Recruitment of FBXO22 for targeted degradation of NSD2

Abstract

Targeted protein degradation (TPD) is an emerging therapeutic strategy that would benefit from new chemical entities with which to recruit a wider variety of ubiquitin E3 ligases to target proteins for proteasomal degradation. Here we describe a TPD strategy involving the recruitment of FBXO22 to induce degradation of the histone methyltransferase and oncogene NSD2. UNC8732 facilitates FBXO22-mediated degradation of NSD2 in acute lymphoblastic leukemia cells harboring the NSD2 gain-of-function mutation p.E1099K, resulting in growth suppression, apoptosis and reversal of drug resistance. The primary amine of UNC8732 is metabolized to an aldehyde species, which engages C326 of FBXO22 to recruit the SCF^FBXO22 Cullin complex. We further demonstrate that a previously reported alkyl amine-containing degrader targeting XIAP is similarly dependent on SCF^FBXO22. Overall, we present a potent NSD2 degrader for the exploration of NSD2 disease phenotypes and a new FBXO22-recruitment strategy for TPD.

Extended Data Fig. 1 |. UNC8732 inhibits cell growth and restores glucocorticoid sensitivity of NSD2 p.E1099K mutant ALL cells.

(a): Total proteome analysis (label-free quantification) of U2OS cells treated with 2 μM UNC8732 or an equivalent volume of DMSO for 3 hours. Volcano plot of all quantified proteins (8240) with log2 fold change shown on the x-axis (dashed line equivalent to a 2-fold change) and −log₁₀ adjusted p values shown on the y-axis (dashed line equivalent to p.adjusted = 0.05). Adjusted p-values were derived from Limma and DEP packages (see methods) and were adjusted using the Benjamini-Hochberg procedure. N=5 independent experiments. (b-d): Viability of isogenic RCH-ACV determined by CellTiter-Glo assay after treatment with varying concentrations of UNC8732 and UNC8884 for (b) 12 days, (c) 15 days, or (d) 21 days. (e-f): Apoptosis of isogenic RCH-ACV detected using annexin V/PI staining by flow cytometry after treatment with varying concentrations of UNC8732 and UNC8884 for (e) 15 days or (f) 21 days. (g-h): Viability of NSD2 mutant RCH-ACV cells determined by CellTiter-Glo after the pretreatment of varying concentrations of UNC8732 and UNC8884 for (g) 12 days or (h) 15 days followed by dexamethasone (1 μM) for 72 hours. (i-j): Apoptosis of NSD2 mutant RCH-ACV cell line detected using annexin V/PI staining by flow cytometry after the pretreatment of varying concentrations of UNC8732 and UNC8884 for (i) 12 days or ( j) 15 days followed by dexamethasone (1 μM) for 72 hours. Data represents the mean ± SEM from three biological replicates. The statistical significance was evaluated using the Two-way ANOVA test and the p-values are displayed. WT, NSD2 WT; Mut, NSD2 p.E1099K; Dex, Dexamethasone. See Supplementary Figs. 3, 4 for representative flow cytometric analysis plots.

Extended Data Fig. 2 |. UNC8732 metabolism to the corresponding aldehyde drives NSD2 degradation.

(a): Structure of UNC8153 prodrug and its corresponding aldehyde metabolite. (b): MS peak area ratio representing the relative levels of UNC8153 and its associated aldehyde species in cell-free DMEM + 10% FBS at indicated time points. * For Extended Data Fig. 2b–d, the experimental treatment was designed for 0h. However, due to the sample preparation process for MS, there’s a brief period of simultaneous presence of the compounds in their respective conditions. (c): MS peak area ratio representing the relative levels of UNC8153 and its associated aldehyde species in cell-free DMEM with no FBS at indicated time points. (d): MS peak area ratio representing the relative levels of UNC8732 and its associated aldehyde species in cell-free DMEM + 10% FBS + 10 mM aminoguanidine (AG) at indicated time points. (e): Structure of UNC8153 prodrug and its corresponding aldehyde metabolite. (f): MS peak area ratio representing the relative level of the aldehyde adduct of UNC9801 upon addition of UNC9801 to DMEM + 10% FBS or DMEM only at indicated time points in hours.

Extended Data Fig. 3 |. FBXO22 is responsible for compound-mediated degradation of NSD2.

(a): U2OS cells were co-treated with DMSO control or 2 μM of UNC8732 and indicated concentrations of MLN4924 neddylation inhibitor for 24h. Representative immunoblot shown. Experiment repeated independently three times with consistent results. (b): U2OS cells were co-treated with DMSO control or 2 μM of UNC8732 and indicated concentrations of MG-132 proteosome inhibitor for 3h. Representative immunoblot shown. Experiment repeated independently three times with consistent results. (c): T-Rex 293 cells stable cell lines with tetracycline inducible NSD2-FLAG-miniTurbo was treated with 10 μM MG-132 and the indicated combination of DMSO, 5 μM UNC8732, tetracycline (1 μg/mL) and biotin (50 μM). (d): Principal Components analysis (PCA) of the BioID spectral counts data from DMSO (blue) or 5 μM UNC8732 (red) treated samples, each containing three independent experiments with 2 technical replicates for each independent experiment. (e): Uniform Manifold Approximation and Projection (UMAP) analysis of the BioID data of DMSO (blue) or 5 μM UNC8732 (red), each containing 3 independent experiments with 2 technical replicates for each independent experiment. (f): AlphaFold protein structure prediction of the SKP1-FBXO22 fusion protein. SKP1 is predicted to interact with the F-box domain of FBXO22 and the FIST_C domain remains some-what separate for putative substrate interactions. (g): SDS-PAGE analysis of the recombinant SKP1-FBXO22 fusion protein and the NSD2-PWWP1 post-purification. (h): Left: Recombinant proteins of SKP1-FBXO22 fusion apo or SKP1-FBXO22 fusion + NSD2-PWWP1 + UNC10088 analyzed using a Superdex 200 Increase 10/300 GL column in the indicated combinations. Right: SDS-PAGE gel showing the eluted proteins of the SKP1-FBXO22 + UNC10088 + NSD2-PWWP1 condition at the indicated peaks. (i): Left: Co-expressed SKP1/FBXO22 apo or co-expressed SKP1/FBXO22 + NSD2-PWWP1 + UNC10088 was analyzed using a Superdex 200 Increase 10/300 GL column in the indicated combinations. Right: SDS-PAGE gel showing the eluted proteins of the SKP1-FBXO22 + UNC10088 + NSD2-PWWP1 condition at the indicated peaks.

Extended Data Fig. 4 |. Aldehyde degraders bind FBXO22 in a cysteine 326-dependent manner.

(a): Representative BLI sensorgrams upon the addition of increasing concentrations of UNC10088 and a fixed concentration of NSD2-PWWP1 (2 μM). WT or C326A-mutant SKP1-FBXO22 was loaded on SA biosensors to an average response of 1 nm. Curves are shown as the average of three independent experiments. (b): In vitro NSD2-PWWP1 ubiquitination with the indicated sets of substrate priming machinery, UBCH5B or UBCH7 with HHARI. Each reaction also contained CUL1-RBX1 and CDC34B. The reaction mixture was analyzed by SDS-PAGE and fluorescence scanning. NSD2-PWWP1 was subject to a sortase reaction for fluorescent labeling. The gel presented is representative of three independent experiments. (c): Immunoblotting following transfection of empty vector pcDNA, WT FBXO22-HT, or C326A-mutant FBXO22-HT in U2OS cells for 24h. (d): Experimental flow chart for the designed ‘wash’ experiment to test for reversibility of the compound-FBXO22 interaction. *0 hour denotes the estimated concentration based on the series of dilutions performed. (e): Normalized fluorescence data from DSF for samples obtained from the experiment described in panel E. The data represents the mean ± SD of 3 technical replicates from one independent experiment. (f): Schematic of the reaction between the aldehyde-containing degrader compound and the C326 of FBXO22. (g): Percent Deuterium Exchange of C326-spanning peptides of co-purified SKP1/FBXO22 in the presence (orange) and absence (black) of UNC10088 over 12 hours. The data represents the mean ± 3σ (3 × SD) of a triplicate technical measurement (n=3).

Fig. 1 |. UNC8732 inhibits cell growth and restores GC sensitivity of NSD2 p.E1099K mutant ALL cells.

a, RCH-ACV cells were treated for 11 days with 0–10 μM UNC8732 or UNC8884. Cells were collected and analyzed by immunoblotting. H3 and HDAC2 are loading controls. NSD2 protein level is normalized to HDAC2. H3K36me2 and H3K27me3 levels are normalized to H3. b, Viability of isogenic RCH-ACV ALL cell lines determined by CellTiter-Glo after treatment with varying concentrations of UNC8732 and UNC8884 for 18 days. c, Apoptosis of isogenic RCH-ACV cell lines detected using annexin V/PI staining by flow cytometry after treatment with varying concentrations of UNC8732 and UNC8884 for 18 days. See Supplementary Fig. 3b for representative flow cytometric analysis plots. d, Viability of NSD2 mutant RCH-ACV cells determined by CellTiter-Glo after pretreatment with varying concentrations of UNC8732 and UNC8884 for 18 days, followed by dexamethasone (1 μM) for 72 h. e, Apoptosis of NSD2 mutant RCH-ACV cells detected using annexin V/PI staining by flow cytometry after pretreatment with varying concentrations of UNC8732 and UNC8884 for 18 days followed by dexamethasone (1 μM) for 72 h. See Supplementary Fig. 4c for representative flow cytometric analysis plots. The data represent mean ± s.e.m. from three biological replicates. The statistical significance was evaluated using the two-way analysis of variance test, and the P values are displayed. WT, NSD2 WT; Mut, NSD2 p.E1099K; Dex, dexamethasone.

Fig. 2 |. UNC8732 metabolism to the corresponding aldehyde drives NSD2 degradation.

a, Structure of UNC8732 prodrug and its corresponding aldehyde metabolite. b–d, MS peak area ratio representing the relative levels of UNC8732 and its aldehyde metabolite: in cell lysates generated from U2OS cells treated with UNC8732 and cultured with DMEM and 10% FBS (b), in cell-free DMEM and 10% FBS (c) and in cell-free DMEM without FBS (d), at indicated time points. *’0 h’ denotes mass spectrometric analysis performed immediately after compound treatment, which includes a very brief sample preparation time to process for MS. e, U2OS cells were treated with 2 μM UNC8732 and varying concentrations of AG for 6 h. NSD2 levels were measured by immunoblot. Vinculin is a loading control. f, U2OS cells were cultured in DMEM + 10% FBS or Opti-MEM with no FBS while being treated with the indicated concentrations of UNC8732 for 24 h, followed by immunoblotting analysis for NSD2 levels. Vinculin is a loading control. g, Structure of UNC10088 prodrug and its corresponding hydrolyzed aldehyde product. h, U2OS cells were treated with the specified compounds in a dose–response format for 24 h. NSD2 levels were measured by an ICW assay. The data represent the mean ± s.d. from four technical replicates in one representative experiment. i, U2OS cells were treated for the indicated amount of time with UNC8732 (2 μM), UNC10088 (2 μM), UNC8153 (10 μM) or UNC9801 (10 μM) and were subsequently analyzed by immunoblotting. Vinculin is a loading control. j, Cells were cultured in DMEM + 10% FBS or Opti-MEM with no FBS and were subsequently treated with indicated concentrations of UNC10088 for 24 h, followed by immunoblotting analysis for NSD2 levels. Vinculin is a loading control.

Fig. 3 |. FBXO22 is responsible for compound-mediated degradation of NSD2.

a, Schematic showing a representation of the NSD2-miniTurbo fusion protein (brown and green), putative UNC8732-mediated interaction with an unknown E3 ubiquitin ligase (blue) and the approximate radius (10 nm) of biotinylation events generated by the NSD2-miniTurbo (pink). b, Volcano plot of BioID results showing log₂ fold change (FC) and −log₁₀ adjusted P values derived from normalized protein spectral counts. P values were adjusted for multiple testing using the Benjamini–Hochberg procedure. Proteins were previously filtered for SAINT BFDR ≤0.01 in either DMSO control or UNC8732 treated conditions. The red data points represent proteins in the GO:BP protein ubiquitination term (GO:0016567), and the blue data points represent proteins included in the GO:CC proteasome core complex term (GO:0005839). Significance cutoff set at adjusted P value <0.05 and fold change >1.5. c, Histogram showing the spectral counts of NSD2 and FBXO22 in either DMSO control or UNC8732-treated conditions. The data presented are the mean ± s.d. of three independent experiments. d, A NanoBRET protein–protein interaction assay with NanoLuc-NSD2 and FBXO22-HT in U2OS cells reveals ternary complex formation. U2OS cells were treated with 0.05–40 μM of UNC8732 24 h post-transfection for 3 h. BRET ratios (mBU) are shown relative to DMSO treatment control (= 1.00). The data presented are the mean ± s.d. of three independent experiments. e, Knockdown of SCF^FBXO22 components by siRNA in U2OS cells. UNC8732 and UNC10088 degrader treatments were performed 4 days post-siRNA transfection for 4 h. f, Representative BLI sensorgrams upon the addition of increasing concentrations of UNC10088 and a fixed concentration of NSD2–PWWP1 (2 μM). Curves are shown as the average of three independent experiments. g, Binding curve with the average BLI response calculated on the basis of steady-state approximation. The data presented are the mean ± s.e.m. of three independent experiments. No response is detected for UNC10088 alone (dark gray) owing to its very small molecular weight relative to SKP1–FBXO22. h, In vitro NSD2–PWWP1 ubiquitination with the indicated sets of substrate priming machinery, UBCH5B or UBCH7 with HHARI. Each reaction also contained CDC34B. The reaction mixture was analyzed by SDS–PAGE and fluorescence scanning. NSD2–PWWP1 was subject to a sortase reaction for fluorescent labeling.

Fig. 4 |. Aldehyde degraders bind FBXO22 in a cysteine 326-dependent manner.

a, ICW quantification of NSD2 levels in U2OS cells co-treated with 1 μM of UNC8732 and 6.25–100 μM of UNC9360 (E3-recruiting handle of UNC8732) or UNC9631 (monomethylated E3-recruiting handle of UNC8732) for 6 h. The data represent the average ± s.e.m. of four technical replicates from two independent experiments. b, Normalized fluorescence data from DSF upon incubation of co-expressed SKP1/FBXO22 with 0–100 μM UNC10088. The data represent the average ± s.d. of three independent experiments. c, The change in melting temperature (T_m) as measured by DSF of co-expressed SKP1/FBXO22 treated with 0–100 μM UNC8732 and UNC10088. The data represent the average ± s.d. of three independent experiments. d, AlphaFold-predicted structure of FBXO22 highlighted to indicate the cumulative HDX difference signal in an HDX-MS experiment (Supplementary Fig. 5). Blue, reduced uptake; red, increased uptake; black, no FBXO22 sequence coverage. For clarity, SKP1 is not shown in the AlphaFold structure, and no changes in deuterium uptake were observed for SKP1 (Supplementary Fig. 6). e, Normalized fluorescence data from DSF upon incubation of co-expressed SKP1/FBXO22 with 0.1–100 μM UNC10088. The error bars represent the s.d. of four technical replicates from one independent experiment. f, Binding curve of the average BLI response for both SKP1–FBXO22 WT and C326A calculated on the basis of steady-state approximation. The data represent the average ± s.e.m. of three independent experiments. g, NanoBRET assay results with NanoLuc-NSD2 and FBXO22-HT WT or C326A mutant performed in U2OS cells. U2OS cells were treated with the indicated concentration of UNC8732 24 h post-transfection for 3 h. BRET ratios (mBU) are shown relative to the DMSO treatment control (= 1.00). The data represent the average ± s.d. of three independent experiments.

Fig. 5 |. A primary amine-containing degrader recruits the SCF^FBXO22 complex for degradation of XIAP.

a, The chemical structure of primary amine-containing XIAP degrader, Compound 10, reported by den Besten et al. b, MCF7 cells were treated with compound 10 and varying concentrations of AG for 24 h, followed by immunoblotting analysis for XIAP levels. Actin is a loading control. c, MCF7 cells were cultured in DMEM + 10% FBS or Opti-MEM with no FBS and subsequently treated with the indicated concentrations of compound 10 for 24 h, followed by immunoblotting analysis for XIAP levels. Actin is a loading control. d, MCF7 cells were treated with compound 10 and varying concentrations of UNC10088 and UNC8732 for 24 h, followed by immunoblotting for XIAP and NSD2. Actin is a loading control. e, MCF7 cells were treated with Compound 10 and varying concentrations of MLN4924 (a selective neddylation inhibitor) for 24 h, followed by immunoblotting for XIAP levels. CUL1 was blotted as a representative Cullin. Vinculin is a loading control. f, siRNA knockdown of core SCF^FBXO22 components in MCF7 cells was performed. Compound 10 treatments were performed 72 h post-siRNA transfection for 24 h. Vinculin is a loading control.