Main

TP53 is the most mutated gene in human cancer (36% according to the Pan-Cancer Analysis of Whole Genomes, 37% according to The Cancer Genome Atlas (TCGA) and 39% according to the American Association for Cancer Research GENIE project)1,2,3. TP53 mutations have been shown to induce resistance to cytotoxic chemotherapy and are associated with significantly worse prognosis across nearly all cancers4,5,6. In total, TP53-mutant cancers are estimated to be associated with ~46% of cancer deaths (Supplementary Fig. 1; analysis across cBioPortal7,8). In the past, attempts have been made to therapeutically target TP53-mutant cancers in several ways9: delivery of wild-type (WT) TP53 genes to cells with gene therapy or mRNA10,11, immunotherapy12,13,14 or small-molecule refolders of p53 protein15,16. Attempts to enact these strategies have invariably failed, and the activity of purported mutant-agnostic p53 refolders has consistently been shown to be off-target17,18. Furthermore, small-molecule refolding of p53 may not be possible for contact mutants of p53 specifically affecting residues that interact with the negatively charged phosphate backbone of DNA. New generalizable strategies are needed to specifically kill cancer cells harboring these mutations.

In contrast to mutations of most tumor suppressor genes, such as APC, RB1 or PTEN, TP53 mutations are typically missense mutations. These mutations act as dominant negative mutations that disallow DNA binding at consensus sequences and, therefore, poison the tetramer19. Because they act as dominant negatives, they function as a shorter evolutionary path to inactivating most p53 functions within the cells that harbor these mutations20,21,22. p53 typically acts as a central signal integrator for various cellular stress signals such as hypoxia, DNA damage or excessive oncogenic signaling23,24. When these mutations occur, the mechanisms that normally sense the above cellular stressors are typically intact and jointly work to activate p53 by increasing its abundance and promoting its nuclear translocation. Meanwhile the mutant p53 that accumulates is impotent to drive the transcription of its own destroyer, MDM2. The negative feedback loop that usually serves to tightly regulate p53 protein is broken and the half-life (t1/2) of p53 stretches from its typical ~5–20 min to several hours. Thus, p53 protein accumulates in cells with TP53 missense mutations25,26,27,28,29.

Several recent examples have revealed the powerful effects of gain-of-function-induced proximity pharmacology such as transcriptional and epigenetic chemical inducers of proximity, velcrins, regulated induced proximity-targeting chimeras (RIPTACs) or cancer mutation-specific effectors of targeted protein degradation30,31,32,33. Application of gain-of-function approaches might provide a generalizable method to target deadly cancers associated with mutant TP53. Here, we develop gain-of-function small molecules that target the abundance of p53 protein, a characteristic of cancers harboring TP53 missense mutants, to selectively kill cancer cells with high levels of mutant p53.

Results

TP53-mutant cells lack synthetic lethal dependencies

Large-scale functional genomic studies have recently been undertaken to systematically determine the genetic vulnerabilities of nearly 1,000 cancer cell lines34,35. We sought to test the hypothesis that there exist synthetic lethal interactions with TP53-mutant cancer cells. We performed a genome-wide analysis of CRISPR dependencies in cells with TP53 mutations versus WT cell lines. Unfortunately, there is no genetic dependency that is enriched in cancer cells with TP53 loss-of-function mutations (Fig. 1a), consistent with a prior report22, nor any profiled small molecules selectively killing TP53-mutant cells (Extended Data Fig. 1a,b). One might instead hope that a subset of proteins are enriched in cells with TP53 mutations, particularly cell surface proteins that might be amenable to targeting with a number of effector modalities such as chimeric antigen receptor T cells or antibody–drug conjugates. Unfortunately, no such cell-surface proteins are enriched in TP53-mutant cancers (Fig. 1b and Extended Data Fig. 1c).

Fig. 1: p53 protein abundance is the only genetic or proteomic distinction between TP53-mutant and WT cancer cells.

a, Mean CRISPR dependency scores for genes across DepMap in TP53 WT (x axis) versus TP53-mutant cells (y axis). There are no genes with (mean DepScore)Mut < −0.5 and Δ(mean DepScore)Mut-WT < −0.2. b, Proteome-wide volcano plot of proteins that are enriched or depleted in TP53-mutant cells versus TP53 WT cells across DepMap (difference in mean z-scored quantitative proteomics abundance; t-test). c, TP53 mRNA expression (log2(transcripts per million + 1)) versus p53 protein abundance (log2(signal) from RPPA) across DepMap. Cell lines are colored by TP53 mutation status (WT, gray; truncating, red; missense, yellow). d, Representative immunohistochemistry of p53 in the Human Protein Atlas66 from resected tumors.

TP53-mutant cells strongly upregulate p53 protein

Quantitative proteomics of the Cancer Cell Line Encyclopedia (CCLE)36 shows that the only protein whose abundance is significantly increased in TP53-mutant cancers is p53 itself (Fig. 1b; Δ(z-scored abundance)Mut-WT > 2, q < 0.05). p53 protein is expressed at a low level in TP53 WT cancer cells but is elevated in cells with TP53 missense mutations (Fig. 1c; >2-fold increase in mean reverse-phase protein array (RPPA) signal), for both monoallelic and biallelic mutations, as well as structural (for example, Y220C) and contact (for example, R273H) mutants (Extended Data Fig. 1d,e). Correspondingly, immunohistochemistry of p53 in tumor tissues shows abundant staining in cancer cells but not adjacent normal tissue (Fig. 1d). Taken together, these data demonstrate that the only genetic dependency or substantial proteomic difference between TP53-mutant cells and WT cells, detected to date, is the overabundance of p53 protein in TP53-mutant cells.

The increased abundance of the p53 protein in cancer cells with TP53 missense mutations provides a unique therapeutic opportunity: a p53 concentration-dependent toxin could selectively kill TP53-mutant cancer cells in a way that is generalizable to multiple TP53 missense mutations.

Bifunctional molecules kill cells with p53 overabundance

We, therefore, hypothesized that, if the abundance of p53 protein could be translated into a proportionate cell death signal, cells with TP53 missense mutations could be selectively killed. We hypothesized that bifunctional small molecules consisting of a p53 binder and a small-molecule toxin may be able to concentrate the toxin in cells overexpressing mutant TP53 (ref. 33). Unfortunately, no high-affinity ligands for WT p53 exist. In the absence of such ligands, we used the Halo tag as a surrogate for a small-molecule binder of p53. Mutant p53-R273H was stably expressed in 293T cells fused to a Halo tag and mCherry for visualization. The fusion protein showed nuclear expression similar to native p53 (Extended Data Fig. 2a).

To leverage the abundance of p53 protein for therapeutic purposes, a bifunctional molecule would ideally be designed such that its binding partner is less abundant than overexpressed mutant p53 protein and highly essential for cell proliferation. To determine which cytotoxin to append to a potential bifunctional molecule, we compared CRISPR gene essentiality scores across Dependency Map (DepMap)37 and absolute protein abundance profiled in OpenCell38 (Fig. 2a). We identified five targets with high essentiality and low abundance: DNA2, WEE1, LRR1, PRELID1 and PLK1 (Fig. 2a and Methods). Of these, WEE1 and PLK1 have ligands previously used to synthesize bifunctional small molecules for targeted protein degradation39,40,41. We functionalized WEE1 inhibitor adavosertib42 (half-maximal inhibitory concentration (IC50) = 5.2 nM for cell-free WEE1 inhibition) and an analog of PLK1 inhibitor BI2536 (ref. 43) (IC50 = 0.83 nM for cell-free PLK1 inhibition), both of which display potent antiproliferative effects across DepMap (adavosertib median IC50: 348 nM, BI2536 median IC50: 11.8 nM; Extended Data Fig. 2b,c), with a Halo tag ligand and PEG2 linker, yielding the bifunctional small molecules Halo-PEG2-adavosertib (1) and Halo-PEG2-BI2536 (2) (Fig. 2b and Extended Data Fig. 2d).

Fig. 2: Bifunctional molecules liganding p53 fusion proteins selectively inhibit proliferation of TP53-mutant cells.

a, Protein concentration in HEK293T cells profiled in OpenCell versus mean CRISPR dependency score across DepMap for all genes and protein products. b, Structure of small molecule (Halo-PEG2-BI2536) and fusion protein (Halo–p53-R273H (FL)–mCherry) constructs used in experiments. c, Cell viability of 293T cells stably expressing Halo–p53-R273H (FL)–mCherry versus parental 293T cells treated with Halo-PEG2-BI2536, measured by CTG after 5 days (n = 8 technical replicates per condition). d, Crystal violet staining of 293T cells stably expressing Halo–p53-R273H (FL)–mCherry versus parental 293T cells treated with Halo-PEG2-BI2536, analyzed after 4 days. The experiment was performed in triplicate; a representative replicate is shown. e, Competition experiment in which a 1:1 mixture of Halo–p53-R273H (FL)–mCherry and parental 293T cells were treated with Halo-PEG2-BI2536 (200 nM) for 7 days and analyzed for mCherry expression by FACS. The experiment was performed in triplicate; a representative replicate is shown. f, Calu-1 cell lines stably expressing Halo–p53 WT (FL)–mCherry–2A–mTagBFP2-V5 and Halo–p53-R273H (FL)–mCherry–2A–mTagBFP2-V5 were established and analyzed by RT–qPCR for expression of the transgene. Transcript abundance is normalized to GAPDH expression using the 2−∆∆Ct method and is plotted relative to Calu-1 cells stably expressing Halo–p53 WT (FL)–mCherry–2A–mTagBFP2-V5. mTagBFP2 expression in these cell lines was also assessed by FACS and the MFI is plotted. g, mCherry expression in Calu-1 cell lines established in f assessed by FACS; the MFI is plotted. h, Cell viability of Calu-1 cell lines established in f treated with Halo-PEG2-BI2536 (left) or BI2536 (right) measured by CTG after 4 days (n = 6 technical replicates per condition for Halo-PEG2-BI2536 and n = 4 technical replicates per condition for BI2536). i, Modeling of final intracellular concentration at steady state for a freely diffusing bifunctional compound binding to p53 (Methods) based on p53 protein t1/2 and initial extracellular concentration (held constant); purple indicates a longer t1/2.

Source data

Halo-PEG2-BI2536 inhibited proliferation of 293T cells stably expressing Halo–p53-R273H (full length (FL))–mCherry at doses between 20 nM and 500 nM, with little impact on the proliferation of parental 293T cells (Fig. 2c,d; p53-R273H 293T IC50: 23 nM, parental 293T IC50: 1,143 nM). When competing Halo–p53-R273H (FL)–mCherry 293T cells with parental 293T cells in the presence of Halo-PEG2-BI2536 (200 nM), we observed a 1,200-fold decline in the population of mCherry+ cells after 1 week (0.04% versus 47.94% for DMSO control; Fig. 2e). These results were reproducible in a competition assay between the parental TP53-null (TP53 homozygous deletion) Calu-1 cell line and Calu-1 cells stably expressing Halo–p53-R273H (FL)–mCherry (Extended Data Fig. 2e; half-maximal effective concentration (EC50) = 13 nM). Furthermore, treatment of Halo–p53-R273H (FL)–mCherry 293 T cells with low doses of Halo-PEG2-BI2536 resulted in selection for low expressors of the fusion protein (Extended Data Fig. 2a). Halo-PEG2-adavosertib also selectively inhibited the proliferation of Halo–p53-R273H (FL)–mCherry 293T cells in the competition assay at doses between 100 nM and 1 μM (Extended Data Fig. 2f; EC50 = 41 nM).

To assess the therapeutic window obtained by this strategy, we established Calu-1 cell lines stably expressing Halo–p53-R273H (FL)–mCherry and mTagBFP2 from the same transcript using 2A peptides. We established versions of this cell line for TP53 WT and TP53R273H. We confirmed that these cell lines displayed approximately equal expression of the transgene by reverse transcription (RT)–qPCR for the transcript and fluorescence-activated cell sorting (FACS) for mTagBFP2 (Fig. 2f; mean [TP53 WT mRNA]/[TP53R273H mRNA] = 1.14; mean mTagBFP2 mean fluorescence intensity (MFI) p53 WT/p53-R237H = 0.77). We then compared expression of Halo–p53-R273H (FL)–mCherry between the cell lines using FACS for mCherry and found 30.2-fold greater expression of Halo–p53-R273H (FL)–mCherry compared to Halo–p53 WT (FL)–mCherry (Fig. 2g). We next assessed whether the 30-fold increase in p53 protein abundance was targetable. Indeed, Halo-PEG2-BI2536 had no effect on the proliferation of Calu-1 p53 WT cells at doses ≤ 2 μM. However, in Calu-1 p53-R237H cells, we observed notable viability declines starting at 125 nM, with an IC50 = 402 nM after 4 days of treatment. BI2536 alone lacked a differential proliferation effect (Fig. 2h). We repeated this experiment for the TP53Y220C mutation. Calu-1 TP53Y220C cells displayed 18.6-fold greater mCherry expression than p53 WT counterparts with similar mTagBFP2 expression levels, corresponding to a ~4-fold therapeutic window for Halo-PEG2-BI2536 (Extended Data Fig. 2g,h). This change in protein abundance is consistent with the reported p53-Y220C in cellulo t1/2 (that is, in the presence of cellular chaperones) of 10–13 h from cycloheximide chase experiments44 (compared to the WT p53 t1/2 of minutes in quiescent cells), implying that Y220C destabilizes an active conformation of p53 rather than reducing p53 protein abundance.

Prior reports suggested that the enhanced efficacy of similar bifunctional compounds can occur because of the accumulation of small molecules in cells expressing high concentrations of the protein target33. We, therefore, modeled reaction–diffusion kinetics of Halo-PEG2-BI2536 using ordinary differential equations (Methods) to gain more insights into the enhanced toxicity observed in the presence of a ligandable p53 fusion protein. Previously reported increases in p53 protein t1/2 upon mutation greatly influence the expected accumulation of the bifunctional small molecule28,29. In particular, an increase in t1/2 from the WT p53 t1/2 of 16 min to a mutant p53 t1/2 of 17 h is expected to increase final intracellular concentration of the small molecule at steady state by 2–3 orders of magnitude depending on the initial concentration (Fig. 2i). Together, these data suggest that, if a bifunctional molecule combining a cytotoxic molecule with a small-molecule ligand of p53 were developed, the high protein abundance of mutant p53 would allow for differential cell killing.

Bifunctional molecules selectively kill p53-Y220C+ cells

To assess the impact of bifunctional molecules containing p53-binding and cytotoxic moieties in an endogenous cellular context, we synthesized bifunctional molecules that bind to p53-Y220C. This mutant indirectly inhibits DNA binding activity through the loss of DNA-binding domain (DBD) thermal stability45. The removal of a bulky tyrosine residue, creating a pocket, and its replacement with a ligandable cysteine make it an attractive target for small-molecule refolders meant to restore transcriptional activity46,47,48,49,50,51, as well as bifunctional cytotoxic strategies. We initially used acrylamide KG5 as the p53-Y220C binder in a PLK1-directed bifunctional; KG5 is a carbazole-based covalent fragment liganding > 95% of p53-Y220C at 10 μM (ref. 49). We synthesized KG5-PEG4-BI2536 (3) and did not observe any difference in cell survival between 293T cells stably expressing Halo–p53-Y220C ΔTAD–mCherry (transactivation domain deleted to remove confounding protein stabilization effects) and parental 293T cells in an 8-day competition assay (Fig. 3 and Extended Data Fig. 3a).

Fig. 3: p53-Y220C–PLK1 bifunctional molecules selectively inhibit proliferation of TP53Y220C-mutant cells.

a, Structure of p53-Y220C–PLK1 bifunctional compound (PMV6-PEG4-BI2536, p53-01) used in experiments. b, Nanoluciferase signal 1 day after dose titration of p53-01 in 293T cells cotransfected with NLS–LgBiT–p53-Y220C (DBD) and mEGFP–PLK1–SmBiT (n = 6 technical replicates per condition). c, Live-cell imaging of 293T cells cotransfected with Halo–p53-Y220C ΔTAD–mCherry and mEGFP–PLK1–SmBiT, treated with p53-01. The white arrow indicates the extranuclear region containing PLK1; the red arrow indicates the nucleus. Dividing and nondividing cells are shown. mCherry (p53) and EGFP (PLK1) channels are shown separately. Two replicates of this experiment were performed; a representative replicate is shown. d, Competition of 293T cells stably expressing Halo–p53-Y220C ΔTAD–mCherry versus parental 293T cells in the presence of various compounds (PMV6, KG5-PEG4-BI2536, selinexor, BI2536, AP1867-C8-BI2536 and p53-01). Right, competition of 293T cells stably expressing Halo–p53-R273H (FL)–mCherry versus parental 293T cells in the presence of p53-01. The mCherry+ percentage on day 8 of competition normalized to that of DMSO-treated cells is shown. Two replicates of this experiment were performed; a representative replicate is shown. e, Crystal violet staining of 293T cells stably expressing Halo–p53-Y220C ΔTAD–mCherry versus parental 293T cells treated with p53-01, analyzed after 5 days. The experiment was performed in triplicate; a representative replicate is shown. f, Caspase 3/7 glo of Halo–p53-Y220C ΔTAD–mCherry versus parental 293T cells treated with p53-01 after 1 day (n = 4 technical replicates per condition). g, Proportion of early apoptotic cells (annexin+, PI−) following 1 day of treatment with p53-01, BI2536 or PMV6 at the indicated concentrations in Halo–p53-Y220C ΔTAD–mCherry versus parental 293T cells. NS, not significant. h, Competition of 293T cells stably expressing Halo–p53-Y220C ΔTAD–mCherry versus parental 293T cells in the presence of p53-01 (250 nM) alone or in combination with PMV6 (2.5 μM). i, DNA content (assessed by flow cytometry of DAPI-stained cells) following treatment of palbociclib-synchronized Halo–p53-Y220C ΔTAD–mCherry versus parental 293T cells with p53-01, BI2536 or PMV6 at the indicated concentrations for 1 day. The experiment was performed in triplicate; one representative replicate is shown. j, Western blot for PLK1 targets including cyclin B1 and phospho-PLK-binding motif of double-thymidine-block-synchronized Halo–p53-Y220C ΔTAD–mCherry versus parental 293T cells treated with p53-01, BI2536 or PMV6 at the indicated concentrations for 8.5 h. Two replicates of this experiment were performed; a representative replicate is shown.

Source data

We surveyed the patent literature and identified a p53-Y220C-binding pharmacophore consisting of an o-anisidine moiety linked to an indole and piperidine (US20230024905A1, US10138219B2, WO2021262483A1, WO2021061643A1, WO2022213975A1 and WO2023016434A1). These binders are exemplified by PMV6 (4), a compound identified in a structure–activity relationship series by PMV Pharma, with close similarity to the recently disclosed structure of the investigational compound, rezatapopt, which has a Kd of ~2.5 nM for p53-Y220C (PMV6, p. 96: US20230024905A1)52. PMV6 induces an 8 °C thermal shift of p53-Y220C (p. 21: US20230024905A1). The cocrystal structure with p53-Y220C showed that the compound binds to a shallow groove created by the Y220C substitution, with the piperidine facing the solvent. To construct a bifunctional molecule containing PMV6 and BI2536, we functionalized PMV6 at the solvent-exposed piperidine, which was functionalized in other analogs that continued to bind p53-Y220C using a thermal shift assay (compounds 7 and 10, p. 21 and 96: US20230024905A1).

We synthesized PMV6-PEG4-BI2536 (5, p53-01; Fig. 3a) and confirmed that p53-01 but not binders PMV6 or BI2536 alone induce ternary complex formation between p53-Y220C and PLK1 using a NanoBiT assay in 293T cells cotransfected with mEGFP–PLK1–SmBiT and NLS–LgBiT–p53-Y220C (DBD) (Fig. 3b and Extended Data Fig. 3b; EC50 = 1.4 μM). As an alternative method to demonstrate ternary complex formation, we assessed protein colocalization upon compound treatment53,54,55,56. Live-cell imaging of 293T cells cotransfected with mEGFP–PLK1–SmBiT and Halo–p53-Y220C ΔTAD–mCherry revealed diffuse PLK1 localization throughout the cell, with strong enrichment in an extranuclear region (possibly the centrosome as previously reported57), whereas p53-Y220C was constitutively nuclear. In nonmitotic cells, p53-01 treatment resulted in PLK1 nuclear enrichment and colocalization with p53-Y220C, whereas p53-Y220C also entered the extranuclear region colocalizing with PLK1. In untreated mitotic cells, p53-Y220C appeared localized to chromatin, whereas PLK1 was localized elsewhere in the cell. Upon p53-01 treatment, PLK1 became colocalized with p53-Y220C on chromatin; binders alone had no impact on localization (Fig. 3c and Extended Data Fig. 3c). We speculate that this mislocalization could disrupt the function of PLK1 as a mitotic kinase beyond simple steric blockade of its active site.

We performed an 8-day growth competition experiment between Halo–p53-Y220C ΔTAD–mCherry and parental 293T cells and observed strong selection against p53-Y220C+ cells (mCherry+) when p53-01 was dosed between 250 nM and 2 μM (Emax: 21.5-fold decline in the mCherry+ fraction; EC50 = 443 nM). We did not observe any activity from nonfunctionalized binders or control compounds, including PMV6, BI2536, AP1867-C8-BI2536 (6) or selinexor. We also did not observe any activity of p53-01 in either 293T or Calu-1 cells expressing Halo–p53-R273H (FL)–mCherry (Fig. 3d and Extended Data Fig. 2e). The differential antiproliferative activity of p53-01 in Halo–p53-Y220C ΔTAD–mCherry 293T cells was validated with crystal violet staining (Fig. 3e and Extended Data Fig. 3d). An analogous bifunctional molecule constructed with a WEE1 inhibitor (7, PMV6-PEG4-adavosertib) was unable to selectively inhibit proliferation of p53-Y220C+ cells (Extended Data Fig. 3e,f).

We next monitored for induction of apoptosis using caspase 3/7 glo and annexin–PI staining. p53-01 but not either binder alone differentially induced apoptosis in Halo–p53-Y220C ΔTAD–mCherry versus parental 293T cells (Fig. 3f,g and Extended Data Fig. 3g,h). To confirm the effects of our system were dependent on protein dosage, we used a doxycycline-inducible Halo–p53-Y220C ΔTAD–mCherry construct in 293T cells and were able to increase the potency of p53-01 approximately twofold without impacting the potency of either PMV6 or BI2536 (Extended Data Fig. 3i). To confirm that the effects were mediated by compound binding to p53-Y220C, we repeated the Halo–p53-Y220C ΔTAD–mCherry and parental 293T competition experiment using a tenfold excess of PMV6, observing partial rescue (p53-01 alone at 250 nM: 8.9-fold decline in mCherry+ percentage; p53-01 + PMV6 at 2.5 μM: 2.1-fold decline in mCherry+ percentage; P < 0.001; Fig. 3h).

PLK1 activity is critical for mitotic progression and its inhibitor, BI2536, was previously shown to induce G2/M arrest and genome doubling58. We monitored DNA content following treatment of synchronized Halo–p53-Y220C ΔTAD–mCherry and parental 293T cells with p53-01 or binders for 1 day. Low-dose p53-01 (0.75 μM) resulted in the accumulation of Halo–p53-Y220C ΔTAD–mCherry 293T cells with 4 N DNA content, leaving parental cells unaffected, whereas a higher dose (1.5 μM) induced genome doubling (accumulation of cells with 8 N DNA content) in ~60% of Halo–p53-Y220C ΔTAD–mCherry 293T but <10% of parental cells, phenocopying the high-dose BI2536 (10 nM) treatment (Fig. 3i). To confirm the effect was because of differential PLK1 inhibition, we performed western blots for PLK1 substrates including cyclin B1 (which accumulates upon PLK1 inhibition) and using an antibody to the phospho-PLK1-binding motif (PLK1-binding partners and substrates). p53-01 once again phenocopied BI2536 specifically in Halo–p53-Y220C ΔTAD–mCherry 293T but not in parental cells, with PMV6 having no effect (Fig. 3j). Together, these data indicate that a bifunctional molecule that both binds to p53-Y220C and binds to and inhibits PLK1 can selectively inhibit proliferation of cells expressing TP53Y220C because of enhanced PLK1 inhibition, leading to mitotic arrest.

Activity is p53-Y220C dependent without reactivation

We next analyzed whether the mechanism of action for p53-01 is distinct from that for PMV6, a reactivator of p53-Y220C. We performed RNA sequencing (RNA-seq) of Huh7 hepatocellular carcinoma cells (homozygous TP53Y220C/Y220C) treated with p53-01 (4 μM) and PMV6 (4 μM) for 1 day. We observed strong induction of a p53 signature by PMV6, including >10-fold upregulation of p53 targets such as MDM2, CDKN1A, GDF15 and PUMA (BBC3) and downregulation of genes essential for proliferating cells such as TOP2A. Manual inspection of a set of high-confidence p53 transcriptional targets and gene set enrichment analysis (GSEA) of a larger group of consensus p53 targets confirmed strong upregulation of these genes by PMV6 (normalized enrichment score (NES) = 2.68) but not p53-01 (NES = 1.33) (Fig. 4a–c and Extended Data Fig. 4a). Thus, p53-01 does not appear to act through a transcriptional mechanism of p53 reactivation.

Fig. 4: p53-Y220C–PLK1 bifunctionals are active in endogenous settings and do not reactivate p53.

a, Huh7 cells homozygous for the TP53Y220C mutation were treated with PMV6 (4 μM) or p53-01 (4 μM) for 24 h and subject to RNA-seq. The log2(fold changes) in transcript abundance and log10(adjusted P values) computed by DESeq2 are plotted. The top differentially expressed genes are labeled (n = 3 replicates per condition). b, Preranked GSEA using DESeq2 t statistic as rank metric for PMV6 versus DMSO (left) and p53-01 versus DMSO (right) comparisons on MSigDB Fischer gene set of direct p53 targets. NESs are listed. c, Gene expression analysis of p53 targets from high-confidence p53 target set. Samples are arranged in columns by hierarchical clustering (Euclidean distance) and rows scaled to z scores (colors). d, Luciferase activity following PMV6, BI2536 or p53-01 treatment (16 h) of Huh7 (left) or MFE319 (right) cell lines stably expressing a p53 luciferase transcriptional reporter. Luciferase signal is normalized to untreated cells (n = 6 technical replicates per condition for p53-01 and PMV6; n = 3 technical replicates per condition for BI2536). e, Cell viability of MFE319 cells treated with PMV6 or p53-01, measured by CTG after 4 days (n = 4 technical replicates per condition). f, Donut plot of cancers in TCGA Pan-Cancer Atlas studies in cBioPortal with TP53Y220C mutations with or without additional TP53 mutations. g, Cell viability of MFE319 cells treated with p53-01 alone or in combination with PMV6 (2.5 μM), measured by CTG after 4 days (n = 4 technical replicates per condition). h, Competition of MFE319 cells stably expressing sgTP53–Cas9–EGFP versus parental MFE319 cells in the presence of p53-01, PMV6 or BI2536. The EGFP+ percentage on day 9 of competition normalized to that of DMSO-treated cells is shown. i, Halo–p53-Y220C ΔTAD–mCherry 293T cells were cultured in the presence of p53-01 for 3 weeks. The percentage of mCherry− cells was analyzed by FACS. Two replicates of this experiment were performed; a representative replicate is shown.

Source data

We subsequently established p53-Y220C+ cell lines stably expressing a p53 luciferase reporter59. In Huh7 cells (TP53Y220C/Y220C), PMV6 was able to induce activity of the reporter at doses of 1.25–10 μM (EC50 = 2.4 μM), whereas p53-01 and BI2536 had no such effect. We repeated the p53 reporter experiments in the p53-Y220C+ endometrial cancer cell line MFE319, where no induction of luciferase activity for any compound (p53-01, PMV6 or BI2536) was observed (Fig. 4d). Note that MFE319 has both TP53Y220C and TP53R273C mutations reported in the CCLE dataset35. Analysis of paired RNA-seq reads in this cell line revealed that the mutations were in trans (Extended Data Fig. 4b). It is possible that the additional dominant negative TP53 mutation in this cell line prevents p53 reactivation by p53-Y220C refolders (for example, PMV6), plausibly because of poisoning of the tetramer even in the presence of correctly functioning p53-Y220C proteins. The strategy presented in this paper, using p53-Y220C to concentrate a toxin in cells, is unlikely to be impacted by the additional TP53R273C mutation. This would result in a ~20% greater scope of treatable cancers for PLK1-p53-Y220C bifunctional molecules compared to p53-Y220C reactivators (Fig. 4e,f; as determined by analysis of TP53Y220C-mutant cancers in TCGA60 with additional TP53 mutations).

Lastly, we wanted to assess whether the effects of p53-01 in MFE319 were dependent on its interaction with p53-Y220C. Indeed, we were able to partially compete the decline in cell viability induced by p53-01 in MFE319 using PMV6 (Fig. 4g). Additionally, we established MFE319 TP53-knockout cells by stably expressing sgTP53–Cas9–EGFP. In competition with parental MFE319 cells, p53-01 was able to significantly enrich an EGFP+ population at doses between 800 nM and 1.25 μM. BI2536 and PMV6 had no such activity (Fig. 4h). As further evidence of on-target activity, we cultured Halo–p53-Y220C ΔTAD–mCherry 293T cells for 3 weeks in the presence of low doses of p53-01 (≤250 nM) and observed the expansion of an mCherry− population (Fig. 4i; 6.9-fold expansion of mCherry− fraction at 250 nM). Furthermore, p53-01 trended toward superior activity across a panel of p53-Y220C+ cancer cell lines compared to p53-Y220C− counterparts (Extended Data Fig. 4c,d). Altogether, these data indicate that p53-01 has activity in cell lines expressing TP53Y220C, distinct from PMV6 (not through p53 reactivation), and its activity is dependent on its interaction with p53-Y220C.

Linker optimization improves efficacy of bifunctionals

We created a library of nine PMV6-BI2536 bifunctional compounds with varying linker lengths (4.33–22.22 Å, compounds 5 and 8–15) and assessed their efficacy in competition assays between parental and Halo–p53-Y220C ΔTAD–mCherry 293 T cells (Table 1). We observed no consistent effect of linker length on compound efficacy in general, although shorter alkyl linkers outperformed longer alkyl linkers. PEG(2,4,6) linkers all performed similarly. The most potent compound, PMV6-C3-BI2536 (10, β-alanine linker), was over 1.5 orders of magnitude more potent than p53-01 in the competition assay (PMV6-C3-BI2536 EC50: 25 nM, p53-01 EC50: 850 nM; Fig. 5a,b). Activity of compounds strongly correlated with their ability to form ternary complexes in the mEGFP–PLK1–SmBiT and NLS–LgBiT–p53-Y220C (DBD) NanoBiT system (Fig. 5a,c). These data indicate that substantial improvements to the efficacy of p53-Y220C–PLK1 bifunctional compounds are possible going forward.

Fig. 5: Linker optimization yields more potent p53-Y220C–PLK1 bifunctionals.

a, Left, mCherry percentage versus dose for individual compounds from the growth competition assay between Halo–p53-Y220C ΔTAD–mCherry versus parental 293T cells on day 10. Right, normalized nanoluciferase signal from NanoBiT assay versus dose for individual compounds. NanoBiT experiments wer