Characterizing the Monomer–Dimer Equilibrium of UbcH8/Ube2L6: A Combined SAXS and NMR Study

Interferon-stimulated gene-15 (ISG15) is an interferon-induced protein with two ubiquitin-like (Ubl) domains linked by a short peptide chain and is a conjugated protein of the ISGylation system. Similar to ubiquitin and other Ubls, ISG15 is ligated to its target proteins through a series of E1, E2, and E3 enzymes known as Uba7, Ube2L6/UbcH8, and HERC5, respectively. Ube2L6/UbcH8 plays a central role in ISGylation, underscoring it as an important drug target for boosting innate antiviral immunity. Depending on the type of conjugated protein and the ultimate target protein, E2 enzymes have been shown to function as monomers, dimers, or both. UbcH8 has been crystallized in both monomeric and dimeric forms, but its functional state remains unclear. Here, we used a combined approach of small-angle X-ray scattering (SAXS) and nuclear magnetic resonance (NMR) spectroscopy to characterize UbcH8’s oligomeric state in solution. SAXS revealed a dimeric UbcH8 structure that could be dissociated when fused N-terminally to glutathione S-transferase. NMR spectroscopy validated the presence of a concentration-dependent monomer–dimer equilibrium and suggested a back-side dimerization interface. Chemical shift perturbation and peak intensity analysis further suggest dimer-induced conformational dynamics at the E1 and E3 interfaces, providing hypotheses for the protein’s functional mechanisms. Our study highlights the power of combining NMR and SAXS techniques to provide structural information about proteins in solution.


INTRODUCTION
Interferon-Stimulated Gene 15 (ISG15), also known as hUCRP or IP17, is a 15 kDa ubiquitin-like, type I interferon (IFN) inducible protein [1]. ISGylation is an ubiquitin-like (Ubl) post-translational modification (PTM) that involves the covalent attachment of ISG15 to target proteins [2]. Similar to other Ubls, ISGylation plays important roles in various cellular processes such as innate antiviral immunity, protein degradation, and signal transduction [3]. Free, unconjugated, ISG15 also serves immunoregulatory functions as a cytoplasmic and secreted signaling protein in eukaryotic organisms [4]. Inherited ISG15 deficiency dramatically reduces the innate immune system's ability to fight viruses in mice yet only appears to cause immunoregulatory issues against mycobacterial, not viral diseases, in humans [3]. Thus, the role of ISG15 in human viral pathogenesis is not clearly understood.
The ISGylation cascade requires the sequential action of three enzymes: Ube1L as the E1 enzyme, UbcH8 as the E2 enzyme, and HERC5 as the E3 enzyme. First, ISG15 binds the catalytically active cysteine of the Ube1L activating enzyme (E1) in an ATP-dependent reaction.
Then, E1 interacts with UbcH8 conjugating enzyme (E2) through its ubiquitin folding domain (UFD), which facilitates the transesterification of active ISG15, and results in an intermediate ISG15-UbcH8 complex joined by a thioester bond [5]. Finally, HERC5 ligase enzyme (E3) interacts with the intermediate ISG15-UbcH8 complex to mediate ligation of ISG15 to the target protein. UbcH8 plays a central role in ISGylation as it interacts with both E1 and E3 enzymesmaking it a key target for the regulation of the ISGylation pathway [6].
Under reducing conditions, E2 enzymes can spontaneously form dimers when a crosslinker is added [7], and apart from a few exceptions, E2 enzymes are capable of preserving their dimer form [8,9]. Nevertheless, both the dimer and the monomer forms of E2 enzymes are capable of recruiting E3 enzymes and conjugating ubiquitin [10]. Although dimeric E2 enzymes are perceived as more advantageous because one of the monomers can remain associated while the ubiquitin conjugation continues with the other.
The Protein DataBank (PDB) contains both dimeric (PDB ID:1WZV) and monomeric (PDB ID:1WZW) crystal structures of UbcH8. Yet, it is unknown whether UbcH8 dimerizes naturally or as a consequence of non-specific crystal packing contacts. In this study, we aimed to characterize the oligomeric state of human UbcH8 (Ube2L6) in solution using Small Angle X-ray Scattering (SAXS). We first used a fusion protein approach with the goal of producing a high-resolution scattering envelope to properly place the UbcH8 protomers. Surprisingly, UbcH8 formed stable dimers upon removal of the N-terminal fusion protein. We next used solution nuclear magnetic resonance (NMR) spectroscopy to first validate and then further characterize the monomer-dimer equilibrium. Our results indicate that UbcH8 contains a substantial dimer population at 150 μM concentration and that dimerization may induce conformational changes at the distal ISG and E3 interaction interfaces.

GST fusion guides SAXS protein structural modeling.
To determine the state of UbcH8 in solution, we expressed and purified it fused to an N-terminal glutathione-S transferase (GST) tag herein termed GST-UbcH8. We hypothesized that the 28 kDa GST molecule should be easily discernible from the smaller (18 kDa) UbcH8, and would dramatically improve fitting SAXS scattering data to the structural model. The sample was concentrated to 280 μM and six, 10 min SAXS frames were collected for a total of one hour. Superimposition of each 10 min frame confirmed that the X-ray beam produced little to no detectable radiation damage (data not shown). The medium to high q region, which is emphasized in the q vs l(q) plot, is consistent with a folded sample ( Figure 1A). The Kratky plot possessed a bell-shaped curve that approaches zero after reaching a maximum at~3 sRg; this result is consistent with a properly folded, globular protein ( Figure 1B). While slight deviations between the typical Kratky plot and dimensionless Kratky plot can aid in the assessments of flexibility, no apparent differences were observed.
The pair-distance distribution function, P(r), is a measure of the frequency of interatomic distances that can also provide information about the protein shape. The presence of a shoulder in the P(r) suggests a multidomain protein as expected for the GST-UbcH8 fusion ( Figure 1C).
The largest distance (D max ) in the P(r) histogram was 8 nm ( Figure 1C). The GST-UbcH8 crystal structures were then fitted into the final 3D DAMMIF dummy atom model ( Figure 1D,E). Both GST and UbcH8 proteins, as well as the linker peptide, are clearly visible fitting a monomeric model. The fact that even the linker region can be detected with SAXS analysis and be observed this clearly, underscores the power of SAXS in structure determination. After GST cleavage, UbcH8 was observed to form a dimer based on Size Exclusion Chromatography (data not shown), hence we hypothesized that GST mayblock the dimerization site.

FreeUbcH8 is a Dimer in Solution.
To test how well the GST-fusion improves the modeling of UbcH8 into the SAXS scattering, we prepared a second UbcH8 sample with the GST protein removed. Again, we concentrated UbcH8 to 280 μM and collected six, 10 min frames for a total of 1 h (Figure 2A). Similar to GST-UbcH8, the Kratky plot possessed a bell-shaped curve that approaches zero ( Figure 2B). We estimated a slightly larger R g~4 .40 nm, compared to GST-UbcH8, from the low q region, whereas the P(r) D max was reduced to 6.2 nm ( Figure 2C). Surprisingly, the free UbcH8 P(r) also contained a shoulder suggesting homodimerization ( Figure 2C). ATSAS molecular weight analysis predicts a 39.5 kDa particle, which is approximately double the expected 18 kDa UbcH8. We then fitted the UbcH8 dimer crystal structure (PDB 1WZV; Figure 2D) to the dummy atom model the scattering envelope ( Figure   2E). The best-fit model (X 2 = 1.7) possesses a dimerization interface with the active site cysteines of each protomer pointed outwards ( Figure 2D,E). The consistency between the previously published dimer crystal structure and the dummy atom model obtained by SAXS analysis, supports the homodimer formation of UbcH8 protein in solution in absence of a GST-tag.
NMR analysis of UbcH8 monomer-dimer equilibrium. To further establish dimerization of UbcH8 in solution, we performed Transverse Relaxation Optimized Spectroscopy (TROSY) for rotational correlation times (TRACT) experiments [11,12] to estimate the rotational correlation time ( ) of UbcH8 at two different concentrations: 300 μM and 150 μM ( Figure 3). The signal τ intensity ranging from 8.6 to 9.2 ppm was integrated to maximize signal to noise and emphasize well-structured regions of the protein that are representative of global tumbling. We estimated 15 N relaxation rates for the TROSY and anti-TROSY integrated signals using Bayesian Parameter Estimation of a two-parameter single-exponential decay model. This method produces a distribution of decay rates, which encompass uncertainty, that were then used to determine the cross-correlated relaxation (CCR) rate. The rotational correlation time was estimated from CCR 6 according to an algebraic solution [12] of the modified Goldman relation [13], assuming an order parameter (O 2 ) of 0.8. We determined a~16 ns at 300 μM and~13 ns at 150 μM ( Figure 3), τ which demonstrates a concentration dependence on molecular rotation diffusion times. We then used hydroNMR [14] to model rotational diffusion of monomeric and dimeric UbcH8 from the PDB 1WZV dimeric crystal structure. hydroNMR reported a = 20.5 ns for the dimer and 7.4 τ ns for the monomer at 25 ºC. Taken together, this confirms that UbcH8 undergoes monomer-dimer exchange and indicates a substantial dimer population even at 150 μM. Data could not be collected at lower concentrations due to the sensitivity limit of the room temperature NMR probe.
We next collected 15 N heteronuclear single quantum coherence (HSQC) solution NMR spectra at 150 μM and 300 μM to identify UbcH8's dimerization interface. Resonances were assigned by visual inspection using BMRB Entry ID 16321 as a reference list. The NH resonances of all residues except for the 18 prolines were assigned (79.85% completion). We then assessed both concentration-dependent chemical shift perturbations (CSPs) and peak intensity differences. The concentration-dependent CSPs were of relatively low magnitude and located far from the crystallographic dimerization interface ( Figure 4). All of the perturbed residues except for N23, which resides at the dimerization site, are situated at either the E1 or the ISG15 interaction surfaces. Residues E80, N81, and G82 are clustered on a loop near the catalytic C85 residue where ISG15 is covalently attached. Whereas F56, K99, V103, L104, and N108 are proximal to the E1 binding region on the UbcH8 surface; interestingly, these residues are arranged towards the UbcH8 core rather than at the surface ( Figure 4). Given that ISG15 and E1 involve distinct interfaces, we hypothesize a conformational change or allosteric pathway influences the transfer 7 or binding of ISG15. Our results suggest that dimerization may play an additional role in ISGylation. We hypothesize that the weak CSPs could reflect a mostly sidechain-mediated interface and/or that the ensemble is predominantly dimeric even at 150 μM concentration.
Thus, we also measured the concentration-dependent changes in peak intensity. We hypothesize these intensity differences result from monomer-dimer exchange on the intermediate Furthermore as the overall structure gets bigger with the dimerization, decreased signals from some peaks were expected due to line broadening. Although unlike the CSP analysis which showed that most of the conformational changes occurred away from the dimerization site, delta chemical shift intensity analysis revealed that most of the affected residues were on the dimerization site. In fact, the peak intensity of N23 and D149 residues (Supp. Figure 4) from opposing protomers, which are within 3.3 Å distance in the crystal structure, deviated from the mean peak height by 44.7% and 62.2%, at 150µM, and 37.6% and 35.6%, at 300 µM, respectively. Taken together, this indicates that dimerization is ipso facto involved in defining interaction dynamics between E1 and E2 enzymes. enzymes form dimers in solution regardless if an active ubiquitin is present [7]. While the monomer form is also active for the acquisition of the ubiquitin, dimer form of the E2 is found to be more advantageous as while one monomer site is binding the ubiquitin molecules, the other site is capable of remaining associated to the target protein and thus facilitates efficient polyubiquitination. The acting mechanism of E2 enzymes proposed in this study suggests that the E2 enzymes function as dimers while catalyzing polyubiquitination process [7]. Our results demonstrate that UbcH8, the E2 enzyme specific for ISGylation, can also form dimers at near

SAXS Data Collection.
All SAXS data were collected at home source SAXSpoint 5.0 (Anton Paar GmbH) as described before [18]. Sample/detector distance (SDD) was 1600 mm for SAXS experiments. All measurements took place at 10°C. Data was collected in one hour session(1-minutes long 6 frames) for each measurement. The scattering curves were checked for radiation damage and no damage was detected after the superimposition of each 10 minute data collection intervals..

SAXS Data Processing and Modeling.
At first, the scattering pattern of all samples were visually inspected in the Primus program of ATSAS 3.0 for any possible issues with the measurement [11]. The radius of gyration (Rg) was calculated using Guinier's equation and inverse Fourier transform by Primus. Distance distribution function P(r) and the maximum particle diameter (Dmax) was calculated by GNOM [19]. After estimating the molecular weight of the model DAMMIF (ab initio) is used to generate 5 independent low resolution models from the data. [20]. DAMAVER and DAMMIN then averaged, clustered, and optimized these 5 distinct solutions to form the final ab-initio shape [21]. SASpy plug-in for PyMOL was used to superimpose the homology modeled structure of the protein [22,23]. Harvested cells were resuspended in lysis buffer (500mM NaCl, 50mM Tris, 0.1% (v/v) Triton X-100, 5% (v/v) glycerol, 1mM DTT, pH=7.5), sonicated, and centrifuged at 20K RCF for 1 hour to remove insoluble debris. The obtained supernatant was loaded to a GST affinity column equilibrated with 38.39 mM Na 2 HPO 4 , 11.61 mM KH 2 PO 4 (pH 7.4), 100 mM NaCl, 1mM DTT.
Non-specific proteins were washed with the same buffer and the protein was eluted with 30mM  [26].
1D TRACT experiments [11] were collected with 1024 complex points and 1.5 s recycle delay.
Relaxation rates for 15     UbcH8 were calculated for each residue. The residues colored green possessed chemical shift perturbations larger than the threshold (red line). Residues with no bars were not observed at either concentration. B) Residues with chemical shift perturbations larger than the threshold were mapped onto the UbcH8 dimeric crystal structure (PDB 1WZV). These residues cluster to three distinct regions: the dimer interface (N23), the ISG15 conjugation site (E80, N81, and G82), and the E1 binding surface (F56, K99, V103, L104, and N108).

Figure 5. Concentration dependent chemical shift signal intensity changes
A)Concentration-dependent changes in 1 H-15 N peak intensity mapped onto UbcH8 dimeric crystal structure. The residues colored yellow possessed peak intensity changes that were larger than the threshold (red line); these residues are D12, K16, N30, V33, E141, L144, and D149.
Residues with no bars were not observed at either concentration. B) Residues with peak intensity changes larger than the threshold were mapped onto the UbcH8 dimeric crystal structure (PDB 1WZV