Abstract
Protein ubiquitination is a multi-functional post-translational modification affecting all cellular processes. Its versatility arises from architecturally complex polyubiquitin chains, in which individual ubiquitin moieties may be ubiquitinated on one or multiple residues, and/or modified by phosphorylation and acetylation1–3. Individual ubiquitin modifications generating the ubiquitin code have been mapped through advances in mass spectrometry, but the architecture of polyubiquitin signals has remained largely inaccessible.
We here introduce Ub-clipping as a methodology to understand polyubiquitin signals and architectures. Ub-clipping utilises an engineered viral protease, Lbpro*, that removes ubiquitin incompletely from substrates, such that the signature C-terminal GlyGly dipeptide remains attached to the modified residue, simplifying direct assessment of protein ubiquitination on substrates and within polyubiquitin. Lbpro*-generated monoubiquitin retains GlyGly-modified residues, enabling quantitation of multiply GlyGly-modified branch-point ubiquitin. Strikingly, a large amount (10-20%) of ubiquitin in polymers appears to exist as branched chains. Moreover, Ub-clipping enables assessment of coexisting ubiquitin modifications. Analysis of depolarised mitochondria reveals that PINK1/Parkin-mediated mitophagy predominantly exploits mono- and short chain polyubiquitin, in which phosphorylated ubiquitin moieties are not further modified. Ub-clipping hence gives unprecedented insight into the combinatorial complexity and architecture of the ubiquitin code.
The last decade has seen spectacular advances in the analysis of ubiquitin modifications, mostly due to mass spectrometry (MS)-based techniques4. Digestion of a ubiquitinated protein with trypsin generates peptides that feature a 114 Da GlyGly modification on residues where ubiquitin had been attached, and which is easily detectable by MS4. Antibodies detecting GlyGly-modified Lys residues enrich these peptides, enabling the mapping of tens of thousands of ubiquitination sites5,6. Moreover, labelled ubiquitin-derived GlyGly-modified peptides facilitate absolute quantitation (AQUA) techniques to quantify polyubiquitin linkage composition7–9, ubiquitin phosphorylation and acetylation10–13.
A disadvantage of tryptic digestion is the loss of architectural information for polyubiquitin chains. Limited trypsinolysis14,15, ubiquitin chain restriction analysis16 and an antibody recognising co-existing Lys11/Lys48 linkages present in branched chains17 can provide some insights into chain architecture18. However, such approaches require experimental optimisation, specialist expertise, or are limited to particular chain combinations. As a result, the combinatorial context of co-existing ubiquitin modifications (e.g. double or triple ubiquitinated ubiquitin species, or phosphorylation in particular chain contexts) has remained largely inaccessible.
We recently demonstrated that the leader protease (Lbpro) of foot-and-mouth disease virus (FMDV) hydrolyses the ubiquitin-like modifier ISG15 at the peptide bond between Arg155 and the C-terminal Gly156-Gly157 motif, generating GlyGly-modified proteins and incapacitating ISG1519. At higher enzyme concentrations Lbpro targets all types of diubiquitin (Fig. 1a), cleaving each ubiquitin moiety after Arg74 (Fig. 1b, Extended Data Fig. 1a), thus generating two products: A truncated ubiquitin spanning residues 1-74 (8450.6 Da) originating from the distal moiety, and a GlyGly-modified ubiquitin 1-74 (8564.6 Da) (Fig. 1c). Prolonged incubation does not lead to further proteolysis (Extended Data Fig. 1b).
Modelling of ubiquitin onto ISG15 from our recent Lbpro ~ISG15 complex structure19 (Extended Data Fig. 1c) rationalises the ubiquitin cleavage site20 and enables enzyme engineering. An Lbpro point mutant, L102W, had improved capability to target all types of diubiquitin, showed improved catalytic efficiency, and excelled at cleaving branched triubiquitin (Extended Data Fig. 2). We refer to either variant as Lbpro, but note that Lbpro* will be a preferred enzyme for studying ubiquitination.
Lbpro converts ubiquitinated proteins into GlyGly-modified proteins (Fig. 1d). This unusual ‘clippase’ activity collapses complex polyubiquitin samples to GlyGly-modified monoubiquitin species that can be further analysed, e.g. by AQUA methods. In vitro analysis of various E3 ligase systems by SDS-PAGE purification and AQUA MS revealed expected linkage compositions, with reduced background from non-ubiquitin peptides (Extended Data Fig. 3).
Importantly, the intact mass of the generated monoubiquitin yields fascinating additional insights into chain architecture (Fig. 2a). In RING-ligase reactions performed with the E2 enzyme UBE2D3, ~10% of ubiquitin had been modified with two, three or even four GlyGly groups (Fig. 2b, c, Extended Data Fig. 4). Branched ubiquitin species were not a result of incomplete cleavage (i.e. GlyGly-modified ubiquitin with an intact C-terminus) since the most abundant transition ions of an isolated di-GlyGly-modified ubiquitin species comprised a ‘clipped’ C-terminus (Extended Data Fig. 4e). UBE2D family enzymes are promiscuous with respect to their target Lys21. Ub-clipping reveals that the promiscuity of the UBE2D3 can generate highly branched, tree-like polyubiquitin architectures.
One complication from these studies is that unassembled monoubiquitin in the reaction artificially increases the apparent amount of unmodified ubiquitin, skewing results on chain composition. Tandem ubiquitin-binding entity (TUBE) pulldowns22 prior to Lbpro treatment remove free monoubiquitin which then enables estimation of average chain length (Extended Data Fig. 5).
The advantages and robustness of Lbpro-mediated Ub-clipping became even more striking when applied to complex mixtures such as cell lysates, opening new possibilities for ubiquitome analysis. Lbpro was active in conditions containing 1 M urea, which inhibits assembly and cleavage of polyubiquitinated proteins by endogenous ligases and deubiquitinases (Fig. 3a, Extended Data Fig. 6a-c). Thus, Lbpro treatment of cell lysates collapses high molecular weight ubiquitin conjugates, generating a monoubiquitin species (Fig. 3a, Extended Data Fig. 6a, b). In-gel trypsin digestion and AQUA MS of the single 8 kDa monoubiquitin product revealed the global linkage composition of the entire cell lysate (Fig. 3a), with identical results as compared to whole cell lysate tryptic digests6,23.
Lbpro treatment of cell lysates generates GlyGly-modified proteins, which can be visualised by the anti-GlyGly antibody5,6,24,19 (Fig. 3b). Previously modified proteins that appear as high-molecular weight smears are collapsed to more discrete bands close to their original size (Fig. 3b). This may prove useful for global ubiquitination site analysis and for directed ubiquitination studies of specific proteins.
We purified the monoubiquitin band to study ubiquitin chain branching in cells (Extended Data Fig. 6d-f). Samples showed ubiquitin linkage compositions as previously reported23,25 (Extended Data Fig. 6g) and minimal background (Extended Data Fig. 6h). Next, intact MS analysis was used to derive relative amounts of differently modified ubiquitin species. Globally, most ubiquitin was unmodified; the pool originates from free ubiquitin, monoubiquitinated proteins, and terminal ubiquitin moieties in chains25. Also, as reported previously25, roughly 10% of ubiquitin was mono-GlyGly-modified (Fig. 3c, d). Interestingly, low levels of di-GlyGly and even tri-GlyGly modified ubiquitin were also detected (Fig. 3c, d, Extended Data Fig. 7a). We confirmed the removal of the GlyGly C-terminus from branched ubiquitin species by fragmentation (Extended Data Fig. 7b-d).
In whole cell lysates, the levels of branch-point ubiquitin accounted for ~0.5% of all ubiquitin, in three cell lines analysed (Fig. 3d). Next, the amount of branching in polyubiquitin was assessed with TUBE pulldowns. TUBE-enriched polyubiquitin showed a similar overall linkage composition as compared to whole cell lysates (Fig. 3e, compare Extended Data Fig. 6g). Intact MS analysis and quantitation by peak integration (Extended Data Fig. 8, also see Methods) revealed that ~4-7% of all ubiquitin in TUBE pulldowns was modified with two GlyGly modifications (Fig. 3f). This suggests that a significant fraction (~10-20%) of ubiquitin in polymers exist in the context of a branched chain, consistent with data from limited trypsinolysis18 (Extended Data Fig. 8e, see further discussion therein).
It is becoming increasingly clear that cells employ special context-dependent ubiquitin signals for particular tasks; a prime example for this is PINK1/Parkin-driven mitophagy, the organised destruction of mitochondria via autophagy26,27. In this process, PINK1 generates Ser65-phosphorylated ubiquitin (phospho-ubiquitin) leading to activation of the E3-ligase Parkin that further ubiquitinates damaged mitochondria to initiate mitophagy26,27. While substrates and linkage composition of Parkin-directed ubiquitination events have been studied in great detail10,28–31, less is known about the architecture of the mitophagy signal.
We used Ub-clipping to study mitophagy. Lbpro was able to hydrolyse phospho-ubiquitin chains (Extended Data Fig. 9a, b) and TUBE-based in vitro analysis of activated Parkin yielded expected polyubiquitin linkage composition with Lys6 (38-43%), Lys11 (11-12%), Lys48 (27-33%) and Lys63-linkages (16%)10,30. Interestingly, intact mass analysis showed that Parkin produces predominantly short chains (ratio between GlyGly-modified and unmodified ubiquitin ~1:2) that also included branched ubiquitin species (Extended Data Fig. 9c-e). Ub-clipping also revealed multiple GlyGly modifications on the E2 enzyme UBE2L3 (Extended Data Fig. 9f), highlighting the utility in characterizing substrate ubiquitination.
In cells, mitochondrial depolarisation triggers Parkin activity26,27. This can be controlled using Parkin-lacking HeLa cell lines that express wild-type (WT) Parkin, or dysfunctional Parkin mutants (C431S or S65A, respectively)10,29. Detailed proteomic studies have characterised this experimental system10,29,32. TUBE pulldowns and Ub-clipping of enriched polyubiquitin (Fig. 4a, b, Extended Data Fig. 10a, b) reproduced expected global ubiquitome changes and appearance of phospho-ubiquitin in response to depolarising agents (Extended Data Fig. 10c). Intact ubiquitin analysis revealed isotopic distribution corresponding to unmodified phospho-ubiquitin (Fig. 4c). Expression of WT Parkin did not lead to global changes in architecture composition in the whole cell lysate, except that phospho-ubiquitin was detected at ~7% of total ubiquitin, in a species that was primarily not GlyGly-modified (i.e. mono or endcap ubiquitin) (Fig. 4d, Extended Data Fig. 10d).
These results were corroborated in mitochondrial preparations. Sodium carbonate-treatment of mitochondrial fractions that enrich (phospho-)ubiquitinated integral membrane proteins33 alleviated requirement of TUBE pulldown as it depleted free ubiquitin from samples (Fig. 4e, Extended Data Fig. 10e). Under these conditions, mitochondrial depolarisation and Parkin activity led to expected increase in amount and change of linkage composition of ubiquitin on mitochondria (Fig. 4f, g)10. Intact mass analysis recovered a high level of phospho-ubiquitin (~10% with Parkin C431S, and ~33% with WT Parkin), the vast majority of which (8% and 27%, respectively) was not further modified (Fig. 4h, Extended Data Fig. 10f, g). In both cell lines, ~30% of ubiquitin was single-GlyGly-modified (Fig. 4h).
These and recent data on ubiquitination site occupancy during mitophagy32, allow modelling of the Parkin-assembled ubiquitin coat on mitochondria (Extended Data Fig. 10h). First, a large relative amount of non-ubiquitinated ubiquitin indicates predominantly monoubiquitination and short chains. Secondly, the levels of monoubiquitin and single-GlyGly-modified ubiquitin are similar in C431S and WT cells indicating the overall ubiquitin chain length is not significantly altered upon expression of active Parkin, under conditions of chemical mitochondrial depolarisation. Thirdly, most phospho-ubiquitin exists at the tip of chains, which is consistent with previous findings that Parkin attaches only unphosphorylated ubiquitin11, and that PINK1 prefers to phosphorylate distal ubiquitin moieties34. Together with the observed overall increase of mitochondrial ubiquitin, this indicates that PINK1 and Parkin generate a carpet of short phospho-ubiquitin chains to trigger mitophagy (Extended Data Fig. 10h).
The ability of Lbpro to turn the ubiquitome into a GlyGly-modified proteome eases access to ubiquitination sites and chain linkage analysis by reducing background from cellular samples and by enabling new applications for well-established tools, including TUBE pulldowns and anti-GlyGly antibodies. More importantly, Ub-clipping provides unprecedented and unbiased insights into ubiquitin chain architecture, which has remained by-and-large inaccessible. Intact mass analysis of the Lbpro-generated monoubiquitin species reveals and quantifies relative amounts of, for example, branch-point ubiquitin, and can access co-modified ubiquitin, such as the chain context of a phosphorylated ubiquitin. This will eventually help to model polyubiquitin architecture, the ‘grammar’ of the ubiquitin code, in greater detail, and unveil the regulatory potential of the vast number of independent polyubiquitin signals.
Online Methods
Protein expression and purification
Lbpro (construct: 29-195 aa) and Lbpro* (L102W) were expressed and purified according to 35. Ubiquitin E1, E2 and E3 enzymes, and PINK1 were purified as described previously 11,36–41.
Diubiquitin cleavage assays
Cleavage assays were performed according to 42 by incubating 1.25 μM of each diubiquitin (Lys6, Lys11, Lys27, Lys29, Lys33, Lys48, Lys63, Met1) with 10 μM Lbpro in reaction buffer (50 mM Tris pH 8.0, 50 mM NaCl, 2 mM DTT). Reactions were incubated at 37 °C and quenched by adding LDS sample buffer. Samples were resolved on NuPAGE 4-12% Bis-Tris gels (Invitrogen) run in MES buffer, and silver stained.
Kinetic cleavage measurements: Fluorescence polarization assays
Fluorescent substrate (0.1 μM of ubiquitin-KG-TAMRA or Lys63-diUb-FlAsH) was incubated with the indicated concentrations of enzyme in reaction buffer (50 mM Tris pH 8.0, 10 mM DTT, 0.05 mg/mL BSA). Measurements were performed with a PheraStar plate reader (BMG Labtech) using either the FlAsH optic module (λex = 485 nm, λem = 520 nm) or the TAMRA optic module (λex = 540 nm, λem = 590 nm). The reaction was carried out in a total volume of 20 μL at 25 °C in a black, round-bottom, nonbinding surface 384-well plate (Greiner) and probed every min for a time-course of 90 min. Raw data were background corrected using a polarization time course from an uncleaved substrate. All experiments were performed as technical triplicate and data were analyzed in Microsoft Excel and GraphPad Prism. Non-linear curve fitting to extract a one-phase exponential decay constant was carried out without a constraint of the plateau position. Rate constants observed in two independent experiments were plotted against enzyme concentration, where the slope of a linear fit yielded the catalytic efficiencies corresponding to the reported kcat/KM ratio.
Kinetic cleavage measurements: Gel-based assays
Lys6/Lys48 branched triubiquitin substrate (10 μM) was incubated with varying enzyme concentrations (1.5, 1.125, 0.75 μM Lbpro or 0.25, 0.188, 0.125 μM Lbpro*) in reaction buffer (50 mM Tris pH 8.0, 10 mM DTT). Samples were removed and quenched in DTT-containing LDS sample buffer every 2 min over a 10 min time course and resolved on NuPAGE 4-12% Bis-Tris gels (Invitrogen). Gels were Coomassie stained (Instant Blue, Expedeon) and band intensities quantified with ImageJ. Monoubiquitin band intensities were normalized to the substrate intensity and data from the linear portion of each time course were used to determine initial rates. The slope of a linear fit of initial rates observed in three independent experiments against enzyme concentration yielded catalytic efficiencies corresponding to the reported kcat/KM ratio.
MS analysis of ubiquitin cleavage
Assays were performed similarly to diubiquitin cleavage assays but with a diubiquitin concentration of 10 μM. Reactions were incubated at 37 °C for 24 h and quenched with 50% acetonitrile (v/v), 0.1% formic acid (v/v). Ubiquitin was diluted to 1 pmol/μL in quenching solution prior to intact MS analysis. Samples were injected at a flow rate of 5 μL/min and ionized using a heated electrospray ionization (HESI-II) source before being analysed with a Q Exactive Orbitrap mass analyser (Thermo Fisher Scientific). Ionization, data collection, and deconvolution of spectra were performed identically to 11.
Ubiquitin chain assembly reactions
Assays were performed by mixing 25-50 μM ubiquitin, 100 nM mouse E1, 2.5-10 μM E2, 2.5-10 μM E3, 10 mM ATP in 40 mM Tris pH 7.4, 10 mM MgCl2, 0.6 mM DTT. In the AREL1 assembly reaction the buffer was supplemented with 10% (v/v) glycerol, 100 mM NaCl, and Tris pH 8.5 (instead of pH 7.4). Assembly reactions were incubated at 37 °C for 1 h, with the exception of UBE2S ΔC and AREL1 in which the reaction times were 16 h and 4 h, respectively. Ubiquitin and phospho-ubiquitin chains were assembled using UBE2D3/cIAP1 assembly machinery according to 11. Ser65-phosphorylated human Parkin (pParkin) was purified as in41. For Parkin assembly reactions, pParkin (4 μM) was incubated for 2 h at 37 °C in the presence of human E1 (0.2 μM), UBE2L3 (2-4 μM as indicated), ubiquitin (15 μM) and where indicated Ser65-phosphorylated ubiquitin (1.5 μM) in reaction buffer (50 mM Tris, pH 8.5, 200 mM NaCl, 10 mM MgCl2, 10 mM ATP, 10 mM DTT). Staurosporine (10 μM) and Nb696 (0.8 μM) were added to inhibit kinase activity43. The assembly reactions were terminated through the addition of 2 mU of apyrase (Sigma-Aldrich) for 1 h at 37 °C.
Lbpro treatment of in vitro assembly reactions
Ubiquitin chain assembly reactions were incubated with 10 μM Lbpro in 50 mM Tris pH 8.0, 10 mM DTT. The reactions were incubated at 37 °C and stopped by mixing with LDS sample buffer at indicated time points. The protein samples were resolved on NuPAGE 4-12% Bis-Tris gels and visualised by anti-ubiquitin Western blots (cat. no. 07-375, Millipore), silver staining (cat. no. 161-0481, Bio-Rad), or Coomassie staining using Instant Blue Coomassie SafeStain (Expedeon).
Western blot analysis
Protein from ubiquitin assembly reactions or whole cell lysates was transferred to nitrocellulose membrane. For additional visualisation of monoubiquitin (e.g. Fig. 3a), the lower portion of the gel (below 14 kDa) was transferred instead to a PVDF membrane. Samples from Fig. 4e and Extended Data Fig. 10e were transferred to PVDF membrane. Western blots were incubated at 4 °C in a 1:1000 primary antibody dilution overnight in PBS-T + 5% (w/v) BSA and 1:5000 secondary antibody dilution overnight in PBS-T + 5% milk prior to being visualised by chemiluminescence. Primary antibodies used include anti-ubiquitin (cat. no. 07-375, Millipore), anti-pSer65 phospho-ubiquitin (cat. no. ABS1513-I, Millipore), anti-Parkin (cat. no. ab77924, Abcam) and anti-GlyGly (cat. no. 30-1000, Lucerna Inc.). The primary anti-ubiquitin antibody (cat. no. NB300-130, Novus Biologicals) was used for Western blots in Figure 4e and Extended Data Fig.10e. Secondary horseradish peroxidase-linked antibodies included anti-mouse and anti-rabbit (cat. no. NXA931V and NA934V, GE Healthcare).
AQUA MS
Following Lbpro cleavage and separation by SDS-PAGE, the molecular weight range surrounding monoubiquitin (5-10 kDa) was excised from the gel and diced into 1 mm3 cubes with a scalpel. Quantitative ubiquitin linkage analysis using AQUA was performed according to 11. For in vitro Parkin assembly reactions, AQUA analysis was performed on TUBE-purified chains. These samples were separated 2 cm into the gel before the lane was processed and analysed by AQUA MS.
Intact MS on ubiquitin from assembly reactions and quantitation by protein deconvolution
Lbpro-treated assembly reactions were diluted to 1 μM ubiquitin with quenching solution (50% acetonitrile (v/v), 0.1% formic acid (v/v)). Samples were ionized using a Heated Electrospray Ionization source (HESI-II, Thermo Fisher Scientific) at a flow rate of 5 μL/min. Ionization settings were performed according to 11. Data were collected on a Q Exactive at a resolution of 140,000 over a period of 1 min. Spectra were averaged and subsequently deconvoluted using the Xtract node of Thermo Xcalibur Qual Browser version 2.2 (Thermo Fisher Scientific). The intensities of un- (8451.65 Da), single- (8565.69 Da), double- (8679.73 Da), triple- (8793.78 Da), and quadruple-modified (8907.82 Da) ubiquitin were exported from Xcalibur into Microsoft Excel for further analysis. For quantitation of phospho-ubiquitin containing samples by spectra deconvolution, the intensities of unmodified phospho-ubiquitin (8531.61 Da), single-GlyGly-modified phospho-ubiquitin (8645.66 Da), and double-GlyGly-modified phospho-ubiquitin (8759.70 Da) were also exported and analysed.
Tissue culture
HeLa, HCT116, HEK293 were obtained from ATCC and exhibited expected cell morphology. Doxycycline-inducible HeLa Flp-In T-REx cells expressing Parkin proteins were a kind gift from Alban Ordureau and Wade Harper (Harvard)10. All cell lines were routinely checked for mycoplasma contamination using the MycoAlert™ detection kit (Lonza). Cells were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) plus GlutaMAX™ (Gibco) supplemented with 10% (v/v) fetal calf serum and PenStrep (100 U/mL Penicillin and Streptomycin). Cell lines were grown to 90% confluency prior to lysate preparation. Cells were washed in PBS, harvested, and immediately frozen in liquid nitrogen.
Protein extraction
Cell pellets were lysed in extraction buffer (50 mM Tris pH 8.0, 50 mM NaCl, 10 mM chloroacetamide, 5 mM EDTA, 1 mM PMSF, 0.1% (v/v) NP-40, 25 mM sucrose, 5% (v/v) glycerol, 4 M urea). To supplement lysis and shear DNA, lysates were sonicated using a microtip. Sonication settings were the following: amplification 10%, 10 sec on, 10 sec off, total sonication time 1 min. The lysate was incubated on ice for 1 min and sonication repeated. The lysate was pelleted at 21,000 x g, after which the supernatant was removed and total protein quantified with Bradford reagent (Bio-Rad Laboratories) using Bovine γ-globulin as a standard.
Generation of fluorescently labelled ubiquitin and diubiquitin
For maleimide labelling, ubiquitin containing an S20C mutation or diubiquitin containing an S20C mutation in the distal moiety were used. Lys11-linked diubiquitin was assembled overnight using UBE2S ΔC38, ubiquitin (K11R, S20C, K63R) as the distal and ubiquitin (K63R, ΔLRGG) as the proximal moiety and purified by size-exclusion chromatography (HiLoad 16/60 Superdex 75, GE Healthcare). Lys48-linked diubiquitin was assembled for 90 min using Cdc3444, ubiquitin (K6R, S20C, K48R) as the distal and ubiquitin (ΔLRGG) as the proximal moiety and purified using cation-exchange chromatography (MonoS, GE Healthcare) followed by size-exclusion chromatography as above.
Prior to labelling, S20C-containing ubiquitin and diubiquitin variants were first incubated in 2 mM β-mercaptoethanol-containing buffer for 45 min at 25°C and then buffer exchanged into 50 mM Tris, pH 7.4 using PD Spintrap G-25 (GE Healthcare). Variants were incubated with a ~6-fold excess of Alexa Fluor maleimide dye (ThermoFisher Scientific; ubiquitin with Alexa Fluor 546 C5 Maleimide, Lys11-linked diubiquitin with Alexa Fluor 488 C5 Maleimide and Lys48-linked diubiquitin with Alexa Fluor 647 C2 Maleimide) at 25°C overnight in the dark. After the labelling step, ubiquitin variants were separated from unreacted dye using illustra NAP-10 columns (GE Healthcare) according to the manufacturer.
Purification of total ubiquitin
All steps were performed at 4 °C unless denoted otherwise. For purification of total ubiquitin, a single confluent 15 cm dish of cells was used per replicate. The cell pellet was resuspended in 300 μL of extraction buffer and processed as described above. In order to standardise cleavage of ubiquitin chains by Lbpro the final concentration of total protein was adjusted to 5 mg/mL using lysis buffer without urea; this typically reduced the final concentration of urea to ~1 M, a concentration in which Lbpro retains activity. Cell lysates were incubated with 100 μM Lbpro for 5 h at 37 °C. After incubation, samples were dialysed in ice-cold ultra-pure water overnight. Following dialysis, precipitated proteins were removed by centrifugation at 21,000 x g. The remaining, ubiquitin-containing fraction was subjected to gel filtration on a Superdex 75 3.2/300 gel filtration column using an ÄKTAmicro System (GE Healthcare). Fractions containing ubiquitin were identified by Western blotting, concentrated and used for mass spectrometry.
To test for retention of ubiquitin ligase or deubiquitinase activity following dilution of the lysates in lysis buffer without urea, lysates were incubated with fluorescently labelled monoubiquitin (Alexa Fluor 546), Lys11-linked di-ubiquitin (Alexa Fluor 488) and Lys63-linked di-ubiquitin (Alexa Fluor 647), whereby the respective fluorophores are chemically attached to S20C ubiquitin. Lysates (0.75 μg total protein per sample) were incubated with 90 ng of each fluorescent ubiquitin at 37 °C for the indicated times, with or without Lbpro, then run on a NuPAGE 4-12% Bis-Tris gels. Fluorescence emission for each fluorophore was measured using a Chemidoc MP (Bio-Rad) using the instrument’s optimal exposure settings. Another exposure was taken at approximately three times the optimal exposure time to detect any lowly abundant forms. To test that the fluorescently modified ubiquitin species were still competent for assembly, 5 μM Alexa Fluor 546-labelled monoubiquitin was assembled using 0.1 μM human E1, 0.5 μM UBE2D3 and 0.5 μM TRAF6. Cleavage of each labelled diubiquitin was confirmed by treating 5 μM diubiquitin with 1 μM USP21 for 1 h at 37 °C.
TUBE pull-down assays
A confluent 10 cm dish was used in each replicate experiment or in the instance of in vitro chain assemblies, 30 μL of the assembly reaction. For doxycycline-inducible Parkin-expressing cell lines, cells were pre-treated with 0.1 μg/mL doxycycline for 16 h, followed by a 1h treatment with 10 μM CCCP. All steps were performed at 4 °C unless stated otherwise. Frozen cell pellets were resuspended in 300 μL of TUBE lysis buffer (20 mM Na2HPO4, 20 mM NaH2PO4, 1 % (v/v) NP-40, 2 mM EDTA 1 mM DTT, 10 mM chloroacetamide, cOmplete EDTA-free Protease Inhibitor Cocktail (Roche) and PhosSTOP (Roche)) supplemented with 100 μg of GST 4x ubiquilin 1 TUBE22 (15 μg for assembly reactions). Samples were incubated on ice for 20 min prior to centrifugation at 21,000 x g. The cleared lysate was incubated with 25 μL of pre-washed Glutathione Sepharose 4B resin (GE Healthcare). Samples were incubated on a rotating wheel for 2 h. Glutathione beads were washed three times with 500 μL of PBS + 0.1% (v/v) Tween20 and a final wash with PBS. Enriched ubiquitin chains were incubated with 4 column volumes of 100 μM Lbpro* for 16 h at 37 °C. After Lbpro cleavage, the supernatant containing ubiquitin species was removed for MS sample preparation.
Mitochondria isolation and enrichment of ubiquitinated proteins
For mitochondrial enrichment experiments, doxycycline-inducible Parkin-expressing cell lines were used10. Prior to harvest, cells were pre-treated for 16 h with 0.5 μM doxycycline, followed by a 2 h treatment with 0.5 μM doxycycline, and 10 μM Oligomycin / 4 μM Antimycin (OA). Mitochondria were isolated as previously described45, with minor modifications. OA-treated cells were resuspended in 20 mM HEPES (pH 7.6), 220 mM mannitol, 70 mM sucrose, 1 mM EDTA, supplemented with 1x cOmplete protease inhibitor (Roche) and 1x PhosStop (Roche). Following homogenisation with a dounce homogeniser, samples were pelleted for 5 min at 1,000 x g, 4 °C. The post-nuclear supernatant was centrifuged for 10 min at 10,000 x g, 4 °C, to pellet the mitochondrial fraction. Further purification of ubiquitinated integral mitochondrial membrane proteins was achieved by sodium carbonate treatment33,46. For this, mitochondrial pellets were resuspended in 100 mM sodium carbonate, incubated at 4 °C for 30 min with occasional vortexing, and then centrifuged for 30 min at 21,000 x g at 4 °C. The pellets were resuspended in 50 mM Tris pH 7.4 and centrifuged for 30 min at 21,000 x g at 4 °C to remove residual sodium carbonate.
Lbpro treatment of mitochondrial ubiquitin
Sodium carbonate-treated mitochondria were resuspended in 50 mM NaCl, 50 mM Tris (pH 8.0), 10 mM DTT containing 20 μM Lbpro and incubated overnight at 37 °C. Where indicated, 10 pmol of 15N-phospho-ubiquitin was spiked in to the sample to detect contaminating phosphatase activity. Membranes were pelleted at 21,000 x g for 30 min at 4 °C. The supernatant containing cleaved ubiquitin was further purified by perchloric acid extraction (see section below).
MS sample preparation of TUBE-purified ubiquitin
After purification and Lbpro cleavage, ubiquitin species were further purified by perchloric acid precipitation. Following cleavage, perchloric acid was slowly added to a final concentration of 0.5% (v/v). After incubation for 10 min on ice, the samples were centrifuged at 21,000 x g in a pre-cooled microcentrifuge. This step retained all GlyGly-modified ubiquitin and GlyGly-modified phospho-ubiquitin in the soluble fraction (data not shown). Unwanted precipitated proteins were pelleted at 21,000 x g for 10 min at 4 °C. The supernatant was removed, placed in a pre-soaked Slide-A-Lyzer MINI 3.5 K MWCO (cat. no. 69550) and dialyzed in 50 mM Tris pH 7.4 for 4 h, and subsequently ultra-pure water for 16h. Samples were then lyophilised and stored at -80 °C.
LC-MS of intact ubiquitin from cells
Ubiquitin was resuspended in 15 μL of reconstitution buffer (5% (v/v) acetonitrile, 0.1% (v/v) formic acid). Insoluble material was removed by centrifugation. The supernatant was transferred to a glass vial and placed in the auto-sampler. Using a Dionex Ultimate 3000 HPLC system (Thermo Fisher Scientific), 10 μL of the sample was loaded onto a C4 PepMap300 precolumn trap (Thermo Fisher Scientific) at a flow rate of 30 μL/min. Trapped proteins were eluted using an acetonitrile gradient (5-40%) over 45 min. Immediately prior to ionization, proteins were separated using an EASY-Spray C4 reverse-phase column (Thermo Fisher Scientific). Proteins were analysed on a Q Exactive mass spectrometer (Thermo Fisher Scientific). MS analysis was performed using the following settings: resolution, 140,000; AGC target, 3E6; maximum injection time, 200 ms; and scan range, 150-2,000 m/z.
LC-MS quantitation by peak integration
Raw files generated from LC-MS runs were analysed in Thermo Xcalibur Qual Browser version 2.2 (Thermo Fisher Scientific). The charge state of ubiquitin species with the highest intensity, corresponding to z= +12 and the nominal mass, i.e. the most naturally abundant 13C isotope, within this charge state were used for quantification. The masses used for quantitation are the following 705.22731 (unmodified), 714.73089 (1xGG), and 724.23446 (2xGG). The total ion current of each mass was integrated using a mass tolerance of 10 or 15 p.p.m.. Additional Qual Browser settings included the following: Gaussian smoothing enabled (15 points) and a mass precision of five decimal places.
PRM assays of branched ubiquitin
Branched ubiquitin molecules (i.e. ubiquitin with the addition of two GlyGly motifs and minus the C-terminal GlyGly amino acids) were isolated and fragmented in a Q Exactive mass spectrometer (Thermo Fisher Scientific). Ubiquitin with a mass-to-charge ratio of 724.24 was isolated in the quadrupole using an isolation window of 2 m/z and fragmented using the following settings: resolution, 17,500; AGC target, 1E5; maximum injection time, 120 ms; normalized collision energy, 28. Fragment ions were inspected manually using Thermo Xcalibur Qual Browser version 2.2, and assigned using Expert System 47.
Shotgun proteomics
Discovery proteomics assays were performed according to 48. In short, a Top10 analysis was performed on peptides generated from the purified monoubiquitin band, and precursor masses were screened using the following settings: mass range, 200-2000 m/z; resolution, 70,000, AGC target, 1E6; maximum ion trap time, 250 milliseconds; scan-type, positive. Data-dependent settings include the following: resolution, 17,500, AGC target, 5E4; maximum ion trap time, 80 milliseconds; isolation window, 2.0 m/z; collision energy, 28.0; data type, centroid; exclusion of unassigned charge states and masses with a charge state of 1. Dynamic exclusion enabled, 30 s. Raw files were searched and spectra assigned using SEQUEST against a human genome database (UniProt) in Proteome Discoverer (Thermo Scientific) with a false-discovery rate of 1%.
Extended Data
Supplementary Material
Acknowledgments
We thank Mark Skehel, Sarah Maslen (MRC LMB), Andrew Webb (WEHI), Wade Harper and Alban Ordureau (Harvard) for reagents, help and discussion on mass spectrometry, and members of the DK lab for reagents and discussions. This work was supported by the Medical Research Council [U105192732], the European Research Council [724804], the Lister Institute for Preventive Medicine (DK), and P 24038 and P 28183 grants from the Austrian Science Fund (TS). JU is funded by a Gates Cambridge Scholarship.
Footnotes
Data Availability Statement
All data have been deposited to the Mass Spectrometry Interactive Virtual Environment (MassIVE) (ftp://MSV000083662@massive.ucsd.edu). Source data for all gels have been included in Supplementary Figure 1. Data to regenerate any graph and described reagents can be obtained upon reasonable request from the corresponding author. Plasmids for Lbpro and Lbpro L102W are available from Addgene.
Author contribution
K.N.S. conceived and designed the study, performed most experiments, interpreted results, and wrote the manuscript. J.L.U. performed mitochondrial experiments, and helped annotate fragmentation patterns of branched ubiquitin, and performed control experiments. A.F.K. helped perform the quantitation of ubiquitin species from whole cell lysates. C.G. contributed to kinetic characterization of Lbpro*. T.E.T.M. provided critical reagents. J.N.P. contributed to characterization of Lbpro and Lbpro*, and provided structural and ubiquitin biology expertise. T.S. provided Lbpro, protocols, and acquired funding. D.K. conceived and designed the study, interpreted results, wrote the manuscript, and acquired funding.
Conflict of Interest Statement
The authors declare no competing interests.
References
- 1.Komander D, Rape M. The ubiquitin code. Annu Rev Biochem. 2012;81:203–229. doi: 10.1146/annurev-biochem-060310-170328. [DOI] [PubMed] [Google Scholar]
- 2.Swatek KN, Komander D. Ubiquitin modifications. Cell Res. 2016;26:399–422. doi: 10.1038/cr.2016.39. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Yau R, Rape M. The increasing complexity of the ubiquitin code. Nature Cell Biology. 2016;18:579–586. doi: 10.1038/ncb3358. [DOI] [PubMed] [Google Scholar]
- 4.Ordureau A, Münch C, Harper JW. Quantifying ubiquitin signaling. Mol Cell. 2015;58:660–676. doi: 10.1016/j.molcel.2015.02.020. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Kim W, et al. Systematic and quantitative assessment of the ubiquitin-modified proteome. Mol Cell. 2011;44:325–340. doi: 10.1016/j.molcel.2011.08.025. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Wagner SA, et al. A proteome-wide, quantitative survey of in vivo ubiquitylation sites reveals widespread regulatory roles. Mol Cell Proteomics. 2011;10 doi: 10.1074/mcp.M111.013284. M111.013284. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Peng J, et al. A proteomics approach to understanding protein ubiquitination. Nat Biotechnol. 2003;21:921–926. doi: 10.1038/nbt849. [DOI] [PubMed] [Google Scholar]
- 8.Kirkpatrick DS, et al. Quantitative analysis of in vitro ubiquitinated cyclin B1 reveals complex chain topology. Nature Cell Biology. 2006;8:700–710. doi: 10.1038/ncb1436. [DOI] [PubMed] [Google Scholar]
- 9.Phu L, et al. Improved quantitative mass spectrometry methods for characterizing complex ubiquitin signals. Mol Cell Proteomics. 2010 doi: 10.1074/mcp.M110.003756. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Ordureau A, et al. Quantitative Proteomics Reveal a Feedforward Mechanism for Mitochondrial PARKIN Translocation and Ubiquitin Chain Synthesis. Mol Cell. 2014;56:360–375. doi: 10.1016/j.molcel.2014.09.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Wauer T, et al. Ubiquitin Ser65 phosphorylation affects ubiquitin structure, chain assembly and hydrolysis. EMBO J. 2015;34:307–325. doi: 10.15252/embj.201489847. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Ohtake F, et al. Ubiquitin acetylation inhibits polyubiquitin chain elongation. EMBO Rep. 2015;16:192–201. doi: 10.15252/embr.201439152. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Swaney DL, Rodríguez-Mias RA, Villén J. Phosphorylation of ubiquitin at Ser65 affects its polymerization, targets, and proteome-wide turnover. EMBO Rep. 2015;16:1131–1144. doi: 10.15252/embr.201540298. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Xu P, Peng J. Characterization of polyubiquitin chain structure by middle-down mass spectrometry. Anal Chem. 2008;80:3438–3444. doi: 10.1021/ac800016w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Valkevich EM, Sanchez NA, Ge Y, Strieter ER. Middle-down mass spectrometry enables characterization of branched ubiquitin chains. Biochemistry. 2014;53:4979–4989. doi: 10.1021/bi5006305. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Hospenthal MK, Mevissen TET, Komander D. Deubiquitinase-based analysis of ubiquitin chain architecture using Ubiquitin Chain Restriction (UbiCRest) Nat Protoc. 2015;10:349–361. doi: 10.1038/nprot.2015.018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Yau RG, et al. Assembly and Function of Heterotypic Ubiquitin Chains in Cell-Cycle and Protein Quality Control. Cell. 2017;171:918–933.e20. doi: 10.1016/j.cell.2017.09.040. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Crowe SO, Rana ASJB, Deol KK, Ge Y, Strieter ER. Ubiquitin Chain Enrichment Middle-Down Mass Spectrometry Enables Characterization of Branched Ubiquitin Chains in Cellulo. Anal Chem. 2017;89:4428–4434. doi: 10.1021/acs.analchem.6b03675. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Swatek KN, et al. Irreversible inactivation of ISG15 by a viral leader protease enables alternative infection detection strategies. Proceedings of the National Academy of Sciences. 2018;115:2371–2376. doi: 10.1073/pnas.1710617115. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Steinberger J, Skern T. The leader proteinase of foot-and-mouth disease virus: structure-function relationships in a proteolytic virulence factor. Biol Chem. 2014;395:1179–1185. doi: 10.1515/hsz-2014-0156. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.David Y, Ziv T, Admon A, Navon A. The E2 Ubiquitin-conjugating Enzymes Direct Polyubiquitination to Preferred Lysines. J Biol Chem. 2010;285:8595–8604. doi: 10.1074/jbc.M109.089003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Hjerpe R, et al. Efficient protection and isolation of ubiquitylated proteins using tandem ubiquitin-binding entities. EMBO Rep. 2009;10:1250–1258. doi: 10.1038/embor.2009.192. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Dammer EB, et al. Polyubiquitin linkage profiles in three models of proteolytic stress suggest the etiology of Alzheimer disease. J Biol Chem. 2011;286:10457–10465. doi: 10.1074/jbc.M110.149633. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Elia AEH, et al. Quantitative Proteomic Atlas of Ubiquitination and Acetylation in the DNA Damage Response. Mol Cell. 2015;59:867–881. doi: 10.1016/j.molcel.2015.05.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Kaiser SE, et al. Protein standard absolute quantification (PSAQ) method for the measurement of cellular ubiquitin pools. Nat Methods. 2011;8:691–696. doi: 10.1038/nmeth.1649. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Harper JW, Ordureau A, Heo J-M. Building and decoding ubiquitin chains for mitophagy. Nat Rev Mol Cell Biol. 2018;19:93–108. doi: 10.1038/nrm.2017.129. [DOI] [PubMed] [Google Scholar]
- 27.Pickles S, Vigié P, Youle RJ. Mitophagy and Quality Control Mechanisms in Mitochondrial Maintenance. Curr Biol. 2018;28:R170–R185. doi: 10.1016/j.cub.2018.01.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Sarraf SA, et al. Landscape of the PARKIN-dependent ubiquitylome in response to mitochondrial depolarization. Nature. 2013;496:372–376. doi: 10.1038/nature12043. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Ordureau A, et al. Defining roles of PARKIN and ubiquitin phosphorylation by PINK1 in mitochondrial quality control using a ubiquitin replacement strategy. Proceedings of the National Academy of Sciences. 2015;112:6637–6642. doi: 10.1073/pnas.1506593112. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Durcan TM, et al. USP8 regulates mitophagy by removing K6-linked ubiquitin conjugates from parkin. EMBO J. 2014;33:2473–2491. doi: 10.15252/embj.201489729. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Cunningham CN, et al. USP30 and parkin homeostatically regulate atypical ubiquitin chains on mitochondria. Nature Cell Biology. 2015;17:160–169. doi: 10.1038/ncb3097. [DOI] [PubMed] [Google Scholar]
- 32.Ordureau A, et al. Dynamics of PARKIN-Dependent Mitochondrial Ubiquitylation in Induced Neurons and Model Systems Revealed by Digital Snapshot Proteomics. Mol Cell. 2018;70:211–227.e8. doi: 10.1016/j.molcel.2018.03.012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Fujiki Y, Hubbard AL, Fowler S, Lazarow PB. Isolation of intracellular membranes by means of sodium carbonate treatment: application to endoplasmic reticulum. J Cell Biol. 1982;93:97–102. doi: 10.1083/jcb.93.1.97. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Gersch M, et al. Mechanism and regulation of the Lys6-selective deubiquitinase USP30. Nat Struct Mol Biol. 2017;510:370–930. doi: 10.1038/nsmb.3475. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Guarné A, et al. Structure of the foot-and-mouth disease virus leader protease: a papain-like fold adapted for self-processing and eIF4G recognition. EMBO J. 1998;17:7469–7479. doi: 10.1093/emboj/17.24.7469. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Wauer T, Simicek M, Schubert AF, Komander D. Mechanism of phospho-ubiquitin-induced PARKIN activation. Nature. 2015;524:370–374. doi: 10.1038/nature14879. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Michel MA, et al. Assembly and specific recognition of k29- and k33-linked polyubiquitin. Mol Cell. 2015;58:95–109. doi: 10.1016/j.molcel.2015.01.042. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Bremm A, Freund SMV, Komander D. Lys11-linked ubiquitin chains adopt compact conformations and are preferentially hydrolyzed by the deubiquitinase Cezanne. Nat Struct Mol Biol. 2010;17:939–947. doi: 10.1038/nsmb.1873. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Hospenthal MK, Freund SMV, Komander D. Assembly, analysis and architecture of atypical ubiquitin chains. Nat Struct Mol Biol. 2013;20:555–565. doi: 10.1038/nsmb.2547. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Keusekotten K, et al. OTULIN antagonizes LUBAC signaling by specifically hydrolyzing Met1-linked polyubiquitin. Cell. 2013;153:1312–1326. doi: 10.1016/j.cell.2013.05.014. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Gladkova C, Maslen SL, Skehel JM, Komander D. Mechanism of parkin activation by PINK1. Nature. 2018;559:410–414. doi: 10.1038/s41586-018-0224-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Mevissen TET, et al. OTU Deubiquitinases Reveal Mechanisms of Linkage Specificity and Enable Ubiquitin Chain Restriction Analysis. Cell. 2013;154:169–184. doi: 10.1016/j.cell.2013.05.046. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Schubert AF, et al. Structure of PINK1 in complex with its substrate ubiquitin. Nature. 2017;552:51–56. doi: 10.1038/nature24645. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Komander D, et al. The structure of the CYLD USP domain explains its specificity for Lys63-linked polyubiquitin and reveals a B box module. Mol Cell. 2008;29:451–464. doi: 10.1016/j.molcel.2007.12.018. [DOI] [PubMed] [Google Scholar]
- 45.Lazarou M, McKenzie M, Ohtake A, Thorburn DR, Ryan MT. Analysis of the assembly profiles for mitochondrial- and nuclear-DNA-encoded subunits into complex I. Mol Cell Biol. 2007;27:4228–4237. doi: 10.1128/MCB.00074-07. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Fujiki Y, Fowler S, Shio H, Hubbard AL, Lazarow PB. Polypeptide and phospholipid composition of the membrane of rat liver peroxisomes: comparison with endoplasmic reticulum and mitochondrial membranes. J Cell Biol. 1982;93:103–110. doi: 10.1083/jcb.93.1.103. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Neuhauser N, Michalski A, Cox J, Mann M. Expert system for computer-assisted annotation of MS/MS spectra. Mol Cell Proteomics. 2012;11:1500–1509. doi: 10.1074/mcp.M112.020271. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Michel MA, Swatek KN, Hospenthal MK, Komander D. Ubiquitin Linkage-Specific Affimers Reveal Insights into K6-Linked Ubiquitin Signaling. Mol Cell. 2017;68:233–246.e5. doi: 10.1016/j.molcel.2017.08.020. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Mevissen TET, et al. Molecular basis of Lys11-polyubiquitin specificity in the deubiquitinase Cezanne. Nature. 2016;538:402–405. doi: 10.1038/nature19836. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Maspero E, et al. Structure of a ubiquitin-loaded HECT ligase reveals the molecular basis for catalytic priming. Nat Struct Mol Biol. 2013;20:696–701. doi: 10.1038/nsmb.2566. [DOI] [PubMed] [Google Scholar]
- 51.Kirisako T, et al. A ubiquitin ligase complex assembles linear polyubiquitin chains. EMBO J. 2006;25:4877–4887. doi: 10.1038/sj.emboj.7601360. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.