Ubiquitin transfer by a RING E3 ligase occurs from a closed E2~Ub conformation

Ubiquitination is a eukaryotic post-translational modification that modulates a host of cellular processes1. Modification is mediated by an E1 activating enzyme (E1), an E2 conjugating enzyme (E2) and an E3 ligase (E3). The E1 catalyses formation of a highly reactive thioester linked conjugate between ubiquitin and E2 (E2~Ub)2. The largest class of ubiquitin E3 ligases, which is represented by RING E3s, bind both substrate and E2~Ub and facilitate transfer of ubiquitin from the E2 to substrate. Based on extensive structural analysis3–5 it has been proposed that RING E3s prime the E2~Ub conjugate for catalysis by locking it into a “closed” conformation where ubiquitin is folded back onto the E2 exposing the restrained thioester bond to attack by a substrate nucleophile. However the proposal that the RING dependent closed conformation of E2~Ub represents the active form that mediates ubiquitin transfer is a model that has yet to be experimentally tested. Here we use single molecule Förster Resonance Energy Transfer (smFRET) to test this hypothesis and demonstrate that ubiquitin is transferred from the closed conformation during an E3 catalysed reaction. Using Ubc13 as an E2, we designed a FRET labelled E2~Ub conjugate, which distinguishes between closed and alternative conformations. Firstly, we defined the high FRET state as the closed conformation using a stable isopeptide linked E2~Ub conjugate, while the low FRET state represents more open conformations. Secondly, we developed a real-time smFRET assay to monitor RING E3 catalysed ubiquitination with a thioester linked E2~Ub conjugate and determined the catalytically active conformation. Our results demonstrate that the reaction proceeds from the high FRET or closed conformation and confirm the hypothesis that the closed conformation is the active form of the conjugate. These findings are not only relevant to RING E3 catalysed ubiquitination but are also broadly applicable to E3 mediated ligation of other ubiquitin-like proteins (Ubls) to substrates.

2 ubiquitin is transferred from the closed conformation during an E3 catalysed reaction.
Using Ubc13 as an E2, we designed a FRET labelled E2~Ub conjugate, which distinguishes between closed and alternative conformations. Firstly, we defined the high FRET state as the closed conformation using a stable isopeptide linked E2~Ub conjugate, while the low FRET state represents more open conformations. Secondly, we developed a real-time smFRET assay to monitor RING E3 catalysed ubiquitination with a thioester linked E2~Ub conjugate and determined the catalytically active conformation. Our results demonstrate that the reaction proceeds from the high FRET or closed conformation and confirm the hypothesis that the closed conformation is the active form of the conjugate. These findings are not only relevant to RING E3 catalysed ubiquitination but are also broadly applicable to E3 mediated ligation of other ubiquitin-like proteins (Ubls) to substrates.

Results and Discussion
Ubiquitin in an E2~Ub conjugate can exist in multiple conformations including a "closed" form in which ubiquitin is folded back onto the E2. E3 ligases for ubiquitin and Ubl proteins appear to trap the E2~Ub/Ubl conjugate in this conformation [3][4][5][6][7][8][9][10][11] ( Fig. 1a). We have previously crystallized Ubc13~Ub in complex with the RNF4 RING domain dimer without the Ubc13 binding partner, UEV, in the same closed conformation 7 (Fig. 1b) even although RNF4 fails to transfer ubiquitin from the Ubc13~Ub conjugate in the absence of UEV (Fig. 1c). Here we found that RNF4 has a higher binding affinity for the Ubc13~Ub conjugate when in complex with UEV ( Fig. 1d) suggesting both UEV and RNF4 are required to stabilize the active conformation of the conjugate. We therefore decided to use smFRET to interrogate 3 the structural dynamics of the Ubc13~Ub conjugate in response to RNF4 and UEV binding and RING E3 catalysed ubiquitin transfer.
Single molecule fluorescence approaches have been applied to the ubiquitin system [12][13][14] ; but not to follow RING catalysed transfer of ubiquitin from E2 to substrate. We thus designed a Ubc13~Ub conjugate labelled with the Cy3B-AlexaFluor 647 FRET pair (R 0 = 60 Å) where proximity of the FRET labels in the closed conformation yields a high FRET efficiency, while the large distance between the labels in the more open conformation results in a low FRET efficiency (Fig. 1e, Extended Data Fig. 1a Fig. 4a-c). Increased temperature is anticipated to destabilize the high FRET state or closed conformation due to breakage of the hydrogen-bonding network in the active site groove that is required for adopting the closed conformation and drives specificity for catalysis in the active site 3 . As a result the C-terminal tail of ubiquitin will sit outside the active site groove similar to a previous NMR model 15 Fig. 5a). An L106A mutation in Ubc13 that is part of the same hydrophobic interface has a similar, but less pronounced, effect on the conformation of the Ubc13~Ub conjugate (Fig. 2 a, 16 that perturbs binding of the acceptor ubiquitin (Fig. 2a), was defective in RNF4 dependent substrate ubiquitination (Fig 2b), but retained the ability to activate the Ubc13~Ub conjugate in an RNF4 dependent lysine discharge assay 7 .
To determine if the S32A mutation in UEV could co-operate with RNF4 to restrain the Ubc13~Ub conjugate in the closed conformation we carried out smFRET analysis of the wild type and S32A versions of UEV. This revealed that the conformation of the Ubc13~Ub conjugates were indistinguishable and indicated that stabilization of the closed conformation was independent of acceptor ubiquitin binding to UEV (Fig.   2e). These data also indicate that none of the FRET states observed are a consequence of ubiquitin binding to UEV.
Using the stable isopeptide linked Ubc13~Ub conjugate allowed us to identify the high FRET state as the closed conformation and indicated that both UEV and RNF4 were required to restrain the Ubc13~Ub conjugate in this conformation. However, to directly test the hypothesis that the closed conformation is the active form of the Ubc13~Ub conjugate, requires a method that reports in real-time the FRET state(s) from which ubiquitin transfer to substrate occurs. This necessitated attaching the same FRET dyes at identical positions used for the isopeptide linked conjugate to an unstable thioester linked conjugate (Fig. 3a, Extended Data Fig. 6a, b). The labelled conjugate modified the substrate in a biochemical assay with similar efficiency to the WT conjugate (Extended Data Fig. 7). In the smFRET experiment we immobilized the unstable thioester linked Ubc13~Ub conjugate onto the surface for imaging and injected the reaction components for ubiquitination including UEV, RNF4 and 6 Ub~4xSUMO-2 substrate. We monitored, in real-time, transfer of Cy3B-labelled ubiquitin from the E2~Ub conjugate onto substrate, which diffuses into solution and results in simultaneous loss of both donor and acceptor signals, and therefore, the FRET signal (Fig. 3a). This is evident from the decrease in number of FRET pairs before and after the reaction has occurred (Fig. 3b) and by following the cumulative intensity of fluorescent spots over time with and without RNF4 catalysed ubiquitination (Fig. 3c). Single molecule trajectories demonstrate stable Cy3B and AlexaFluor 647 intensities lasting hundreds of seconds following a UEV only injection versus the rapid simultaneous loss of all fluorescent signals following the ubiquitination reaction (Fig. 3d). Combined analysis of single molecule traces shows a significant portion of E2~Ub conjugates react upon injection of reaction components (~60%), compared to injection of UEV only (~14%) (Fig. 3e). A small portion of the sample contains only Cy3B due to under-labelling with AlexaFluor 647 or prior photobleaching of AlexaFuor 647, however, a reaction was also observed in these conjugates (Fig. 3e).
Contour-plots of FRET trajectories in a 90 second time-window following showed rate of FRET loss due to ubiquitin transfer to substrate depends on RNF4 concentration and FRET loss was not observed in the absence of RNF4 or with an RNF4 mutant unable to bind the Ubc13~Ub conjugate ( Fig. 4a). Single molecule analysis demonstrated that ubiquitin transfer rates can be distinguished from slow Cy3B and AlexaFluor 647 photobleaching events and quantified via dwell time of the FRET signal ( Fig. 4b-c). The smFRET population histogram derived from only those molecules undergoing ubiquitin transfer, accurately shows the reaction proceeded from the high FRET state (E FRET~0 .71) similar to that assigned to the closed conformation of the isopeptide linked E2~Ub 7 conjugate (Fig. 4d, e). A low fraction of molecules reacting at a slower rate and from a low FRET state were also present in the UEV only injection, indicating that this was a background reaction that is not dependent on RNF4 (Fig. 4d-

Data Availability
All data are available from the authors upon request.

Author Information
Correspondence and requests for materials should be addressed to R.T.H.

Competing Financial Interests
The authors declare no competing financial interests.

Cloning, expression and purification of recombinant proteins. Expression and
purification of RNF4 and linear fusion RNF4 constructs was described previously 20 .
Ub~4xSUMO-2 was also expressed and purified as described previously 21 . Human Ubc13 (also known as Ube2N) was previously subcloned into the pHISTEV30a vector using NcoI and HindIII restriction sites 7 . A sequence encoding an avitag (Gly-Leu-Asn-Asp-Ile-Phe-Glu-Ala-Gln-Lys-Ile-Glu-Trp-His-Glu) followed by a linker (Gly-Gly-Ala) was inserted between the TEV cleavage site and Ubc13 using the NcoI restriction site. A K24C mutation was introduced into Ubc13 using site-directed mutagenesis. Additional C87K and K92A mutations in Ubc13 were used as described previously 7 to generate the stable isopeptide linked Ubc13~Ub conjugate. Ubc13 and Ube2V2 variants were expressed and purified as described previously 7 . As a result of cloning, Ubc13 has four extra residues (Gly-Ala-Met-Ser) before the N-terminal avitag after cleavage with TEV protease. His 6 tagged Ubiquitin M-2C was expressed and purified as described previously 3,22 . An additional K63R mutation in ubiquitin was used to generate the unstable thioester linked Ubc13~Ub conjugate. Fluorescence polarization was measured using a PHERAstar FS microplate reader with 485 nm excitation and 520 nm emission wavelengths.
The substrate single turnover assay to validate labelled proteins was performed as previously described 20  an N-terminal linker that is not present in these PDB files, the closest surface accessible amino acid, Gln2, was used for attachment and modelling the dye on ubiquitin. This is likely to result in a distance measurement that is shorter than expected; therefore, the resulting distance calculations were used as a guide.
Slide passivation. Aminosilane treated slides were passivated with PEG-SVA and Biotin-PEG-SVA as described previously. Slides were assembled into four channels for stable isopeptide linked conjugate studies. Single channel slides with tubing and syringe attached for injection purposes were assembled for real-time reaction studies using the unstable thioester linked conjugate. Channels were coated with 0.2 mg ml -1 neutravidin for 10 minutes prior to addition of biotinylated FRET labelled proteins.
Single molecule total internal reflection. smFRET experiments were performed as described previously 25 . All experiments were performed on a prism-type total-internal reflection microscope using an inverted microscope (Olympus IX71). A 532 nm laser (Crystalaser) was used for Cy3B excitation and images were collected on a back illuminated Ixon EMCCD camera (Andor, 512x512 pixels). Cy3B and AlexaFluor 647 fluorescence were split by dichroic mirrors (DCRLP645, Chroma Technology) into two channels allowing simultaneous imaging with Cy3B on the left and AlexaFluor 647 on the right of the EMCCD camera.
Isopeptide linked conjugate smFRET experiments. 50 pM of the stable isopeptide linked FRET labelled Ubc13~Ub conjugate was bound to a slide passivated with biotinylated PEG and neutravidin for 10 minutes. Excess free AlexaFluor 647 labelled ubiquitin was washed from the surface using 50 mM Tris, 150 mM NaCl, 0.5 mM