Elaboration of molecular glues that target TRIM21 into TRIMTACs that degrade protein aggregates

Abstract

Approaches for the discovery of molecular glues remain limited. Here we report a phenotypic screening approach in which cytotoxins whose mechanisms require ubiquitination show a gain of viability following pharmacological inhibition of the Ubiquitin-like modifier activating enzyme (UBA1/UAE). This approach reveals PRLX-93936 and BMS-214662 as molecular glues that directly target the E3 ligase TRIM21 to induce degradation of nucleoporin proteins and inhibit nuclear trafficking. The cytotoxicity of these agents correlates strongly with TRIM21 expression, suggesting re-evaluation of these clinically tested agents in patients with TRIM21-high cancers. Relative to recently disclosed TRIM21-targeting glues, PRLX-93936 and newly-synthesized analogs represent a distinct structural series, lack known cellular off-targets, and offer greatly enhanced potency. Additionally, we elaborate PRLX-93936 to a heterobifunctional degrader that uses wild-type TRIM21 to degrade a multimeric protein. Together, our work creates opportunities for targeted protein degradation and enables the design of additional TRIM21-targeting glues and Proteolysis-Targeting Chimeras (PROTACs).

Fig. 1. A synthetic rescue screen identifies PRLX-93936 as inducing cell death via a ubiquitin-dependent mechanism.

a Schematic of the synthetic rescue screening strategy. Desired molecules show reduction of cytotoxicity when E1 Ubiquitin Activating Enzyme activity is impaired by the UBA1 inhibitor TAK-243, indicative of a mechanism of cell death reliant on ubiquitination. Created in BioRender. Adams, D. (2025) https://BioRender.com/66b3l1h. b Dot–plot representing the performance of 1754 bioactive small molecules in the synthetic rescue screen performed using OCI-AML-3 cells treated for 10 h. x-axis, viability measured for each library molecule with only DMSO vehicle treatment; y-axis, viability measured for each library molecule co-treated with TAK-243 (200 nM). Hits (red oval) were significantly cytotoxic as single agents (viability <50%, x-axis) and showed at least 30% greater viability in combination with TAK-243. Red circle, PRLX-93936. Each point represents data from a single well. c Structures of PRLX-93936 and erastin. d Cell viability (CellTiter-Glo) following treatment with PRLX-93936 alone or in combination with TAK-243 (200 nM) or Bortezomib (500 nM) for 10 h in OCI-AML-3 cells (n = 2 independent experiments, each with at least 2 wells per concentration). e, f Pearson correlation plots (www.depmap.org) establishing that TRIM21 expression (log₂(Transcripts per million+1) is correlated with PRLX-93936 cytotoxicity (log₂(Fold Change) (e) and that the cytotoxicity of PRLX-93936 and BMS-214662 is uniquely well correlated with TRIM21 among the small molecules within the database (f). Pearson correlation p-values were derived from two-sided tests with no correction for multiple comparisons. Top 1000 correlates are shown. Each correlation represents data from at least 453 cell lines. g Structures of BMS-214662 and BMS-225975. h Cell viability (CellTiter-Glo) following treatment with PRLX-93936 alone or in combination with TAK-243 (200 nM) or Bortezomib (500 nM) for 10 h in OCI-AML-3 cells (n = 2 independent experiments, each with at least 2 wells per concentration). Source data are provided as Source Data files.

Fig. 2. TRIM21 loss prevents cell killing by PRLX-93936 and BMS-214662.

a, b Western blot following CRISPR/Cas9 targeting of TRIM21 in OCI-AML-3 cells (a) and JURKAT cells (b). Representative of n = 1 independent experiments. c, d Cell viability following treatment with PRLX-93936 for 24 h in OCI-AML-3 cells (c) or JURKAT cells (d). e, f Cell viability measurements as described for c, d but using BMS-214662. g Cell viability following treatment with erastin for 72 h in OCI-AML-3 cells. h Cell viability following treatment with BMS-225975 for 72 h in JURKAT cells. All cell viability measurements used CellTiter-Glo. All CTG data represents n = 2 independent experiments, each with at least 2 wells per condition, except h which is n = 1 with 2 wells per condition. Source data are provided as Source Data Files.

Fig. 3. Overexpression of TRIM21 drives sensitivity to PRLX-93936 and BMS-214662, which directly target TRIM21.

a, d, g Western blot following lentiviral overexpression of TRIM21-FLAG or inactive mutant TRIM21CA-FLAG in OCI-AML-3 cells (a), C33A cells (d), or HEK293T cells (g). Representative of n = 1 independent experiment. b, e, h Cell viability following treatment with PRLX-93936 for 24 h in TRIM21-expressing OCI-AML-3 cells (b) or 72 h in TRIM21-expressing C33A (e) or HEK293T cells (h) (n = 2 independent experiments, each with at least 2 wells per concentration, except (h), which shows two replicate wells from n = 1 independent experiment). c, f, i Cell viability following treatment with BMS-214662 for 24 h in TRIM21-FLAG-expressing OCI-AML-3 (c), C33A cells (f), or HEK293T cells (i). N = 2 independent experiments, each with at least 3 wells per concentration, except (i) which shows two independent replicate wells from one experiment. j, k CETSA experiment monitoring the impact of PRLX-93936 (j; n = 1) or BMS-214662 (k; n = 2 independent experiments) on the thermal denaturation of TRIM21-FLAG in OCI-AML-3 cells. l Immunoprecipitation assay in which TRIM21 is captured by an IgG-coated resin in the presence of PRLX-93936 analog 1 (see Supplementary Table 2 and Fig. 6a for characterization of this analog) at indicated concentrations. IP = Immunoprecipitation, FT = Flow-through. IP and FT/Input samples from the same experiment were run on separate gels in parallel. Representative of n = 1 independent experiments. Source data are provided as Source Data Files.

Fig. 4. PRLX-93936 and BMS-214662 induce degradation of nucleoporin protein and impair nuclear transport.

a Label-free LC/MSMS quantitation (LFQ) of protein levels following PRLX-93936 treatment (500 nM, 6 h) in Jurkat cells (x-axis) and OCI-AML-3 sgScramble (sgSCR) cells (y-axis), with nucleoporin proteins highlighted in orange. b, c Volcano plot highlighting alterations in protein levels by LFQ following treatment with PRLX-93936 in OCI-AML-3 sgSCR cells (b) and OCI-AML-3 sgTRIM21 cells (c) with nucleoporin proteins labeled in blue. FC = fold change. d, e LFQ intensity values for specific nucleoporins noted panels a-c in both sgSCR and sgTRIM21 OCI-AML-3 cells. f Heatmap showing fold change in nucleoporin abundance by LFQ after treatment of OCI-AML-3 cells with PRLX-93936 (500 nM) for 1, 2, or 4 h. Represents mean of n = 3 independent replicates. g Volcano plot as in b except OCI-AML-3 sgSCR cells are treated with BMS-214662 (1 μM, 4 h). h, i LFQ intensity values for specific nucleoporins in both sgSCR and sgTRIM21 OCI-AML-3 cells following treatment with BMS-214662 (1 μM, 4 h). Scatter and volcano plot points represent mean of n = 3 independent replicates. Volcano plot P-values derived from two-sided t tests, no correction for multiple comparisons. Bar graph points represent one of n = 3 independent replicates. j Western blot demonstrating reduction of NUP214 and NUP88 protein levels with increasing concentrations of PRLX-93936 treatment (4 h, OCI-AML-3 cells). k Western blot evaluating the impact of pre-treatment (2 h) with the proteasome inhibitor bortezomib (5 μM) or UBA1 inhibitor TAK-243 (0.5 μM) prior to addition of PRLX-93936 (1 μM, 4 h). Western blot results in (j, k) representative of n = 1 independent experiments. l Quantitation of immunofluorescence imaging of RANBP1 reported as the ratio of nuclear to cytoplasmic signal intensity in the indicated C33A overexpression cell lines. LMB = Leptomycin B; PRLX = PRLX-93936; BMS = BMS-214662. All treatments 6 h. Points represent average of 2 wells/condition with >900 cells quantified/condition from n = 2 independent experiments. m Representative images of RANBP1 subcellular localization following the treatments in l (n = 2 independent experiments). Panels (a, b) and (g, h) include data from the same experiment. Source data are provided as Source Data Files.

Fig. 5. PRLX-93936 and BMS-214662 phenotypes can be abrogated by disrupting the NUP98 autoproteolysis domain.

a Structure of hydroxy-acepromazine. b, c Viability (CellTiter-Glo) of OCI-AML-3 cells (WT or TRIM21 KO) following 24 h treatment with the indicated concentrations of hydroxy-acepromazine (b) or PRLX-93936 (c). d–f A549 cells were subjected to CRISPR/Cas9 targeting using 5 sgRNAs targeting the autoproteolysis domain of NUP98 and then exposed to PRLX-93936 (1 μM) for 7 days. Surviving cells were then exposed to the indicated concentrations of PRLX-93936 (d), BMS-214662 (e) or TAK-243 (f) and cell viability was measured after 72 h with CellTiter-Glo. All CellTiter-Glo experiments represent n = 2 independent experiments with at least 2 wells per condition. g Western blot of NUP98 in WT A549 as well as A549 cells targeted with the five independent NUP98 sgRNAs. A NUP98 proteoform (~200 kDa) indicative of impaired autoproteolysis is highlighted. h Western blot of NUP214 and NUP88 following PRLX-93936 treatment (1 μM /12 h) in A549 cells resistant to PRLX-93936 following treatment with four of the NUP98 sgRNAs reported in d-g. Western blotting results in (g, h) representative of n = 1 independent experiments. Source data are provided as Source Data Files.

Fig. 6. Identification of PRLX-93936 analogs with enhanced potency and a PRLX-93936-derived TRIMTAC.

a–c Viability (CellTiter-Glo) of OCI-AML-3 cells following 24 h treatment with the indicated concentrations of the drawn PRLX-93936 analogs. n = 3 independent biological replicates are shown, each representing the mean of two independent wells. d Representative images of an A549 PML-eGFP-BRD4(BD2) clone expressing TRIM21-FLAG treated with indicated compounds for 8 h. 31–35 each 10 μM; see Supplementary Fig 3d for structures of 31–33. Representative of n = 1 independent experiments. e Quantification of EGFP foci in A549 PML-eGFP-BRD4(BD2) clone expressing TRIM21-FLAG treated with indicated compounds in dose for 8 h. Points indicate two independent wells derived from n = 1 independent experiment with >50 quantified cells per dose. f Representative images as in (d). 34, 10 μM; HGC1g, 10 μM; BTZ, Bortezomib, 2 μM. Representative of n = 2 independent experiments. g Image quantification as in (e) for the experiment described in (f). N = 2 independent experiments are shown, each point represented as the mean of two independent wells and >100 quantified cells per dose. h Structures of 34 and 35. Absolute configuration of the chiral axis of 34/35 is unknown, and the assignment is arbitrary. Source data are provided as Source Data Files.