Unbiased mapping of cereblon neosubstrate landscape by high-throughput proteomics

Abstract

Molecular glue degraders (MGDs) are small molecules that co-opt the ubiquitin-proteasome system to induce degradation of target proteins, including those considered undruggable. Their discovery remains challenging due to the lack of rational design strategies and limited throughput of unbiased proteome-wide screening approaches. To address this gap, we develop a high-throughput proteomics platform based on label-free, data-independent acquisition mass spectrometry (DIA-MS), enabling integrated proteomics and ubiquitinomics profiling. Screening a diverse set of 100 cereblon (CRBN)-recruiting ligands on this platform leads to identification of a broad array of novel degraders and neosubstrates. Subsequent hit validation and structure-degradation relationship analyses guided by global proteomics reveal highly selective and potent phenyl glutarimide-based degraders targeting previously uncharacterized neosubstrates such as KDM4B, G3BP2 and VCL; none of which contain the classical CRBN β-hairpin degron. These findings underscore the power of unbiased high-throughput proteomics in MGD drug discovery and reveal a substantially expanded CRBN neosubstrate landscape beyond that defined by classical immunomodulatory imid drugs (IMiDs).

Fig. 1. The molecular glue degrader (MGD) proteomics screening workflow.

a Schematic of the proteomics screening workflow. A subset of 100 CRBN-based molecular glue degraders (MGDs) from a 5000-compound library was screened in Huh-7 and NB-4 cells (10 µM, 24 h) in 96-well plates. Each plate included 25 MGDs in triplicate. Proteins were quantified using Bruker timsTOF MS and processed with DIA-NN. Significantly downregulated proteins (>25%) were selected for 6-h re-testing with or without MLN4924, a neddylation inhibitor. Ubiquitinomics (K-GG peptide profiling) was used to validate MLN4924-dependent neosubstrates. Created in BioRender. BioRender.com/h02z778. b Count of significant downregulation events (LIMMA, FDR = 1%) for known neosubstrates in Huh-7 or NB-4 cells after 24-h compound treatment (see Supplementary Data 2) for details. c Scatter plot showing suppression of MGD-induced downregulation (6 h) of known CRBN neosubstrates by MLN4924 (5 µM). d Volcano plots showing protein-level effects of SJ11322 (10 µM), from left to right: (i) 24 h, (ii) 6 h, and (iii) 6 h with MLN4924 (5 µM). Log₂ fold-change (x-axis) and log₁₀ standard error (y-axis) are shown. The chemical structure of SJ11322 is shown on the left side. Up- and downregulated proteins or ubiquitination sites (LIMMA, FDR = 1%/5%) are colored in blue and red. P-value adjustments were made for multiple comparisons using the Benjamini-Hochberg procedure. n.s. not significant. The chemical structure of SJ11322 is shown. e Left: Volcano plot of SJ11322-induced ubiquitinated peptides (30 min). Upregulated sites (LIMMA, FDR = 5%) are in blue, including four FIZ1 sites. Right: Correlation of proteome (x-axis) and ubiquitinome (y-axis) t-statistics. Significantly regulated proteins/sites are colored; turquoise highlights proteins downregulated at 24 h with increased ubiquitination at 30 min. n.s. not significant. f Heatmap of log₂ fold-changes in ubiquitination for CSNK1A1 and ZFP91. Asterisks (*) indicate statistically significant regulations (LIMMA, FDR = 5%). g Correlation between site-specific ubiquitination—CSNK1A1-K51 (left; Spearman’s ρ = −1, p-value = 0.02) and ZFP91-K357 (right; Spearman’s ρ = −1, p-value = 0.08)—with corresponding 6-hour protein downregulation. Asterisks (*) indicate significant protein downregulation (LIMMA, FDR = 1%); linear regression line and Spearman coefficient are shown. Source Data are provided as a Source Data file.

Fig. 2. Discovery of novel neosubstrates in Huh-7 cells.

a Scatter plot illustrating newly identified CRBN neosubstrates in Huh-7 cells. The plot displays all proteins that showed significant downregulation following treatment with 10 µM MGD for 6 h (x-axis) alongside their corresponding fold-changes under co-treatment with 5 µM MLN4924 (y-axis). Proteins regulated in a neddylation-dependent manner are highlighted in blue, while those regulated in a neddylation-independent manner are indicated in red. b Volcano plots depicting protein (left) and ubiquitinated peptide (right) quantifications for the KDM4B degrader SJ41564 (10 µM), with log₂ fold-change on the x-axis and standard error on the y-axis (log₁₀ scale). Significant up- and downregulations of proteins (LIMMA, FDR = 1%) or ubiquitination sites (LIMMA, FDR = 5%) are colored in blue and red, respectively. c Summary of proteome (x-axis) and ubiquitinome (y-axis) t-statistic comparisons for all newly validated neosubstrates in Huh-7 cells. T-statistics for ubiquitination sites mapped to the same protein were averaged, and only significantly upregulated sites were included in the analysis. Volcano plots of protein quantifications for SJ42101 (d) and SJ06986 (e) after 24 h of treatment. Significantly up- and downregulated proteins (LIMMA, FDR = 1%) are colored in blue and red, respectively. The proteome (x-axis) and ubiquitinome (y-axis) t-statistic comparison plots are shown (bottom). Statistically significant up- and downregulations in the proteome (LIMMA, FDR = 1%) and the ubiquitinome (LIMMA, FDR = 5%) are colored in blue and red, respectively. P-value adjustments were made for multiple comparisons using the Benjamini-Hochberg procedure. Proteins that are both significantly downregulated after a 24-h compound treatment and harbor upregulated ubiquitination sites at 30 min of compound treatment are colored in turquoise. n.s. not significant. Source Data are provided as a Source Data file.

Fig. 3. Neosubstrate discovery in HEK293-CRBNoe cells.

a Volcano plots showing protein quantifications of HEK293 (left) and HEK293-CRBNoe (right) in response to the degrader compound SJ10040. The x-axis represents log₂ fold change, while the y-axis shows standard error on a log₁₀ scale. Both cell lines were treated with the compound for 6 h. Significantly up- and downregulated proteins, as determined by LIMMA (FDR = 1%), are indicated in blue and red, respectively. The chemical structure of SJ10040 is shown on the right side. n.s. not significant. b Comparison plot of the t-statistics of the proteome (x-axis) and the ubiquitinome (y-axis) for the two MGDs SJ10040 (top) and SJ42101 (bottom) in HEK293-CRBNoe cells. The t-statistics of significantly upregulated ubiquitination sites mapping to the same gene were averaged. Statistically significant up- and downregulations of the proteome (LIMMA, FDR = 1%) and the ubiquitinome (LIMMA, FDR = 5%) are colored in blue and red, respectively. Proteins that were significantly downregulated after 6 h of compound treatment and exhibited upregulated ubiquitination sites at 30 min of compound treatment are highlighted in turquoise. P-value adjustments were made for multiple comparisons using the Benjamini-Hochberg procedure. Neosubstrates that were identified and validated in HEK293-CRBNoe are indicated in bold. n.s. not significant. c Correlation between site-specific ubiquitination—CSNK1A1-K51 (left; Spearman’s ρ = −0.9, p-value = 0.0002), MNAT1-K176 (middle; Spearman’s ρ = −0.9, p-value = 0.0004), and ZFP91-K357 (right; Spearman’s ρ = −0.9, p-value = 0.0003) on the x-axis—and corresponding protein downregulation after 6 h of compound treatment (y-axis). The different colors indicate the cell line, and asterisks (*) statistically significant protein downregulation, and a best-fitting linear regression line, along with the Spearman correlation coefficient, is shown. d Sankey diagram summarizing all novel neosubstrates identified and validated through ubiquitinomics in Huh-7, NB-4, and HEK293-CRBNoe cells. Compounds are displayed on the left side of the diagram, and the corresponding cell lines are on the right side. Asterisks (*) denote predicted G-loop motifs, while † and ‡ proteins previously reported as CRBN binders in Refs. and , respectively. Source Data are provided as a Source Data file.

Fig. 4. Structure-degradation relationship analysis of screened MGDs.

a t-SNE plot of MGDs clustered by core structure: lenalidomide (Lena), thalidomide (Thal), and phenyl-glutarimide (PG). Each dot represents a compound; black dots indicate the 100 analogs selected for proteomics. Representative core structures are shown. b Heat map of log₂ fold-changes (MGD vs. DMSO) for known and novel neosubstrates in Huh-7 cells. Compounds are grouped by IMiD-like or PG-based cores. The top bar chart shows the number of significantly downregulated neosubstrates per compound. Red boxes highlight G3BP2 degradation by SJ42229 and SJ41824. Proteins with predicted G-loop motifs are indicated. c Volcano plots depicting protein quantifications for the G3BP2 SJ42229 (IMiD-like), SJ41824 and SJ41813 (PG-based) in NB-4 cells (10 µM, 24 h). The x-axis shows log₂ fold-changes, and the standard errors are plotted on the y-axis (log₁₀ scale). Significant changes (LIMMA, FDR = 1%) are in red (down) and blue (up). IMiD-derived neosubstrates are highlighted for SJ42229. The compound structures are shown. n.s. not significant. d MS quantified SALL4 levels in SUSA cells after treatment with IMiD-like (blue) and PG-based (red) compounds (10 µM, 24 h). Box plots represent log₂-transformed intensity values for four biological replicates per compound, with 15 DMSO-treated samples included as controls. Horizontal lines indicate the median (bold) and interquartile range (thin). e Western blot analysis of the indicated proteins in SK-N-BE(2)-C wild-type and CRBN knockout (CRBN^KO) cells treated with SJ41813 (10 µM, 24 h). Shown are the results of one experiment. f Volcano plot of biotinylated proteins from TurboID-CRBN HEK293 cells treated with SJ41813, showing log₂ fold-changes (x-axis) and log₁₀ standard error (y-axis). Significantly enriched proteins (LIMMA, FDR = 1%, n = 3) are labeled in blue. n.s. not significant. g Immunoblots for G3BP2 and USP10 in SK-N-BE(2)-C cells treated with increasing doses (in µM) of SJ41813 (24 h). Blots are representative of three independent experiments. h Quantification of G3BP2 and USP10 blots from (g). DC₅₀ values were calculated from band intensities (mean ± S.D., n = 3) and normalized to DMSO controls. i Volcano plot of ubiquitinated peptides in NB-4 cells treated with SJ41813 (30 min). Significantly upregulated peptides (LIMMA, FDR = 5%; n = 4) are in blue; G3BP2 sites are highlighted. n.s. not significant. Source Data are provided as a Source Data file.

Fig. 5. In-depth characterization of KDM4B and VCL as CRBN neosubstrates.

a Volcano plot of the proteome from SJ41564-treated NB-4 cells (6 h). Log₂ fold-change (x-axis) and log₁₀ standard error (y-axis) are shown. KDM4 family members are highlighted; KDM4B is significantly downregulated (red, LIMMA, FDR = 1%). n.s. not significant. b Immunoblots of KDM4B and α-tubulin in SK-N-BE(2)-C cells treated with increasing concentrations of SJ41564 (nM, 24 h). Representative of three experiments. c Quantification of KDM4B levels from (b). The N-methylated negative control SJ48108 was included. DC₅₀ values were calculated from band intensities (mean ± S.D., n = 3), normalized to DMSO. d Top: Schematic of KDM4B protein architecture and truncated fragments (A–F). Bottom: Immunoblots of FLAG-tagged KDM4B fragments in CRBN-overexpressing SK-N-BE(2)-C cells treated with SJ41564 (10 µM, 24 h), shown from one experiment. Fragments E and F are labeled in red and indicate degradation by SJ41564. e MS quantification of KDM4B in HEK293-TurboID-CRBN cells treated with bortezomib (0.5 µM) or SJ41564 + bortezomib (120 min). KDM4B was significantly enriched (LIMMA, FDR = 1%, n = 3 per condition). f AlphaScreen assay showing dose-dependent ternary complex formation between CRBN–DDB1 and KDM4B (PHD–DTD) upon SJ41564 treatment. EC₅₀ = 4.7 µM. No signal was observed with the negative control SJ48108. DMSO controls are shown in gray. Data represent mean ± S.D. of three technical replicates. g Heat map of log₂ fold-changes (vs. DMSO) for KDM4B and VCL in response to various MGD compounds by global proteomics. h Chemical structures of selected KDM4B and VCL degraders. i MS quantification of VCL in HEK293-TurboID-CRBN cells treated with bortezomib or SJ41564 + bortezomib (120 min, n = 3). VCL was significantly enriched (LIMMA, FDR = 1%, n = 3 per condition). j Volcano plots of proteome (6 h) and ubiquitinome (30 min) in NB-4 cells treated with SJ46479. Significantly regulated proteins (LIMMA, FDR = 1%) and ubiquitinated peptides (LIMMA, FDR = 5%) are shown in red (down) and blue (up). VCL ubiquitination sites are highlighted. n.s. not significant. Source Data are provided as a Source Data file.