Defining the Antitumor Mechanism of Action of a Clinical-stage Compound as a Selective Degrader of the Nuclear Pore Complex

Abstract

Cancer cells are acutely dependent on nuclear transport due to elevated transcriptional activity, suggesting an unrealized opportunity for selective therapeutic inhibition of the nuclear pore complex (NPC). Through large-scale phenotypic profiling of cancer cell lines, genome-scale functional genomic modifier screens, and mass spectrometry-based proteomics, we discovered that the clinical drug PRLX-93936 is a molecular glue that binds and reprograms the TRIM21 ubiquitin ligase to degrade the NPC. Upon compound-induced TRIM21 recruitment, the nuclear pore is ubiquitylated and degraded, resulting in the loss of short-lived cytoplasmic mRNA transcripts and the induction of cancer cell apoptosis. Direct compound binding to TRIM21 was confirmed via surface plasmon resonance and X-ray crystallography, whereas compound-induced TRIM21-nucleoporin complex formation was demonstrated through multiple orthogonal approaches in cells and in vitro. Phenotype-guided optimization yielded compounds with 10-fold greater potency and drug-like properties, along with robust pharmacokinetics and efficacy against pancreatic cancer xenografts and patient-derived organoids.

Significance: This study establishes the cancer therapeutic potential of optimized TRIM21 molecular glues to degrade the NPC and underscores the value of reexamining drugs with previously unknown mechanisms using current technologies.

Figure 1.

TRIM21 promotes PRLX-induced cytotoxicity. A, Chemical structures of PRLX and its parent compound, erastin. B, PRISM pooled barcoded cell line screen overview. C, Scatter plot of erastin vs. PRLX PRISM viability profiles at 2.5 μmol/L demonstrating a lack of correlation. Each point is a single cancer cell line. D, Volcano plot depicting predictive mRNA expression biomarkers of PRLX activity by linear modeling. E, Scatter plot of TRIM21 mRNA expression vs. PRISM viability dose response (AUC). Each point is a single cancer cell line. P values were adjusted by the Benjamini–Hochberg procedure. F, Lineage-enrichment testing for PRLX PRISM activity. Fisher’s exact test was performed to identify cancer types enriched within the most sensitive quartile of cell lines. G, Cell viability of individual cancer cell lines with high or low TRIM21 expression following treatment with PRLX or vehicle for 72 hours (n = 3–6; error bars, SD). H, Cell proliferation of H661 WT or TRIM21 overexpression (OE) cells treated with PRLX vs. vehicle for 72 hours (n = 3; error bars, SD). I, CRISPR-Cas9 KO and activation library screening overview. J, Gene KO enrichment in the CRISPR-Cas9 KO screen (21 days; PRLX vs. DMSO). K, Gene activation enrichment in the CRISPR-dCas9 activation screen (13 days; PRLX vs. DMSO). L, Cell proliferation of WT or TRIM21 KO PANC-1 cells treated with PRLX or erastin vs. vehicle for 72 hours (n = 3; error bars, SD). sgNT, single guide RNA non-targeting control; TPM, transcripts per million.

Figure 2.

Structure–activity relationship of PRLX analogues targeting TRIM21. A, Chemical structures of PRLX analogues. B, Chemical structure of optimized lead compound JWZ-8-103. C, Cell viability IC₅₀ values for WT and TRIM21 KO PANC-1 cells treated with each analogue vs. vehicle for 72 hours. D, Cell proliferation of WT or TRIM21 knockout PANC-1 cells treated with each analogue vs. vehicle for 72 hours (n = 3; error bars, SD). E, Cell proliferation of H661 WT or TRIM21 overexpression (OE) cells treated with each analogue vs. vehicle for 72 hours (n = 3; error bars, SD). F, Scatter plot of PRISM dose–response viability profiles (AUC) for PRLX vs. JWZ-8-103. Each point is a single cancer cell line. G, Volcano plot depicting predictive mRNA expression biomarkers of PRLX activity by linear modeling. TRIM21 is indicated.

Figure 3.

PRLX and related analogs bind TRIM21. A, TRIM21 primary structure showing RING, B-box, coiled-coil, and PRYSPRY domains. B, Predicted structure of the TRIM21 dimer with bound E2 and ubiquitin. A single protomer is shown in color. Mutated residues conferring resistance in the base editor screen are shown as spheres (yellow, carbon; blue, nitrogen; red, oxygen). C, SPR sensorgrams showing PRLX, JWZ-8-103, or JWZ-8-60 binding to immobilized TRIM21 PRYSPRY when tested at 32 μmol/L. D–F, Crystal structure of TRIM21 PRYSPRY bound to JWZ-8-103. Close-up views show key contacts. Fo–Fc compound omit map contoured at 3.5 σ is shown G, TRIM21 complementation experiments: cell proliferation of TRIM21 KO PANC-1 cells expressing the indicated TRIM21-mutant cDNAs treated with the indicated PRLX analogue vs. vehicle for 72 hours (n = 3; error bars, SD).

Figure 4.

PRLX induces TRIM21-mediated ubiquitylation and degradation of the NPC. A and B, Global proteome profiling of WT PANC-1 cells (A) or TRIM21 KO PANC-1 cells (B) treated with 500 nmol/L PRLX for 6 hours, showing differential statistics of LC-MS/MS quantified proteins (n = 3–4 biological replicates). Nuclear pore proteins are colored blue. C, Immunoblotting of key proteins in WT or TRIM21 KO PANC-1 cells treated with 500 nmol/L PRLX (6 hours). D, Ubiquitylation proteomics of WT PANC-1 cells pretreated with 1 μmol/L MG132 for 1 hour before adding 500 nmol/L PRLX for 0.5 hours showing differential statistics of LC-MS/MS quantified ubiquitylation sites (n = 3 biological replicates). Nuclear pore protein ubiquitylated peptides are colored purple. E, Proximity proteomics in TRIM21-TurboID–expressing PANC-1 cells following 1 μmol/L MG132 (0.5 hours) pretreatment before adding 500 nmol/L PRLX (0.5 hours) or DMSO, followed by 100 nmol/L biotin (0.5 hours), showing differential statistics of LC-MS/MS quantified proteins (n = 3 biological replicates). Nuclear pore protein-labeled proteins are colored green. F, Immunoblotting validation of TurboID-based proximity enrichment following 500 nmol/L PRLX and 100 nmol/L biotin treatment of PANC-1 cells containing TRIM21-TurboID (C-terminal fusion), TurboID-TRIM21 (N-terminal fusion), or TRIM21 KO. G, Heatmap of nuclear pore proteins organized by annotated structural location, showing log₂ fold changes in global (blue), ubiquitylation (purple), or proximity (green) proteomics in PANC-1 cells treated with 500 nmol/L PRLX (*, Benjamini–Hochberg adjusted P value < 0.01). H, Schematic of the NPC with structural features labeled. All displayed adjusted P values (A, B, D, E, and G) were calculated using a LIMMA-based moderated t test with Benjamini–Hochberg adjustment. [H, Created in BioRender. Dwyer, B. (2025) https://BioRender.com/wdnlhkr]. TM, transmembrane.

Figure 5.

NUP98 is the nuclear pore receptor for TRIM21. A, Schematic of NUP98–NUP96 proteins (NCBI gene 4928). The site disrupted by NUP98 sgRNA3 in the CRISPR-Cas9 KO screen is marked. B, Enrichment of NUP98 sgRNA3 reads in the CRISPR screen. C, Cell viability of WT or sgRNA3 PANC-1 cells treated with the indicated compounds (n = 4; error bars, SD). D, Immunoblots of indicated proteins in PANC-1 cells (WT, TRIM21 KO, or sgRNA3) treated with PRLX or JWZ-8-103 for 6 hours. E, Schematic for the TRIM21-NUP98 NanoBiT split luciferase assay (top). NanoBiT assay showing dose-dependent induced proximity between TRIM21 and NUP98 (bottom, isoform 2/3; n = 3, error bars, SD). F, NUP98 APD mutations (top) and induced proximity measurements as in (E) between TRIM21 and NUP98 (n = 3, area fill, SD). G, Coimmunoprecipitation (IP) assay in TRIM21 KO PANC-1 cells expressing 3× Flag-tagged TRIM21 treated with JWZ-8-103. H, SPR sensorgrams showing immobilized NUP98 APD exposed to the indicated analytes: JWZ-8-103 alone (left), TRIM21 PRYSPRY (center left), TRIM21 PRYSPRY with increasing JWZ-8-103 (center right), or JWZ-8-103 with increasing TRIM21 PRYSPRY (right). I, TR-FRET assay measuring ternary complex formation between the NUP98 APD and full-length TRIM21.

Figure 6.

NPC destruction impairs mRNA export and induces cancer cell death. A, Caspase-3/7 activation in WT and TRIM21 KO PANC-1 cells treated with PRLX for 6, 12, or 24 hours (n = 3; error bars, SD). B, Immunoblots of indicated proteins in WT or TRIM21 KO PANC-1 cells treated with PRLX for 6 hours. C, qRT-PCR analysis of cytoplasmic and nuclear MCL-1 and NUP98 mRNA. PANC-1 cells were treated with PRLX for 4 or 6 hours and fractionated as described in the methods section (n = 3–4). Target gene transcript abundance in each fraction was normalized to GAPDH using the 2 (−ΔΔCt) method and analyzed by one-way ANOVA, comparing with the DMSO control (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; error bars, SD). D–E, Immunoblots of indicated proteins in KP3 cells (WT, TRIM21 KO) or H661 (WT, TRIM21 OE) treated with PRLX, JWZ-8-103, or JWZ-8-60 for 6 hours. F, smFISH for MCL1 (red puncta) and c-MYC (yellow puncta) transcripts in PANC-1 cells (box, enlarged at the bottom) treated with DMSO, PRLX, JWZ-8-103, or JWZ-8-60 for 6 hours to assess mRNA expression and localization. Nuclei are outlined with dashed circles. Arrowheads denote PANC-1 cells in which MCL1 and c-MYC transcripts were primarily localized in the nucleus. Scale bars, 50 μm. OE, overexpression.

Figure 7.

Efficacy of TRIM21 molecular glues against PDO and in vivo tumor models. A, TRIM21 expression distribution across HCMI pancreaticobiliary organoids by mRNAseq (n = 65). Models selected for testing are shown in orange. B, Immunoblot to compare TRIM21 protein expression levels between organoids selected for compound testing. C, Cell proliferation of selected HCMI organoids treated with the indicated compounds (n = 3–6; error bars, SEM). D, Immunoblots for nuclear pore proteins and MCL1 in HCMI organoid models. E, Plasma concentrations of PRLX and JWZ-8-103 following a single 3 mg/kg IP injection (n = 3; 8 hours of experiment shown; error bars, SD). F, Immunoblot analysis of indicated PANC-1 xenograft tumor tissue collected 2 hours after the second daily dose of 50 mg/kg PRLX, 15 mg/kg JWZ-8-103, 50 mg/kg JWZ-8-103, or vehicle. G and H, Tumor growth in NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ mice injected intraperitoneally with compound (50 mg/kg PRLX, 15 mg/kg JWZ-8-103, 50 mg/kg JWZ-8-103, or vehicle) daily for 3 weeks. Compound treatments started on day 21 (G) or day 19 (H) after tumor inoculation (n = 9 or 10). Tumor volumes at the end of the experiment were analyzed by one-way ANOVA followed by Dunnett’s multiple comparisons test, comparing with the vehicle (*, P < 0.05; ****, P < 0.0001; error bars, SEM).