Structural and functional insights into the E3 ligase, RNF126

RNF126 is an E3 ubiquitin ligase that collaborates with the BAG6 sortase complex to ubiquitinate hydrophobic substrates in the cytoplasm that are destined for proteasomal recycling. Composed of a trimeric complex of BAG6, TRC35 and UBL4A the BAG6 sortase is also associated with SGTA, a co-chaperone from which it can obtain hydrophobic substrates. Here we solve the solution structure of the RNF126 zinc finger domain in complex with the BAG6 UBL domain. We also characterise an interaction between RNF126 and UBL4A and analyse the competition between SGTA and RNF126 for the N-terminal BAG6 binding site. This work sheds light on the sorting mechanism of the BAG6 complex and its accessory proteins which, together, decide the fate of stray hydrophobic proteins in the aqueous cytoplasm.


Results
NMR solution structure of the RNF126 N-terminal zinc finger. The N-terminal region of RNF126 (residues 1-100 previously defined as the BAG6-interacting module 4 ) yields a 1 H-15 N HSQC spectrum indicative of a partially folded protein. Backbone assignment of residues 1-66 shows that residues 1-40 constitute the structured region while the remainder presents the chemical shift dispersion typical of disordered proteins. A new construct (residues 1-40, hereafter named RNF126_NZF) displays comparable spectra to the longer version, showing that the additional truncation has no effect on the folding of the zinc finger region ( Figure S1).
We acquired the complete battery of triple resonance experiments using the shorter construct, RNF126_NZF, to facilitate full backbone and side-chain assignment (BMRB Accession Number: 25913). We solved the solution structure using standard methods analogous to our previous approach 28,29 (Fig. 1A; PDB Accession Number: 2N9O; statistical parameters in Table 1). The folded domain comprises residues Arginine 10 to Leucine 40. This region contains a three-stranded short antiparallel β -sheet (β 1 = Y11-C13; β 2 = V18-I20; β 3 = I37-E39) and a zinc finger involving four cysteine residues; one pair from the β 1 -β 2 short loop (C13; C16) and the other pair from the long structured region between strands β 2 and β 3 (C29; C32); we confirmed the equimolar presence of zinc cation by ICP/MS. The first zinc coordination shell contains the four cysteine side-chains in tetrahedral conformation (Fig. 1B) whose chemical shift values for 13 Cα (~59 ppm) and 13 Cβ (~31 ppm) are consistent with their zinc-binding character 30 . The second coordination shell is defined by the formation of three hydrogen bonds between the cysteine sulphur atoms (C13, C29 and C32) and the amide group of the residue at position + 2 (C15, R31 and S34 respectively) (Fig. 1B); we summarise all zinc coordination parameters in Table S1. The long structured loop β 2 -β 3 folds against one side of the β -sheet burying several non-polar residues and creating a stable hydrophobic core; by contrast the other side of the β -sheet is solvent exposed, containing some glutamic acid side-chains and aromatic rings and constituting a putative platform for protein-protein interaction (Fig. 1C).

Biophysical characterisation of RNF126-BAG6 interaction.
To analyse the binding interface between RNF126 and BAG6, we carried out reciprocal chemical shift perturbation (CSP) studies by titrating unlabelled RNF126_NT (both 1-40 and 1-100 constructs) into 15 N-labelled BAG6_UBL and vice versa. We used our backbone assignments described above in concert with the assignments for BAG6_UBL deposited in the BMRB (Accession Number: 11263). Both RNF126 constructs show comparable shifts upon BAG6_UBL binding and have the same effect on the BAG6_UBL spectrum confirming that residues 1-40 are necessary and sufficient for the interaction ( Figure S2). We therefore adopted the 1-40 construct for all subsequent experiments. The spectra show binding in a slow exchange regimen (Figs 2A and S3A), suggesting a high-affinity complex with a saturation point close to a 1:1 molar ratio. The assignment of the bound state spectra showed that the β -sheet (predominantly the β 3 strand) of RNF126_NZF is clearly affected upon titration, together with some residues from the β 2 -β 3 loop. The binding region in BAG6_UBL is located along its exposed β -sheet; β 3 , β 4 , β 5 strands, the loop between β 1 and β 2 and also includes part of the C-terminal region (Fig. 2B). In addition, we confirmed and characterised the binding between RNF126_NZF and BAG6_UBL by native gel electrophoresis, microscale thermophoresis (MST; Figure S4) and isothermal titration calorimetry (ITC) (Fig. 2C). ITC results indicate 1:1 stoichiometry with a dissociation constant of 0.40 ± 0.05 μ M in agreement with the slow exchange regime of the NMR titration data. The favourable enthalpy and entropy values obtained from ITC (Δ H = − 4.52 ± 0.04 kcal/mol; Δ S = 14.1 ± 0.4 cal/mol·K) suggest that the complex formation between RNF126 and BAG6 is driven by the establishment of both hydrogen bonds and hydrophobic interactions.
Finally to further characterize the interface, we designed some RNF126_NZF mutants based on the largest chemical shift perturbations (His14/Ala, Phe36/Ala and double mutant for charge inversion Glu38/Arg Glu39/ Arg) and analysed their effect on BAG6_UBL interaction using ITC (Fig. 2C), native PAGE mobility shift assay and NMR ( Figure S5). All mutations result in a drop in binding affinity of at least an order of magnitude, proving that they are involved in the interaction with BAG6. The wild-type structure is preserved in each mutant, as judged by 1D NMR ( Figure S5). RNF126_NZF/BAG6_UBL complex structure. The stability of the complex and the high quality NMR data facilitated experimental structure solution of the RNF126_NZF/BAG6_UBL complex with no need to resort to in silico methods. We acquired complete triple resonance experiments for backbone and side-chain assignment of the bound state for each protein (BMRB Accession Number: 25914) and we used 3D NOESY experiments to calculate the complex structure (PDB Accession Number: 2N9P; statistical parameters in Table 2). We unambiguously assigned reciprocal intermolecular NOEs between RNF126 and BAG6 ensuring the efficacy of the structure calculation. The complex shows overall conservation of the original fold for each protein and a perfect match between the binding interface of the obtained structure with our CSP analysis (Fig. 3A). RMSD values between the free form of RNF126_NZF (this work) and the bound state are low (backbone RMSD = 0.62) and the zinc coordination parameters are conserved (Table S2); BAG6 also presents low values of RMSD comparing free (PDB: 1WX9) and bound states (backbone RMSD = 1.43), but in this case there is a minor modification of loop β 1 -β 2 configuration ( Figure S6). A variety of interactions is observed in the interface between both proteins including hydrogen bond formation and electrostatic and hydrophobic interactions, consistent with our ITC data (Fig. 2C). A complex hydrogen bond network is detected involving the guanidinium groups of three arginine residues in BAG6 and the carboxyl group of two glutamic acids in RNF126: Arg64 side-chain forms hydrogen bonds with Glu39 and Glu38 is similarly stabilised by Arg58 and Arg87; moreover the Glu38 amide is in hydrogen bond formation with the Gly63 backbone carbonyl group from BAG6. Tyr27 from RNF126, which is aligned with Gln62 from BAG6 stabilising its aromatic ring, interacts with Arg64 through the hydroxyl group. Interactions also occur for imidazole groups of His14 (RNF126) and His83 (BAG6) (Fig. 3B). Hydrophobic contacts are detected between the two components: the aliphatic side-chain of Glu38 from β 3 of RNF126 packs against Ile60, Val65 and side-chain of Arg58 from BAG6, and Phe36 in β 2 of RNF126 fits neatly within a hydrophobic pocket created by Leu24, His83 and Val85 from BAG6 (Fig. 3C).
RNF126 and SGTA compete for the same binding site on BAG6. Since the RNF126 binding site on BAG6 defined here showed some overlap with its SGTA interaction interface 15,31 , we explored the possibility of simultaneous or competitive binding between these three proteins. First direct binding between SGTA and RNF126_NZF was excluded by reciprocal NMR titrations, which showed unperturbed spectra of each protein upon addition of the other, to a molar ratio of 1:4 (SGTA dimer: RNF) ( Figure S7 for the 15 N-labelled SGTA perspective and data not shown for the reciprocal titration). Next we analysed the CSP of 15 N-labelled SGTA_NT (C) Detailed view of the solvent-exposed β -sheet interface. Polar and hydrophobic residues present at this interface are depicted using ball-and-stick representation. Carbon atoms are coloured in grey, oxygen atoms in red, nitrogen atoms in dark blue and sulphur atoms in yellow; zinc cation in magenta.
upon titration with BAG6_UBL and subsequent saturation with RNF126_NZF; the peaks that are perturbed upon BAG6_UBL binding incrementally return to their free state upon RNF126_NZF addition (Figs 4A and S8). Finally, 15 N-labelled BAG6_UBL protein was titrated with SGTA and then saturated with RNF126_NZF with the characteristic shifts for SGTA binding heading towards the RNF126_NZF binding state upon saturation. We also performed reverse saturation by adding RNF126_NZF followed by SGTA, with equivalent results (Fig. 4B) although the BAG6/RNF126 complex signals were never completely lost in this case. We employed native PAGE mobility shift assays to further analyse this competition for BAG6_UBL binding. We saturated SGTA_NT with BAG6_UBL and detected the band shift indicative of complex formation. We added GFP-tagged RNF126 into the protein mixture at increasing concentrations which caused characteristic band shifts for the unbound form of SGTA and subsequent complex formation between RNF126 and BAG6_UBL. Conversely, reverse titration (i.e. BAG6_UBL/RNF126_NZF complex formation followed by SGTA_NT addition) did not show the release of RNF126 from the RNF126/BAG6_UBL complex upon competition with SGTA_NT (Fig. 4C).
These experiments consistently indicate that RNF126_NZF competes with SGTA for the same binding region in BAG6_UBL and eliminate the possibility of RNF126 and SGTA binding to BAG6 simultaneously. In addition, BAG6_UBL binds more tightly to RNF126_NZF than it does to SGTA_NT as the former interaction has K d ≈ 0.4 μ M as calculated from ITC whereas the the ITC data for the latter interaction could not be adequately fitted, despite trying numerous experimental conditions, and was only demonstrable by NMR chemical shift perturbation in fast exchange and size exclusion chromatography 31 .
Interaction of RNF126 with UBL4A_UBL. Since BAG6_UBL and UBL4A_UBL each bind to SGTA 31 and are both displayed by the BAG6 complex, we speculated that there might be an interaction between RNF126_NZF and UBL4A_UBL and first investigated this possibility using CSP. Upon titration with unlabelled UBL4A_UBL, 15 N-labelled RNF_NZF showed chemical shift perturbations within the first forty residues that compared well with those shifts that occur upon BAG6_UBL binding (Fig. 5A). The analysis of the reciprocal titration with 15 N-labelled UBL4A_UBL was performed using assignments from the BMRB database (Accession Number: 11279). They revealed that the binding interface on UBL4A is equivalent to the one obtained for BAG6: β 3 , β 4 , β 5 strands and the loop between β 1 and β 2 are perturbed upon binding RNF126 (Figs 5B and S3B). Both titrations occupy a fast exchange timescale implying that this interaction has a lower affinity than that of BAG6_UBL binding; in fact, plotting the CSP data against the concentration of RNF126_NZF and fitting the curve to a simple 1:1 equimolar binding model yields a dissociation constant (K d ) in the order of 20 μ M, around 50 times higher than that of BAG6 (Fig. 5C). We also investigated this interaction using ITC but, in the same experimental conditions as were used for BAG6, the observed transition was small and impossible to fit, in keeping with the reduced affinity that we observed (data not shown). MST and native PAGE mobility shift assays also demonstrated a lower affinity or no binding, respectively, between UBL4A and RNF126 as compared to the BAG6 binding ( Figures S4).
Finally, we generated a structural model of the RNF126_NZF/UBL4A_UBL complex using our CSP data in HADDOCK-based semi-rigid, data-driven docking 32,33 . We analysed the three lowest energy HADDOCK clusters of RNF126_NZF/UBL4A_UBL ( Figure S9A and Table S3). The top-ranked cluster is highly similar to our experimentally-derived RNF126_NZF/BAG6_UBL structure ( Figure S9B) showing that residues making contacts with RNF126_NZF are conserved between UBL4A_UBL and BAG6_UBL (Figs 6A, S9C and S9D). The UBL4A_ UBL residues displayed in the binding interface with RNF126 are Leu8, Arg42, Leu44, Lys48, Ala49, Asn68 and Val70 of UBL4A ( Figure S9C), which align with residues Leu24, Arg58, Ile60, Arg64, Val65, His83 and Val85 of BAG6_UBL ( Figure S9D). Both of these sets of UBL residues make contact with His14, Tyr27, Phe36, Glu38 and Glu39 in RNF126_NZF. The positively charged Lys48, Arg42 and Lys72 of UBL4A (equivalent to Arg64, Arg58 and Arg87 in BAG6) form electrostatic interactions with Glu39 and Glu38 in RNF126_NZF (Fig. 6B). Phe36 of RNF126_NZF inserts into a pocket formed by Leu8, Asn68 and Val70 of UBL4A making hydrophobic contacts between aliphatic portions of the side-chains (Fig. 6C,D). We conclude that, like SGTA, RNF126 is capable of binding to both of the UBL domain containing subunits of the BAG6 complex.

Discussion
There are hundreds of E3 ubiquitin ligases in mammalian cells to process the wide variety of protein degrons 34,35 , but the way in which E3 ubiquitin ligases select their specific substrates remains elusive. RNF126 belongs to the large RING family of ligases, whose defining feature is a zinc-binding RING finger domain which performs the ubiquitin ligase activity through interaction with an E2 (ubiquitin-conjugating) enzyme. RNF126 has been specifically connected with the BAG6 sortase, ubiquitinating hydrophobic substrates to send them down the route towards proteasomal degradation 4 . Many of the RING E3 ligases contain additional zinc-finger motifs which can act as recognition sites for substrates or substrate-bearing delivery systems such as the BAG6 complex. Here we define and characterise a molecular-level interaction between RNF126 and BAG6 and report a further, lower-affinity RNF126 interaction with UBL4A, another subunit of the BAG6 sortase complex.
The structure of the complete heterotrimeric BAG6 complex has yet to be solved and its large size, dynamic behaviour and transient interactions present substantial challenges to the structural biologist. Structures of several isolated domains from different species 31,36,37 and partial complexes 19,38,39 have shed some light on its function at a molecular level. Nevertheless, much of BAG6 itself is completely unsolved and there are still many unknowns. In this work we determine that RNF126, although capable of binding to two different UBL domains on the BAG6 complex, has a clear preference for the BAG6_UBL.
By comparing the RNF126 binding sites on the BAG6 and UBL4A subunits at a structural level we can gain some insight into the likely basis for these differential binding affinities (Fig. 6C,D) with the caveat that we are comparing a solved NMR complex structure (BAG6) to a solid HADDOCK model derived from solved component structures and chemical shift perturbation data (UBL4A). Notably, the conformation of the loop between the β 1 and β 2 strands of UBL4A_UBL differs from that of the BAG6_UBL such that its Leu8 sidechain points away from RNF126_NZF Phe36 in the binding cavity. Furthermore the backbone NH group of the UBL4A Leu8 is solvent-exposed and lines this pocket. In contrast, the corresponding backbone NH group of Leu24 in the BAG6_UBL is buried in the hydrophobic core. The sidechain NH group of Asn68 in the UBL4A_UBL also lines this binding cavity whereas it is the aliphatic part of the His83 sidechain that occupies the equivalent space in the BAG6_UBL (Fig. 6C,D). Taken together, these structural differences reduce the hydrophobicity of the RNF126 Phe36 binding cavity in the UBL4A_UBL relative to the BAG6_UBL, which likely decreases the affinity of the former interaction. A further reduction may be caused by the weaker interaction of Lys48 in UBL4A_UBL with the RNF126_NZF Glu38 compared with Arg64 of BAG6_UBL which has the space to form more hydrogen bonds (Fig. 6B). Earlier studies 14,31 identified a comparable situation to the one described here for RNF126, in which the cochaperone SGTA can also bind to the UBLs of both BAG6 and UBL4A, in the same region as RNF126 ( Figure S10), but, in this case, with the opposite preference i.e. more tightly with UBL4A than BAG6. The UBL binding site on SGTA looks significantly different to that of RNF126 with the UBL-binding capability of SGTA seemingly more dependent on electrostatic interactions, which are stronger for UBL4A_UBL 29,31 , providing a molecular basis for its relative preference. This delicate balance of binding affinities might be key in determining the dynamic process of decision-making at the BAG6 complex (Fig. 7).

NMR distance and dihedral constrains
In contrast to the analogous Get4/Get5 complex in yeast which displays one type of UBL (located at the centre of each Get5 subunit), the mammalian BAG6 complex has two types of UBL to which SGTA can deliver hydrophobic substrates, and it now appears that these may be functionally distinct 19,38 . In particular, it seems that only a small C-terminal portion of the BAG6 protein that brings together the UBL4A and TRC35 subunits is required to enable the transfer of TA proteins from SGTA to the TRC40 delivery complex 19 . Given that SGTA is the first port of call for newly synthesised TA proteins 36 , and SGTA seems to preferentially bind to the UBL4A_UBL subunit of the BAG6 complex, how are hydrophobic polypeptide clients that are destined for quality control passed on to the BAG6 subunit in order to enable their RNF126 mediated ubiquitination? Both SGTA and the BAG6 complex can interact with the same mislocalised membrane protein (MLP) model 16 , confirming the possibility of substrate transfer between them (Fig. 7). Therefore, based on the analogy of TA proteins, one possibility is that SGTA can present quality control clients to the BAG6 complex via the UBL4A_UBL, and it is an inability of TRC40 to collect these substrates that enables their redirection to the BAG6 subunit 40 . In this case, the well-defined TMD of  TA proteins would ensure that they are quickly passed along from the C-terminal domain of SGTA via a UBL4A mediated interaction with the BAG6 complex to TRC40 and hence taken onwards for ER delivery. The failure of substrates such as MLPs to be efficiently collected from SGTA by TRC40 would instead direct them towards the BAG6 subunit and thereby elicit entry into the alternative quality control pathway (Fig. 7).
Alternatively, it is possible that SGTA can "sense" the nature of the bound client, for example, by detecting unfolded or non-native hydrophobic regions. In the case of aberrant clients such as MLPs, such recognition might increase the preference of SGTA for the BAG6_UBL, thereby ensuring the direct transfer of these substrates to the BAG6 subunit for quality control. The BAG6 subunit is certainly capable of substrate discrimination as moderately hydrophobic polypeptides are known to bind close to its N-terminal UBL domain 41 while membrane proteins and heat denatured luciferase bind its central proline rich region 15,42 . In the case of MLPs, we predict that competition between SGTA and RNF126 for binding to the BAG6_UBL influences the stability of these substrates since any increase in quantity or residence time of SGTA would reduce the access of RNF126 to its BAG6-bound clients, thereby reducing their ubiquitination and inhibiting their proteasomal degradation. This model is consistent with both the reduction in BAG6 client polyubiquitination and delay in MLP degradation that are observed upon overexpression of wild type SGTA but not of an SGTA mutant that is defective in binding to the BAG6 complex 16 .
The importance of the N-terminal UBL of BAG6 in recruiting quality control factors is underlined by the observation that it is also wholly (in the case of Hrd1) or partly (in the case of Gp78) responsible for its interaction with two, membrane-bound E3 ligases, which are involved in ERAD 43 . We were therefore struck by the fact that another soluble E3 ligase RNF115 shares near identical NZF and RING domains with RNF126 (with 42.3% overall sequence identity and 52.6% similarity) ( Figure S11). We speculate that RNF115 is highly likely to bind to the BAG6_UBL, perhaps conferring some functional redundancy on its role in the selective ubiquitination of MLPs as suggested by the incomplete effects that are observed following RNF126 depletion both in vitro and in vivo 4 .
This study therefore contributes towards our understanding of the way in which hydrophobic proteins are sorted between distinct cytosolic quality control pathways. Further work will help us fathom the process of selection between all the possible interactions that can occur at the BAG6 complex. Furthermore, the client specificity of E3 ligases is proposed to limit the potential side-effects of drugs that target them, and the new RNF126 structure presented here provides a potential starting point for developing a targeted anti-cancer therapy 25,44,45 .

Methods
Plasmid preparation. Gene fragments encoding human RNF126 (1-40 and 1-100) and N-terminal SGTA (1-86) were PCR amplified from cDNA (Life Technologies) and cloned into BamHI/XhoI restriction sites of a home-modified pET28 vector which encodes an N-terminal thioredoxin A fusion protein followed by a hexahistidine tag and tobacco etch virus (TEV) protease cleavage site. GFP versions of RNF124 (1-40) were cloned following a similar strategy but using a pET28 vector variant fused with N-terminal GFP instead of thioredoxin A. The RNF126 mutants (H14A, F36A, E38K and E39K) were obtained by PCR mutagenesis reaction using pET28-GFP-RNF126 (1-40) vector as the template and different oligonucleotides carrying the mutated codons. Sequences corresponding to the UBL domains of BAG6 (17-101) and UBL4A (1-74) were amplified and inserted via ligation-independent Ek/LIC cloning (Novagen) into a pET46 vector.
Harvested cells were resuspended in lysis buffer (20 mM potassium phosphate, pH 8.0, 300 mM NaCl, 10 mM Imidazole, 250 μ M TCEP), supplemented with protease inhibitors (0.3 μ M Aprotinin, 10 μ M Leupeptin and 1 μ M Pepstatin A) and 1mM PMSF, and lysed by sonication or using a cell disruptor (Constant Systems Ltd). Cell debris and insoluble material were removed by centrifugation and overexpressed protein recovered from soluble fractions was purified using nickel affinity chromatography (HisTrap TM HP 5 ml, GE Healthcare). Recombinant proteins were eluted with buffer containing 300 mM imidazole, then dialyzed against cleavage buffer (20 mM potassium phosphate, pH 8.0 and 300 mM NaCl) and simultaneously digested with homemade TEV protease (≈ 100 μ g/ml) at 4 °C overnight. After TEV cleavage a second nickel affinity chromatography was performed to remove fusion protein, histidine tags, undigested protein and TEV protease; the desired protein was then recovered in the flow through and loaded into a HiLoad 16/60 Superdex 75 column (GE Healthcare), previously equilibrated in buffer containing 10 mM potassium phosphate pH 6.0, 100 mM NaCl and 250 μ M TCEP, for a final gel filtration step. Proteins were concentrated using Vivaspin concentrators (Sartorius Stedin) and sample purity and homogeneity was checked by SDS-PAGE, mass spectrometry and NMR. lane 15 -free BAG6_UBL. Right panel: lane 1 -free SGTA, lanes 2 and 3 -SGTA_NT titrated with BAG6_UBL at 0.3 and 1 molar equivalents, lanes 4-8 -SGTA_NT in complex with BAG6_UBL (1:1) titrated with GFPtagged RNF126_NZF at 0.2, 0.5, 1, 3 and 4 molar equivalents, lane 9 -free GFP-tagged RNF126_NZF, lane 10 -GFP-tagged RNF126_NZF in complex with BAG6_UBL (1:1), and lane 11 -free BAG6_UBL. Free GFP-tagged RNF126_NZF (green), free BAG6_UBL (red) and free SGTA_NT (yellow) are marked with asterisks; GFP tagged RNF_126_NZF/BAG6_UBL complex with a green square and SGTA_NT/BAG6_UBL complex with a red square.
Scientific RepoRts | 6:26433 | DOI: 10.1038/srep26433 NMR structural experiments. Protein samples at concentrations between 100 and 1000 μ M were prepared in 10% or 100% D 2 O (Sigma Aldrich), 10 mM potassium phosphate pH 6.0, 100 mM NaCl and 250 μ M TCEP buffer (also containing 10 μ M DSS for proton chemical shift referencing). All NMR experiments were acquired in 5 mm NMR tubes at 25 °C on Bruker Avance spectrometers operating at 500 MHz and 700 MHz equipped with cryoprobes, controlled by the TopSpin 3.1 software package. Backbone assignments were carried out using 3D experiments (HNCO, HN(CA)CO, CBCA(CO)NH, and CBCANH) 46 and side-chain resonances were assigned using 3D HCCH-TOCSY experiments for both RNF126_NZF and for the complex of RNF126_NZF together with BAG6_UBL. All NMR spectra were processed with NMRPipe 47 and analysed with CcpNMR Analysis 48 .
NOE distance restraints were derived from different NOESY experiments: 2D NOESY (in 90% H 2 O and 100% D 2 O) for RNF126_NZF; and two pairs of 15 N-edited NOESY-HSQC, 13 C-edited NOESY-HSQC and 12 C, 14 N-filtered-13 C-edited NOESY-HSQC experiments for the complex using either 15 N, 13 C labelled BAG6_UBL or RNF126 1-40. Dihedral constraints (φ and Ψ angles) were extracted from the chemical shift values using TALOS+ program 49 . The structure calculation was performed using the ARIA2 program 50 by generating 100 conformers in the final iteration and selecting the 20 best structures with lowest restraint energies for a final step of refinement in water. The coordination sphere of the zinc atom in RNF126 structure was obtained in a first round of calculations using only NOE based distance restraints and once it was confirmed, the zinc atom coordination was incorporated as an additional restraint in calculations that followed using ARIA2 tools. The final ensemble of each structure (PDB: 2N9O; RNF126_NZF, 2N9P; RNF_NZF/BAG6_UBL) was analysed and represented using MOLMOL 51 and PyMOL (DeLano Scientific LLC, Palo Alto, CA, USA).  (20,21,22,23,24,25,26,28,57,58,59,61,62,63,64,81,82,83,84,85,86) upon titration with different concentrations of RNF126_NZF. Fitting was performed using the DynaFit program 52 yielding a K d of 16.6 ± 2.7 μ M.
Scientific RepoRts | 6:26433 | DOI: 10.1038/srep26433 NMR titrations. Proteins used for NMR titrations (RNF126 constructs, UBL4A_UBL, BAG6_UBL and SGTA) were dialysed overnight in the same buffer (10 mM potassium phosphate pH 6.0, 100 mM NaCl and 250 μ M TCEP) and mixed in different molar ratios keeping the concentration of the labelled protein constant. Typically, 1 H-15 N HSQC experiments were recorded for each titration point at 25 °C and CSP calculated for each amide signal using the following formula, where Δ δ 1H and Δ δ 15N are the chemical shift differences for the same amide in its free and bound spectra (δ free -δ bound ) and for proton and nitrogen values respectively: CSP results were mapped onto the structures using the PyMOL software. For the RNF126 titration with UBL4A_UBL, chemical shift perturbation data were analysed and fitted using the Dynafit software 52 . with RNF126_NZF (experimentally calculated using ARIA) with the HADDOCK generated complex of UBL4A_UBL and RNF126_NZF highlighting putative differences in electrostatic interactions at the binding interface. BAG6 is coloured in cyan, UBL4A in magenta and RNF126 in green, with polar residues at the interface shown in ball and stick representation. (C) Vacuum electrostatics view of UBL binding pockets (BAG6 on the right and UBL4A on the left) indicating higher hydrophobicity of BAG6_UBL compared to that of UBL4A_UBL. RNF126 is shown in cartoon representation (green) with the zinc cation shown as a sphere (blue); RNF126 residues involved in interactions are shown as sticks. (D) Detailed view of RNF126_NZF/UBL hydrophobic interactions showing key differences between residues lining the hydrophobic pocket. The backbone NH group of Leu24 and the aliphatic part of the His83 sidechain that lines the cavity in BAG6_UBL suggests stronger hydrophobic interactions with Phe36 in RNF126_NZF (right panel). In contrast, the sidechain NH group of Asn68 lines the cavity in UBL4A_UBL and the backbone amino group of Leu8 is solvent exposed (left panel).

ITC. ITC experiments were performed at 25 °C using an ITC-200 MicroCal microcalorimeter (GE Healthcare)
following the standard procedure as reported previously 31 . Proteins were prepared in 10 mM potassium phosphate pH 6.0, 100 mM NaCl, 250 μ M TCEP. In each titration, 20 injections of 2 μ L of RNF126_NZF (wild-type or mutant), each at a concentration of 500 μ M, were added to a sample of BAG6_UBL or UBL4A_UBL at 50 μ M in the reaction cell. Integrated heat data obtained for the titrations, corrected for heats of dilution, were fitted using a nonlinear least-squares minimization algorithm to a theoretical titration curve, using the MicroCal-Origin 7.0 software package. Δ H (reaction enthalpy change in Kcal/mol), K b (equilibrium binding constant per molar), and n (molar ratio between the proteins in the complex) were the fitting parameters. The reaction entropy, Δ S, was calculated using the relationships Δ G = − RT⋅ lnK b (R = 8.314 J/(mol⋅ K), T 298 K) and Δ G = Δ H− TΔ S. Dissociation constants (K d ) are shown for each interaction.

MST.
Microscale thermophoresis protein-protein interaction studies were performed on the Monolith NT.115 (Nanotemper Technologies, Munich, Germany) using GFP tagged version of RNF126_NZF under these buffer conditions: 10 mM potassium phosphate, pH 6.0, 100 mM NaCl, 250 μ M TCEP. Samples were prepared mixing GFP tagged RNF126_NZF with either BAG6_UBL or UBL4A_UBL, keeping RNF126 at constant concentrations of 0.125 or 0.250 μ M, while UBL domain concentrations ranged from 25 μ M to 0.8 nM (BAG6_UBL), or 50 μ M to 1.5 nM (UBL4A_UBL) using 1:1 serial dilutions. After a 15 minute incubation, ~5 μ l of each solution was loaded into Monolith NT Standard Capillaries (NanoTemper Technologies GmbH). Thermophoresis rates were measured at an ambient temperature of 25 °C with 5s/30s/5s laser off/on/off times, respectively. Instrument parameters were adjusted with 10-20% LED power and 20% IR-laser power. Data from three independently pipetted measurements were analyzed (NT Analysis software version 1.2.101, NanoTemper Technologies) using the signal from Thermophoresis + T-Jump.
Native PAGE mobility shift assay. Native PAGE mobility shift assay was used in the analysis of protein-protein interactions. The experiment was performed by incubating 25 μ M GFP tagged RNF126_NZF (wild type and mutant variants) with increasing amounts of BAG6_UBL or UBL4A_UBL (at a concentration range between 0 to 4 molar equivalents) for 10 min at room temperature in a final volume of 20 μ l. In the case of competition experiments, GFP tagged RNF126 was added to the SGTA_NT and BAG6_UBL protein mixture at increasing concentrations (between 0 to 4 molar equivalents). In reverse titrations, increasing amounts of SGTA_NT were added to the RNF126/BAG6_UBL complex. Native gels were prepared using 10% polyacrylamide in 0.45 M Tris-HCl pH 8.8, 1% ammonium persulphate and 0.02% TEMED and run under native conditions, in buffer containing 25 mM Tris, 200mM glycine at 100 V for 180 min. Fluorescent bands were visualized using an Amersham Imager 600 (GE Healthcare) with a light source of Epi-RGB green light at 520 nm and Cy3 emission filter. The gels were later stained using Coomassie to visualise all protein bands.
RNF126_NZF/UBL4A complex assembly using HADDOCK. Chemical shift perturbation studies defined clear interaction surface areas in the RNF126_NZF/UBL4A_UBL complex. This data was used for complex structure calculation using the HADDOCK approach. For the calculation, PDB-deposited structures of UBL4A_ UBL (2DZI) and the lowest energy NMR structure from our family of RNF126_NZF structures were used. Ambiguous Interaction Restraints (AIRs) were implemented according to the standard protocol. The chemical shift perturbation data enabled identification of 14 amino acid residues in RNF126_NZF and 21 in UBL4A_UBL with chemical shift changes greater than 0.15 ppm. After filtering for a relative solvent accessibility higher than (A) SGTA bearing a hydrophobic substrate reaches the BAG6 complex, likely binding the UBL4A_UBL domain, but possibly the BAG6_UBL. The substrate (TA proteins and/or MLPs) could remain bound to SGTA (and passed directly to heat-shock chaperones that bind the TPR) or be transferred to BAG6. (B) TA substrates are transferred to TRC40 for ER membrane insertion likely through SGTA bound to UBL4A which binds TRC35 and recruits TRC40. (C) At the same time, BAG6 is involved in protein quality control of hydrophobic substrates. RNF126_NZF binds BAG6_UBL, possibly replacing SGTA and, in concert with E2 enzymes could ubiquitinate the hydrophobic substrates from BAG6 or SGTA, leading them to proteasomal degradation. This process is regulated by the opposite process of deubiquitination in which SGTA may have a recruiting role at least at the proteasome level. 45%, calculated using the program Naccess, 10 residues in RNF126_NZF and 14 in UBL4A respectively, were identified as active. These were RNF126_NZF residues 14,27,34,36,38,39 and UBL4A residues 8,44,48,49,68, 70. Solvent exposed residues juxtaposed to the active residues were automatically termed passive residues by the HADDOCK protocol. One thousand initial complex structures were generated by rigid body energy minimization, and the best 200 (lowest total energy) were selected for torsion angle dynamics and subsequent Cartesian dynamics in an explicit water solvent. Default scaling for energy terms was applied. Following the standard benchmarked protocol, cluster analysis yielded 175 structures in 5 cluster ensembles. The top scoring cluster (lowest energy) was taken as the most reliable result as shown by HADDOCK benchmark testing.