A novel polyubiquitin chain linkage formed by viral Ubiquitin prevents cleavage by deubiquitinating enzymes

The Baculoviridae family of viruses encode a viral Ubiquitin gene. Although the viral Ubiquitin is homologous to eukaryotic Ubiquitin (Ub), preservation of this gene in the viral genome indicates a unique function that is absent in the host eukaryotic Ub. We report the structural, biophysical, and biochemical properties of the viral Ubiquitin from Autographa Californica Multiple Nucleo-Polyhedrosis Virus (AcMNPV). The structure of viral Ubiquitin (vUb) differs from Ub in the packing of the central helix α1 to the beta-sheet of the β-grasp fold. Consequently, the stability of the fold is lower in vUb compared to Ub. However, the surface properties, ubiquitination activity, and the interaction with Ubiquitin binding domains are similar between vUb and Ub. Interestingly, vUb forms atypical polyubiquitin chain linked by lysine at the 54th position (K54). The K54-linked polyubiquitin chains are neither effectively cleaved by deubiquitinating enzymes, nor are they targeted by proteasomal degradation. We propose that modification of proteins with the viral Ubiquitin is a mechanism to counter the host antiviral responses.


INTRODUCTION
Ubiquitin (Ub), a 76 residue protein, regulates multiple cellular pathways and is a major way of regulating protein levels by the Ubiquitin-proteasome pathway (1)(2)(3). Ubiquitin is covalently attached to the target substrate via an isopeptide bond to generate a post-translation modification (PTM). The process of conjugating Ubiquitin to different substrates is a three-step process involving enzymes viz. Ubiquitin-activating enzyme E1, Ubiquitin-conjugating enzymes E2s, and finally, Ubiquitin ligase enzymes E3s. At the end of conjugation, the C-terminal glycine residue of Ubiquitin forms an isopeptide bond with the ɛ-amino group of the substrate's lysine residue. Furthermore, Ubiquitin can build polyubiquitin chains via seven lysine residues, and depending on chain linkage, the function of the polyUb chain is decided (4). For e.g., a substrate with K63 polyUb chains is involved in DNA repair pathways (5) while K11 and K48 polyUb chains eliminate the substrate via proteasomal degradation (6).
The Ubiquitin pathway plays a major role in host-virus interactions (7). Often viruses coopt the host ubiquitination machinery for effective infection (8). In some cases, viruses also encode genes that express enzymes involved in ubiquitination reaction (9). These enzymes help the virus to degrade host immune response, avoid detection during latency, enhance viral The structure of vUB forms the conserved beta-grasp fold like ubiquitin, including five beta sheets (β1-β5) and one helix (α1) (Fig 2C). The structure of vUb and Ub superimposes well with an rmsd of 2.0 Å (Fig 2D). A significant difference in structures of Ub and vUb is observed in the orientation of helix α1 (Fig 2E). Due to the substitutions in -strands 2 and 5, and helix α1, there are differences in the side-chains present at the buried core of the protein (Fig S1A & S1B). These differences give rise to altered packing, and consequently, the helix α1 in vUb is rotated by twenty degrees away from the helix α1 in Ub (Fig 2E). The helix α1 in Ub is closer to the 1-5 beta-sheet than α1 in vUb. As a result, a cavity is observed in vUb, and the two salt-bridges present in helix α1 of Ub are absent in vUb (Fig 2F and S1C). The electrostatic surface potential distribution is similar, except a minor additional negative charge is observed in the loop between 1 and α1 of vUb ( Fig S1D). The negatively charged residue D21 is buried in Ub, but due to differences in packing of the loop between 1 and α1, the corresponding residue E21 is exposed in vUb giving rise to the minor difference in surface charge. The surface of betasheet that presents the critical hydrophobic patch is similar between Ub and vUb ( Fig 2G). Apart from the seven conserved lysines, an extra lysine K54 is present in the loop between 4 and α2 in vUb ( Fig 2H). Overall, the structural studies suggest that due to the cavity at the buried core and the absence of salt-bridges, the stability of vUb may be different from Ub. On the other hand, since the hydrophobic surface patch in conserved in vUb, the activation of vUb by enzymes in the Ubiquitination reaction and interaction of vUb with Ubiquitin binding domains may be indifferent between vUb and Ub.

The backbone dynamics of vUb is similar to Ub
The backbone dynamics in the ps-ns timescale of vUb and Ub was compared by measuring the Longitudinal (T1) and transverse (T2) relaxation and the heteronuclear nuclear Overhauser enhancement (het-NOE) of the two proteins. The dynamics experiments were carried out at 298K and pH 6.0 (Fig 3A-3D). These measured values were used to analyze the order parameter (S 2 ) using the model-free approach (23,24). The values of S 2 are limited between 0 and 1, where S 2 close to 0 means a higher degree of disorder in N-H bond vector, and value close to 1 indicates restricted N-H bond vector. The average S 2 value of vUb is 0.824, indicating that the protein is significantly rigid. A minor decrease in the S 2 values was detected in the loop between β1 and β2, and the loop between β4 and β5, suggesting that these regions are more dynamic than the rest (Fig 3E). For comparison, the T1, T2, and het-NOE experiments were repeated on 15 N-Ub, and its S 2 was calculated (Fig S2A-S2E). A comparison between the S 2 values of Ub with vUb is given in Fig 3F. The regions in vUb that have values higher than standard deviation include the region in between 2 and α1, and the region from 4 to β5, indicating that these regions are more dynamic in vUb than Ub (Fig 3F). The lower rigidity in the loops is in synch with the differences in packing observed in the structure of vUb. Interestingly, the loop between 3 and β5 is dynamic and disordered in a near-native folding intermediate of and Table 2). The stability of the proteins was also measured by thermal denaturation. The melting point (Tm) for vUB was 74ºC. Ub did not melt up to 90ºC, confirming it is more stable than vUb (Fig 4B).
The local stability of vUb and Ub were probed using NMR by measuring and comparing the temperature coefficients per residue. The 2D-HSQC experiment was recorded for 15 N-vUb and 15 N-Ub from 283K till 323K at every 5K increment (Fig 4C and 4D). The resonance shifts were fit linearly with respect to the temperatures to calculate the temperature coefficients ( Fig   4E and Fig 4F). The average temperature coefficient ΔδNH/ΔT values of almost all the secondary structures of vUb and Ub were nearly the same. The ΔΔδNH/ΔT (Ub-vUb) values are plotted in Fig 4G, where negative and positive values indicate residues in vUb with higher and lower stability than Ub residues. Intriguingly, the of ΔΔδNH/ΔT values of non-conserved residues like V23, L26, and A31 in α1, T55, and M57 in α2, Q62, S65 and M69 in β5 strand were positive, indicating that these substitutions cause local instability in vUb ( Fig 4G). Several residues had high positive values in α2, 5, and in the loop in between, implicating vUb is relatively unstable in these loop regions ( Fig 4G).
As mentioned above, the difference in stability between vUb and Ub could be due to the difference in hydrophobic core packing. Alanine scanning studies on Ubiquitin has suggested that hydrophobic contacts at the buried core are essential to maintain high stability and rigidity (26). Mutations of buried residues I30A-Ub and L43A-Ub created a cavity and significantly reduced the stability of Ub. I30 on α1 tightly contacts L15 on β2 strand in Ub ( Fig S1A).
However, L15 is substituted to A15 in vUb, which disrupts the contact. The I36 residue present in the C-terminal loop of α1 packs against β5 strand by making contacts between its side-chain and the sidechains of L69 and L71 of β5 strand (Fig S1D). L69 is substituted to M69 in vUb, which may loosen contacts between the β5 strand and α1. The higher temperature coefficients of E16 and M69 confirm local instability in these regions due to the substitutions. Altogether, Ub is more stable than vUb. Measurement of backbone dynamics and local stability indicates that the region between 2 and α1, and the region between α2 and 5 are largely dynamic and unstable in vUb when compared to Ub. Often stability of a protein determines its molecular recognition by co-factors (27). The interaction of vUb with Ubiquitin binding domains in proteasome receptors was examined to assess if the differences in stability are reflected in the interactions with co-factors.

vUb interacts with proteasome receptors Rpn10/S5a and Rpn13
The Ubiquitin-26S proteasome pathway targets misfolded or tightly regulated proteins for degradation by conjugating them with a poly-Ub chain (28). The ubiquitinated substrate interacts with non-ATPase subunits such as Rpn1, Rpn10/S5a, and Rpn13 of the proteasome via Ub.
Consequently, the substrate enters the core particle for degradation, and the Ub chains are cleaved off by Deubiquitinases (DUBs) present in the proteasome (29)(30)(31). Since the insect proteasome subunits have a high degree of similarity to the mammalian system (32), and the proteins of mammalian proteasome receptors were available, the interaction between vUb and the ubiquitin-binding domains in proteasome receptors (S5a and Rpn13) were tested to examine if substrates conjugated with poly-vUb chains can be recognized by the proteasome receptors. CSPs, suggesting that these residues are present at the interface of S5a/vUb complex ( Fig 5C).
Additionally, the hydrophobic residues like A15, I30, L43, G47 and L67 and charged residues like D32, R42, and R72 are also present at the interface. The chemical shifts values were fit against the ligand concentration in the NMR titration experiments to yield the dissociation constant (Kd) of 78+12 µM.
Residues in the conserved hydrophobic patch (L 216 ALAL 220 ) have high CSPs in UIM1 and interacts with vUb. Additionally, several polar and charged residues in UIM1 (R221, S223, E225, and E226) also interact with vUb. Similarly, apart from the hydrophobic patch (I 287 AYAM 291 ) in UIM2, a series of polar and charged residues (S279, S280, E284, E285, Q292, S294, Q296, and E299) are also involved in the interaction with vUb ( Fig 5F). Overall, the affinity and the interface of the vUb/S5a complex is similar to the Ub/S5a complex, indicating that S5a can identify substrates conjugated with vUb. The binding of vUb and S5a-UIMs was further confirmed by ITC (Fig 5G). The binding isotherm was fitted to a 1:2 complex to obtain the overall Kd of 85.0+9.9 µM, which agrees well with the values as calculated by 15 (Table 3), which suggests that proteins conjugated with poly-vUb chains can be identified by the proteasome.
The degradation of vUb conjugated substrates in cellular conditions was studied by transfection of HEK293T cells with FLAG-Ub and FLAG-vUb, followed by detection of the polyUb and poly-vUb conjugated substrates in the presence/absence of proteasome inhibitor MG132. Similar to the Ub-conjugated substrates, the vUb conjugated substrate accumulate significantly upon addition of proteasome inhibitor, indicating that these conjugates are recognized and degraded by the proteasome (Fig 5I). These results indicate that vUb can bind to Ubiquitin binding domains in the proteasome receptors, and can act as a signal for proteasomal degradation.

Polyubiquitination activity of vUb is similar to Ub
It is important to investigate if the differences in structure and stability can affect the activation of vUb by enzymes in the Ubiquitin pathway. The poly-ubiquitination activity of vUb was measured using multiple E2:E3 pairs. Initially, E2s that form specific Ubiquitin linkages were used in the ubiquitination reactions. E2-25K specifically forms K48-linked poly-Ub chains whereas, Ubc13 (along with its co-factor Mms2) specifically forms K63 linked poly-Ub chains. E2-25K and Ubc13 can synthesize poly-Ub chains without any E3 in-vitro (34,35). Ubiquitination reactions were carried out with Ube1, E2-25K, and either Ub or vUb in the reaction mix. The intensity of polyubiquitin chains observed in the western blots was similar between Ub and vUb, indicating that vUb forms K48-linked chains with equal efficiency as Ub ( Fig 6A). The amount of K63linked Ubiquitin chains catalyzed by the Ubc13/Mms2 complex was less in vUb than Ub ( Fig   6B).
E3s catalyze the rate of ubiquitination by several folds. The in-vitro reactions were repeated in the presence of two different E3s, RNF38, and RNF4. The synthesis of K48-linked polyubiquitin chains was indifferent between vUb and Ub when either RNF4 or RNF38 is used ( Fig 6C and Fig 6F). The synthesis of K63-linked vUb chains was, however, slightly lower in the presence of E3s (Fig 6D and Fig 6G). Finally, E2D2 was used as the E2 in the ubiquitination reaction. E2D2 synthesizes poly-Ub chains of various linkages. Polyubiquitination by E2D2/RNF38 or E2D2/RNF4 complexes were similar between Ub and vUb ( Fig 6E and Fig 6H).
Analysis of the polyubiquitin chains may fail to detect minor differences in the activation of Ub and vUb. As mono-Ub is activated by the enzymes to form poly-Ub chains in ubiquitination reaction, the rate of mono-Ub depletion reflects the efficiency of activation by Ubiquitin enzymes. Hence, the depleting mono-vUb or monoUb species was quantified over time in a typical polyubiquitination reaction. These experiments were performed with the E2s E2-25K, E2D2, and Ubc13/Mms2, using RNF38 as the E3. Interestingly, minor differences were observed in the activation of Ub and vUb, depending on the E2 used in the reaction. The activation of vUb is slightly better than Ub in the case of E2D2 and E2-25K (Fig 6I and 6K).
However, in the presence of Ubc13/Mms2, activation of vUb was slower than Ub ( Fig 6J). In all the assays, the activity of Ubc13/Mms2 complex was lower for vUb than Ub. Unlike the E2-25K and E2D2, the activity of Ubc13/Mms2 depends on the interaction between Mms2 and acceptor Ub, which could be different between Ub and vUb. Overall, the rate of depletion in Ub did not differ by more than 2-fold over vUb, indicating that the activity of Ub was similar to vUb ( Fig 6L).
The activation of Ub by E1 and the trans-thiolation to E2 depends on the hydrophobic patch and the C-terminal end residues (36). Since these are conserved in vUb, the E1 activation and transthiolation activity is expected to be similar between Ub and vUb. R54 in Ub contributes to the intrinsic affinity of Ub with E1. Replacing this charged residue with Leucine affects the binding of E1 with Ub (37). However, if replaced with a charged residue like lysine, the binding affinity is not affected (38). Hence, although vUb has lysine at 54 th position, and conjugation capability with E1 and E2 should be similar. Finally, the activation of E2~Ub for transfer of Ub to substrates is induced by the closed conformation of E2~Ub, where the central helix in the UBC domain of E2 contacts the hydrophobic patch of Ub. Since the hydrophobic patch is conserved in structure, it is reasonable that the ubiquitination activity of vUb is similar to that of Ub. K6-, K11-, K48-, and K63- (Table 4 and Fig S4A-S4D). Apart from these known isopeptide linkages, a peptide containing the novel K54-linkage in di-vUb was also detected (Fig 6N).

Polyubiquitin chains formed by vUb includes a novel K54-linked chain
To further confirm the K54-linked chains, a mutant was designed in vUb, where all the lysines except K54 were substituted to arginines (K54only-vUb). The far UV-CD scan of the K54only-vUb protein confirmed that secondary structure is retained (Fig S5A). Structurally, K54 lies in between K48 and K63. Given that K54-linked chains are unique, whether the E2s that build specific type chains can assemble K54-type chains is unknown. The ubiquitination reactions with E2-25K or Ubc13/Mms2 were repeated with K54only-vUb. Intriguingly, K54-linked di-or poly-vUb were not detected in this reaction (Fig 7A and 7B). The K54-linked chains were not detected even when the E3s RNF4 was used (Fig 7C and 7D). Hence, the E2s that catalyze specific Ubiquitin linkages cannot catalyze K54-linked chains, indicating that these chains may have a distinct topology from K48-and K63-linked chains. Being promiscuous in catalyzing a variety of linkages, E2D2 could catalyze K54-linked poly-vUb chains (Fig 7E, S5C). Apart from K54 linked, these chains could also be N-terminal linked. However, no polyubiquitin chains were observed when K0-vUb was used in the ubiquitination reaction, indicating that the chains in Fig   7E are linked by K54 (Fig 7F). Moreover, the K54-linked di-vUb was analyzed by MS/MS, which confirmed the isopeptide linkage at K54 (Fig 7G). Altogether, vUb can form an entirely new K54-linked polyubiquitin chain, whose topology could be different from the previously known linkages.

K54-linked Ubiquitin chains are not cleaved by DUBs
Since K54-linked Ubiquitin chains are not formed by the eukaryotic Ub, they may not be susceptible to cleavage by the host DeUbiquitinases (DUBs). The poly-vUb chains and K54linked poly-vUb chains formed by E2D2 and RNF38 were treated with three different DUBs (USP2CD, AMSH, and OTUBAIN). USP2CD is a non-specific DUB, AMSH is specific for K63linked chains, and OTUBAIN is specific for K48-chains. USP2CD could efficiently cleave poly-vUb chains, but not K54-linked poly-vUb chains (Fig 8A). The treatment with K63-linkage specific DUB, AMSH, reduced poly-vUb chains partially. This is expected given that AMSH will specifically cleave the K63-linked chains, leaving the other linkages intact. However, the K54 poly-vUb chains were not cleaved (Fig 8B). The similar result was observed when the K48linkage specific DUB OTUBAIN1 was used (Fig 8C). Overall, the results indicate that the K54linked poly-vUb chains are fairly resistant to DUBs.
Whether K54only-polyvUb chains participate in the Ubiquitin-proteasomal pathway was tested by transfecting HEK293T with FLAG-vUb and FLAG-vUb-K54. Cells were subsequently treated with proteasome inhibitor MG132 to detect the accumulated proteasomal substrates.
High molecular weight K54only-polyvUb conjugates could be observed in the cells (Fig 8D).
However, the amount of K54only-polyvUb conjugates did not differ in the presence/absence of MG132, suggesting that these conjugates may not be targets for proteasomal degradation ( Fig   8D). Hence, we uncovered atypical linkage in vUb, that is neither effectively cleaved by DUBs, nor are they targeted by the proteasomal pathway. Conjugation of viral proteins by the vUb is important for assembly of virions (15,16). Ubiquitination of nucleocapsid proteins specifically with vUb determines whether they will egress from the nucleus to form budded virus or not (17).  (39). However, the retention of vUb indicates that it holds unique properties different from the eukaryotic Ub. We report that differences in the sequence between vUb and Ub have resulted in minor structural differences at the buried core of vUb. Consequently, the stability of vUb is reduced compared to Ub. However, the surface properties of Ub is retained in vUb, and its polyubiquitination activity and interaction with co-factors are indifferent from Ub. Additionally, vUb can form a unique K54-linked chain that is not regulated by either the host DUBs or the host proteasomal pathway. Hence, modification of host/viral proteins by the K54-linked viral Ubiquitin may be an effective way to counter antiviral responses that regulate the viral protein.
The molecular details of K54-linked chain, how it resists proteasomal degradation and DUBs, and the function of these chains, are now exciting questions for future research.

Reconstruction of vUb gene tree
39 Baculoviruses containing vUb encoding gene were selected for tree reconstruction, and only the amino acid sequences for Ubiquitin-like protein were selected, whereas the non-ubiquitin amino acid sequences were discarded. The gene tree was reconstructed using the Maximum Likelihood (ML) method. Sequences were aligned using muscle algorithm provided by MEGA X (40) . In order to reconstruct the ML phylogeny, we used Partition finder (41) to obtain the evolutionary model for codon evolution (GTR+I+γ, GTR+γ, and HKY+I+γ for three codon positions, respectively). Additionally, RaxML Blackbox (42) was used in the CIPRES platform (43) to get the ML tree.

Cloning
The DNA sequence encoding vUb and its K54-vUb mutant (in which all lysine were substituted to Arginine, except K54) were synthesized commercially and then cloned into pET3a vector without any tag using NdeI and BamHI restriction sites. The ligation reaction was carried out by T4 DNA ligase, and the plasmid, containing the insert, from the ligated colonies were further verified by Sanger sequencing method using T7 promoter and terminator primers.

Purification of vUb and vUb-K54
Plasmid encoding vUb and K54-vUb mutant were transformed into BL21 DE3* E. coli expression strain. Cells were grown at 37ºC initially, followed by IPTG induction and incubated further at 37ºC for 5-6 hrs. Harvested cell pellet of vUb was dissolved in Sodium Acetate buffer (50mM) without any salt, followed by sonication and centrifugation. The soluble fraction containing vUb was loaded onto the SP FF column (GE), and the protein was eluted at a 300-400mM salt gradient. Eluted fractions were concentrated and injected in a Superdex SD75 16/600 Prep grade (GE) column using AKTA pure machine (GE). The pure fractions of vUb were concentrated and stored at -20ºC in 25mM Sodium Acetate (pH 5.0) and 100mM NaCl.
NMR samples of vUb were prepared by growing BL21-DE3* cells in M9 media containing isotopic Ammonium chloride ( 15 NH4Cl) or isotopic Glucose ( 13 C-Glucose). Protein was purified in the same manner as mentioned above. The final buffer for NMR structural experiments contained 25mM Sodium Acetate (pH 5.0) and 100mM NaCl.
Unlike vUb, K54-vUb mutant protein was insoluble after sonication. Hence the pellet was incubated in 8M urea buffer at 4ºC for 12-14 hrs. After centrifugation the unfolded supernatant fraction was kept for serial dialysis from 6M, 3M, 1.5M, 0.75M, 0.375M and finally at 0M, for 4-6hrs each, for gradual refolding of protein. The refolded protein was loaded in SP FF column and eluted at a 300-400mM salt gradient. Finally, size exclusion chromatography was carried out to get the pure protein which was stored at -20ºC in 25mM Sodium Phosphate buffer (pH 7.5) and 100mM NaCl.

Purification UbE1, E2s, E3s, and Ub proteins
UbE1 clone was a generous gift from Dr. Cynthia Wolberger lab, and their respective purification protocol was followed for purifying UbE1 protein (44).
All the E2s, E2d2, E225k, Ubc13, and its cofactor Mms2 used in this study were tagged with 6XHis at their N-terminus. Ube2d2/UbcH5a plasmid was a gift from Wade Harper (Addgene: 15782). All his-tagged proteins were purified using IMAC FF (GE) columns followed by their elution with 200-300mM Imidazole linear gradient. Eluted fractions were loaded on Superdex SD75 16/600 Prep grade column, which was pre-equilibrated in 50mM Tris buffer (pH7.5) and 250mM NaCl, to get the pure protein and finally stored at -20ºC.
Purification of Ubiquitin proceeded in the same manner as described for vUb.
GST-RNF38 clone was a generous gift from Prof. Danny Huang. Purification of GST-RNF38 and GST-RNF4 was done using GST beads (Protino) and eluted out with 10mM reduced Glutathione. Pure protein was obtained after SEC in 50mM Tris buffer with 250mM NaCl and store at -20ºC.

Purification of Proteasomal receptors
Clones of proteasome receptors Rpn10/S5a (UIM motifs: 196-306aa) and hRpn13 (Pru domain: 1-150aa) were a kind gift from Dr. Kylie Walters lab. Both the clones were tagged with Nterminus 6XHis-tag, and their respective purification was performed as followed for other his tagged E2 proteins mentioned above. with a mixing time of 100ms were used to get NOE restraints. All NMR data were processed in NMRpipe (45) and analyzed by NMRFAM-SPARKY software (19). Following peak-picking of the backbone and side-chain experimental data in SPARKY, the peaks were assigned by PINE and confirmed manually. Automatic peak picking was performed for NOESY spectra. NOESY peak assignments were performed in CYANA (20) and corrected over multiple cycles of structure building. TALOS+ (18) was used to predict the phi and psi torsion angles from the assigned chemical shifts of backbone atoms. Phi/psi torsion angles and NOESY based distance restraints were used to determine the solution structure of vUb by CYANA software. 100 starting structures were calculated, and out of which 20 lowest energy structure models were chosen to represent the final structural ensemble. The backbone dihedral angles of the final converged structures were evaluated by the Molprobity (21) and PSVS (22) suite of programs. Details of the NMR restraints and structure evaluation are provided in Table 1.

NMR Dynamics
For NMR dynamics studies, uniformly labeled 15  b. Thermal Melt 40µM of protein in Sodium Phosphate buffer and 100mM Salt was used for thermal melt. The mean residual ellipticity at 222nm was measured using Jasco J-815 spectropolarimeter, which is connected to a Peltier. The data were recorded from 30ºC to 95ºC after every 1ºC per minute rise in temperature with 32s of data integration time and finally fitted with a two-state folding curve to obtain Tm.

Temperature Coefficient
Temperature coefficients experiments were performed using uniformly labeled 15 N-vUb and 15 N-Ub samples dialyzed in 25mM NaOAc, 50mM NaCl, and pH 5.8. A total of 9 HSQC spectra were collected for both the proteins from 283K to 323K after every 5K increment. All the NMR HSQC experiments were processed by NMRPipe and analyzed in Sparky. Only shifts in 1 Hdimension were considered and later analyzed using linear regression by MATLAB. The slope of each plot between ΔδNH and temperature is termed as temperature coefficients (ΔδNH/ΔT).

ITC Binding Assay
vUb and S5a were co-dialyzed in 20mM sodium phosphate buffer (pH 7.5) containing 100mM NaCl. 4mM vUb was injected into the sample cell containing 80µM of S5a at 25ºC. The measurement was performed on a Microcal ITC200 (GE Healthcare), and binding isotherm was plotted and analyzed using Origin (v7.0). Integrated interaction heat values from individual experiments were normalized, and the data were fit, omitting the first point, using a single binding site model.

Affinity measurements by fluorescence spectroscopy
Change in the intrinsic fluorescence of Tryptophan in Rpn13 upon addition of vUb was monitored, and hence the Kd value was determined. The intrinsic fluorescence emission scan from 300nm to 400nm was measured using Horiba Fluromax-4 after excitation at 295nm. The slit width for excitation and emission was kept at 3nm. The decrease in fluorescence was observed after adding vUb with increasing molar quantities from 0.5nM till 10uM. The change in fluorescence intensity at 340nm at different vUb molar ratios was used to generate the binding curve ( Fig 5H). The total bound fraction was calculated and normalized and then fitted to onesite specific binding curve in Graph pad Prism to calculate Kd.

In-vitro Ubiquitination
In-vitro ubiquitination reactions were carried out using 1µM of UbE1, 5µM of E2s, 10µM of E3s and 50µM of Ub and vUb. The reaction was activated by reaction buffer containing 50mM Tris (pH 7.5), 100mM NaCl, 2.5mM MgCl2 and 5mM ATP and incubated for 1hr at 37ºC. The ubiquitination reaction was quenched with 25mM EDTA and further ran on 15% SDS gel. For probing poly-ubiquitin chains, immunoblotting was performed using Ubiquitin monoclonal antibody (P4D1 from ENZO) at 1:5000 dilution and HRP conjugated anti-mouse secondary antibody at 1:10000 dilutions. The chemiluminescence was observed by clarity ECL (Bio-Rad) staining in Image Quant LAS 4000 (GE). The time point based assay was carried out with 5µM of Ub and vUb. The SDS-PAGE gel was stained overnight with SYPRO® Ruby protein gel stain and imaged in UVITEC Cambridge.

In-vitro Deubiquitination Assay
For the Deubiquitination assays, 5µl of Ub and vUb and 10µl of vUb-K54 of the quenched ubiquitination reaction mixture, which was carried out by E2D2 and RNF38, was incubated with different DUBs such as USP2CD, OTUBAIN, and GST-AMSH. USP2CD and OTUBAIN were activated in DUb activation buffer (25mM Tris, 150mM NaCl, and 10mM DTT) for 15 minutes at RT before initiating their respective deubiquitination reactions. The deubiquitination reaction was carried out at 37ºC for 1hr, quenched with SDS loading dye and immunoblotted onto PVDF membrane and finally probed using Ubiquitin antibody (as mentioned above).

Mass Spectrometry
In-vitro ubiquitination reaction of vUb and K54-vUb was carried out using UbE1 (0.5µM), E2D2 Approximately 50,000 cells were seeded in each well of 12 well plates and incubated for 12hrs at 37ºC with 5% CO2, before transfecting them with 1XFLAG-Ub, 1XFLAG-vUb and 1XFLAG-vUb-K54 respectively using Lipofectamine 3000 reagent and further, the cells were incubated for 24hrs. The cells were then treated with MG132 for 12hrs before lysing the cells in 1XTBS containing 0.5% NP40 and 1X protease inhibitor cocktail. Total protein was estimated using BCA reagent (Thermo), and the final amount of 3µg of FLAG-Ub and 15µg of FLAG-vUb and FLAG-vUbK54 were loaded on 10%SDS-PAGE gel. Immunoblots were probed with Flag antibody.         The difference in temperature coefficients of Ub and vUb is plotted against residue numbers.

Acknowledgments-
The standard deviation is shown as dashed black lines. Red bars represent the residues in vUb that are significantly destabilized upon increasing temperature. The secondary structure elements of vUb are drawn on top of the plot.