Targeted MDM2 Degradation Reveals a New Vulnerability for p53-Inactivated Triple-Negative Breast Cancer

Cancers that inactivate the p53 tumor suppressor, through mutation or deletion, are more aggressive and have increased resistance to many therapies (1, 2). Triple-negative breast cancer (TNBC) is an aggressive malignancy with high rates of p53 inactivation (∼65%–88%; refs. 3–5). TP53 mutations commonly arise in TNBC, with missense mutations crippling its transcription factor function and nonsense mutations causing loss of p53 protein (6). Compared with other breast cancer types, TNBC patients have lower survival rates due to increased metastasis and relapse (7) and occur more frequently in African Americans (8). Because of their high p53 inactivation rates, compounds that inhibit p53 from binding to its negative regulator, MDM2, are ineffective in TNBC (9). Current standard-of-care treatment for TNBC is complex and dependent on several factors, but all approaches involve DNA-damaging chemotherapy that can result in cardiac and other toxicities (10). Unfortunately, efforts to treat this aggressive malignancy have made minimal progress in recent years, indicating innovative therapeutic strategies are needed.

Mouse developmental genetic studies showed deletion of MDM2, a negative regulator of p53, was embryonic lethal unless p53 was concomitantly deleted (11, 12). This established the view that if p53 was absent, cells were not affected by MDM2 loss. MDM2, an oncogene, is frequently overexpressed in cancer, including those with inactivated p53, as compared with normal cells, which have low levels of MDM2 (13, 14). This suggested MDM2 may be important in cancer cells lacking functional p53. Recently, we reported that MDM2 deletion induced apoptosis of p53-null murine sarcoma and T-cell lymphoma cells (15). However, whether a dependency on MDM2 for cancer cell survival in mice translates to p53-inactivated human cancers, and particularly those with mutant p53, is unknown.

PROteolysis TArgeting Chimeras (PROTAC) are heterobifunctional molecules composed of a targeting ligand and an E3 ubiquitin ligase recruiting ligand to induce selective degradation of specific cellular proteins (16). PROTACs provide pharmacologic approaches for protein silencing, potentially mimicking genetic knockdown (17). Moreover, with their catalytic activity, PROTACs should be able to be used at lower concentrations than small-molecule inhibitors, making them less toxic. We sought to pharmacologically target MDM2 through the design and synthesis of a new MDM2-targeted PROTAC, YX-02-030, in order to evaluate the effects of MDM2 degradation on p53-inactivated TNBC. Our PROTAC potently binds MDM2 and recruits the von Hippel−Lindau (VHL) E3 ubiquitin ligase to initiate MDM2 degradation, and effectively killed p53 mutant or deleted TNBC cells in two-dimensional (2D) and three-dimensional (3D) culture models, patient explants, and in tumor xenografts through activation of the p53 family member TAp73. We show MDM2 is required for TNBC cell survival when p53 is inactivated, and PROTAC-targeted MDM2 degradation is a new potential therapeutic approach for this deadly cancer, demonstrating superiority to existing MDM2 inhibitors, which have been ineffective.

Although MDM2 levels are reported as elevated in multiple human cancers that have inactivated p53 (14), it was unclear whether MDM2 was an essential, required protein for any human malignancy that had inactivated p53. Therefore, we designed and synthesized a new MDM2-targeting PROTAC, YX-02-030, to explore the requirements of MDM2 in human cancers, particularly those with high rates of TP53 mutation such as TNBC (3–5). The MDM2-PROTAC, YX-02-030, replaces the 3-(methylsulfonyl)-propyl tail on the piperazine motif of the clinically available MDM2 inhibitor RG7112 (18) with acetic acid to generate RG7112D (Fig. 1A; chemical synthesis Supplementary Fig. S1A). An amide bond couples RG7112D to VHL-Amine, which is the VHL E3 ligase recruiting ligand VH032 (ligand-7; ref. 19) that is coupled to a hydrophilic tripolyethylene glycol linker (ref. 20; Fig. 1A). The resulting bifunctional molecule YX-02-030 potently inhibited MDM2-p53 binding [homogeneous time-resolved fluorescence (HTRF) IC₅₀ = 63 ± 3 nmol/L; Fig. 1B], bound with high affinity to MDM2 [surface plasmon resonance (SPR) K_D = 35 nmol/L; Fig. 1C; Supplementary Fig. S1B], inhibited VHL-HIF1α binding (HTRF IC₅₀ = 1,350 ± 181 nmol/L; Fig. 1B), and efficiently formed a ternary complex between GST-MDM2 and His-VHL complex (proximity-based AlphaScreen; ref. 21; Fig. 1D). Thus, our MDM2-targeted PROTAC has the binding and ternary complex formation abilities necessary to be an effective degrader.

PROTACs work by forming a ternary complex consisting of the target protein, the PROTAC, and the recruited E3 ubiquitin ligase-E2 complex, causing ubiquitination of the target protein and marking it for 26S proteasome-mediated degradation (16). Our MDM2-directed PROTAC (YX-02-030) showed a concentration (Fig. 1E) and time-dependent (Fig. 1F) loss of MDM2 protein in TNBC cells with either mutated p53 (MDA-MB-231) or deleted p53 (MDA-MB-436). VHL protein levels remained unchanged, indicating MDM2, also an E3 ubiquitin ligase, was not ubiquitinating VHL (Fig. 1E and F). Because MDM2 can be phosphorylated, which can block antibody binding sites (22), we tested multiple MDM2 antibodies, and all showed loss of MDM2 protein with PROTAC exposure (Supplementary Fig. S1C). Moreover, the loss of MDM2 protein was not due to decreased MDM2 mRNA levels, as they remained unchanged with PROTAC treatment (Supplementary Fig. S1D). Testing the contribution of the ubiquitin cascade with MLN4924, a NEDD8-activating enzyme inhibitor critical for ubiquitin transfer (23), showed it antagonized PROTAC-mediated MDM2 protein degradation in TNBC cells (Fig. 1G). Similarly, when we inhibited the proteasome with MG132, MDM2 protein levels were maintained despite MDM2-PROTAC exposure (Fig. 1H). Therefore, our MDM2-directed PROTAC targets MDM2 protein for 26S proteasome-mediated degradation in TNBC cells, and this requires the ubiquitin cascade.

To determine whether the ternary complex was required for MDM2 degradation, we competed for binding to MDM2 and VHL by adding excess RG7112 or VHL-Amine, respectively. PROTAC-induced MDM2 degradation was prevented with either RG7112 or VHL-Amine (Fig. 1I and J, respectively). Analogous results were obtained using VH298, a VHL-specific small-molecule inhibitor (ref. 24; Fig. 1K). Thus, ternary complex formation of MDM2, the PROTAC, and VHL is required for PROTAC-induced MDM2 degradation.

To characterize the biological effects of our MDM2-PROTAC, we first tested it head-to-head with MDM2 inhibitors that prevent MDM2-p53 binding in p53 wild-type breast cancer cells with reported sensitivity to the MDM2 inhibitor Nutlin (25–27). Treatment of p53 wild-type MCF7 and DU4475 cells with MDM2 inhibitors (Nutlin, RG7112, RG7112D) or our MDM2-PROTAC resulted in the expected stabilization of MDM2 protein with the inhibitors and loss of MDM2 with the PROTAC (Fig. 2A). All compounds increased p53 protein levels (Fig. 2A) and significantly reduced cell survival in a concentration-dependent manner (Fig. 2B; Supplementary Fig. S2A). Of note, the IC₅₀ of our MDM2-PROTAC in MCF7 and DU4475 cells was lower than Nutlin (2.8 μmol/L and 2.3 μmol/L versus 11 μmol/L and 4 μmol/L, respectively) and lower than or equal to RG7112D and RG7112. Similarly, increased caspase-3 activity was detected in MDM2-PROTAC—treated cells compared with those that received the MDM2 inhibitors at the same concentrations (Fig. 2C; Supplementary Fig. S2B). Over time, there was more cleaved PARP (Fig. 2D) and sub-G₁ apoptotic DNA (Fig. 2E) with MDM2-PROTAC treatment compared with the MDM2 inhibitors. Furthermore, similar to Nutlin and its derivatives, MDM2-PROTAC treatment of MCF7 cells resulted in increased mRNA and protein levels of proapoptotic p53 target genes BAX, NOXA, and PUMA (Fig. 2F and G, respectively).

We then evaluated whether the MDM2-PROTAC also affects cell survival in 3D culture. MCF7 cells were treated with the MDM2-PROTAC or MDM2 inhibitors the same day they were placed in 3D culture. A significant decrease in the formation of mammospheres was observed in the PROTAC-treated cells compared with those exposed to MDM2 inhibitors or vehicle control (Fig. 2H; Supplementary Fig. S2C). For already established MCF7 mammospheres, our MDM2-PROTAC significantly decreased their survival (Fig. 2I) due to increased apoptosis (caspase-3/7 activity) at half the concentration as the MDM2 inhibitors (Fig. 2J). Moreover, there was disaggregation of the mammospheres with MDM2-PROTAC treatment, but not with MDM2 inhibitor treatment (Fig. 2 images). Therefore, our MDM2-PROTAC activates p53 in p53 wild-type breast cancer cells, causing apoptosis in both 2D and 3D cultures, and was equally or more potent than Nutlin and its derivatives.

We next evaluated whether our MDM2-PROTAC could kill TNBC cells that had inactivated p53 by mutation or deletion. We tested the effects of our MDM2-PROTAC on TNBC lines (MDA-MB-231, HCC-1143, and HCC-1395) with three different gain-of-function p53 point mutations (R280K, R248Q, and R175H, respectively; refs. 28–30). All three mutant p53 lines showed sensitivity to the MDM2-PROTAC with a narrow IC₅₀ range (4.0–5.3 μmol/L), and no sensitivity to MDM2 inhibitors (Fig. 3A). Two p53-deleted TNBC cell lines (MDA-MB-436 and MDA-MB-453) showed a dose-dependent decrease in survival with our MDM2-PROTAC with an IC₅₀ (4.5–5.5 μmol/L) similar to the p53-mutant lines, whereas RG7112 and RG7112D had no effect (Fig. 3A). These data reveal that in contrast to MDM2 inhibitors, p53 deletion or mutation does not affect sensitivity to our MDM2-PROTAC.

Although there are no reports that the VHL recruiting ligand alone or as part of a PROTAC can function as a molecular glue, we tested the requirements of MDM2 and non-MDM2 lethal effects of our PROTAC. Because MDM2 is a highly conserved protein, we took advantage of p53-null mouse sarcoma cells from mice born with or without Mdm2 deletion. Mdm2^+/+p53^−/− sarcoma cells were as sensitive to our MDM2-PROTAC as the human TNBC cells (5.8 μmol/L IC₅₀; Fig. 3B), whereas Mdm2^−/−p53^−/− sarcoma cells were completely resistant to MDM2-PROTAC treatment, yet still sensitive to other compounds (e.g., ABT-263; Fig. 3B). Therefore, MDM2 must be present for our MDM2-PROTAC to be cytotoxic, indicating its lethal effect is not due to molecular glue.

To further test MDM2-targeted degradation as the cause for p53-inactivated TNBC cell death, we performed compound competition experiments as in Fig. 1I–K. Increasing concentrations of RG7112 or VHL-Amine to compete with the MDM2-PROTAC for binding MDM2 or VHL, respectively, blocked MDM2-PROTAC–induced TNBC cell death (Fig. 3C; Supplementary Fig. S3A). Analogous results were obtained with the VHL inhibitor VH298 (Fig. 3C; Supplementary Fig. S3A). Therefore, our MDM2-PROTAC kills p53-inactivated TNBC cells by targeting MDM2 and requires ternary complex formation.

To characterize the type of cell death caused by our MDM2-PROTAC, we performed kinetic experiments using the IC₅₀ concentrations determined in Fig. 3A. The MDM2-PROTAC inhibited p53-mutant or deleted TNBC cell expansion over time (Fig. 3D; Supplementary Fig. S3B). Within 24 hours of MDM2-PROTAC treatment, there was significantly increased Annexin-V positivity (Fig. 3E; Supplementary Fig. S3C) and caspase-3 activity (Fig. 3F; Supplementary Fig. S3D), both markers of apoptosis, and increased sub-G₁ apoptotic DNA (Fig. 3G). PARP cleavage was also evident but was prevented when MDM2 degradation was inhibited with MG132, VHL binding to the PROTAC was blocked, or when VHL activity was inhibited (Fig. 3H; Supplementary Fig. S3E). These data show our MDM2-PROTAC induces p53-mutant or deleted TNBC cell apoptosis, and this requires ternary complex formation and MDM2 degradation.

To verify the apoptotic effects of our MDM2-PROTAC and its specificity for MDM2, we knocked down MDM2 with shRNA in both p53-mutant and deleted TNBC cell lines. Effective MDM2 knockdown (Fig. 3I) significantly reduced TNBC cell expansion (Fig. 3J) due to apoptosis; there was increased Annexin-V positivity (Fig. 3K), cleaved PARP (Fig. 3I), and sub-G₁ apoptotic DNA (Fig. 3K). Because MDM2 knockdown mirrors the cytotoxic effects of our MDM2-PROTAC on p53-inactivated TNBC cells, this provides evidence supporting the MDM2 specificity of our PROTAC. Moreover, we conducted whole-transcriptome analysis, and it showed there was a significant positive correlation between the genes with differential expression in MDA-MB-231 and MDA-MB-436 cells treated with the MDM2-PROTAC versus MDM2 knockdown (Spearman ρ = 0.756, P = 0 for MDA-MB-231 and ρ = 0.836, P = 0 for MDA-MB-436; Fig. 3L). These data provide independent evaluations of the specificity of our PROTAC for MDM2 and indicate it is targeting MDM2.

We next determined whether PROTAC-mediated degradation of MDM2 would alter the clonogenic potential of p53-inactivated TNBC cells by performing colony formation assays with MDA-MB-231 and MDA-MB-436 cells. Following MDM2-PROTAC treatment, there was a significant reduction in colony numbers with 1 μmol/L of MDM2-PROTAC (Fig. 4A). There was little/no effect with 2 μmol/L of MDM2 inhibitors, but ≤2 colonies formed with 2 μmol/L of MDM2-PROTAC (Fig. 4A).

We then evaluated in 3D culture whether our MDM2-PROTAC was effective at preventing TNBC cell mammosphere formation and eliminating already formed mammospheres. For both MDA-MB-231 and MDA-MB-436 cells, significantly fewer mammospheres formed at each MDM2-PROTAC concentration tested, compared with the MDM2 inhibitors and vehicle control (Fig. 4B). In already established mammospheres, there was a significant decrease in mammosphere area following MDM2-PROTAC treatment (Fig. 4C). Notably, a 3-fold lower concentration of MDM2-PROTAC than the MDM2 inhibitors reduced mammosphere area by 49.5% and 43.8% in MDA-MB-231 and MDA-MB-436 mammospheres, respectively, and equal concentrations reduced mammosphere area even more (Fig. 4C). This reduction was due to apoptotic cell death, as we detected significantly increasing caspase-3/7 activity over time (Fig. 4D) and decreased survival (Fig. 4E) in the mammospheres following MDM2-PROTAC treatment compared with the MDM2 inhibitors. Together, our data provide strong evidence that our MDM2-PROTAC induces apoptosis in p53-inactivated TNBC cells in both 2D and 3D culture models.

Prior to testing our MDM2-PROTAC on tumors in vivo, we performed mouse liver microsome and pharmacokinetic (PK) studies to determine its stability in vivo. Our MDM2-PROTAC was moderately metabolically stable with a 25.9-minute half-life (Fig. 5A) and had excellent in vivo stability with plasma levels stable over 6 hours in mice after a single 10 mg/kg intraperitoneal dose (Fig. 5B). Due to its stability in vivo, we then evaluated the effectiveness of our MDM2-PROTAC at killing TNBC tumors in mice. Xenografts of MDA-MB-231 and MDA-MB-436 TNBC cells grew to ∼80 mm³, and then tumor size–matched mice were treated with our MDM2-PROTAC, RG7112D control compound, or vehicle control. Fourteen days of MDM2-PROTAC treatment significantly extended mouse survival (Fig. 5C) and decreased tumor volume (Fig. 5D; Supplementary Fig. S4A), compared with control mice for both xenograft models. To confirm our MDM2-PROTAC was hitting its target (MDM2) in the TNBC tumors, we harvested tumors from a cohort of mice after 72 hours of MDM2-PROTAC treatment. Mice that received MDM2-PROTAC showed loss of MDM2 protein (Fig. 5E) and increased cleaved PARP (Fig. 5E), Annexin-V positivity (Fig. 5F), caspase-3 activity (Fig. 5G), sub-G₁ apoptotic DNA (Fig. 5H), and nonviable cells (Fig. 5I). Notably, no signs of overt toxicity in immune-competent C57Bl/6 mice (Supplementary Fig. S4B and S4C) or immune-deficient mice from the xenograft experiments (Supplementary Fig. S4D–S4F) were observed following MDM2-PROTAC treatment. Specifically, mouse weight was maintained and complete blood counts and the histology and cellular content of the spleen, bone marrow, and intestine were normal (Supplementary Fig. S4). Therefore, our MDM2-PROTAC showed clear in vivo efficacy against p53-inactivated TNBC tumors and no obvious toxicity to normal tissues.

Because clinical trials of RG7112 showed hematopoietic cell toxicities (31, 32), to further investigate the lack of MDM2-PROTAC–induced cytotoxicity of normal cells, we tested whether p53 was stabilized and its target genes induced in normal hematopoietic cells upon MDM2-PROTAC treatment. First, following 3 days of treatment (same regimen as Fig. 5E–I), splenocytes and bone marrow cells from C57Bl/6 mice were evaluated. In the mice that were irradiated (positive control), p53 was stabilized, and there were increased protein and mRNA levels of p53 targets (BAX, NOXA, and PUMA), but not in the MDM2-PROTAC–treated mice (Supplementary Fig. S4G and S4H). Second, we tested normal human CD34⁺ hematopoietic cells. CD34⁺ cell viability remained largely unaffected by MDM2-PROTAC treatment, but viability significantly decreased after RG7112 treatment and to a lesser extent with RG7112D (Supplementary Fig. S4I; MDM2-PROTAC–treated MDA-MB-231 TNBC cells were a positive control). Also, in CD34⁺ cells, mRNA levels of the p53 target genes, BAX, NOXA, and PUMA were unchanged following MDM2-PROTAC treatment, but were significantly increased with RG7112 and less strongly with RG7112D (Supplementary Fig. S4J). Thus, our MDM2-PROTAC shows specificity for cancer cells, as it did not activate p53-mediated apoptotic gene upregulation in normal mouse or human hematopoietic cells. However, RG7112, which is known to cause bone marrow cell toxicity in humans (31, 32), did negatively affect human CD34⁺ cells by activating p53.

To assess the effects of our MDM2-PROTAC on TNBC patient samples, fresh, surgically resected tumors from five TNBC patients were obtained, and 80% were from Black/African-American patients (Supplementary Table S1). p53 is mutated in the vast majority of TNBC (3–5). After sequencing TP53 in the five TNBC samples we were provided, all had TP53 missense mutations with two having the same hot-spot (R248Q) p53 mutation and one that had 3 TP53 mutations (Fig. 6A). While preserving tumor architecture, pieces of each patients’ tumor were placed into explant cultures (33, 34) and subjected to treatment with our MDM2-PROTAC and vehicle control, and when enough tissue was provided, also RG7112D. Within 4 days of treatment, MDM2 protein was lost in the patient tumors that received the MDM2-PROTAC compared with vehicle and RG7112D controls, and as expected, no change in VHL protein (Fig. 6B). MDM2-PROTAC–treated patient tumors were undergoing apoptosis, as those explants showed increased levels of cleaved caspase-3 and cleaved PARP by Western blot (Fig. 6B). IHC also showed a significantly increased number of cleaved caspase-3–positive tumor cells in MDM2-PROTAC–treated explants, but normal breast epithelial and stromal cells were unaffected (Fig. 6C). Thus, the data from the patient-derived TNBC explants show that our MDM2-PROTAC effectively and specifically kills TNBC patient tumor cells.

For three of the patient TNBC tumor samples, we had enough tissue to also establish 3D mammosphere cultures, and for one of these patients, we also received normal breast epithelial tissue that generated loose cell clusters. Our MDM2-PROTAC significantly increased caspase-3/7 activity (Fig. 6D) and reduced survival (Fig. 6E) of the mammospheres from all three patient tumors, whereas MDM2 inhibitors had no effect. The normal breast epithelial clusters remained largely unaffected by MDM2-PROTAC treatment (Fig. 6D and E). These data provide significant evidence that targeting MDM2 for degradation may be a viable, nontoxic therapeutic strategy for TNBC.

To gain insight into the mechanism by which our MDM2-PROTAC was killing p53-mutant and deleted TNBC cells, we evaluated RNA sequencing (RNA-seq) data on cells treated with MDM2-PROTAC and cells expressing MDM2 shRNA and each of their controls. As expected, evaluation of gene signatures from the Hallmark database (35) that were significantly enriched showed genes linked to apoptosis in both cell lines whether they were treated with the MDM2-PROTAC or MDM2 was knocked down (>2 fold change, FDR < 0.05; Fig. 7A; Supplementary Fig. S5A). Additionally, our analysis also revealed that p53 pathway gene signatures were also significantly enriched (>2 fold change, FDR < 0.05; Fig. 7A; Supplementary Fig. S5A). More than 25 genes targeted by the p53 family, as reported in the IARC TP53 Database (36), showed significantly elevated expression in both p53-mutant and deleted TNBC cells with MDM2 degradation or knockdown (FDR<0.05; Fig. 7B; Supplementary Fig. S5B). Notably, there was a highly significant correlation between p53 family targeted genes with differential expression in MDA-MB-231 and MDA-MB-436 cells treated with the MDM2-PROTAC versus MDM2 knockdown (Spearman ρ = 0.873, P = 0 for MDA-MB-231 and ρ = 0.931, P = 0 for MDA-MB-436; Fig. 7C; Supplementary Fig. S5C), further illustrating the specificity of our PROTAC for MDM2.

To independently validate our RNA-seq results, we performed qRT-PCR on a subset of the genes regulated by the p53 family that mediate apoptosis. Following MDM2 degradation or MDM2 knockdown in MDA-MB-231 and MDA-MB-436 cells, levels of apoptotic genes BAX, NOXA, PUMA, AEN, APAF1, TP53I3, and PIDD1 were significantly increased in both TNBC cell lines compared with controls (Fig. 7D; Supplementary Fig. S5D).

MDM2 can bind and regulate p53 family members p73 (37–40) and p63 (41, 42), but the conditions in which this occurs, particularly in p53-inactivated cells, remain unresolved. Because the transcriptionally active forms of p73 (TAp73) and p63 (TAp63) are capable of transactivating most of the same genes as p53 (6), we evaluated their expression. Upon MDM2-PROTAC treatment or MDM2 knockdown, levels of TAp73 protein increased in all five TNBC cell lines assessed (Fig. 7E and F; Supplementary Fig. S6A–S6C); however, TAp63 was lowly expressed and its levels remained unchanged (Fig. 7E and F; Supplementary Fig. S6A and S6B). Additionally, levels of TAp73 protein, but not TAp63, were increased in the xenograft tumors harvested 72 hours after PROTAC treatment and in the TNBC patient-derived explants following MDM2-PROTAC treatment (Fig. 7G; Supplementary Fig. S6C). In contrast, MDM2 inhibitor treatment did not result in increased TAp73 levels. Note that to effectively detect TAp73 and TAp63, proteins were isolated under conditions designed to extract transcription factors, which tend to be positively charged and/or chromatin-bound (Supplementary Fig. S6D). Levels of ∆Np73, which lacks the N-terminal transactivation domain and can inhibit TAp73 (43), in MDA-MB-231 and MDA-MB-436 cells were low and unaffected by MDM2 degradation or knockdown (Fig. 7E and F; Supplementary Fig. S6A and S6B). Increased TAp73 protein was not due to increased transcription, as TAp73 mRNA levels were unaltered upon MDM2 loss (Supplementary Fig. S6E). Both mRNA (Fig. 7D; Supplementary Fig. S5D) and protein (Fig. 7E and F; Supplementary Fig. S6A and S6B) levels of TAp73 apoptotic targets were upregulated with MDM2 loss in the TNBC cell lines, suggesting TAp73 was activated. Additionally, we evaluated TAp73 after PROTAC binding to MDM2 or VHL was blocked with excess RG7112 or VHL-Amine, respectively, or when VHL was inhibited with VH298. TAp73 protein was only stabilized, its apoptotic target genes upregulated, and apoptosis induced (cPARP) when the PROTAC was able to successfully target MDM2 for degradation, but not when the ternary complex was disrupted (Fig. 7H; Supplementary Fig. S6F). Moreover, chromatin immunoprecipitation of TAp73 after MDM2-PROTAC treatment showed significant enrichment of TAp73 at the promoters of its apoptotic target genes (BAX, NOXA, PUMA, AEN, APAF1, TP53I3, and PIDD1), but not at the promoter of AchR, which is not targeted by p53/TAp73 (ref. 44; Fig. 7I; Supplementary Fig. S6G). These data indicate that MDM2-PROTAC treatment causes TAp73 to be stabilized and transcriptionally activated.

To better understand how MDM2 loss results in TAp73 stabilization and transcriptional activation, we began by assessing whether our MDM2-PROTAC binding in the p53/p73 binding pocket of MDM2 disrupts interactions with TAp73 in p53-inactivated TNBC cells. We performed immunoprecipitations of both MDM2 and TAp73 in the presence of the MDM2-PROTAC or vehicle control in p53-mutant and deleted TNBC cells. The proteasome inhibitor MG132 was included to prevent the proteasomal degradation of MDM2. In the MDM2-PROTAC–treated samples, there was markedly less TAp73 associated with MDM2, in both MDM2 and TAp73 immunoprecipitations (Fig. 7J; Supplementary Fig. S6H), indicating the PROTAC was interfering with MDM2:TAp73 binding. Because TAp73 was binding to MDM2 in the TNBC cells, disruption of this interaction would be expected because a derivative of RG7112 is the MDM2-targeting molecule in the MDM2-PROTAC and binds tightly to the p53/p73 binding pocket of MDM2 (Fig. 1B). However, for TAp73 to be an active transcription factor, our data indicate that it needs more than losing its association with MDM2. Because TAp73 requires specific posttranslational modifications to be transcriptionally active, we evaluated the phosphorylation of tyrosine-99 in TAp73, the site phosphorylated by c-Abl and a requisite to be activated, leading to TAp73 stabilization (45). Following treatment of MDA-MB-231 and MDA-MB-436 cells with our MDM2-PROTAC, TAp73 Tyr-99 was phosphorylated, but not after treatment with RG7112D or RG7112 (Fig. 7K; Supplementary Fig. S6I). In addition, c-Abl was phosphorylated at Tyr-412, a mark of its activation, in the cells exposed to MDM2-PROTAC, but not with RG7112D or RG7112 (Fig. 7K; Supplementary Fig. S6I). TAp73 Tyr-99 phosphorylation is a signal that should call in the CBP/p300 lysine acetyltransferase complex, which is necessary for TAp73 to become transcriptionally activated (45). Immunoprecipitation of TAp73 showed its association with CBP and p300 and acetylation of TAp73 after MDM2-PROTAC treatment in MDA-MB-231 and MDA-MB-436 cells, but not following treatment with the MDM2 inhibitors (Fig. 7K; Supplementary Fig. S6I). Together, these data show that in p53-inactivated TNBC cells, MDM2 degradation results in the transcriptional activation of TAp73 and that simply releasing TAp73 from MDM2 is insufficient to induce TAp73-mediated transcription.

To assess the requirements of p73 in mediating the effects of the MDM2-PROTAC, we knocked down TAp73 using two independent shRNA in p53-mutant and deleted TNBC cells. These two shRNAs were TAp73-specific, as they did not affect levels of mutant p53, TAp63, or ∆Np73 (Fig. 7L; Supplementary Fig. S6J). Although MDM2 protein was lost following MDM2-PROTAC treatment, knockdown of TAp73 in both TNBC lines prevented upregulation of its apoptotic target genes (Fig. 7M; Supplementary Fig. S6K), which was also reflected at the protein level (Fig. 7L; Supplementary Fig. S6J). TAp73 knockdown also largely rescued the decrease in cell growth (Fig. 7N; Supplementary Fig. S6L) and the cleavage of PARP (Fig. 7L; Supplementary Fig. S6J), resulting from PROTAC-mediated MDM2 degradation. Collectively, our results indicate that our MDM2-PROTAC effectively targets MDM2 for degradation, activating TAp73-dependent apoptosis in p53-inactivated TNBC, and provides a future avenue for the treatment of this deadly cancer.

Based largely on developmental data with p53 and Mdm2 knockout mice, it was believed that once p53 was inactivated, cells no longer needed Mdm2 (11, 12). However, our data here challenge this dogma, revealing that MDM2 is required in p53-mutant and deleted human cancer cells, specifically TNBC cells, for their continued growth and survival. We have capitalized on this identification of a targetable vulnerability in p53-inactivated TNBC by designing and synthesizing a new MDM2-PROTAC. The MDM2-PROTAC, YX-02-030, specifically targets MDM2, is stable, and eliminates TNBC cells in 2D and 3D cultures, patient TNBC explants, and xenografts in mice, while sparing normal cells, by activating TAp73, which induces apoptosis (Supplementary Fig. S7). With Blacks/African Americans having an increased frequency of TNBC (8), 80% of our patient explants were from this demographic. Our results open an entirely new avenue for treating patients with aggressive TNBC that harbor inactivated p53 that has allowed them to evade many therapies. Moreover, because PROTACs have catalytic activity, this allows them to be able to be used at lower concentrations than conventional small-molecule inhibitors, which should make our MDM2-PROTAC less toxic.

PROTACs require high-affinity binding to the targeted protein and recruitment of an E3 ligase that ubiquitinates the target, resulting in proteosome-mediated degradation of the protein (16). Both HTRF and SPR analyses showed high-affinity binding of our PROTAC to MDM2 and binding to VHL, and AlphaScreen analysis confirmed the formation of the ternary complex. Using inhibitors and competition assays, we showed MDM2 underwent proteosome-mediated degradation that required binding to MDM2 and recruitment of VHL. We utilized the VHL recruiting ligand, which has not been reported to function as a molecular glue, unlike the Cereblon recruiting ligand (46–48). Our MDM2-PROTAC required MDM2 and did not function as a molecular glue, and formation of the ternary complex did not result in MDM2-mediated ubiquitination and degradation of VHL. There are multiple possible reasons why VHL is not affected by our MDM2-PROTAC; for example, the length and flexibility/rigidity of the linker that affects the orientation of the ternary complex only allowing MDM2 to be ubiquitinated by VHL (49), and the inaccessibility of the three lysines of VHL when complexed with the proteins necessary to transfer ubiquitin to MDM2 (50). Importantly, comparisons between our MDM2-PROTAC and MDM2 knockdown in every assay consistently showed analogous results. Moreover, Spearman correlation coefficients were strongly positive in two p53-inactivated TNBC lines comparing globally differentially expressed genes after MDM2-PROTAC treatment to MDM2 knockdown. Taken together, the data indicate our MDM2-PROTAC specifically targets MDM2 for degradation.

Previous reports, largely limited to acute lymphoblastic leukemia, have used compounds being asserted as MDM2-specific degraders, but oddly, they only kill wild-type p53 cancer cells and not p53-inactivated cancer cells (47, 48, 51–53). This is inconsistent with multiple lines of evidence that we and others have obtained. Our evaluation of the effects of Mdm2 deletion and MDM2 knockdown in two species (mouse and human) resulted in data mirroring the data we obtained with our MDM2-PROTAC, showing that loss of MDM2 kills p53-inactivated cancers. Moreover, inhibiting MDM2 E3 ubiquitin ligase activity with the compound MEL23 was recently shown to induce death in some p53-mutant and deleted cancer cells (54), suggesting MDM2 E3 ligase function is necessary for cancer cell survival. One compound designed as an MDM2-PROTAC used a Cereblon recruiting ligand and was shown to function as a molecular glue, independent of the MDM2 ligand, and therefore was not an MDM2-PROTAC (48). Cereblon has also been shown to bind and degrade neosubstrates (55), so PROTACs that use Cereblon may also lead to the degradation of these neosubstrates that may contribute to the effects observed. PROTAC studies have used a phosphorylation-sensitive MDM2 antibody that does not bind MDM2 when it is phosphorylated at that site (22, 47, 48, 53), possibly masking its presence; we tested multiple MDM2 antibodies to rule this out. It is also possible that there are cancer cell–type or context-specific effects for MDM2-PROTACs. For example, the E3 ligase being recruited could have very different expression in different cell types and conditions. Although further investigation will be needed to resolve the differences between compounds, our data consistently show that MDM2 is required for cancer cell survival and that targeted MDM2 degradation with our MDM2-PROTAC kills TNBC.

The multiple MDM2 inhibitors available were designed to bind MDM2 with high affinity, disrupt p53:MDM2 interactions, and were shown to stabilize MDM2 protein, leading to significantly increased MDM2 levels (18, 56–61). The use of these inhibitors in patients has resulted in neutropenia and thrombocytopenia (32, 62), raising concerns for their use. The high levels of MDM2 caused by these inhibitors, which engages both p53-dependent and p53-independent MDM2 functions (15, 59, 63–65), are thought to contribute to the toxic effects observed in patients. A distinct advantage of our MDM2-PROTAC is that it degrades MDM2, preventing the toxic buildup of MDM2 observed with current inhibitors (18, 59, 62, 66). Moreover, as PROTACs function in a catalytic manner, it is expected that lower doses than those used for MDM2 inhibitors would reduce possible cytotoxicity. Concerns about globally or locally activating wild-type p53 in normal tissues with MDM2 degradation is significantly reduced with our data showing normal human breast epithelial or stromal cells and CD34⁺ hematopoietic cells were not affected by our MDM2-PROTAC, and similarly in mice, there was no overt toxicity detected in hematopoietic or other tissues. Of note, we designed our PROTAC to recruit VHL, and VHL levels are especially low in platelets compared with cancer cells (67) and would, therefore, lower the risk of patients developing thrombocytopenia, as is observed with MDM2 inhibitors (31, 32). In contrast to RG7112, our MDM2-PROTAC caused no toxicity to human CD34⁺ bone marrow cells, providing further evidence that our PROTAC should be safer than the MDM2 inhibitors have shown to be. Moreover, recently, a compound (PC14586) that stabilizes the Y220C destabilizing mutant p53 was tested in phase I trials in patients with solid tumors. Surprisingly, Li–Fraumeni patients with this germline p53 mutation were also included in this testing cohort, and notably, did not show increased toxicity to this compound (68). These results indicate that the normally low levels of wild-type p53 in tissues are not activated with loss of MDM2 or stabilization of a destabilizing mutant p53, most likely due to another stress needed to activate p53.

Our MDM2-PROTAC effectively killed both p53-mutant and deleted TNBC cells in vitro and in vivo with a similar IC₅₀ for both. RNA-seq data in both p53 mutant and deleted TNBC revealed the effects of our MDM2-PROTAC were an apoptotic mechanism mediated by TAp73 transcriptional upregulation of proapoptotic genes. MDM2 binding TAp73 inhibits its transcriptional function by preventing it from binding chromatin and is not thought to ubiquitinate TAp73 as it does p53 (37, 39, 40, 63), but this is controversial (69, 70). Instead, MDM2 can regulate TAp73 by blocking it from binding to transcriptional cofactors (39). We determined that MDM2 degradation, but not MDM2 inhibition, resulted in the phosphorylation of TAp73, which stabilizes it, and the association of TAp73 with CBP/p300 and its acetylation. Therefore, our MDM2-PROTAC results in an increased amount of TAp73 present at the chromatin competent to transcribe genes, which our data showed. Specifically, data from multiple human TNBC cell lines in culture, xenograft tumors in vivo, and TNBC patient explants treated with our MDM2-PROTAC all showed TAp73 activation. Similarly, we also observed p73 activation in p53-null mouse cancers when Mdm2 was deleted (15). Although it has been reported that specific p53 gain-of-function mutants can inhibit TAp73 transcriptional activity (71–73), that did not appear to occur to an extent that would block the induction of TAp73-mediated transcription of apoptotic target genes and the resulting apoptosis we observed. In addition, reduced levels of TAp73 via knockdown largely blocked the effect of our MDM2-PROTAC, suggesting that loss of p73 could confer resistance to our PROTAC. Collectively, our data show that in both human and mouse cancer cells with p53 inactivation, TAp73 is activated with MDM2/Mdm2 loss by shRNA, our PROTAC, or deletion, and that TAp73 compensates for the lack of functional p53 under these conditions in TNBC cells.

The failure of current preclinical cancer models to predict patient responses continues to significantly affect the ability to leverage targeted therapies for precision medicine (74). By utilizing patient-derived explants and 3D culture models of cell lines and patient-derived mammospheres, we observed that our MDM2-PROTAC killed p53-inactivated TNBC cells, including those from patients who had relapsed, and not normal cells. We also determined that our MDM2-PROTAC induced apoptosis of p53 wild-type breast cancer cells, indicating the applicability of our PROTAC beyond p53-inactivated cancers. Currently, two PROTACs, ARV-110 (NCT03888612) and ARV-471 (NCT04072952), that target the androgen receptor and the estrogen receptor, respectively, were determined safe and have advanced to phase II clinical trials (75, 76). Multiple other PROTACs target proteins, such as BTK (NX-2127), BCL-X_L (DT2216), IRAK4 (KT-474), BRD9 (CFT8634/FHD-609), and STAT3 (KT-333), that are in phase I trials for different cancer types (75) and more on the way to the clinic, highlighting that PROTACs are a rapidly growing treatment approach. Because TNBC typically grows and spreads quickly, has few treatment options, and patients tend to have a worse prognosis, our preclinical data have highlighted our MDM2-PROTAC as a new potential therapeutic approach for treating p53-inactivated TNBC.

Detailed synthetic procedures, nuclear magnetic resonance–spectra, and characterization are in Supplementary Methods.

Compounds were purchased from SelleckChem: RG7112 (#S7030), MG132 (#S2619), ABT-263/navitoclax (#S1001), MLN4924 (#S7109), VH298 (#S8449), and doxorubicin (#S1208) or Sigma (Nutlin-3 #N6287 and Etoposide #E1383). YX-2-23 (RG7112 derivative) and VHL-Amine were synthesized (see Supplementary Methods for details).

For MDM2:p53 HTRF binding assays, GST-MDM2 and HIS-p53 were expressed in E. coli and purified by affinity chromatography. pGEX-4T-MDM2-WT (RRID:Addgene_16237) and human p53-(1–393; RRID:Addgene_24859) were from Addgene. HTRF assays contained GST-MDM2, HIS-p53, anti-GST-Tb HTRF donor (RRID:AB_2927626), and anti-HIS-d2 HTRF acceptor (RRID:AB_2884027) antibodies (PerkinElmer-CisBio). HTRF assays testing the inhibition of VHL peptide binding to VHL complex contained VHL complex (R&D Systems), biotinylated VHL peptide, anti-HIS-Tb-Gold HTRF donor antibody (RRID:AB_2716834), and anti-streptavidin-d2 HTRF acceptor (RRID:AB_2928111; Perkin­Elmer-CisBio). IC₅₀ values were determined using nonlinear regression with four-parameter dose–response curve equation. Details are in Supplementary Methods.

Association and dissociation rate constants for MDM2-PROTAC and control compounds were determined using a Biacore-T200 SPR instrument using HIS-SUMO-MDM2 (expressed and purified) immobilized on a multidentate linear carboxylate high-affinity nickel chip (Xantec-Bioanalytics). Kinetic data were globally fit to a single site binding model using Biacore evaluation software (RRID:SCR_008424). Details are in Supplementary Methods.

GST-MDM2 and HIS-VHL complex were preincubated with AlphaScreen donor or acceptor beads, respectively. Dose-response assays were initiated by combining the GST-MDM2/AlphaScreen donor bead mix with the HIS–VHL complex/AlphaScreen acceptor bead mix. AlphaScreen signals were measured using the ClarioStar plate reader (BMG-Labtech). Data were fit to a bell-shaped dose–response curve equation in GraphPad Prism (RRID:SCR_002798). Details are given in Supplementary Methods.

DU4475 (HTB-123; RRID:CVCL_1183), MCF7 (HTB-22; RRID:CVCL_0031), MDA-MB-231 (HTB-26; RRID:CVCL_0062), MDA-MB-436 (HTB-130; RRID:CVCL_0623), MDA-MB-453 (HTB-131; RRID:CVCL_0418), HCC-1143 (CRL-2321; RRID:CVCL_1245), HCC-1395 (CRL-2324; RRID:CVCL_1249), and HCC-1937 (CRL-2336; RRID:CVCL_0290) cell lines were obtained from ATCC and cultured as they directed. Cell lines were authenticated using short tandem repeat profiling. Murine sarcoma primary cell cultures were derived from sarcomas spontaneously arising in Mdm2^−/−p53^−/− (from Dr. G. Lozano, MD Anderson) and Mdm2^+/+p53^−/− mice. All lines were confirmed Mycoplasma-free (MycoSensor PCR Assay, Agilent #302109).

Lentiviral shRNA constructs for MDM2 were provided by Dr. L. Mayo (IUSM) and purchased for TAp73 (TRCN6508, TRCN6511) and pLKO.1 nontargeting shRNA control (Millipore-Sigma; RRID:Addgene_8453). Lentivirus was produced by the calcium phosphate transfection method into 293T cells (RRID:CVCL_0063). For infection of MDA-MB-231 and MDA-MB-436 cells, 1 × 10⁶ cells were placed into 10-cm plates 16 hours prior to infection. Following 16 hours of exposure to lentiviral particles, cells were harvested and subsequently plated (considered time 0 hours) for experimentation.

Cells were placed in 96-well plates (2,500–3,000 cells/well, quadruplicate) and MTT (562 nm; Millipore-Sigma #M2128) or MTS (DU4475 and MDA-MB-453 only; 492 nm; Promega #G3580) survival/growth assays performed according to the manufacturer's protocol at 24-hour intervals (kinetic analyses) or 48 hours (dose responses) after adding compounds. Annexin-V positivity (Annexin-V Apoptosis Detection Kit; BD-Biosciences #556547), Caspase-3 activity (Caspase-3 Assay Kit; BD-Biosciences #556485), and Caspase-3/7 activity (Live Caspase-3/7 Green Detection Reagent; Invitrogen #C10723) were measured according to the manufacturer's instructions. Cell viability was determined by Trypan Blue dye (Gibco #15250061) exclusion. Propidium iodide (Millipore-Sigma #P4170) marked apoptotic sub-G₁ DNA content, as we previously reported (77). FlowJo (BD-Biosciences; RRID:SCR_008520) was used for flow-cytometric analyses.

Assays were performed as previously described (78). Briefly, MDA-MB-231 (250 cells/well) and MDA-MB-436 (300 cells/well) were placed, in triplicate, in 6-well plates. After 12 days in culture, colonies were stained with 0.5% crystal violet in methanol for 10 minutes at room temperature and colonies (≥50 cells) were counted using a dissecting microscope. Representative pictures were taken using an LG-G6 phone camera with 13MP standard-angle lens (no magnification).

Mammospheres were established and maintained as previously described (79). Cells were placed in low-attachment 24-well plates (1,000 cells/well) for formation assays or 96-well plates (200 cells/well) for survival/ATP production assays (CellTiter-Glo 3D Cell Viability Assay, Promega #G9681) and caspase-3/7 activity assays (Live Caspase-3/7 Green Detection Reagent, Invitrogen #C10723). IncuCyte Live-Cell Analysis System (Sartorius) was used for the real-time quantitative live-cell imaging of caspase-3/7 activity. Luminescence (survival/ATP production) and fluorescence (caspase-3/7 activity) were measured, and representative images were captured using the Cytation-5 Cell Imaging Multimode Reader (BioTek). Mammosphere (≥50 cells) number and area were determined using the Gen5 image analysis tool on the Cytation-5.

For Western blotting, whole-cell protein lysates were prepared as described (80) using ARF lysis buffer (50 mmol/L HEPES pH 7.5, 150 mmol/L NaCl, 1 mmol/L EDTA, 2.5 mmol/L EGTA, 0.1% Tween-20) with sonication (0.5 power level, two 7-second pulses separated by 2 minutes on ice; VirTis VirSonic-600). For lysis comparisons, half of the cells were lysed using RIPA lysis buffer (50 mmol/L Tris pH 7.4, 150 mmol/L NaCl, 1% sodium deoxycholate, 1% Triton X-100, 0.1% SDS). For immunoprecipitations, cells were lysed in EBC buffer (50 mmol/L Tris pH 7.5, 1 mmol/L EDTA, 1 mmol/L EGTA, 1% Triton X-100) using Dounce homogenizers as described (81). Antibodies are in Supplementary Table S2.

As we previously described (77), total RNA was isolated using TRIzol (Invitrogen #15596026) according to the manufacturer's instructions. SuperScript III First-strand Synthesis System (Invitrogen #18080051) generated cDNA prior to analysis of mRNA expression (triplicate) using RT²-SYBR Green ROX qPCR Mastermix (Qiagen #330521) on the StepOnePlus RT-PCR system (Applied Biosystems #4376600). Values were first normalized to β-ACTIN then made relative to vehicle or nontargeting shRNA controls and are presented as 2^−∆∆CT. Primer sequences for human genes are in Supplementary Table S3; mouse primers were previously published (15).

MDA-MB-231 and MDA-MB-436 cells treated with MDM2-PROTAC or DMSO vehicle control (6 μmol/L, 16 hours; triplicate) or expressing MDM2 shRNA or nontargeting shRNA control (48 hours; triplicate) were harvested and RNA isolated (described above). Quality control assessment by Azenta/GENEWIZ included the Agilent TapeStation System and Qubit assay. Paired-end RNA-seq profiles of triplicate samples for each condition were generated from Illumina HiSeq-4000 and obtained from Azenta/GENEWIZ. RNA-seq data were analyzed as we previously described (82). Details are in Supplementary Methods. Differentially expressed genes [>2 fold change with Benjamini–Hochberg (83) adjusted P < 0.05] were used to identify in which cancer Hallmark gene signatures they were significantly (FDR < 0.05) enriched. We utilized WebGestalt (RRID:SCR_006786; ref. 84) to conduct the enrichment analysis using the cancer Hallmark gene set available in the Molecular Signature DataBase (MSigDB; RRID:SCR_016863; ref. 35).

Chromatin immunoprecipitation (ChIP) was performed as we previously reported (85), except DNA was sheared using a Q800R3 sonicator (Qsonica #Q800R3-110). ChIP products were quantified using RT²-SYBR Green ROX qPCR Mastermix (Qiagen #330521) on the StepOnePlus RT-PCR system (Applied Biosystems #4376600). Each immunoprecipitation was first normalized to the amount of DNA in the input and then made relative to the appropriate IgG control. Antibody and primer information is given in Supplementary Tables S2 and S3, respectively.

MDM2-PROTAC and controls (0.5 μmol/L) were incubated with 0.5 mg/mL of mouse liver microsomes and an NADPH-regenerating system (cofactor solution) in potassium phosphate buffer (pH 7.4). Aliquots were taken at 0, 5, 15, 30, and 45 minutes and reactions were quenched with an acetonitrile solution containing an internal standard. Controls not containing the cofactor solution were also measured. Samples were analyzed by LC-MS/MS, and results were reported as peak area ratios of each analyte to the internal standard. Intrinsic clearance (CL_int) and half-life (t_1/2) were determined from the first-order elimination constant by nonlinear regression. Alliance Pharma conducted this study.

PK study (1-arm) in 18 CD-1 mice (n = 3 mice/time point; RRID:IMSR_CRL:022) administered a single intraperitoneal injection of the MDM2-PROTAC at 10 mg/kg dissolved in 10% DMSO, 10% solutol, and 80% PBS. Plasma samples were collected at 0.25, 0.5, 1, 2, 4, and 6 hours following injection and analyzed by LC-MS/MS. Mean plasma concentrations of the MDM2-PROTAC were calculated using linear regression analysis. Alliance Pharma conducted this study.

Mouse experiments were approved by the Thomas Jefferson University Institutional Animal Care and Use Committee and followed all state and federal rules and regulations. For xenografts, 10 × 10⁶ MDA-MB-231 or MDA-MB-436 cells were injected (subcutaneously) into one flank of 6- to 8-week-old female athymic nude mice (Envigo; RRID:RGD_5508395). Mice were randomized (tumor size–matched) into treatment groups once tumor volumes reached 80 mm³, and daily 50 mg/kg intraperitoneal injections of MDM2-PROTAC, RG7112D, and vehicle control began and continued for 14 consecutive days for the survival studies or 3 days for the tumor response to treatment studies. Compounds were dissolved in 10% DMSO, 10% solutol, and 80% PBS. Tumors were measured using digital calipers, and volumes were calculated using the ellipsoid volume formula. A blood sample was collected and complete blood counts were determined (GENESIS Veterinary Hematology Analyzer, Oxford Science) 13 days after treatment began. Mice were euthanized once tumors reached 2,000 mm³. Tissues harvested were formalin-fixed, paraffin-embedded, sectioned, H&E stained, and histologically evaluated. For the tumor response to treatment studies, after 72 hours of treatment, apoptotic analyses, as described above, of single-cell suspensions of tumors harvested from euthanized mice were performed.

Deidentified, fresh surgically resected TNBC tumor and normal breast tissue from adjuvant-treated patients (Supplementary Table S1) were obtained with patient written informed consent (IRB protocol #20D.826) from the Thomas Jefferson University biorepository, which is a College of American Pathologists (CAP)–certified lab that follows International Ethical Guidelines for Biomedical Research Involving Human Subjects. Patient-derived explants were established as previously reported (33, 34). Following 48 or 96 hours of treatment with the MDM2-PROTAC, control compound RG7112D, or DMSO vehicle control, explants were harvested and stained with H&E for histopathologic review or processed for Western blots, respectively. A board-certified pathologist (Dr. Juan Palazzo) reviewed each blinded case for viable tumor and/or normal benign breast tissue. IHC for cleaved caspase-3 (Supplementary Table S2) was performed on 4-μm sections using the Vectorlab ABC-HRP detection Kit, HIER (pH 6.0) with Biocare Medical intelliPATH Autostainer. Blinded samples were scored for percent cleaved caspase-3–positive cells by Dr. Palazzo and representative images were taken. Mammosphere cultures from patient tumor samples were established, as described above. Following treatment with the MDM2-PROTAC, RG7112, RG7112D, or DMSO vehicle control, mammospheres were evaluated for caspase-3/7 activity (0–72 hours) and survival/ATP production (96 hours), as described above. Sequencing of TP53 cDNA was completed during or after experimentation, as published (86), except RNA was isolated using TRIzol and cDNA was generated, as described above. Primer sequences are in Supplementary Table S3.

All experiments were performed with a minimum of technical triplicates and at least two biological replicates per condition per cell line. Data are displayed as either mean ± standard deviation or mean ± standard error of the mean (indicated in the figure legend). For biological experiments, statistical significance was determined using unpaired, two-tailed Student t tests when comparing two groups, one-way ANOVA with Tukey multiple comparisons test when comparing more than two groups at one time point, two-way ANOVA with Tukey multiple comparisons test when comparing more than two groups at multiple time points, or by log-rank test for survival analysis using GraphPad Prism software (RRID:SCR_002798). Longitudinal tumor growth analysis using two-way ANOVA with Bonferroni correction, using the tool TumGrowth (87) determined significance of the difference in xenograft tumor volume over time. To not distract from the data, a single P value corresponding to the highest P value in the figure panel (*, P < X) is provided in the legend, unless otherwise noted; specific P values are in the Supplementary spreadsheet. Spearman rank-correlation coefficients determined the association between MDM2-PROTAC—treated and MDM2 knockdown samples; the coefficient (ρ) and corresponding P value are indicated in each plot. For RNA-seq data containing read counts, edgeR (RRID:SCR_012802; ref. 88) was used to determine significance.

RNA-seq datasets are available on the Gene Expression Omnibus (RRID:SCR_005012) under accession number GSE214101.

C.M. Adams reports other support from AbbVie outside the submitted work. J.M. Salvino reports NCI grants from R01 CA272645, S10OD030245, and P30CA010815 during the conduct of the study; other support from Alliance Discovery, Context Therapeutics, the Barer Institute, and Syndeavor Therapeutics outside the submitted work; and a patent for MDM2 PROTACs pending. C.M. Eischen reports grants from the NCI, the Department of Defense, the AACR, and the Ovarian Cancer Research Alliance and other support from Jefferson University philanthropy during the conduct of the study; other support from AbbVie outside the submitted work; and a patent for MDM2 PROTACs pending. No disclosures were reported by the other authors.

C.M. Adams: Conceptualization, formal analysis, investigation, writing–original draft. R. Mitra: Formal analysis, investigation, writing–original draft. Y. Xiao: Investigation. P. Michener: Investigation, writing–review and editing. J. Palazzo: Investigation. A. Chao: Scale up of compounds. J. Gour: Investigation, writing–original draft. J. Cassel: Investigation. J.M. Salvino: Conceptualization, formal analysis, supervision, funding acquisition, writing–original draft. C.M. Eischen: Conceptualization, formal analysis, supervision, funding acquisition, investigation, methodology, writing–original draft.

We thank the members of the Eischen and Salvino labs for their helpful comments. We also thank Dr. Anne van Harten for assistance with the IncuCyte, Dr. Matthew Schiewer for the patient-derived tumor explant protocol, Dr. Bruno Calabretta for human CD34⁺ cells, Nikkole James for technical assistance, and Lily Lu for help in the Molecular Screening Shared Resource at Wistar. This work was supported by the NCI MPI-R01 CA272645 (J.M. Salvino and C.M. Eischen), NCI R01 CA181204 (C.M. Eischen), DOD W81XWH-19-1-0212 (C.M. Eischen), AACR–Bayer Innovation and Discovery Award (C.M. Eischen), Ovarian Cancer Research Alliance (C.M. Adams), NCI Cancer Center grant P30CA056036 that supports the Flow Cytometry, Translational Pathology, Laboratory Animals, and MetaOmics Shared Resource Cores in the Sidney Kimmel Cancer Center, NCI Cancer Center grant P30CA010815 and S10OD030245 (J.M. Salvino) that supports Molecular Screening and Protein Expression Shared Resource Cores at The Wistar Institute, T32CA9171 (A. Chao), the Steinfort Fund (C.M. Eischen), and the Herbert A. Rosenthal endowed chair fund (C.M. Eischen).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.