Biophysical Studies on Interactions and Assembly of Full-size E3 Ubiquitin Ligase

Background: The component subunits of CRL E3 ligases assemble into specific complexes. Results: Components of CRL5SOCS2 were identified from human cell lysate, the full-size complex was reconstituted in vitro, and protein-protein interactions were biophysically characterized. Conclusion: CRL5SOCS2 components exist in a monomeric state, and proposed structural models are supported by ion mobility mass spectrometry. Significance: We provide structural insights into the assembly of full-size CRL5SOCS2 that can aid development of small molecules targeting CRL complexes.

The N-terminal domain (NTD) of cullin proteins consists of three five-helix bundles ("cullin repeats") that form a long stalk architecture, and a globular C-terminal domain (CTD, or "cullin homology domain"). Cullin NTD recruits variable substrate receptors either directly or via an adaptor protein, whereas cullin CTD serves as a docking site for RING domain proteins that in turn recruit a cognate ubiquitin-loaded E2 (6). RING domain proteins contain a distinct Zn 2ϩ -binding domain characterized by a canonical RING motif. Selection of the substrate receptors for a particular CRL occurs through a specific receptor LPXP motif that forms a minor yet crucial supplementary interaction with cullin NTD (7,8).
SOCS2 is a member of the SOCS box protein family that, in association with the adaptor elongin B-elongin C complex (EloBC), cullin 5 scaffold, and Rbx2, constitutes a CRL5 SOCS2 E3 ligase. SOCS2 contains three structural domains: a conserved C-terminal SOCS box domain that binds to adaptor EloBC; a central SH2 domain mediating recruitment of phosphorylated tyrosine-containing sequence of the substrate; and a variable N-terminal region that facilitates interaction with the substrate. CRL5 SOCS2 negatively regulates growth hormone signaling by targeting growth hormone receptor (GHR) for ubiquitination and proteasomal degradation (9). Phosphorylated tyrosine 595 serves as the key structural determinant of GHR recognition by the SH2 domain of SOCS2 (10). Crystal structures of SOCS2-EloBC (PDB code 2C9W) (11) and SOCS2-EloBC-Cul5 NTD (PDB code 4JGH) (12) describe structural features of the protein-protein interfaces within these leftarm complexes. However, the details of assembly of the full-size CRL5 SOCS2 E3 complex both in vivo and in vitro remain missing.
In recent years, interest in studying the structure, function, and assembly of CRLs has been growing, notably driven in part by their potential role as drug targets in a number of human diseases (13)(14)(15)(16). However, only a few studies have investigated the full-size CRL complexes biophysically, primarily due to difficulties in obtaining some of the protein components recombinantly, in particular full-length cullins. Furthermore, large heteromeric protein complexes such as CRLs are notoriously difficult to crystallize into diffraction quality crystals. Therefore, it seems promising to engage the strengths of diverse biophysical methods in order to facilitate characterization of both the individual subunits and the full-size complexes as well as to provide a means for examining their association and interactions.
Here, we show that all components of the CRL5 SOCS2 could be pulled down from a cell lysate via SOCS2-mediated recognition of the phosphorylated GHR_pY595 peptide immobilized on beads. The full-length E3 ligase complex was then reconstituted in vitro using purified recombinant proteins and characterized biophysically. Investigations of assembly and interactions within the complex were carried out using size exclusion chromatography and multiangle light scattering (SEC-MALS), isothermal titration calorimetry (ITC), and nanoelectrospray traveling wave-ion mobility mass spectrometry (TWIM-MS).

EXPERIMENTAL PROCEDURES
Pull-down Experiments-Pull-down experiments were performed using biotinylated GHR-derived 11-mer peptides phosphorylated (GHR_pY595) or not (GHR_Y595) on tyrosine 595, harboring an aminohexanoic acid as spacer after the biotin (Biotin-aminohexanoic acid-PVPDpYTSIHIV-amide), and immobilized on high capacity Neutravidin beads. Competition experiments were performed by incubating human K562 total cell lysate with 100 M non-biotinylated phosphorylated peptide (GHR_pY595) and the immobilized (Biotin-aminohexanoic acid-PVPDpYTSIHIV-amide) beads for 2 h at 4°C. After washing, bound proteins were eluted with SDS-sample buffer and prepared for tandem mass tags labeling and MS analysis as described previously (17).
Protein Expression and Purification-Recombinant human SOCS2 (amino acids , elongin C (amino acids 17-112), and elongin B (amino acids 1-118) were co-expressed in Escherichia coli BL21(DE3) from the pLIC (SOCS2) and pCDF_ Duet (EloBC) plasmids (gifts from A. Bullock, Structural Genomics Consortium, Oxford, UK). A starter culture was grown overnight from a single transformant colony using 50 ml of LB medium containing 100 g/ml ampicillin and 50 g/ml streptomycin. Starter culture then was used to inoculate 7 liters of LB medium containing 100 g/ml ampicillin and 50 g/ml streptomycin. The cells were grown at 37°C until A 600 ϳ0.7 and cooled to 18°C, and then protein expression was induced with 1 mM isopropyl ␤-D-1-thiogalactopyranoside for 12 h.
Recombinant human Cul5 NTD (N-terminal domain, residues 1-386) was expressed in E. coli BL21(DE3) from a pNIC plasmid encoding sequence for Cul5 NTD , containing His 6 and FLAG tags at the C-terminal end and a tobacco etch virus (TEV) cleavage site, as described previously (18). Briefly, 50 ml of starter culture was grown overnight using a single transformant colony in LB medium containing 50 g/ml kanamycin and used to inoculate 2 liters of LB medium supplemented with 50 g/ml kanamycin. The cells were grown at 37°C until A 600 ϳ0.7 and cooled to 18°C, and protein expression was induced with 0.5 mM isopropyl ␤-D-1-thiogalactopyranoside for 12 h.
Recombinant SOCS2-EloBC and Cul5 NTD were independently purified using the following protocol. The cell pellets were harvested by centrifugation at 5,000 rpm and 4°C for 30 min and resuspended in binding buffer (50 mM HEPES, pH 7.5, 500 mM NaCl, 5% glycerol, 0.5 mM TCEP). The supernatant was treated with 10 g/ml DNase I, 10 mM MgCl 2 for 30 min and then filtered through a 0.22-m filter. The sample was applied on a HisTrap column (GE Healthcare), and the resin was washed with wash buffer (50 mM HEPES, pH 7.5, 20 mM imidazole, 500 mM NaCl, 5% glycerol, 0.5 mM TCEP) and then the bound proteins were eluted with an incremental gradient of elution buffer (50 mM HEPES, pH 7.5, 500 mM imidazole, 500 mM NaCl, 5% glycerol, 0.5 mM TCEP). Fractions containing protein were pooled, and the His 6 tag was cleaved off by overnight dialysis in the presence of TEV protease at 4°C in binding buffer. The protein was applied to a HisTrap column for a second time, collecting the flow-through, and then concentrated and purified on a HiLoad 16/60 Superdex 75 column with running buffer 25 mM HEPES, pH 7.5, 250 mM NaCl, 0.5 mM TCEP.
Recombinant human Cul5 (amino acids 1-780) and Rbx2 (amino acids 1-113) were co-expressed in Sf21 insect cells using pFastBac TM Dual vector in the Bac-to-Bac baculovirus expression system. In this vector, Cul5 is N-terminally tagged with a fragment of bacterial PBP5 (Dac tag) (19), which can be removed with TEV protease, as described previously (20).
Bacmids for Dac-TEV-Cul5/Rbx2 were generated in DH10BAC cells and transfected into Sf21 cells, using Cellfectin II reagent (Invitrogen). The transfected cells were kept for 7 days at 27°C in Insect Express medium (Lonza), supplemented with ANTI-ANTI (Invitrogen). The cells were sedimented, and the virus-containing supernatant was used to infect 150 ml of Sf21 culture at a density of 1.5 ϫ 10 6 cells/ml. For the expression of RING E3 ligases, we supplemented the Insect Express medium with 5 M ZnCl 2 . After 5 days the cells, were collected under sterile conditions, and the supernatant was used to infect 2 liters of Sf21 cell culture for 3 days. The Cul5-Rbx2 protein complex was purified using the following procedure. Cells were harvested by centrifugation at 3,500 rpm and 4°C for 15 min and then resuspended in 25 ml of 50 mM HEPES, pH 7.4, 0.1 mM EGTA, 1 M ZnCl 2 , 1 mM TCEP, 1 mM Pefabloc, and 20 g/ml leupeptin (both from Apollo Scientific) and incubated for 15 min at 4°C. Cells were sheared using a 50-ml tight fit Dounce homogenizer, and insoluble material was removed by centrifugation at 40,000 rpm and 10°C for 20 min. To perform Dac affinity purification, the supernatant was gently mixed with ampicillin-Sepharose at room temperature for 50 min. The Sepharose was collected by centrifugation and washed six times in 10 volumes of 50 mM HEPES, pH 7.4, 150 mM NaCl, 0.1 mM EGTA, 1 M ZnCl 2 , 1 mM TCEP. To recover untagged Cul5-Rbx2, the Sepharose was incubated with TEV protease (50 g of TEV, 1 ml of Sepharose) overnight at room temperature and drained and washed through Econopac filter units (Bio-Rad). The Cul5-Rbx2 protein was concentrated and further purified by preparative size exclusion chromatography using HiLoad 16/600 Superdex 200 column (GE Healthcare) with 50 mM HEPES, pH 7.4, 150 mM NaCl, 10% glycerol, 0.5 mM TCEP.
The identity and purity of all obtained proteins was confirmed using denaturing electrospray ionization-MS and SDS-PAGE. All proteins were flash-frozen using liquid nitrogen and stored at Ϫ80°C.
SEC-MALS-SEC-MALS experiments were performed using Dionex Ultimate 3000 UHPLC system (Thermo Scientific) with an inline Wyatt miniDAWN TREOS MALS detector and Optilab T-rEXTM refractive index detector. Molar masses spanning elution peaks were calculated using ASTRA version 6.0.0.108 (Wyatt). SEC-MALS data were collected for the following sam- TWIM-MS-The native TWIM-MS experiments were conducted on a Synapt HDMS G2 instrument (Waters, Milford, MA), which has been described previously (21). Samples following gel filtration were buffer-exchanged into 500 mM aqueous ammonium acetate at pH 7.0, using Micro Bio-Spin P-6 columns (Bio-Rad), at concentrations in the range of 5-10 M. Aliquots of 3-5 l were transferred to gold-coated nano-electrospray ionization needles prepared in house. The instrument was tuned to ensure the preservation of non-covalent interactions (22), using the following parameters: capillary, 1.2 kV; sample cone, 40 V; extraction cone, 0.5 V; nanoflow gas pressure, 0.3 bar; trap collision energy, 4.0 V; transfer collision energy, 3.5 V; backing pressure, 4 millibars, trap pressure, 3.4 millibars. For the measurement of the full 148-kDa complex, SOCS2-EloBC-Cul5-Rbx2, the backing pressure was increased to 5 millibars to facilitate the transmission of high m/z signal. Gas pressure in the ion mobility cell was 3.0 millibars, and helium and N 2 gas flows were 180 and 90 ml/min, respectively, with a trap bias of 50 V. The traveling wave velocity was 800 m/s with a traveling wave height of 40 V. The data were acquired and processed with MassLynx version 4.1 software (Waters), and drift times were extracted using Driftscope version 2.3 (Waters). The experimental collision cross-sections (CCS) of the protein complexes were determined by calibration with known protein cross-sections determined under native conditions as described previously (23).
Calculation of Theoretical CCS-Theoretical CCS values of the protein complexes were calculated from model structures, obtained by docking individual protein subunits together, using the program MOBCAL with both the projection approximation (PA) and the exact hard sphere scattering (EHSS) methods (24,25). The PDB files were cleaned (i.e. by resolving dihedral conflicts and adding missing side chains and removing crystal water molecules) prior to the PA or EHSS calculation. The theoretical CCS was compared with the experimental CCS of the lowest available charge state for that species in the mass spectra, which corresponds to the most native-like structure of the protein complex (Fig. 4, A-C, bottom panels).
Molecular Modeling of Protein Complexes-Due to the absence of an Rbx2 crystal structure, its closest homolog, Rbx1, was used for the model construction. The structural model of the SOCS2-EloBC-Cul5-Rbx1 complex was prepared in PyMOL, using the crystal structure of Cul1-Rbx1-Skp1-Skp2 (PDB code 1LDK) (26) as the initial template. To construct the model, SOCS2-EloBC-Cul5 NTD (PDB code 4JGH) (12) was superimposed on the template by aligning its Cul5 NTD subunit with the Cul1 NTD of the template. After that, Cul5 CTD -Rbx1 (PDB code 3DPL) (27) was aligned with Cul1 CTD subunit of the template to generate a model of the full-length E3 ligase. The resulting model of the CRL5 SOCS2 complex was used to obtain the model for Cul5-Rbx1. To generate a model of the "open" neddylated complexes, NEDD8ϳCul5 CTD -Rbx1 (PDB code 3DQV) (27) was aligned with the Cul1 CTD subunit of the template. Alternatively, to prepare a model of the closed neddylated complexes, NEDD8 was simply added from aligned 3DQV onto the non-neddylated Cul5 CTD -Rbx1 and SOCS2-EloBC-Cul5-Rbx1.

Components of CRL5 SOCS2 E3 Ligase Can Be Pulled Down from Human Cell Lysates Using Phosphopeptide-modified
Beads-Specific subunits of E3 ligase SOCS2, EloB, EloC, Cul5, and Rbx2 are known to function as a CRL5 SOCS2 complex (28). The SH2 domain of SOCS2 recognizes and specifically binds a GHR sequence containing the phosphorylated tyrosine Assembly and Interactions of CRL5 SOCS2 E3 Ligase FEBRUARY 13, 2015 • VOLUME 290 • NUMBER 7 Tyr(P) 595 . We envisaged that SOCS2 and components of fulllength CRL5 SOCS2 E3 ligase should be amenable for capturing from cell lysate using substrate peptides immobilized on beads. With this aim, we performed pull-down experiments from human K562 cell lysate using beads decorated with both phosphorylated GHR_pY595 (PVPDpYTSIHIV-amide, positive control) and non-phosphorylated GHR_Y595 (PVP-DYTSIHIV-amide, negative control) peptides. Mass spectrometry analysis revealed a reproducible and limited set of proteins captured and subsequently displaced by the phosphorylated peptide (Fig. 1A, bottom left corner). All components of the CRL5 SOCS2 (SOCS2, EloB, EloC, Cul5, and Rbx2) were among this protein set.
Binding profile (Fig. 1B) shows that endogenous components of CRL5 SOCS2 E3 ligase were only captured by GHR_pY595beads. In contrast, no significant capturing from cell lysates was observed with non-phosphorylated GHR_Y595-containing beads (Fig. 1B). This observation shows that the Tyr(P) residue plays a key role in recognition of the substrate by the CRL SOCS2 complex in the cell and that phosphorylation of the peptide is essential for specific interaction with the E3 ligase. Interestingly, we also detected NEDD8 as a protein specifically pulled down by the phosphorylated peptide. NEDD8 is a ubiquitin-like protein that is known to be covalently attached to cullins and acts as a CRL activator by inducing scaffold dynamics and increasing conformational flexibility of the E3 enzyme (27,29). Identification of NEDD8 suggests that the active neddylated complex is also being pulled down in the assay.
Surprisingly, in addition to the expected CRL5 SOCS2 complex subunits, we detected subunits of the CRL1 FBXO31 E3 ligase, namely FBXO31, Rbx1, Cul1, and Skp1 proteins, as being cap-tured by the beads and displaced by the phosphorylated peptide ( Fig. 1A and Table 1), indicating specific binding. FBXO31 is an F box protein that binds phosphorylated substrates; therefore, it could have been recruited by the GHR_pY595 peptide directly. However, significant recruitment of the four CRL1 FBXO31 subunits was also observed by the non-phosphorylated GHR peptide (Fig. 1C). This would imply a degree of phosphorylation-independent interaction, either directly with the beads or indirectly via binding to the components of the CRL5 SOCS2 E3 ligase complex.
Moreover, CSK (C-terminal Src kinase) and CISH (cytokineinducible SH2-containing) proteins were also recruited ( Fig. 1A and Table 1). Both CSK and CISH contain the SH2 domain; therefore, both proteins were probably directly recruited by the phosphorylated peptide.
SOCS2-EloBC Forms a Weak Interaction with GHR and a Tight Interaction with Cul5 NTD -To determine the affinity of interaction and the thermodynamic parameters of binding between SOCS2-EloBC and substrate GHR or scaffold Cul5 NTD , we performed isothermal titration calorimetry experiments (Fig. 2). SOCS2-EloBC binds GHR_pY595 peptide with K d ϭ 1.8 M, which is consistent with the previously reported value (11), and binds Tyr(P) with K d ϭ 191 M, both at 298 K ( Fig. 2A). The binding affinity for phosphotyrosine is ϳ100-fold weaker than for the phosphorylated peptide, suggesting that other peptide residues make some contribution to interaction with the protein. However, negative control titration using non-phosphorylated GHR_Y595 peptide showed no binding ( Fig. 2A), reinforcing the key contribution of the phosphate group to substrate binding.
We next determined the affinity of SOCS2-EloBC for Cul5 NTD by measuring a K d ϭ 11 nM for the interaction (Fig. 2C). These data are in good agreement with the previously reported K d of 28 nM (by ITC) (12). In addition, two groups independently reported K d ϭ 7 nM (by ITC) (30), K d ϭ 10 nM (by ITC), and K d ϭ 47 nM (by surface plasmon resonance) (31) for this interaction, albeit using SOCS box domain instead of the whole SOCS2 protein in complex with EloBC.
To test the potential cooperativity of interactions at the GHR/SOCS2-EloBC/Cul5 NTD interfaces, we performed titration of GHR_pY595 peptide into SOCS2-EloBC-Cul5 NTD and titration of Cul5 NTD into GHR_pY595-SOCS2-EloBC complex. No change in the K d or ⌬H values was observed in either case, suggesting no cooperativity or cross-talk between these interactions.
The interaction between SOCS2-EloBC and the Cul5 scaffold is high affinity and crucial to the assembly of CRL complex. To provide further insights into the nature of this interaction, we performed temperature-dependent ITC titrations and determined a change in heat capacity ⌬C P ϭ Ϫ450 cal/mol/K (titration curves shown in Fig. 2B). Fig. 2D demonstrates a plot with a temperature-dependent change of thermodynamic parameters of SOCS2-EloBC/Cul5 NTD interaction. The experimental ⌬C P value is calculated from the slope of the ⌬H linear regression. As a comparison, previously reported ⌬C P values for ASB9-EloBC/Cul5 NTD and Vif-EloBC/Cul5 NTD interactions were found to be Ϫ350 cal/mol/K (18) and Ϫ300 cal/ mol/K (30), respectively.

TABLE 1 Proteins enriched by the phosphorylated GHR_pY595-modified beads
The proteins were specifically captured by the beads and displaced by the phosphorylated GHR peptide. We next calculated the theoretical solvent-accessible surface area values in GetArea (32) and NACCESS (33) software using the crystal structure of SOCS2-EloBC-Cul5 NTD complex (PDB code 4JGH) as a model (12) ( Table 2). Theoretical ⌬C P values were calculated using the following equation (34),

Protein UniProt ID Comments
where ⌬ASA is the apolar (ap) and polar (p) surface buried upon interaction of the proteins, and ⌬c is the area coefficient, representing per Å 2 contribution of residues in heat capacity change. The polar and non-polar area coefficients represent values empirically determined from a range of protein data sets by different groups (35-39) (reviewed in Ref. 34). We observe good agreement between theoretical and experimental data when using area coefficients according to Refs. 39 and 37 (Table 3).
SOCS2-EloBC Forms Stable Monomeric Complexes with Cul5 NTD and Cul5-Rbx2-To validate formation of CRL5 SOCS2 and determine the stoichiometry of subunits in the complex, we demonstrated assembly of the full-length E3 ligase in vitro using recombinantly expressed and purified protein components (schematic representation in Fig. 3A). SOCS2 and EloBC were co-expressed in E. coli to obtain the SOCS2-EloBC ternary complex, and

JOURNAL OF BIOLOGICAL CHEMISTRY 4183
Cul5 NTD was independently expressed in E. coli. The Cul5-Rbx2 protein complex was co-expressed in Sf21 insect cells.
The SEC-MALS elution profiles of the different protein components and complexes show that they all exist as monomeric and monodisperse entities (Fig. 3D). The molar mass over elution peaks is shown in corresponding colors. Molecular weight values of eluted proteins are summarized Fig. 3C. The results of SEC-MALS analysis confirm the formation of expected protein complexes with experimentally determined molecular weights that correlate well with theoretical values.
Experimental Collision Cross-sections for Protein Complexes Are in Good Agreement with Theoretical Values-To validate the structural model of SOCS2-EloBC-Cul5-Rbx2, TWIM-MS was used to examine the molecular weight and stoichiometry of the intact protein complexes as well as to confirm their topology by CCS measurements. The protein components SOCS2-EloBC and Cul5 NTD alone were first analyzed using native MS, and the resulting spectra are shown in Fig. 7, A and B. The masses were confirmed as ϳ43 and 45 kDa, respectively. Theoretical and experimental masses for each complex are shown in Table 4.
Combining SOCS2-EloBC with Cul5 NTD produced an 89-kDa complex, which could be detected with charge states ranging from 16ϩ to 20ϩ (Fig. 4A). Some free SOCS2-EloBC was also observed in this spectrum, with the same charge states as the native mass spectrum of SOCS2-EloBC alone, indicating that some dissocia-

TABLE 3
Large negative ⌬C P value for SOCS2-EloBC/Cul5 NTD interaction indicates a highly hydrophobic interface between the proteins; comparison between theorectical and experimental ⌬C P values shows good agreement tion in solution occurs (Fig. 7A). Table 4 shows the experimental CCS values compared with the theoretical ones calculated with the PA and the EHSS methods. It is normally expected that the experimental values would be smaller than the EHSS results and larger than the PA results (24,25). The collision cross-section determined using ion mobility for SOCS2-EloBC-Cul5 NTD is 5,092 Å 2 for the most native charge state (16ϩ; Fig. 4A, bottom), which is reasonably close to the theoretical value calculated for the model (Table 4, PA value 5,306 Å 2 ).
A typical spectrum of the Cul5-Rbx2 complex is shown in Fig.  7C with a predominant 6ϩ charge state for Rbx2 and a series of charge states from 19ϩ to 22ϩ representing the binary complex (104 kDa). Fig. 4B shows the Cul5-Rbx2 complex in more detail in addition to the drift time plot. Moreover, there are less intense peaks to the left-hand side of the predominant peak corresponding to a loss of ϳ1 kDa from the complex that may represent a truncation in the Cul5 subunit. These species are clearly separated, however, by their ion mobility (Fig. 4B), so it is possible to calculate a collisional cross-section for the intact complex. The CCS value from these data for the most native 19ϩ charge state was found to be 6,061 Å 2 (Fig. 4B, bottom), which compared well with the theoretical value (Table 4, PA value 5,988 Å 2 ).
The native mass spectrum of the SOCS2-EloBC-Cul5-Rbx2 showed intense peaks at 3,000 -4,000 m/z, indicating a relative abundance of free SOCS2-EloBC, with charge states 11ϩ to 13ϩ, as described previously (Fig. 7A). It is possible that there is   FEBRUARY 13, 2015 • VOLUME 290 • NUMBER 7 an excess of SOCS2-EloBC in these samples or that this subunit has a greater ionization efficiency compared with the other protein components. Fig. 4C (bottom) depicts the 4,000 -7,000 m/z range of the SOCS2-EloBC-Cul5-Rbx2 spectrum that shows peaks ranging from 4,300 to 4,900 m/z representing a small amount of SOCS2-EloBC dimer. Second, at 4,700 -5,600 m/z, the Cul5-Rbx2 complex is detected with charge states from 19ϩ to 22ϩ. Finally, the peaks representing the full 148-kDa complex, SOCS2-EloBC-Cul5-Rbx2, are in the range of 5,600 -6,600 m/z, with charge states 23ϩ to 26ϩ.

Assembly and Interactions of CRL5 SOCS2 E3 Ligase
The CCS values measured for each charge state of the full 148-kDa complex are displayed in Fig. 4C (bottom). For the lowest charge state of SOCS2-EloBC-Cul5-Rbx2, an experimental cross-section of 7,653 Å 2 was determined compared with a theoretical CCS of 7,918 Å 2 (Table 4), confirming the structural model as shown in Fig. 10D. In this case, the experi- mentally determined value is slightly smaller than the theoretical value, which could indicate that the structure is slightly more compact than the model suggests.
To investigate the effect of neddylation on the complex assembly, we performed in vitro neddylation assays on the purified Cul5-Rbx2 complex (Fig. 8) and used the reaction product NEDD8ϳCul5-Rbx2 to reconstitute neddylated full complex SOCS2-EloBC-Cul5-Rbx2. The masses for neddylated complexes (Table 4, 113 and 157 kDa, respectively) are in agreement with the theoretical mass for the addition of NEDD8. First, the same range of charge states was observed for the non-neddylated and neddylated complexes in the native mass spectra (Fig.   9). This would indicate that no significant conformational rearrangement had occurred. Second, the CCS values measured by TWIM-MS for each charge state of the neddylated complexes were compared with those of the non-neddylated ones (Fig. 5, A  and B), showing an increase in CCS of 150 -200 Å 2 in each case (Table 4). We compared the experimental data with two alternative models: one "open" model that assumes a conformational change upon neddylation, as observed crystallographically for NEDD8ϳCul5 CTD -Rbx1 (27), and a second "closed" model that simply has NEDD8 added onto the non-neddylated complex without any conformational rearrangement, with the aim to distinguish them based on the TWIM-MS data. The  Assembly and Interactions of CRL5 SOCS2 E3 Ligase FEBRUARY 13, 2015 • VOLUME 290 • NUMBER 7 calculated CCS increase for the two alternative neddylated models (NEDD8ϳCul5-Rbx1) versus the non-neddylated one (Fig. 10, A-C), is 350 Å 2 for the closed model and 523 Å 2 for the open model (PA method; Table 4), whereas for SOCS2-EloBC-NEDD8ϳCul5-Rbx1, the neddylation accounts for an extra 312 Å 2 (closed) and 415 Å 2 (open). It would therefore appear that the increase in size is predominantly due to the addition of NEDD8 rather than a significant conformational change.

DISCUSSION
Here, we show that all CRL5 SOCS2 components SOCS2, EloBC, Cul5, and Rbx2 can be specifically pulled down from the human cell lysates with subsequent validation of their identity by MS analysis. These components were recombinantly expressed and purified and then assembled in vitro into different sized complexes up to the full-size E3 ligase.
Biophysical studies of the full-size CRLs are important for better understanding the principles of assembly and to gain insight into their structural architecture. This is particularly relevant for the cases where crystal structures are not available, as for CRL5 SOCS2 . We addressed this by presenting the first report of in vitro assembly of full-size human CRL5 SOCS2 reconstituted from recombinant components and provide a biophysical analysis of the obtained complexes.
The structural model of the SOCS2-EloBC-Cul5-Rbx2 complex was validated by TWIM-MS studies. The experimentally measured CCS values are in agreement with the theoretically calculated ones, although the molecular architecture of SOCS2-EloBC-Cul5 NTD and SOCS2-EloBC-Cul5-Rbx2 appears to be slightly more compact than predicted.
Modification of the cullin scaffold with NEDD8 protein is crucial for the activation of CRLs (40). Previous structural studies showed that NEDD8 promotes a conformational rearrangement of the Cul5-Rbx1 component of CRL5 (27). Such a structural alteration enables Rbx1-E2ϳubiquitin to extend toward the substrate receptor subunit, thereby promoting substrate polyubiquitination. One of the proteins identified in the pull-down experiments was NEDD8, supporting the presence of the active neddylated complex inside cells. Comparison of TWIM-MS data between neddy-lated and non-neddylated Cul5-Rbx2 and SOCS2-EloBC-Cul5-Rbx2 complexes showed an increase of 150 -200 Å 2 in CCS values (Table 4). Interestingly, the increase in calculated CCS values defined by the addition of NEDD8 is ϳ2-3 times larger than the conformational change of the Rbx1 subunit in the open models (Fig. 10, C and F). Therefore, the difference between open and closed neddylated models is not significant enough, and we cannot distinguish them based on the experimental data. The observed change in CCS is therefore largely due to the addition of the extra NEDD8 subunit.
One of the main limitations of biophysical studies of the whole multisubunit CRLs is the difficulty of obtaining all of the components in appropriate amount and quality, particularly  full-length cullins. Expression and purification of stable fulllength cullin scaffolds in complex with RING domain proteins is not trivial and has been previously reported only for Cul1-Rbx1 (26), Cul4A-Rbx1 (41), Cul4B-Rbx1 (42), and Cul5-Rbx2 (20). As a result, there were only a few cases in the literature describing characterization of the full-size CRLs assembled from recombinant subunits (26,42). To purify the Cul5-Rbx2 complex in this study, we used the Dac tag technology, which provides additional stability and solubility to the protein complex and additionally improves the yield of recombinant proteins (19). This approach has also proven to be successful for purification of Cul2-Rbx1 complex 4 and could be further extended to other cullins and large multisubunit complexes.
In certain cases, CRLs exist and function in homo-or heterooligomeric states. The biological implications of CRL oligomer-ization are postulated to include activity regulation, enhancement of substrate ubiquitination, and alternative mechanistic aspects of ubiquitin transfer (6). For example, several studies have shown that CRL3 can dimerize via an adaptor BTB domain (43) or through NEDD8-mediated interaction between two Cul3 scaffolds (44). CRL1 was also demonstrated to be able to dimerize via the receptor Cdc4 (cell division control protein 4), resulting in enhanced ubiquitination of substrate Sic1 (45). Additional examples include other BTB receptor/adaptor subunits of CRL3 (46 -49), F box receptors of CRL1 (45, 50 -53), and the DCAF receptor of CRL4 (54). More recently, a two-site model for substrate recognition was proposed for CRL3 KLHL11 based on the crystal structure of KLHL11-Cul3 NTD (49). However, no evidence for dimerization of elongins, cullin 2 or cullin 5, or SOCS subunits has been reported to date. In this work, using SEC-MALS and native MS techniques, we have established that CRL5 SOCS2 exists in a monomeric state. We provide (open). The models were assembled using available crystal structures. Due to the lack of Rbx2 crystal structure, its closest homolog, Rbx1, was used instead. These models were used for calculation of the theoretical CCS values. FEBRUARY 13, 2015 • VOLUME 290 • NUMBER 7 a structural model validated by TWIM-MS studies that suggests a similar mechanism of ubiquitin transfer to a previously reported monomeric CRL1 Skp2 complex (26). Therefore, according to the proposed model, CRL5 SOCS2 oligomerization does not seem to be necessary for enzyme activity.

Assembly and Interactions of CRL5 SOCS2 E3 Ligase
Our measured K d values for the interaction of SOCS2-EloBC with GHR_pY595 peptide or Cul5 NTD are in good agreement with previously reported data (11,31). Weak SOCS2-EloBC/ GHR_pY595 interaction (K d ϭ 1.8 M) could suggest low selectivity toward a particular substrate and instead the ability to target a variety of phosphorylated proteins. In contrast, the interaction of SOCS2-EloBC with scaffold Cul5 NTD is very tight (K d ϭ 11 nM at 298 K). The large negative ⌬C P value (Ϫ450 cal/mol/K) for the SOCS2-EloBC/Cul5 NTD interaction indicates a major contribution of the hydrophobic interface and further reflects the high affinity (34, 38) (e.g. when compared with other related interactions, such as ASB9-EloBC/Cul5 NTD and Vif-EloBC/Cul5 NTD ) (18,30). Overall, these results indicate the structural importance of the SOCS2-EloBC/Cul5 NTD interface for assembly and stability of the CRL5 SOCS2 .
As the next logical step following the current study, we believe it would be important to develop an assay to measure activity of the recombinant CRL5 SOCS2 against the substrate GHR protein resulting in ubiquitination and the subsequent proteasomal degradation of the latter. Such an assay could be useful for testing the potency of small molecule modulators of CRL5 SOCS2 activity. In accordance with this, a recent example demonstrates in vitro reconstitution of murine CRL5 SOCS3 , containing SOCS3, a close homolog of SOCS2, as a substrate receptor subunit (55). The authors used co-expressed Cul5 NTD , Cul5 CTD , and Rbx2 proteins to form a complex with SOCS3-EloBC, and the assembled E3 ligase then demonstrated activity in the ubiquitination assay against substrates JAK2 and gp130.
In addition, it would be important to obtain the crystal structure of the receptor SOCS2 bound to the substrate GHR depicting the details of the interface between these two proteins. This could substantially advance the development of inhibitors of this interaction (i.e. structural phosphotyrosine analogs or isosteres). The biophysical insights into the interactions and assembly of the full-size CRL5 SOCS2 E3 ligase reported in our study will aid future developments in this direction.