Design of linked-domain protein inhibitors of UBE2D as tools to study cellular ubiquitination

Abstract

Ubiquitin (Ub) is a post-translational modification that largely controls proteostasis through mechanisms spanning transcription, translation, and notably, protein degradation. Ub conjugation occurs through a hierarchical cascade of three enzyme classes (E1, E2, and E3s) involving >1000 proteins that regulate the ubiquitination of proteins. The E2 Ub-conjugating enzymes are the midpoint, yet their cellular roles remain under-characterized, partly due to a lack of inhibitors. For example, the cellular roles of the promiscuous E2 UBE2D/UBCH5 are not well described. Here, we develop a highly selective, multivalent, engineered protein inhibitor for the UBE2D family that simultaneously targets the RING- and backside-binding sites. In HeLa cells, these inhibitors phenocopy knockdown of UBE2D by reducing the IC₅₀ to cisplatin and whole-cell proteomics reveal an increased abundance of ~20% of the identified proteins, consistent with reduced Ub degradation and proteotoxic stress. These precision tools will enable new studies probing UBE2D's central role in proteome management.

Figure 1:. Linked-domain proteins inhibit UBE2D.

A) Schematic for the design of RING-UBL. The structural model of the UHRF1 RING (purple) bound to UBE2D (cyan) was produced by aligning the UHRF1 RING to the RNF4 RING domain in PDB 4AP4 and the UBL/UBE2D model was produced using Rosetta in a previous publication. The expected distance between the C-terminus of the UHRF1 RING and the N-terminus of the UHRF1 UBL is shown. B) IAP2 autoubiquitination assay in the presence of the UBL (50μM), UBL (50μM) and RING (75μM), and RING-UBL (50μM). C) Cul3 autoubiquitination in the presence of the indicated amounts of RING-UBL. D) Schematic for the design of UBOX-UBL. The structural model of the UHRF1 UBL and E4B UBOX (hot pink) domains bound to UBE2D was produced by aligning the UBOX domain from PDB 2KRE to the RING domain of UHRF1 from our previous model. The expected distance between the C-terminus of the UBOX and the N-terminus to the UHRF1 UBL is shown. E) Inhibition of IAP2 autoubiquitination in the presence of the indicated concentrations of UBOX-UBL, RING-UBL, and the UBL domain. Ub assays in B-D were conducted using FLAG-Ub and visualized using anti-FLAG WB. ITC binding isotherm for (F) RING-UBL and G) UBOX-UBL binding to UBE2D1. Thermodynamic parameters and heat per injection shown in Figures S1E–G. H) UHRF1 ubiquitination assay in the presence of the indicated concentrations of UBOX-UBL (90, 45, 15.6, 3.9, 0.975, 0.244, 0.131, 0.087, 0.058, 0.038 μM). This Ub assay was conducted using Cy5-Ub and the * represents a background band in the Ub stock.

Figure 2:. Inhibition mechanism for the linked-domain inhibitors.

A) Crystal structure of E1 (surface; grey)/Ube2D (cyan) complex (PDB code: 4II2) with the E4B UBOX and UBL model superimposed to show the overlap between the UBOX-UBL and the E1. The tested COSMIC mutations in the UBOX and the F46V mutation in the UBL are shown as spheres. Mutations that abrogated inhibition are shown in yellow. The supporting experiments are shown in Figures S2A, S2C, and S2D. B) Crystal structure of the Cul1^nedd8-Rbx1 (CRL) bound (surface; grey) to UBE2D with the UHRF1 UBL and UHRF1 RING superimposed to show the overlap with the Cul1^nedd8-Rbx1 (CRL). The Ub conjugate is omitted from the CRL surface map for clarity. COSMIC mutations tested in the UHRF1 RING are shown as spheres. Ubiquitination assays are shown in Figure S2B, S2E, and S2F. C) E2 loading assay in the presence of UBOX-UBL and RING-UBL. D) Quantification of the E1~Ub (top; n=2) and E2~Ub (bottom) in the assay depicted in panel C. E) E3 competition assay with increasing UHRF1 concentration (0.7, 2, 4μM) in the presence and absence of 15 μM RING-UBL. F) Quantification of normalized H3-Ub and UHRF1 autoubiquitination activity (+inhibitor/−inhibitor) from 0.7μM versus 4μM UHRF1. Statistical significance tested using the repeated-measure one-way ANOVA (**=p-value < 0.01 n=8) G) E1 competition assay with increasing E1 concentration (100nM, 200nM, 400nM) in the presence and absence of 1μM UBOX-UBL. H) Quantification of normalized E2~Ub band (+inhibitor/−inhibitor) from the assay depicted in panel G. Statistical significance tested using a repeated-measure one-way ANOVA (*=p-value <0.05 n=2). I) EDTA-quenched thioester ubiquitin discharge assay in the presence of 23μM UBOX-UBL and RING-UBL. J) Oxyester discharge assay in the presence of UBOX-UBL and RING-UBL at the indicated concentrations. In this assay the bands are detected using Coomassie stain.

Figure 3:. Linked-domain inhibitors are selective for UBE2D.

A) Skp1/CUL1^Nedd8/F-box/Rbx1 (SCF) ubiquitination of P32 β-catenin peptide with either UBE2D3 or UBE2R in the presence of 10μM or 100μM of RING-UBL or UBOX-UBL. B) Quantification of the ubiquitinated products in A (n=2). C) Example APC/C ubiquitylation assay of fluorescent Ub-Cyclin B with either UBE2D2, UBE2C, or UBE2S in the presence of either RING-UBL or UBOX-UBL. D) Quantification of APC/C reaction in Figure S3A with 23μM of each inhibitor (n=2). E) Autoubiquitination of SUMO-UBOX using UBE2D1 or UBE2E1. UBOX-UBL concentrations are 1, 10, and 100 μM. F) Quantification of the assays depicted in panel E (n=2). G) Yeast two-hybrid assay showing growth of yeast co-transformed with the inhibitor and a single E2 from the panel of 24, grown in liquid synthetic dropout media lacking Histidine, Tryptophan, and Leucine and supplemented with Aerobasidin A (n=2). H) E2 loading assay with the indicated recombinant purified E2s (n=3–5 depending on the E2). While there are no significant differences between D1, D2, D3, all other E2s were significantly different from D1, D2, and most from D3 (p-value <0.05). Statistics are tested using a repeated-measure, one-way ANOVA.

Figure 4:. Design of high-affinity UBE2D inhibitors.

A;top) Crystal structure of UbvD1 (PDB: 6D4P) bound to UBE2D1 with UBOX domain superimposed. A;bottom, Architecture of UBOX-UbvD1_short and UBOX-UbvD1_long. B) Autoubiquitination of IAP2 in the presence of 0.1, 1, 10, and 100μM UBOX-UbvD1_short, UBOX-UBL, or UbvD1. This assay was conducted with FITC-Ub. C) Quantification of the assay depicted in panel B (n=2). D) SEC assay showing UBOX-UbvD1_short (sky blue) or UBOX-UbvD1_long (purple) and UBE2D (pink) alone compared to the complexes (UBOX-UbvD1_short/UBE2D1;blue or UBOX-UbvD1_long/UBE2D1;brown). E) ITC binding isotherm showing the binding of UBOX-UbvD1_short(blue) or UBOX-UbvD1_long(brown) with UBE2D1. Heat per injection and the thermodynamics parameters are shown in Figure S4F. F) Autoubiquitination of IAP2 using 3nM UBE2D in the presence of decreasing concentrations of UBOX-UbvD1_short (100, 50, 30, 10, 5, and 1 nM) and UBOX-UbvD1_long (100, 50, 30, 10, 5, 1, 0.5, 0.1 nM). This assay was conducted using Cy5-Ub. G) Quantification of assay depicted in panel F). Statistical significance tested using a repeated-measure one-way ANOVA (*=p-value <0.05, n=3). H) E2 loading assay using UD1_long. There is no statistically significant difference between UBE2D1, UBE2D2, and UBE2D3. UBE2D1 is statistically significant from all other E2s, while UBE2D2, and UBE2D3 are not (p-value >0.05 or greater). These assays were conducted with Cy5-Ub. I) E2-loading assay showing increase in E1~Ub only in the presence of UBE2D1 and not UBE2D2 or UBE2D3. J) Quantification of the assay depicted in panel J. Statistical significance tested using a repeated-measure one-way ANOVA (**=p-value <.01, ***=p-value 0.001 n=5).

Figure 5:. Linked-domain inhibitors rewire the proteome in HeLa cells.

A) Viability of HeLa cells transfected with the indicated constructs and treated with indicated cisplatin concentrations for 72 hours. B) Bar chart of 24-hour cisplatin IC₅₀ (n=6). Viability assays are shown in Figure S5D. C; left) Volcano plot of matched protein abundance for UBOX-UBL versus UBOX-UBL_control. Difference is calculated as Log₂(FC) based on label-free quantification using MaxQuant. C; right) Pie chart of the more abundant, less abundant, and not significantly different proteins from the shotgun proteomics. Enrichr analysis of the enriched/depleted proteins identified in shotgun proteomics tested against the indicated databases, D) MAGMA, E) KEGG, F) GO Biological Pathway G) PANTHER. The left Y-axis is combined score (pink) right Y-axis is adjusted P-value (purple).