ub (mUb) through a common hydrophobic NF-κB signalling, assessed in luciferase reporter patch on the β-sheet surface of Ub.21,22 In contrast, assays, support the role of dimer formation in the UBA domains, which have low sequence modulating the levels of signal activation. homology, show diverse modes of interaction with wide-ranging binding affinities and specificities for both a single Ub motif and for the more compact 23 Results K48-linked Ub2, where the UBA domain is engaged simultaneously with binding surfaces on each of the Ub motifs.17,19 Biophysical analysis supports formation of a Although the majority of UBA domains appear to p62 UBA dimer bind Ub as monomers, the UBA-related CUE domain of Vps9p recognises Ub through an Titrations of p62-UBA with Ub at NMR concen- asymmetric domain-swapped dimer.24 More re- trations of protein (0.2−1 mM) have revealed a cently, dimers have been characterised for the binding mechanism involving a UBA conformer UBA domains of c-Cbl and Cbl-b involving hydro- that is sparsely populated in the absence of Ub, phobic surface interactions at the interface of helices leading to a complex binding process involving both – 2 and 3.25 27 The role of c-Cbl and Cbl-b as scaffold slow- and fast-exchange steps.29 The minor con- proteins, coupled with their capacity to interact former could not be observed directly in hetero- with different downstream signalling proteins, nuclear single quantum coherence (HSQC) spectra; suggests that dimerisation provides a mechanism however, NMR transverse relaxation techniques are for an expanded repertoire of Cbl-dependent suited to detecting the effects of low populations of signals recruited to receptor-tyrosine kinases. Al- different species at equilibrium that are in chemical though, in the case of the c-Cbl homodimer, Ub exchange.31–37 15N relaxation dispersion data col- recognition does not appear to be its primary lected with 0.8 mM p62 UBA at 600 MHz showed function,25 the dimerisation and activation of the strong dispersion effects at 310 K, although these Ub protein ligase Cbl-b are Ub-mediated.26 Within were less evident at lower temperatures where the context of the yeast proteins Rad23 and Ddi1, exchange rates were slower. Well-resolved cross identified in the regulation of the metaphase– peaks were observed for 43 of the 50 residues. Flat anaphase transition of cells through mitosis, the relaxation dispersion profiles were evident for 25 UBA domains were found to be essential for residues; however, the remaining 18 residues homodimerisation of Rad23 and for the heterodi- exhibited clear dispersion effects due to exchange merisation of Rad23 and Ddi1. In both cases, the with a minor conformer. UBA domains bind Ub, with a regulatory role for To identify an appropriate model for data analysis, Rad23 dimerisation in the control of Ddi function including the possibility of a monomer–dimer suggested by the fact that dimerisation appears to equilibrium, we used a number of other biophysical block Ub binding.28 In contrast, the UBA domains techniques sensitive to population shifts at micro- of the human homologue of Rad23 (hHR23a), molar concentrations. The helix-rich p62 UBA which stimulates DNA repair mechanisms, show domain gives a CD spectrum with double minima no evidence of dimerisation.16,29 observed at 208 and 222 nm. The temperature The p62 UBA domain, in line with many other dependence of the signal at 222 nm results in a UBA domains, appears to have a low affinity for Ub reversible sigmoidal unfolding transition (Fig. 1a); when measured by NMR at protein concentrations however, the melting curves show a distinct concen- ∼ − μ 30 in the millimolar range (Kd 500 700 M). The tration dependence of the Tm, as evident from the affinity appears not to be significantly different data collected at concentrations of 3 and 82 μM μ when binding to K48-linked Ub2 and longer chains; shown in Fig. 1a. Above 20 M, the Tm remains neither is there any evidence for apparent chain- invariant with [UBA] at 345 K but decreases by 6 K at linkage specificity. We report NMR studies, includ- the low-concentration end of the range with a ing relaxation dispersion analysis, together with CD, transition midpoint at around 10 μM(Fig. 1b). The isothermal titration calorimetry (ITC) and kinetic data are consistent with self-association modulating fluorescence measurements, which reveal that the the thermal stability of the UBA domain. Author's personal copy

180 p62 UBA Domain Dimerisation Inhibits Ub-Binding

Fig. 1. Biophysical data demonstrating the presence of a p62 UBA dimer in solution. (a) Far-UV CD thermal unfolding curves recorded at 222 nm for the p62 UBA domain at 3 and 82 μM concentration showing the difference in the transition midpoint. The curve for the lower protein concentration sample has been scaled to show approximately the same change in molar ellipticity. (b) Protein concentration dependence of the melting temperature (transition midpoint) for protein in the range 1−50 μM. The higher temperature plateau at 345 K represents unfolding of the dimer to the monomeric ∼ disordered state, but at low concentrations, the transition represents the unfolding of the monomer (Tm 339 K), with the crossover in behaviour at around 10 μM. (c) Kinetic stopped-flow traces monitored by W412 fluorescence following dilution of protein into 50 mM acetate buffer at pH 5.0. Four kinetic traces are shown, for concentrations of UBA pre- dilution of 5, 10, 20 and 50 μM. The data were normalised such that at each concentration, the fluorescence decreases from 5 −1 − 1 − 1 1 to 0. The smooth fits to the data are simulated curves for kon =8×10 M s and koff =3.2 s . (d) ITC data showing endothermic heat pulses from the injection of 5 μM aliquots of a 270 μM stock solution of p62 UBA in buffer solution into the calorimetry cell containing the same buffer. (e) The dimer dissociation isotherm derived from the integrated heat μ absorbance profile fitted to a dimer dissociation model with a Kdim =4.1±0.6 M at 298 K.

Stopped-flow experiments were used to monitor Information).38 The association and dissociation rate the effects of a rapid 11-fold dilution in protein constants were estimated from the best fit to be μ − 1 − 1 − 1 concentration on the time-dependent fluorescence kon =0.8 (±0.4) M s and koff =3.2 (±0.4) s .The intensity of W412 of the UBA domain at 298 K. Raw larger uncertainty in kon results in an estimated kinetic traces and their relative amplitudes were equilibrium dissociation constant Kdim in the range 2 collected at pre-dilution UBA concentrations be- −10 μM at 298 K, reflecting a highly stable dimer tween 1 and 50 μM and are shown in Fig. 1c. The interface. curves in each case were normalised such that at We carried out analogous dilution experiments each concentration, the fluorescence decreases from using ITC with sequential injections at 298 K from a 1to0(Fig. 1d), and simulated curves were stock solution of 270 μM. A series of endothermic calculated from an explicit numerical integration of heat pulses of decreasing intensity were observed, the fundamental rate equations (see Supplementary which reflected the relative population of monomer Author's personal copy

p62 UBA Domain Dimerisation Inhibits Ub-Binding 181 and dimer as the protein concentration in the cell titrations with Ub led to fast-exchange perturba- increases (Fig. 1d). At [UBA] N20 μM, we observe tions to the resonances of monomeric UBA (mUBA) only a small enthalpy of mixing. A dissociation and a decrease in intensity of the signals for the isotherm of the integrated heat absorbance could be dimer. Overlaid portions of the HSQC spectra of a μ 15 fitted to a dimer dissociation model with a Kdim =4.1 10 M sample of N-labelled p62 UBA in the (±0.6) μM at 298 K (Fig. 1e). As a control, we carried presence of various concentrations of Ub (up to 20 out analogous ITC dilution studies on the UBA2 equivalents) are shown in Fig. 2d. Cross peaks for domain of hHR23 under similar conditions (data not the monomer (m) and dimer (d) forms of L398, shown). Previous conclusions that there was no L402 and E409 are highlighted, with the latter two evidence for UBA2 dimerisation were confirmed,16 residues showing significant chemical shift differ- with detection of only a small exothermicity for ences between the two forms. Most noticeably, the solution mixing. Thus, all three techniques described HSQC cross peaks for the dimeric form of the UBA appear to be consistent with the p62 UBA forming a (dL402 and dE409) show no changes in shift but stable dimer (summarised in Table 1). decrease in intensity as the equilibrium is pulled over to the bound form of the monomer. The data NMR chemical shift analysis of the clearly show binding specificity for the monomer conformational equilibria but no evidence for interaction with the dimer. The chemical shifts for the mUBA:Ub complex observed Previous attempts to test for dimer formation by on saturating labelled UBA with Ub were found to NMR explored protein concentrations well above the be essentially identical with those observed at N estimated range of Kdim values (Table 1). We higher UBA concentrations ( 0.5 mM) such that therefore recorded 15N HSQC data on the p62 UBA the Ub titration data readily enabled us to transfer domain at concentrations as low as 10 μM using a ourearlierassignmentsfromthemUBA:Ub cryoprobe on a 600-MHz spectrometer. The high complex30 back to those of the unbound mUBA. sensitivity revealed two sets of HSQC cross peaks in Binding isotherms from the fast-exchange HSQC slow exchange (Fig. 2a−c), with substantial shift titration data of Ub with 15N-mUBA (Fig. 2d) were differences between the two forms. The apparent globally fitted for the large majority of UBA intensities, but not chemical shifts, were concen- residues (42 of 50) and gave a high-affinity μ tration and temperature dependent, consistent with interaction (Kd =40±10 M) (representative curves – a monomer dimer equilibrium. An estimate of Kdim are shown in Fig. 2e). from the population of species visible in the 10 μM With the assignments of the 15N HSQC spectra of μ HSQC spectra at 298 K suggests Kdim =6±3 M (14± all three forms of the UBA (dimer, monomer and 3 μM at 310 K), consistent with the data derived from Ub-bound monomer), it is possible to dissect the CD, stopped-flow and ITC measurements. chemical shift changes into separate dimer dissoci- The low signal-to-noise ratio of the HSQC spectra ation and monomer binding steps (Fig. 3). In collected at ∼10 μM concentration, showing popula- general, the chemical shift changes associated with tions of both the monomer and dimer form of the p62 the dimer dissociation step are larger than those for UBA, precluded collection of high-quality 2D and 3D ligand binding to the monomer. This is most striking data sets for assignment purposes. However, HSQC for 15N shift changes for residues T430–S434 at the C-terminus, where a perturbation larger than 10 ppm is observed for I431 upon dimer dissociation. In contrast, the largest 15N shift changes on Ub Table 1. Estimates of the dimerisation dissociation binding to the UBA monomer are less than 2 ppm. constant for the p62 UBA dimer Such large 15N shift changes suggested that a μ Method Kdim ( M) significant backbone conformational rearrangement a was occurring at the C-terminus. Since full hetero- Far-UV CD Tm measurements 10 Fluorescence stopped-flow kineticsb 2−10 nuclear assignments have been obtained for the 30 ITCc 4.1±0.6 dimeric and Ub-bound forms of the UBA, we used α β HSQC NMR dilution 6±3d the program TALOS to analyse the C ,C ,C′, N and 14 M α e H chemical shifts to predict the backbone dihedral 14±3 ϕ ψ 39 NMR relaxation dispersion analysisf 1−100 angles and . The analysis revealed an increase

a in helical character for the C-terminal residues from Determined from the midpoint of the concentration depen- I431 onwards, with the predicted changes in S434 dence of the Tm value; data collected at pH 7.0 and 298 K in phosphate buffer. and K435 consistent with a change from fully b Rates of dimer dissociation determined in 50 mM acetate extended structure in the dimer to fully helical buffer, pH 5.0 and 298 K. structure in the Ub-bound monomer (Fig. 3). c ITC data collected at 298 K. Since full heteronuclear NMR assignments were d Derived from relative cross-peak intensities for monomer and dimer in the 1H/15N HSQC spectrum of a 10 μM sample of UBA only accessible at higher UBA concentrations where domain at 298 K. the protein is essentially dimeric, it was not possible e Derived from relative cross-peak intensities for monomer and to extend the TALOS analysis to determine whether dimer in the 1H/15N HSQC spectrum of a 10 μM sample of UBA the conformational changes are associated with domain at 310 K. dissociation of the dimer or with Ub-binding to the f Relaxation dispersion data collected at 310 K. monomer. However, the 1H–15N assignments for Author's personal copy

182 p62 UBA Domain Dimerisation Inhibits Ub-Binding

Fig. 2. 1H–15N HSQC spectra of the p62 UBA domain at different protein concentrations (a) 10 μM and (b) 1 mM in buffered solution at pH 7.0 and 298 K. At a higher concentration (b), a single set of peaks is identified for the symmetrical UBA dimer, but twice the number of cross peaks is evident in (a), reflecting a nearly equal population of monomer and dimer. (c) Expanded boxed region of the HSQC spectra in (a) and (b) showing monomer and dimer HSQC peaks for G405, G411, Y422 and G425 in the left-hand panel (monomer peaks labelled) and dimer-only peaks in the right-hand panel. (d and e) NMR analysis of the Ub-binding interaction with monomeric p62 UBA. (d) Portion of the 1H-15N HSQC spectrum of the p62 UBA domain at 10 μM showing the dimer and monomer to be approximately equally populated; the cross peaks for L398, L402 and E409 are illustrated with the prefix ‘m’ or ‘d’ representing unbound monomer or dimer and are connected by arrows. The Ub-induced CSPs for mL402 and E409 are shown in boxes representing the addition of between 0 and 20 equivalents of Ub. The CSPs for mL402 and mE409 are significant, whereas the cross peaks for the dimer are unperturbed by Ub, as is the case for all other residues, demonstrating clear specificity in binding to monomer rather than to dimer. (e) Binding isotherms derived from (d) for a representative set of peaks in the titration of Ub with 15 μ the monomeric form of N-UBA; estimated Kd from globally fitting 42 binding curves is 40±10 M. the monomer could be determined from the titration exchange is on the slow-exchange side of interme- data. The magnitude of the 15N shift changes, diate exchange (explaining why the effects are larger predicted from analysis of the tripeptides to which at higher temperature), and hence, the on-rate TALOS makes the best matches, is in line with the constant k is much less well defined (in the range on − − magnitude of the large experimental 15Nshift 1.4×105 to 1.4×107 M 1 s 1). These rate constants − μ changes that are directly attributable to dimer correspond to a Kdim value in the range 1 100 M, dissociation (Fig. 3). Therefore, it seems most likely which, despite the uncertainty in kon, is consistent that the increase in helical conformation at the C- with the value determined from HSQC cross-peak terminus of helix 3 occurs upon dimer dissociation. intensities and with kinetic parameters determined We subsequently analysed the 15N relaxation by stopped-flow fluorescence. dispersion data in terms of a monomer–dimer We find an excellent correlation between residue- equilibrium in order to test whether the chemical specific 15N shift differences derived from the Δδ shift perturbations (CSPs) determined from HSQC analysis of HSQC spectra ( HSQC) and the differ- Δω spectra at equilibrium at low protein concentration ences ( CPMG) derived from the fitting of the are consistent with the same chemical exchange relaxation dispersion data (Fig. 4b), which further process at higher concentrations, where the mono- confirms that the dispersion effects are due solely to meric UBA domain is only very weakly populated. exchange between the dimer and a sparsely popu- The quality of the fit is illustrated in Fig. 4a for a lated monomeric state.40 Since the two measure- representative set of dispersion profiles from the 0.8- ments are reporting on the same event, we can use μ mM data collected at 310 K where the exchange the value of Kdim =14±3 M derived from the low- contributions were stronger. The data fitted well to concentration HSQC spectra and the well-deter- − 1 − 1 the model, giving a value for koff =11±2 s . The mined value of koff =11±2 s from the relaxation Author's personal copy

p62 UBA Domain Dimerisation Inhibits Ub-Binding 183

stantial perturbations (Fig. 5). The CSP analysis and subsequent calculations of ψ-angle changes from the TALOS analysis show that the most pronounced effects are associated with conformational changes at the C-terminus of the sequence, with the largest 1H shift of 2.2 ppm for the NH of I431, although adjacent residues T430, Q432 and Y433 are all shifted by N0.6 ppm. In contrast, with the exception

Fig. 3. Dissection of CSPs for the dimer to monomer (D–M) and monomer to Ub-bound (M–B) transitions of the p62 UBA from HSQC assignment data. Both 1H(ΔδH) and 15N(ΔδN) shift perturbations are shown for the amide NH groups. The calculated ψ torsion angle changes for the transition from the dimer to the bound state [Δψ (D–B)] were calculated using TALOS (see Materials and Methods); the predicted 15N shifts are also shown for comparison [ΔδNtp (D–B)]. The positions of the helices within the UBA sequence are shown along the bottom of the figure.

dispersion data, to obtain a value for kon =8 (±2)× 15 5 − 1 − 1 Fig. 4. N relaxation dispersion profiles for a repre- 10 M s . There is therefore excellent agreement sentative set of residues showing clear evidence for between the kinetic parameters from the relaxation curvature associated with exchange effects with a minor dispersion and fluorescence stopped-flow data, conformational state (a); the lines of best fit were from which we conclude that the two techniques determined using a dimer–monomer model on a protein are reporting on the same structural changes sample of 0.8 mM at 310 K and at 600 MHz (see Materials inferred from equilibrium methods. and Methods). Exchange parameters and chemical shift differences between the two states were obtained by – Mapping the p62 UBA dimer interface fitting data to the modified Bloch McConnell equations for two-site exchange. (b) Correlation of 15N chemical shift differences derived from relaxation dispersion experi- The detailed dissection of shift perturbations into Δω ments ( CPMG) and from differences in monomer and 1 –15 Δδ dimer dissociation and ligand binding steps has dimer peaks observed in H N HSQC spectra ( HSQC) enabled us to model the dimer interface, with a on a 10 μM protein sample (slope, 1.05; correlation significant number of residues experiencing sub- coefficient, 0.998). Author's personal copy

184 p62 UBA Domain Dimerisation Inhibits Ub-Binding

Fig. 5. (a) CSPs for the monomer–dimer equilibrium showing large perturbations to residues within loop 1 and across helix 2 and at the C-terminus of helix 3 (I431 is truncated from 2.2 ppm). The positions of the helices and connecting loops are shown along the top of the figure. The largest perturbations are mapped to the surface of the UBA domain in (b), showing that the CSPs across loop 1 and helices 2 and 3 form a largely contiguous dimer interface. The position of G410 is highlighted since this residue is clearly perturbed as its resonance broadens and disappears in the monomer. (c) Ribbon drawing of the UBA domain showing the same orientation with key side chains highlighted. (d) Plot of CSPs for Ub binding to mUBA showing a concentration of strong interactions around loop 1 (M404, G405 and E409) and at the C- terminus of helix 3 (T430, I431 and Q432). (e) CSPs mapped to the surface of the UBA domain, and (f) a ribbon diagram of the UBA domain in the same orientation highlighting residues in loop 1 and the C-terminus of helix 3 that are perturbed Δδ Δδ 2 Δδ 2 1/2 Δδ Δδ by the binding of Ub. CSPs were calculated as HSQC =[( H) +( N/5) ] , where H and N are the observed shifts in the 1H and 15N dimension of the HSQC spectrum. of L402, there are few significant effects within helix (residues 413−420), with notable binding shifts for 1 (residues 392−404; generally b0.1 ppm). However, L413, L416 and T419. Mapping the CSPs onto the substantial perturbations (N0.2 ppm) are observed UBA monomer shows that the perturbed residues in across loop 1 (residues 405−412), notably for D408 loop 1 and helix 2 and at the C-terminus of helix 3, and E409, with significant line-broadening effects on including residues in the more flexible C-terminal G410. The perturbations extend along helix 2 loop region, are juxtaposed and form a largely Author's personal copy

p62 UBA Domain Dimerisation Inhibits Ub-Binding 185 contiguous binding surface (Fig. 5b and c). The same defining the relative orientation of the two mono- perturbed residues are shown explicitly on the meric units. The lowest-energy monomer structure, ribbon structure in Fig. 5c, emphasising that the calculated using the full set of NOE and dihedral majority of perturbations are found at the helix 2 angle restraints,30 was docked to form a UBA dimer −helix 3 interface and at the C-terminus of helix 3. guided by the intermolecular NOEs and CSP data using the automated semirigid docking protocol Model of a symmetrical p62 UBA dimer within HADDOCK,42 whilst imposing an overall C2 axis of symmetry. The 10 lowest-energy structures Intermolecular nuclear Overhauser enhancements gave an RMSD of 0.59 (±0.2) Å and were used for (NOEs) were identified at the dimer interface in further analysis (structure statistics are available in 13C/15N-filtered NOE spectroscopy (NOESY) Supplementary Table 1). experiments using a mixed 13C/15N-labelled/unla- The calculations produced a consistent set of low- belled UBA sample.41 A comparison of filtered and energy structural models showing an anti-parallel unfiltered NOESY spectra shows that intramolecu- dimer with regard to the relative alignment along lar NOEs are suppressed, but short-range intermo- helix 2 (Fig. 6), with the C-terminus of helix 3 of each lecular NOEs are observed (1H–1H strips shown in monomer in close proximity in an end-to-end Fig. 6a).12,18 As a consequence, we identified a set of orientation. In the structural models illustrated, a 24 intermolecular NOEs involving a cluster of 11 number of the disordered C-terminal residues are residues at the dimer interface, including side-chain omitted for clarity. The experimental observation contacts between residues at the C-terminus of helix from the TALOS analysis of backbone ψ-angle 2andloop1′ (T419 to E409′/G410′), between changes that the C-terminus of the UBA domain is residues in helix 2 and helix 2′ (W412 to L416′), extended in the dimer can be addressed on steric between residues in helix 2 and helix 3′ (W412 to grounds. The model suggests that further elongation T430′/I431′) and between the C-termini of helices 3 of helix 3 is likely to push apart the dimer interface and 3′ (I431 and Q432 to themselves). These and reduce buried surface area. In contrast, a more contacts, and the symmetry-related counterparts, extended and/or flexible C-terminus is sterically facilitate the modelling of the dimer structure by less demanding and facilitates contacts along helix 2

Fig. 6. Model of the p62 UBA dimer. (a) 1H–1H strips for W412 and T419 showing NOEs from the NOESY spectrum and from the 13C/15N-filtered NOESY spectrum of the half-labelled UBA dimer with some key intermolecular contacts highlighted. (b) Ensemble of 10 lowest-energy HADDOCK structures, calculated using the intermolecular NOE restraints, showing an anti-parallel alignment of α2 and α2′ and end-to-end contacts between the C-termini of α3 and α3′. The N- and C-termini have been removed for clarity. (c) Structures in (b) rotated by 90°. (d) Contacts at the dimer interface showing the burial of the key hydrophobic side chains of W412 and I431 and their symmetry-related counterparts in the other half of the dimer (identified by a prime). The contacts of the loop 1 residues E409 and G410 with T419′ in helix 2 are also evident, as is the proximity of the symmetry-related helix 3 residue side chains of Q432 and Q432′. Author's personal copy

186 p62 UBA Domain Dimerisation Inhibits Ub-Binding and the optimal positioning of helix 3 of each Despite the added complexity of the NMR spectra monomer in an end-to-end orientation. The dimer of E409K and G410K, titration with Ub results in fast- interface is estimated to bury 590 Å2 (13%) of the exchange conversion of peaks for the monomer to solvent-accessible surface area of each monomer. those of the bound form with a decrease in intensity Key residues involved in intermolecular contacts are of both sets of dimer peaks. The corresponding highlighted in Fig. 6d. The side chain of W412 is HSQC spectra of the complexes are of high quality buried at the dimer interface, accounting for the with only a single bound species evident in solution fluorescence enhancement observed in stopped- (spectra for E409K shown in Fig. 7a). The resulting flow dilution experiments and the 20-nm shift in CSP analysis for the fully bound state shows very λ max as the indole ring becomes solvent accessible in similar effects to those for wt-UBA, which suggests the monomer. The expansion of loop 1 with a that the E409K and G410K mutants are using the double-Gly insertion (G410 and G411) is a unique same binding interface and interact in a similar feature of the p62 UBA domain.7 The stabilising fashion to wt-UBA (data not shown). However, intermolecular contacts between G410/E409 of loop binding saturation is reached at much lower Ub: 1 and T419′ at the C-terminus of helix 2′ in the dimer UBA ratios for the E409K and G410K mutants. The partner (Fig. 6d) appear to be facilitated by the plot of fraction of bound monomer as a function of flexible G410/G411 loop. Subsequent effects on Ub concentration (at the same [UBA]=1 mM) is dimer stability of point mutations within this loop shown for wt-UBA and the E409K and G410K further support this suggestion (see below). mutants in Fig. 7b. The wt-UBA requires a 6:1 ratio of Ub:UBA to reach the fully bound state; however, Ub-binding shifts for the monomeric UBA E409K reaches saturation at close to a 1:1 ratio of Ub: UBA, and G410K reaches saturation at a ratio of 2:1. We similarly analysed the CSP data set to identify NMR dilution experiments show that the E409K and the Ub-binding shifts for the monomeric UBA domain G410K mutants alone have a higher population of to locate the Ub interaction surface (Fig. 5d−f). The the monomer at a given protein concentration, most pronounced CSP effects, though much smaller permitting an estimate of Kdim values from cross- than for dissociation of the dimer, are observed for peak intensities in HSQC spectra (E409K∼200 μM M404, G405 and E409 within loop 1 (N0.2 ppm) and and G410K∼60 μM, versus ∼6 μM for wt-UBA at again for the cluster of residues at the C-terminus of 298 K). These are in good agreement with subse- helix 3, including T430, I431 and Q432, which, quent calculations of Kdim values from ITC dilution when mapped to the surface of the UBA monomer, analysis (E409K∼150 μM and G410K∼85 μM, versus form a contiguous binding patch. There is good ∼4 μM for wt-UBA at 298 K, data not shown). The agreement with our previous conclusions based data are consistent with higher affinity Ub-binding upon the more complex analysis of the Ub titration facilitated by mutations in loop 1 that appear to with the UBA dimer in which the CSPs represented selectively weaken the dimer interface. the sum of the larger slow-exchange perturbations We analysed the effects further using pull-down arising from the monomer–dimer structural rear- assays in which glutathione S-transferase (GST)- rangement and the smaller fast-exchange mUBA UBA fusion proteins were immobilised on glutathi- Ub-binding shifts.30 The p62 UBA Ub-binding sur- one-Sepharose beads and assessed the affinity for face shows many similarities to other UBA domains mixtures of K63-linked polyUb chains (Fig. 7c).10,30 that also possess the conserved MGFX loop (MGFS The E409K and G410K mutants precipitate more for p62).16 polyUb than wt-UBA, consistent with the enhanced binding affinity evident in the NMR titration studies. Mutations in loop 1 destabilise the dimer In contrast, and as a control, the PDB-associated interface G425R mutation results in complete loss of polyUb chain binding. On the basis of CD and NMR studies, On the basis of our structural model, we probed we attribute this not to any significant change in the dimer interface in more detail by introducing the stability of the dimer, but to steric interactions at the point mutations E409K and G410K within loop 1. Ub-binding interface from the Gly→Arg substitu- Both mutants gave HSQC spectra at 298 K ([UBA] tion in helix 3 (data not shown). ∼1 mM) that are more complex than observed for wt-UBA, with evidence for at least three species at Effects of mutations on NF-κB signalling equilibrium, including a low population of the monomer with chemical shifts very similar to those The precise biological and physiological signifi- for the wt-mUBA (Fig. 7a). In addition, a new set of cance of p62 UBA dimerisation remains to be peaks is evident, which is not apparent for wt-UBA. established. Previous studies have shown that HSQC spectra at various dilutions of the E409K and over-expression of wt-p62 is associated with repres- G410K mutants showed that the spectra simplify at sion of basal NF-κB activity in luciferase reporter [UBA] b100 μM, with the monomer being the most assays and that certain PDB-associated p62 mutants significantly populated species. The new set of peaks do not repress signalling to the same level as wt-p62; also appears to decrease in intensity, consistent with instead, they increase basal NF-κB activity relative a second, as yet uncharacterised dimeric species to wt-p62.43,44 The E409K and G410K mutations (data not shown). were introduced into a mammalian polyHis-FLAG- Author's personal copy

p62 UBA Domain Dimerisation Inhibits Ub-Binding 187

Fig. 7. Effects of loop 1 p62 UBA mutations (E409K and G410K) on the dimerisation and Ub-binding affinity. (a) 1H–15N HSQC spectrum of the E409K mutant at 1 mM concentration and 298 K showing multiple species corresponding to monomer, wt-dimer and a second dimeric species (unassigned). In contrast, the HSQC spectrum of the complex of E409K with Ub (ratio of 1:2) shows a single bound species. (b) Binding isotherms showing mole fraction of bound monomer versus [Ub] (in millimolar), with both mutants showing enhanced binding affinity over wt-UBA. (c) Top: Western blot (anti-Ub), showing binding of unanchored K63-linked polyUb chains to GST-p62 UBA proteins with wild- type sequence or the E409K, G410K and G425R mutations. Recombinant GST-UBA proteins were expressed in 10 mL cultures of E. coli, captured on glutathione-Sepharose and used to precipitate 1 μg of polyUb chains for 1 h at 277 K. Bound polyUb chains were detected by Western blotting. Both the E409K and G410K mutants precipitate more polyUb than the wt-UBA; the G425R PDB-associated mutant is associated with a complete loss of polyUb binding in pull-down assays. Bottom: prior to immunodetection, the blot was stained with Ponceau S confirming approximately equal levels of GST- UBA expression for the wild-type and mutant proteins. (d) Effects on relative basal NF-κB activity in U20S cells of transfecting polyHis-FLAG-p62 constructs containing the indicated UBA domain mutations. The E409K and G410K mutants are associated with reduced basal NF-κB activity, relative to wt-p62, whereas the PDB-associated G425R mutant shows a greater signalling activity than wt-p62. Western blotting analyses (anti-FLAG) confirmed comparable expression levels of the different p62 constructs under the assay conditions (representative examples are shown). In each experiment, luciferase activity data for the mutants were expressed relative to values for the wt-p62, which was set to a value of 1. p62 expression construct45 and co-transfected into the observation that dimerisation and Ub-complex U20S cells with an NF-κB reporter plasmid to assess formation are mutually exclusive events with the the effects of UBA dimerisation on p62's ability to stable dimer modulating the affinity of the p62 UBA regulate NF-κB signalling. Both mutants resulted in domain for Ub. A common interaction surface is reduced basal NF-κB activity relative to wt-p62, clearly evident at the C-terminus of helix 3 where whereas the PDB-associated G425R mutant showed T430, I431 and Q432 show evidence for substantial a greater signalling activity than wt-p62 (Fig. 7d). perturbations in both the UBA monomer–dimer equilibrium and UBA–Ub complex (Fig. 5a and b and Fig. 5d and e). Our analysis of changes in Discussion backbone torsion angles from NMR chemical shift analysis suggests that the C-terminus of helix 3 is partially disordered in the dimer but becomes more Mutually exclusive binding surfaces for helical in the monomeric active form, which may dimerisation and Ub-binding also facilitate Ub-binding. Previous NMR estimates of Ub-binding affinities at millimolar UBA concen- ∼ NMR analysis of the dimerisation interface and trations suggested a weak interaction (Kd 500 the Ub-binding surface of the p62 UBA domain −700 μM) and was supported by SPR analysis μ 23 identifies overlapping binding sites and rationalises (Kd =750 M). In this respect, the p62 UBA Author's personal copy

188 p62 UBA Domain Dimerisation Inhibits Ub-Binding appeared to have a comparable affinity to other nestling between helices 1 and 2 of the second mUBA and CUE domains. However, we have monomer.24 The CUE dimer wraps itself around the shown that the competitive effects of the mono- bound Ub molecule with the dimer increasing the mer–dimer equilibrium regulate the binding affinity amount of buried surface and enhancing the binding such that the observed equilibrium dissociation affinity (Fig. 8). A number of other UBA motifs have constant for mUBA:Ub complex formation (Kd =40± been suggested to both homo- and heterodimerise, 10 μM) is significantly stronger at concentrations although none of these processes has been char- where dimerisation has been largely eliminated. acterised in structural detail. The homodimerisation The p62 UBA domain provides strong evidence of the UBA domain of Rad23 and heterodimerisa- that dimerisation is an inhibitory process in regulat- tion of Rad23 and Ddi1 have been implicated in the ing Ub interactions, linked to regulation of NF-κB regulation of the anaphase transition of yeast cells signalling pathways. Somewhat unusually, although through mitosis.28 By analogy to the p62 UBA not without precedent,46 the dimeric form of the UBA domain, the Rad23 UBA also appears to block is inactive, fulfilling a regulatory role in controlling Ub-binding, suggesting a potential regulatory role biological function by protecting the active, and less in the control of Ddi function. stable, monomeric state. Moreover, the dimer stabi- Although few other UBA dimers have been lises the classical canonical UBA conformation, identified or characterised, dimerisation of the UBA – which, in the case of p62, appears to be the domains of the scaffold proteins c-Cbl and Cbl-b,25 27 biologically inactive form.30 Subsequent liberation coupled with their capacity to control signalling of the monomer results in binding to Ub in an downstream from many receptor-tyrosine kinases, alternative conformation that involves the repacking suggests that dimerisation provides a mechanism for of the three-helix bundle. By analogy, conformational an expanded repertoire of Cbl-dependent signals. instability and structural reorganisation, explained by NMR titration studies estimated c-Cbl homodimer- afolding–unfolding equilibrium, have been proposed isation to be in the low micromolar affinity range; for the recognition by the monomeric hMARK3 UBA however, binding studies with an excess of Ub failed domain of the Par-1/MARK protein kinase. Notably, to produce any detectable interaction in HSQC the hMARK3 UBA binds the kinase domain through experiments, suggesting that the functional role of a noncanonical structure, which simultaneously the c-Cbl is not in regulating Ub interactions. In 47 μ provides a mechanism for destabilising Ub-binding. contrast, the dimerisation (Kdim =235±60 M) and Previously, the UBA-like CUE domain of Vps9p activation of the Ub protein ligase Cbl-b are was observed to form a quasi-domain-swapped Ub-mediated and are required for tyrosine phos- dimer in complex with Ub, in which the asymmetric phorylation of Cbl-b and ubiquitination of Cbl-b dimer was stabilised by helix 3 of one monomer substrates. In this context, Ub appears to regulate

Fig. 8. Structures of known UBA/CUE dimers, showing (a) the Cbl-b structure with Ub bound via helix 1 and the separate helix 2/helix 3 dimerisation interface in which Ub-binding enhances dimerisation. (b) The p62 UBA dimer structure is largely stabilised by the helix 2/helix 3 interface with the C-termini of helix 3/3′ end-on to each other. (c) The Vps9p–CUE complex with Ub showing a quasi-domain-swapped asymmetric dimer that produces a larger Ub-binding surface. In all cases, the two monomeric components of each dimer are coloured differently (red and blue) for visibility with Ub shown in grey. Only in the case of the p62 UBA domain are dimerisation and Ub-binding mutually exclusive events. Author's personal copy

p62 UBA Domain Dimerisation Inhibits Ub-Binding 189 biological activity by promoting dimerisation. Ub- linked polyUb chain may partially compensate for binding and dimerisation by Cbl-b use quite different the inhibitory effects of UBA dimerisation. Studies binding interfaces (Fig. 8), showing them to be of the localisation of wt-p62 to autophagosomes coupled events.25 The UBA domain of Ub-ligase suggest that mutations within the PB1 domain that Cbl-b lacks the widely conserved MGF motif in loop 1 suppress p62 self-association also abrogate associa- but employs a novel set of contacts on helix 1 to bind tion with these Ub-positive structures. This is Ub.12,25 The dimer interface is stabilised by helices 2 presumed to be normally driven in part by p62 and 3. In contrast to Cbl-b, these phenomena are binding to K63-linked polyUb chains, since UBA mutually exclusive in the context of p62 with mutations that eliminate Ub-binding also abrogate dimerisation providing an effective mechanism for localisation to autophagosomes.4 Studies of the suppressing the interaction of p62 with mUb/polyUb localisation of wt-p62 to autophagosomes suggest chains and regulating cellular signalling events. that mutations within the PB1 domain that suppress A further feature of the ability of p62 to act as a p62 self-association also modulate binding to K63- multifunctional scaffold protein, relevant to K63- linked polyUb chains, whilst UBA mutations elimi- linked polyUb signalling in NF-κB (and probably nate Ub-binding. autophagy) pathways, is the fact that it is highly Thus, the overall binding affinity of polyUb chains oligomeric,48 as a result of interactions mediated by to the oligomeric p62 is a subtle balance between the N-terminal PB1 domain. This has led to the competing effects that are not likely to be readily conclusion that linkage-specific avidity effects may modelled through solution-phase binding experi- favour K63-linked polyUb selectivity by presenting ments. We can speculate that these two opposing an array of UBA domains that enhance binding manifestations of the chelate effect may amplify affinity through multiple UBA interactions with the small changes in UBA–mUb binding affinity and Ub motifs within a polyUb K63-linked chain.49 The UBA dimer stability associated with certain PDB- ability of the UBA domain to dimerise adds a associated mutations into more significant effects on further layer of complexity to this model since the polyUb chain binding in vivo. In the context of the high effective concentration of the UBA domains link between p62 and K63 polyUb-mediated signal- within the oligomeric p62 array is also likely to ling in NF-κB pathways, the combination of avidity promote dimer formation, which is expected to effects from p62 oligomerisation, modulated by the repress Ub-binding (Fig. 9). The avidity effects on repressive effects of UBA dimerisation on Ub-motif UBA dimerisation strongly disfavour binding of a binding, suggests a mechanism for linkage-specific mono-ubiquitinated substrate to oligomeric p62 but recognition of K63-linked polyUb chains. This favour the association of longer Ub chains where contrasts with the proposed mechanism for speci- multiple Ub–UBA interactions on the same K63- ficity in K48-linked polyUb recognition, which

Fig. 9. Schematic representation of avidity effects on the binding selectivity of oligomeric p62 for K63-linked polyUb chains. (a) Dimerisation of p62 through the PB1 and UBA domains strongly represses the binding of mUb. (b) Enhanced binding of K63-linked Ub2 to the p62 dimer with the chelate effect from Ub2 partially compensating for the repression of binding by the dimerisation of the UBA domain. (c) The avidity effects of the oligomeric state of p62 strongly favour UBA dimerisation; however, multiple Ub interactions from a K63-linked polyUb chain may result in high-affinity binding as the chelate effect is amplified. Author's personal copy

190 p62 UBA Domain Dimerisation Inhibits Ub-Binding involves a UBA domain in two binding interfaces 13C/15N-labelled, have been described previously, as was 30 simultaneously.49 the G425R domain mutant. The E409K and G410K UBA mutants were prepared by site-directed mutagenesis of the wild-type GST-UBA-expressing construct using pre- Biological and physiological significance of p62 10,30 UBA dimers viously described protocols. All the proteins were lyophilised, and their concentrations were determined using the monomeric molecular weight. Luciferase reporter assays indicate that dimerisation of the p62 UBA domain may play a role in the CD, stopped-flow fluorescence and ITC experiments regulation of NF-κB signalling in vivo. In contrast to PDB-associated p62 mutants, which exhibit increased basal NF-κB activity relative to wt-p62 Far UV-CD spectra were collected on an Applied 43,45 PhotoPhysics Pi-star-180 spectrophotometer with a Peltier (Fig. 7d), the E409K and G410K mutants, in heating device for melting studies. In equilibrium studies, which dimerisation is impaired, show reduced the temperature was regulated using a Neslab RTE-300 signalling activity. As p62 is required to potentiate circulating programmable water bath and spectra were NF-κB signalling, evidenced from studies of p62 collected using either a 1 mm (20−150 μM solutions) or a knock-out mice,8 the observation that over-expres- 10 mm (1−10 μM solutions) path-length quartz cuvette sion of wt-p62, as occurs in the reporter assays, over a wavelength range from 215 to 340 nm. The Tm results in repression of basal NF-κB activity relative values are the apparent midpoints calculated from the first to empty vector controls is not immediately recon- derivative of the CD melting profiles. ciled. Indeed, it has been suggested that the ability of Kinetic experiments were also performed on an Applied Ub-binding proteins to repress NF-κB signalling in PhotoPhysics Pi-star-180 spectrophotometer with a temperature-regulated cell. Protein was diluted 11:1 into such assays may result from competition of binding 25 mM acetate buffer at pH 5.0 and 298 K, resulting only of the over-expressed proteins for poly-ubiquiti- 44 in a jump in protein concentration and not buffer nated signalling intermediates. In that regard, it concentration or pH. The initial protein concentrations could be argued that the effects of PDB-associated varied between 1 and 50 μM. Rapid-mixing dilution p62 mutants defective in Ub-binding that do not experiments led to fluorescence buildup curves that were repress signalling to the level of wt-p62, and the analysed by calculating numerical solutions to the rate E409K and G410K p62 mutants, which further equation for dimerisation (Supplementary Information), repress NF-κB activity below levels associated with taking into account the large dilution factor in the rapid- wt-p62, simply result from differences in Ub- mixing experiments. For trial values of kon and koff, the binding affinity. However, it should be noted that initial concentrations for M and D were calculated using the pre-dilution value for the total protein concentration a PDB-associated non-UBA domain mutant of p62, κ and then dividing each value by the dilution factor of 11. P364S, activated NF- B signalling (relative to wt- The concentrations were then integrated forwards in time p62) in a similar manner to other PDB-associated using the Numerical Recipes routine odeint.38 The pre- UBA mutants, despite retaining full Ub-binding dicted curves were compared with the experimental data 50 function. Hence, the effects of over-expression of graphically and kon and koff adjusted manually to give the different p62 constructs in luciferase reporter assays best fit simultaneously for the four concentrations. are not simply due to alterations in Ub-binding Likewise, limiting values of kon and koff were then found, affinity of the expressed proteins. which gave an acceptable goodness of fit. The values of kon An alternative interpretation of the effects of the and koff are anti-correlated, in that an increase in kon can be E409K and G410K mutants in our reporter assays compensated to some degree by a decrease in koff. ITC was performed on a MicroCal VP-ITC instrument. is that the monomeric p62 UBA is able to participate UBA stock solution of 270 μM in 50 mM phosphate buffer in interactions with other proteins to potentiate (pH 7.0) was injected sequentially (5 μL) into the ITC cell signal transduction. There is certainly a precedent at 298 K (volume, 1.424 mL) and the endothermic heat for the p62 UBA to participate in Ub-independent pulse was measured. Data were corrected for the effects of interactions.51 It is tempting to speculate that in vivo, buffer mixing by including an enthalpy of mixing as an signal-induced changes in the p62 UBA monomer– iterated variable. The data were analysed using the dimer equilibrium represent a mechanism to control MicroCal Origin software to determine Kdim. these Ub-independent interactions of the p62 UBA, and indeed, such an equilibrium could also be regu- Relaxation dispersion NMR lated by Ub-binding. It is also possible that certain PDB-associated mutations within the C-terminal Experiments were performed with 0.8 mM uniformly UBA domain may impact on the dimer–monomer 15N-labelled p62-UBA samples containing 25 mM potas- equilibrium, thus influencing p62's Ub-binding func- sium phosphate buffer, pH 7.0, 25 mM NaCl, 0.02% tion via ‘indirect’ effects. sodium azide, and 10% D2O, in Shigemi tubes (CortecNet, Paris). NMR spectra were acquired at 600 MHz using a cryoprobe. Temperature calibration of the spectrometer Materials and Methods was performed using perdeuterated methanol.52 15N Carr–Purcell–Meiboom–Gill (CPMG) single-quantum relaxation dispersion data were collected using a relaxation- Protein expression and purification compensated experiment53 that averages the time 15N magnetisation spends in-phase and anti-phase during the The expression and purification of the p62 UBA domain relaxation period. The pulse scheme was produced by (residues 387–436) and Ub, both unlabelled and fully modification of a published transverse relaxation Author's personal copy

p62 UBA Domain Dimerisation Inhibits Ub-Binding 191 experiment with duty-cycle heating compensation and a sample at a total protein concentration of 200 μM. The data pre-scan proton saturation sequence.54 The length of the were identical within error; therefore, the same range of duty-cycle compensation block matched the relaxation best-fit estimates for ka and kb were taken to be valid. The − −1 − − 1 delay, with the CPMG pulse spacing adjusted to give a estimates of ka =9 13 s and kb =20 400 s at both μ μ constant total number of pulses from both the compensa- [P]tot =800 M and [P]tot =200 M convert to estimates for 5 7 − 1 −1 − − 1 tion and relaxation blocks in each experiment. In addition, kon =1.4×10 to 1.4×10 M s and koff =9 13 s . the pre-scan proton saturation sequence is designed to completely destroy all magnetisation and is followed by a NMR titration experiments and CSPs fixed delay for relaxation: it therefore ensures that all experiments start from exactly the same magnetisation. Dilution experiments were carried out over a range of A series of 15 2D spectra (each with an experiment time μ of ca 2 h) was collected with the CPMG field strength, protein concentrations down to 10 M and HSQC spectra ν were collected on 15N-UBA samples containing 25 mM CPMG, covering a range of 50 to 750 Hz, in random order, with repeat experiments performed at 300 and 500 Hz and potassium phosphate buffer and 25 mM NaCl (0.02% a reference spectrum obtained by omitting the CPMG sodium azide, pH 7.0) using a cryoprobe on the DRX600 period in the pulse sequence and doubling the length of and standard pulse sequences with Watergate solvent the compensation block. The 1H–15N assignment was suppression. At this point of the dilution, twice the carried out by reference to previous assignments.30 The number of resonances were visible compared to the relaxation delay, T , was set to 40 ms and the recycle spectra collected at concentrations close to 1 mM at CPMG 298 K. NMR titration studies with unlabelled Ub were delay was set to 2.5 s. The proton carrier was set to the μ frequency of the water resonance. In order to reduce off- conducted at 310 K by collecting HSQC spectra on 10 M 15N-UBA at up to a 20:1 ratio of Ub:UBA. CSPs were resonance effects, we collected each 2D experiment twice, Δδ Δδ 2 Δδ 2 1/2 Δδ 15 calculated as HSQC =[( H) +( N/5) ] , where H with the N carrier frequency set to 111 or 123 ppm. Δδ 1 15 Spectra were processed and chemical shifts and peak and N are the observed shifts in the H and N intensities were measured using topspin (Bruker Biospin) dimension of the HSQC spectrum, and binding isotherms were constructed and analysed from the dependence of and Felix (Felix NMR Inc.). The effective transverse Δδ eff ν HSQC on [Ub]TOT. The data were fitted using the relaxation rate, R2 , was calculated for each CPMG value – from cross-peak intensities, Levenberg Marquardt nonlinear least-squares algorithm within in-house routines using a numerical solution of the four equations (two equilibria, two conservation of effðÞm − 1 ICPMG R2 CPMG = ln species) describing a competition between dimerisation TCPMG I0 μ and ligand binding with a Kdim fixed to 14 M at 310 K. The estimates of K from the 10 μM HSQC spectra were ν τ τ dim where CPMG =1/(2 CPMG)and CPMG is the delay made from the ratio of intensities measured from the peak between successive refocusing pulses in the CPMG pulse heights. There were no significant consistent differences in train, ICPMG is the peak intensity obtained with a given the lineshapes of the monomer and dimer peaks, and thus, CPMG pulse spacing and I0 is the intensity of the peak concentration was taken to be directly proportional to 55 when the relaxation delay is omitted, TCPMG =0. concentration of protomer units. α β α The C ,C ,C′,H and N shifts from NMR assignments Chemical exchange data analysis of the UBA dimer and Ub-bound UBA monomer determined at high concentrations30 were analysed using the program TALOS, which yielded estimates for the ψ and ϕ Data from the relaxation dispersion experiments were torsion angles for each residue in the two states.39 The modelled by a two-site chemical exchange reaction predicted chemical shift changes were obtained by averag- between states A and B with forward and back rate ing the secondary chemical shifts in the TALOS database constants ka and kb, respectively (see Supplementary for the central residue in the best-matching 10 triplets as Information for discussion of conversion between para- found by the TALOS program, using an in-house script. meters in an A/B model and a dimer/monomer model). The data were fitted using the Levenberg–Marquardt nonlinear least-squares algorithm within in-house rou- Detection of intermolecular NOEs tines, which generate dispersion profiles by explicit numerical modelling of magnetisation evolution using For the identification of NOEs at the dimer interface, a the modified Bloch–McConnell equations,35,55 evaluated dimer sample was prepared from a mixture of fully using the numpy and scipy python modules for the labelled and unlabelled protein whereby 1 mg 13C/15N- evaluation of matrix exponentials. The transverse relaxa- labelled p62 UBA and 2 mg unlabelled p62 UBA were tion rate constants in the absence of exchange were dissolved in water to a concentration of 1 μM, before μ assumed to be identical for the major (R2A) and minor lyophilisation. The solid was resuspended in 600 Lof (R2B) sites; measurements of R2A and R2B in the sample at NMR buffer (50 mM potassium phosphate, 50 mM NaCl, μ 13 15 10 M indicates that the values differ by less than 20% 10% D2O, and 0.04% sodium azide, pH 7.0), and C/ N (data not shown). The adjustable parameters were global half-filtered experiments were recorded following the 41 values of ka and kb and residue-specific values of R2 and method of Otting et al. with 432 scans and 512 indirect Δω . The limits on the kb parameter (which is ill- increments and a 200 ms mixing time. The NOEs were determined in the slow to intermediate exchange regime) compared to a NOESY with a 200 ms mixing time were determined by graphically inspecting the quality of collected on a 1 mM unlabelled p62 UBA sample in data fitting when kb was fixed at values away from the NMR buffer. The structure of the monomeric p62 UBA optimal value. A reasonable quality of fit was judged for domain was generated using NOE and TALOS restraints, − − 1 30 kb in the range 20 400 s . The corresponding variation in as previously described. The lowest-energy monomer the best-fit ka values when kb was fixed to its limiting structure was docked to form a dimer using the values was taken as the estimate of the limits on the ka intermolecular NOEs and HADDOCK's automatic semi- parameter. More noisy data were also collected for a rigid docking protocol,42 whilst imposing a C2 axis of Author's personal copy

192 p62 UBA Domain Dimerisation Inhibits Ub-Binding symmetry. Two hundred water-refined structures were the School of Chemistry at Nottingham. Various p62 determined, and the 10 lowest in energy were used for constructs were kindly provided by Dr. J. Cavey, and further analysis (structure statistics available in Supple- the NF-κB reporter construct was provided by Dr. T. mentary Table 1). MOLMOL was used for the display of 56 Hagen. We are grateful for the provision of NMR structures and in generating figures. pulse sequences by E. R. P. Zuiderweg (University of Michigan) for measurement of transverse relaxation PolyUb-binding assays and for measurement of transverse relaxation dispersion via the constant-time relaxation-compen- The polyUb-binding ability of the various GST-UBA sated CPMG experiment by V. Miloushev and A. G. constructs was assessed using pull-down assays as 30 Palmer (Columbia University). We thank M. J. Cliff described previously. Briefly, GST-UBA proteins with for providing his python-based fitting program and wild-type sequence or the E409K, G410K and G425R mutations were expressed in 10 mL cultures of Escherichia for assistance with the fitting of the relaxation coli, captured on glutathione-Sepharose and used to dispersion data and B. Ciani for stimulating the precipitate 1 μg of K63-linked polyUb chains for 1 h at collaboration. Whilst preparing this article, we 277 K. Bound polyUb chains were detected by Western became aware of a crystal structure of the p62 UBA blotting. Prior to immunodetection, the blot was stained dimer (unpublished) from the laboratory of Prof. with Ponceau S to confirm approximately equal levels of Masahiro Shirakawa at Kyoto University (Japan), GST-UBA expression. The experiment was repeated on including related NMR studies that confirmed the three independent occasions, and a representative exam- presence of the p62 UBA dimer in solution. ple is presented.

NF-κB reporter assays Supplementary Data

Reporter assays were performed as described Supplementary data associated with this article 45 previously, except that U20S cells were used in place can be found, in the online version, at doi:10.1016/ of HEK293 cells. Briefly, the E409K or G410K mutations j.jmb.2009.11.032 were introduced into a polyHis-FLAG-p62 pcDNA3.1 construct, which was subsequently co-transfected into cells with an NF-κB firefly luciferase reporter construct (0.3 and 0.2 μg of plasmids, respectively, to each well of a References 12-well plate). An identical p62 construct already available in the laboratory containing the PDB-associated G425R 1. Moscat, J., Diaz-Meco, M. T. & Wooten, M. W. (2007). mutation was also used. Luciferase activity was measured Signal integration and diversification through the p62 30 h after transfection using the Steady-Glo Luciferase scaffold protein. Trends Biochem. Sci. 32,95–100. Assay System (Promega). Western blotting analyses 2. Moscat, J. & Diaz-Meco, M. T. (2009). P62 at the confirmed comparable expression levels of the different crossroads of autophagy, apoptosis and cancer. Cell, p62 constructs under the assay conditions and assays were 137, 1001–1004. repeated on five independent occasions. In each experi- 3. Kirkin, V., McEwan, D. G., Novak, I. & Dikic, I. (2009). ment, firefly luciferase activity data for the UBA domain Role for ubiquitin in selective autophagy. Mol. Cell, 34, mutants were expressed relative to values for the wild 259–269. type, which was set to a value of 1. Data are presented as 4. Bjørkøy, G., Lamark, T., Brech, A., Outzen, H., Perander, mean values±SE. Statistical analyses were performed M., Overvatn, A. et al. (2005). p62/SQSTM1 forms using one-way ANOVA (Prism4) and Dunnett's test to protein aggregates degraded by autophagy and has a determine the level of significance of differences between protective effect on huntingtin-induced cell death. J. Cell values for wild-type p62 and the UBA domain mutants, Biol. 171,603–614. with significance set at pb0.05. 5. Geetha, T. & Wooten, M. W. (2002). Structure and functional properties of the ubiquitin binding protein 512 – Accession number p62. FEBS Lett. ,19 24. 6. Vadlamudi, R. K., Joung, I., Strominger, J. L. & Shin, J. (1996). p62, a phosphotyrosine-independent Coordinates and structure factors have been deposited ligand of the SH2 domain of p56lck, belongs to a new in the Protein Data Bank with accession number 2knv. class of ubiquitin-binding proteins. J. Biol. Chem. 271, 20235–20237. 7. Ciani, B., Cavey, J. R., Sheppard, P. W., Layfield, R. & Searle, M. S. (2003). Structure of the UBA domain of p62(SQSTM1) and implications for mutations which Acknowledgements cause Paget's disease of bone. J. Biol. Chem. 278, 37409–37412. We acknowledge the funding of the Biotechnology 8. Durán, A., Serrano, M., Leitges, M., Flores, J. M., and Biological Sciences Research Council in suppor- Picard, S., Brown, J. P. et al. (2004). The atypical ting J.L. and M.J.P., the Engineering and Physical PKC-interacting protein p62 is an important mediator of RANK-activated osteoclastogenesis. Dev. Cell, 6, Sciences Research Council and the School of Chem- 303–309. istry at Nottingham for supporting a studentship to 9. Layfield, R. (2007). The molecular pathogenesis of T.P.G. and the National Association for the Relief of Paget disease of bone. Expert Rev. Mol. Med. 9,1–13. Paget's Disease for funding R.L. and B.S. P.C. was 10. Cavey, J. R., Ralston, S. H., Hocking, L. J., Sheppard, supported by an Overseas Research Scholarship and P. W., Ciani, B., Searle, M. S. & Layfield, R. (2005). Author's personal copy

p62 UBA Domain Dimerisation Inhibits Ub-Binding 193

Loss of ubiquitin-binding associated with Paget's binding of the UBA domains from human c-Cbl and disease of bone p62 (SQSTM1) mutations. J. Bone Cbl-b: structural and biochemical insights. Protein Sci. Miner. Res. 20,619–624. 17, 1805–1814. 11. Layfield, R. & Shaw, B. (2007). Ubiquitin-mediated 28. Bertolaet, B. L., Clarke, D. J., Wolff, M., Watson, M. H., signalling and Paget's disease of bone. BMC Biochem. Henze, N., Divita, G. & Reed, S. I. (2001). UBA 8, S5. domains mediate protein–protein interactions be- 12. Matta-Camacho, E., Kozlov, G., Trempe, J.-F. & tween two DNA damage- inducible proteins. J. Mol. Gehring, K. (2009). Atypical binding of the Swa2p Biol. 313, 955–963. UBA domain to ubiquitin. J. Mol. Biol. 386, 569–577. 29. Walters, K. J., Lech, P. J., Goh, A. M., Wang, Q. & 13. Zhang, D., Raasi, S. & Fushman, D. (2008). Affinity Howley, P. M. (2003). DNA-repair protein hHR23a makes the difference: nonselective interaction of the alters its protein structure upon binding protea- UBA domain of ubiquilinin-1 with monomeric ubi- somal subunit S5a. Proc. Natl Acad. Sci. USA, 100, quitin and polyubiquitin chains. J. Mol. Biol. 377, 12694–12699. 162–180. 30. Long, J, Gallagher, T. R., Cavey, J. R., Sheppard, P. W., 14. Kozlov, G., Nguyen, L., Lin, T., De Crescenzo, G., Ralston, S. H., Layfield, R. & Searle, M. S. (2008). Park, M. & Gehring, K. (2007). Structural basis of Ubiquitin recognition by the ubiquitin-associated ubiquitin recognition by the ubiquitin-associated domain of p62 involves a novel conformational (UBA) domain of the ubiquitin ligase EDD. J. Biol. switch. J. Biol. Chem. 29, 5427–5440. Chem. 282, 35787–35795. 31. Murphy, J. M., Korzhnev, D. M., Ceccarelli, D. F., 15. Swanson, K., Hicke, L. & Radhakrishnan, I. (2006). Briant, D. J., Zarrine-Afsar, A., Sicheri, F. et al. (2007). Structural basis for monoubiquitin recognition by the Conformational instability of the MARK3 UBA Ede1 UBA domain. J. Mol. Biol. 358, 724–731. domain compromises ubiquitin recognition and pro- 16. Mueller, T. D. & Feigon, J. (2002). Solution structures motes interaction with the adjacent kinase domain. of UBA domains reveal a conserved hydrophobic Proc. Natl Acad. Sci. USA, 104, 14336–14341. surface for protein–protein interactions. J. Mol. Biol. 32. Sugase, K., Lansing, J. C., Dyson, H. J. & Wright, P. E. 319, 1243–1255. (2007). Tailoring relaxation dispersion experiments for 17. Trempe, J. F., Brown, N. R., Lowe, E. D., Gordon, C., fast-associating protein complexes. J. Am. Chem. Soc. Campbell, I. D., Noble, M. E. & Endicott, J. A. (2005). 129, 13406–13407. Mechanism of Lys48-linked polyubiquitin chain rec- 33. Tollinger, M., Skrynnikov, N. R., Mulder, F. A. A., ognition by the Mud1 UBA domain. EMBO J. 24, Forman-Kay, J. D. & Kay, L. E. (2001). Slow dynamics 3179–3189. in folded and unfolded states of an SH3 domain. J. Am. 18. Kang, R. S., Daniels, C. M., Francis, S. A., Shih, S. C., Chem. Soc. 123, 11341–11352. Salerno, W. J., Hicke, L. & Radhakrishnan, I. (2003). 34. Jee, J. G., Ishima, R. & Gronenborn, A. M. (2008). Solution structure of the CUE–ubiquitin complex Characterization of specific protein association by 15N reveals a conserved mode of ubiquitin binding. Cell, CPMG relaxation dispersion NMR: the GB1A34F 113, 621–630. monomer–dimer equilibrium. J. Phys. Chem. B, 112, 19. Varadan, R., Assfalg, M., Raasi, S., Pickart, C. & 6008–6012. Fushman, D. (2005). Structural determinants for 35. Korzhnev, D. M., Salvatella, X., Vendruscolo, M., selective recognition of a Lys48-linked polyubiquitin Di Nardo, A. A., Davidson, A. R., Dobson, C. M. & chain by a UBA domain. Mol. Cell, 18, 687–698. Kay, L. E. (2004). Low-populated folding interme- 20. Ohno, A., Jee, J., Fujiwara, K., Tenno, T., Goda, N., diates of Fyn SH3 characterized by relaxation disper- Tochio, H. et al. (2005). Structure of the UBA domain sion NMR. Nature, 430, 586–590. of Dsk2p in complex with ubiquitin: molecular 36. Mittermaier, A. & Kay, L. E. (2006). New tools provide determinants for ubiquitin recognition. Structure, 13, new insights in NMR studies of protein dynamics. 521–532. Science, 312, 224–228. 21. Hurley, J. H., Lee, S. & Prag, G. (2006). Ubiquitin 37. Kempf, J. G. & Loria, J. P. (2004). Measurement of binding domains. Biochem. J. 399, 361–372. intermediate exchange phenomena. Methods Mol. Biol. 22. Hicke, L., Schuberet, H. L. & Hill, C. P. (2005). 278, 185–232. Ubiquitin-binding domains. Nat. Rev., Mol. Cell Biol. 6, 38. Press, W. H., Flannery, B. P., Teukolsky, S. A. & 610–621. Vetterling, W. T. (1992). Numerical Recipes in FOR- 23. Raasi, S., Varadan, R., Fushman, D. & Pickart, C. M. TRAN: The Art of Scientific Computing, 2nd edition (2005). Diverse polyubiquitin interaction properties of Cambridge University Press, Cambridge, UK. ubiquitin-associated domains. Nat. Struct. Mol. Biol. 39. Cornilescu, G., Delaglio, F. & Bax, A. (1999). Protein 12, 708–714. backbone angle restraints from searching a database 24. Prag, G., Misra, S., Jones, E. A., Ghirlando, R., Davies, for chemical shift and sequence homology. J. Biomol. B. A., Horazdovsky, B. F. & Hurley, J. H. (2003). NMR, 13, 289–302. Mechanism of ubiquitin recognition by the CUE 40. Hill, R. B., Bracken, C., DeGrado, W. F. & Palmer, A. G. domain of Vps9p. Cell, 113, 609–620. (2000). Molecular motions and protein folding: char- 25. Kozlov, G., Peschard, P., Zimmerman, B., Lin, T., acterization of the backbone dynamics and folding α 13 Moldoveanu, T., Mansur-Azzam, N. et al. (2007). equilibrium of 2D using C NMR spin relaxation. Structural basis for UBA-mediated dimerisation of c- J. Am. Chem. Soc. 122, 11610–11619. Cbl ubiquitin ligase. J. Biol. Chem. 282, 27547–27555. 41. Otting, G. & Wüthrich, K. (1989). Extended hetero- 26. Peschard, P., Kozlov, G., Lin, T., Ahmad Mirza, I., nuclear editing of 2D 1H NMR spectra of isotope- Berghuis, A. M., Lipkowitz, S. et al. (2007). Structural labelled proteins, using the X(ω1,ω2) double half basis for ubiquitin-mediated dimerisation and activa- filter. J. Magn. Reson. 85, 586–594. tion of the ubiquitin protein ligase Cbl-b. Mol. Cell, 27, 42. Dominguez, C., Boelens, R. & Bonvin, A. M. (2003). 474–485. HADDOCK: a protein–protein docking approach based 27. Zhou, Z.-R., Gao, H.-C., Zhou, C.-J., Chang, Y.-G., on biochemical or biophysical information. J. Am. Chem. Hong, J., Song, A.-X. et al. (2008). Differential ubiquitin Soc. 125, 1731–1737. Author's personal copy

194 p62 UBA Domain Dimerisation Inhibits Ub-Binding

43. Rea, S. L., Walsh, J. P., Ward, L., Yip, K., Ward, B. K., observed for UBA domains. Nat. Struct. Mol. Biol. Kent, G. N. et al. (2006). Novel mutation (K378X) in the 16, 883–890. sequestosome 1 gene associated with increased NF-κB 50. Rea, S. L., Walsh, J. P., Ward, L., Magno, A. L., Ward, signaling and Paget's disease of bone with a severe B. K., Shaw, B. et al. (2009). Sequestosome 1 mutations phenotype. J. Bone Miner. Res. 21, 1136–1145. in Paget's disease of bone in Australia: prevalence, 44. Fenner, B. J., Scannell, M. & Prehn, J. H. (2009). genotype/phenotype correlation, and a novel non- Identification of polyubiquitin binding proteins in- UBA domain mutation (P364S) associated with volved in NF-κB signaling using protein arrays. increased NF-κB signaling without loss of ubiquitin Biochim. Biophys. Acta, 1794, 1010–1016. binding. J. Bone Miner. Res. 24, 1216–1223. 45. Najat, D., Garner, T., Hagen, T., Shaw, B., Sheppard, 51. Lange, S., Xiang, F., Yakovenko, A., Vihola, A., P. W., Falchetti, A. et al. (2009). Characterization of a Hackman, P., Rostkova, E. et al. (2005). The kinase non-UBA domain missense mutation of sequestosome domain of titin controls muscle gene expression and 1 (SQSTM1) in Paget's disease of bone. J. Bone Miner. protein turnover. Science, 308, 1599–1603. Res. 24, 632–642. 52. Findeisen, M., Brand, T. & Berger, S. (2007). A 1H- 46. Marianayagam, N. J., Sunde, M. & Matthews, J. M. NMR thermometer suitable for cryoprobes. Magn. (2004). The power of two: protein dimerisation in Reson. Chem. 45, 175–178. biology. Trends Biochem. Sci. 29, 618–625. 53. Loria, J. P., Rance, M. & Palmer, A. G. (1999). A 47. Murphy, J. M., Korzhnev, D. M., Ceccarelli, D. F., relaxation-compensated Carr–Purcell–Meiboom–Gill Briant, D. J., Zarrine-Afsar, A., Sicheri, F. et al. (2007). sequence for characterizing chemical exchange by Conformational instability of the MARK3 UBA NMR spectroscopy. J. Am. Chem. Soc. 121, 2331–2332. domain compromises ubiquitin recognition and pro- 54. Yip, G. N. B. & Zuiderweg, E. R. P. (2005). Improve- motes interaction with adjacent kinase domain. Proc. ment of duty-cycle heating compensation in NMR spin Natl Acad. Sci. USA, 104, 14336–14341. relaxation experiments. J. Magn. Reson. 176, 171–178. 48. Seibenhener, M. L., Geetha, T. & Wooten, M. W. 55. Demers, J.-P. & Mittermaier, A. (2009). Binding (2007). Sequestome 1/p62—more than just a scaffold. mechanism of an SH3 domain studied by NMR and FEBS Lett. 581, 175–179. ITC. J. Am. Chem. Soc. 131, 4355–4367. 49. Sims, J. J., Haririnia, A., Dickinson, B. C., Fushman, D. 56. Koradi, R., Billeter, M. & Wüthrich, K. (1996). & Cohen, R. E. (2009). Avid interactions underlie MOLMOL: a program for display and analysis of the Lys63-linked polyubiquitin binding specificities macromolecular structures. J. Mol. Graphics, 14,51–55.