Li–Fraumeni Syndrome–Associated Dimer-Forming Mutant p53 Promotes Transactivation-Independent Mitochondrial Cell Death

Tumor suppressor p53 sits at the regulatory nexus of intrinsic tumor-suppressive responses, including DNA repair, cell-cycle arrest, regulation of metabolism, cell death, and others, that work in concert to preclude malignant transformation and progression (1, 2). Consequently, p53 is mutated in more than 50% of all sporadic cancers, and germline mutations in p53, causative of Li–Fraumeni syndrome (LFS), result in a predisposition toward early-onset cancers, which frequently include breast carcinomas, acute leukemia, brain tumors, sarcomas, and adrenal cortical carcinomas (3–6). Unlike many other tumor suppressor genes, which are inactivated through frameshift or nonsense mutations, ∼75% to 80% of TP53 mutations encode missense mutant proteins. This can result in a mutant p53 protein that has both lost canonical p53 transactivation ability and gained not only dominant repressive activity over wild-type p53 (WT p53) but also neomorphic properties that impart oncogenic ability (7, 8). The spectrum of TP53 mutations, marked by highly frequent “hotspot” mutations at key residues, implicates the selective potential that certain mutations may confer to actively drive tumor development, as the occurrence of hotspot mutations is a hallmark of a gain-of-function (GOF) oncogene. Such GOF activities in p53, however, vary across hotspot mutants with R175H and R273H but not R248W and D281G mutant p53 conferring the ability of SAOS-2 cells to grow in soft agar, for instance (9). Thus, it is essential to consider the specific mutation and context when probing mutant p53 GOF (10–12). Most work in understanding the oncogenic contributions of mutant p53 and WT p53 loss of function (LOF) has focused on hotspot mutations occurring in the DNA binding domain (7). In contrast, mutations that occur within the oligomerization domain (OD) of p53, which generally compromise p53 oligomerization (13), have been poorly studied despite the high frequency of LFS mutations within the OD relative to its length (Supplementary Fig. S1A; ref. 14). Although the most common OD mutant form is the p.R337H variant, which is highly prevalent in Brazil presumably due to founder effect and may thereby be overrepresented in datasets of germline p53 variants (15–17), several other germline p53 OD mutations have been documented in patients with LFS (18). Among these, p.A347D is the most frequent and nearly only dimer-forming variant as identified from the TP53 Database (R20, 2019; ref. 14). Elucidating the oncogenic GOF and tumor-suppressive LOF effects of OD mutant p53 may reveal targetable vulnerabilities that will yield more effective therapies for a previously understudied subset of patients harboring OD mutations.

p53 exists in monomeric, dimeric, and tetrameric states. Under unstressed conditions, p53 primarily exists as a dimer that is preferentially bound and degraded by its negative regulator, MDM2 (19, 20). Upon stress induction, p53 levels rise and the protein transitions into the tetrameric conformation to more efficiently bind DNA and transactivate canonical p53 targets (21, 22). Over the years, much evidence has stressed the importance of the p53 OD as critical in transcriptional regulation, posttranslational modification, degradation, and protein–protein interactions (23–26). Mutations within the OD affect oligomerization and p53 function in a mutation-specific and context-dependent manner. Kawaguchi and colleagues demonstrated that mutations impairing oligomerization vary in their impact on p53 transcriptional activity in yeast-based assays using several canonical targets (13). Furthermore, certain cancer-associated OD mutants have been shown to accumulate at abnormally high cellular levels, yet their impact on tumori­genicity is unknown (19, 27). Previous studies have revealed that OD mutants may still stimulate transcription selectively, including the ability to activate cell-cycle arrest but not proapoptotic genes (28). Moreover, limited evidence has shown that the monomer-forming p53(R337H) mutant carriers may experience increased oxidative damage and elevated antioxidant status (29). Yet, the phenotype of OD mutants, especially in the context of GOF activity, remains largely uncharacterized, and most studies investigating the function of such mutants were performed under nonphysiologic conditions (24).

Here, we focused on the cancer-associated p53(A347D) mutant, a dimer-forming (13), pathogenic p53 variant observed in ∼15% of LFS patients with OD mutations and among a set of multimeric mutant p53 alleles that together result in 80% disease penetrance (18). The prevailing, albeit limited, literature on OD mutant p53 focuses primarily on monomeric mutants, whereas the protumorigenic functions of p53(A347D), the third most frequent OD mutation and the most common dimeric mutant p53, remain unknown. We utilized three patient-derived primary fibroblast cell lines from a single family with one sibling expressing WT p53 and two expressing the germline p53(A347D) mutant. We generated stem cells [induced pluripotent stem cells (iPSC)] from these fibroblasts, which were then differentiated into mesenchymal stem cells (MSC) and osteoblasts, the cell of origin for osteosarcoma, as a model for early LFS tumors. As osteosarcomas are one of the common tumors found in patients with LFS, we engineered via CRISPR–Cas9 U2OS osteosarcoma cells that harbor WT p53 to express either no p53 or are heterozygous (p53^+/AD) or homozygous (p53^AD/AD) for the A347D mutation. These cell lines allowed us to demonstrate novel GOF and LOF activities of dimeric mutant p53.

Single residue substitutions within the OD of TP53 range from completely abolishing p53 oligomerization to allowing tetramerization in a mutation-specific manner. It has been suggested that p53(A347D) specifically yields a dimeric protein unable to form the active, tetrameric structure and consequently lacking in canonical transcriptional activity (13). In concordance, we observed that both p53(A347D) and p53(A347T), when expressed via plasmid transfection, primarily form dimers, whereas the p53(L330A) mutant yields a monomeric protein (Supplementary Fig. S1B). We utilized CRISPR–Cas9 gene editing to generate single clones of U2OS either heterozygous (p53^+/AD #1, p53^+/AD #2, and p53^+/AD #3) or homozygous (p53^AD/AD #1 and p53^+/AD #2) for the p53(A347D) mutation, as well as U2OS cells that are p53 null (KO #1, KO #2, and KO #3). Cells were pooled according to mutation status to account for clonal variation and were used where indicated. Furthermore, primary dermal fibroblast cells isolated from two LFS siblings (L53-M1 and L53-M2) heterozygously expressing p53(A347D) were compared with primary dermal fibroblast cells isolated from a third sibling (L53-WT) with WT p53. We assessed the oligomeric status of p53 in U2OS cells (Fig. 1A) and LFS fibroblasts (Fig. 1B) following protein cross-linking with glutaraldehyde and observed that p53(A347D) preferentially forms dimers under basal conditions. Even after induction with etoposide, p53(A347D) expressed in U2OS p53^AD/AD cells was completely unable to form tetramers, whereas the tetramer was the primary species observed in WT p53 cells (Fig. 1A). Interestingly, p53^+/AD U2OS cells exhibited intermediate tetramer formation (i.e., roughly half that of WT p53 cells) after etoposide treatment, implying that heterotetramers of WT and mutant p53 may not form. Prior studies have suggested that p53 dimers form cotranslationally, whereas p53 tetramers form posttranslationally, thereby resulting in 50% heterotetramer formation when WT and mutant p53 are coexpressed (30). However, we found that ectopically expressed WT and mutant p53 do not coimmunoprecipitate, further suggesting that hetero-oligomers of WT and dimeric mutant p53 do not form (Supplementary Fig. S1C).

Previously published work has revealed that two antibodies can discriminate between WT and mutant forms of p53, namely, PAb240, which can preferentially recognize the unfolded conformation of several common mutant forms of p53 by binding an epitope within the core domain that is normally hidden in WT p53 (31–33), and PAb1620, which appears to bind specifically to WT p53 (34). Surprisingly, while we confirmed increased immunoprecipitation of WT p53 by the 1620 antibody and superior immunoprecipitation of a common tumor-derived mutant form of p53(R175H) by PAb240, we observed that neither antibody preferentially immunoprecipitated p53(A347D), indicating that the conformation of p53(A347D) is neither fully WT nor mutant and implying that p53(A347D) is only partially unfolded (Fig. 1C).

Given the importance of p53 oligomerization for DNA binding and evidence highlighting the varied effects that p53 OD mutations exert on transcriptional activity, we sought to elucidate the transactivation potential of canonical p53 targets by p53(A347D) within an endogenous context. To assess the induction of downstream targets, we treated cells varying in p53 status with nutlin-3a, a small-molecule MDM2 antagonist that disrupts the p53–MDM2 complex and stabilizes p53 (35). We found that p53^AD/AD U2OS cells were completely unable to increase expression of MDM2 and p21 proteins, whereas heterozygous p53^+/AD U2OS cells (Fig. 1D; Supplementary Fig. S1D) and patient fibroblasts (Fig. 1E) revealed partially attenuated expression of these two proteins compared with WT p53–bearing cells, as demonstrated by immunoblot analysis. Furthermore, p53(A347D) in p53^AD/AD U2OS cells appeared to be hyperstable, presumably as a result of the inability of p53(A347D) to activate its negative regulator, MDM2. The observed hyperstability of p53(A347D) was verified kinetically via a cycloheximide (CHX) chase assay (Fig. 1E; Supplementary Fig. S1E). Although we have previously reported that dimeric p53 is preferentially bound and degraded by MDM2, such enhanced degradation requires the presence of sufficient quantities of MDM2 (19). Quantitative RT-PCR similarly revealed the markedly attenuated expression of CDKN1A and MDM2, as well as additional canonical p53 target TIGAR mRNAs, even upon nutlin-3a treatment in p53^AD/AD U2OS cells, whereas p53^+/AD U2OS cells (Supplementary Fig. S2A) and LFS fibroblasts (Supplementary Fig. S2B) exhibited intermediate transcriptional activity between p53^+/+ and p53^AD/AD cells. To further validate if p53(A347D) also loses its transactivation potential in LFS patients’ osteoblasts, the cell of origin of osteosarcoma, we generated WT and LFS-derived iPSCs from L53-WT and L53-M1 fibroblasts, respectively (see Methods). We then differentiated these iPSCs to their corresponding MSCs and then to their osteoblast counterparts (see Methods). In concordance with p53^+/AD U2OS cells, LFS MSCs also demonstrated partially attenuated transcription of MDM2 and CDNK1A (Supplementary Fig. S2C). Additionally, as p53 is known to bind DNA as a tetramer, we sought to assess whether the A347D mutation would disrupt its promoter binding ability. We performed quantitative chromatin immunoprecipitation (ChIP-qPCR) to assess the occupancy of p53(A347D) at select p53 targets and, as expected, found that p53(A347D) is unable to bind promoter sequences of CDNK1A and PUMA even when treated with etoposide(Supplementary Fig. S2D).

In order to gain insight into the global transcriptional profiles of dimeric mutant p53–bearing cells, we performed bulk RNA sequencing (RNA-seq) of U2OS cells varying in p53 status treated with DMSO or etoposide as well as of LFS and WT osteoblasts (Supplementary Data S1–S3). Consistent with prior observations, RNA-seq analysis revealed significantly attenuated expression of high-confidence p53 target genes overall in p53^AD/AD U2OS cells even after etoposide treatment, whereas p53^+/AD U2OS cells retained partial ability to induce p53 targets upon activation (Fig. 2A; ref. 28). Not surprisingly, p53 target gene expression was found to be higher overall in WT U2OS cells compared with p53^AD/AD cells, both basally and following etoposide treatment (Supplementary Fig. S3A). Gene set enrichment analysis (GSEA) confirmed that the p53 signaling pathway is significantly enriched in U2OS p53^+/+ and WT osteoblasts but not mutants (Supplementary Fig. S3B and S3C). Collectively, these results demonstrate the transcriptional impairment, a decidedly LOF phenotype, and the associated hyperstability of p53(A347D).

RNA-seq revealed not only that p53(A347D) has lost canonical transcriptional activity but also that the transcriptomes of mutant p53–bearing U2OS cells and LFS osteoblasts are decidedly altered, as indicated by the presence of significantly upregulated and downregulated genes in such cells relative to their WT counterparts (Supplementary Fig. S4A–S4E). Pathway enrichment analysis via WebGestalt further demonstrated coordinate changes in biological processes within p53^AD/AD U2OS cells under basal conditions and after etoposide treatment relative to both WT p53 and p53 knockout (KO) U2OS cells (Fig. 2B). Pathway alterations were similarly observed in p53^+/AD LFS osteoblasts compared with WT (Fig. 2B). Interestingly, fewer highly enriched pathways were observed in U2OS p53^AD/AD cells relative to p53 KO cells after etoposide treatment than in the p53^AD/AD versus p53 KO etoposide-treated gene set, suggesting that transcriptional response to etoposide treatment may not be significantly different between mutant p53 and p53 KO cells (Fig. 2B). Amidst the myriad altered processes, glycolysis and electron transport chain pathways were strikingly found to be consistently enriched in both mutant p53 U2OS cells and osteoblasts under basal conditions. This concordance not only strongly implicates the potential GOF role of p53(A347D) in driving metabolic reprogramming, as the same enrichment was detected in two completely different cell types, but also suggests the validity of CRISPR-generated U2OS p53(A347D) cells as a model for what occurs in a patient-derived system.

Given that a significant proportion of genes involved in glycolysis were observed to be upregulated in p53^AD/AD as compared with WT p53 cells (Supplementary Fig. S4F) and that GOF activities of mutant p53 in driving increased glucose uptake and glycolysis have been demonstrated in the recent literature (although never with OD mutants), we sought to validate our findings (36, 37). Metabolic flux analysis of U2OS cells varying in p53 status revealed that cellular glycolytic flux, as measured by the proton efflux rate, was significantly elevated in both p53^+/AD cells and p53^AD/AD cells as compared with WT p53 and p53 KO cells (Fig. 2C). Additionally, ATP generated from glycolysis was highest in both p53^+/AD cells and p53^AD/AD cells, whereas ATP generated in mitochondria remained similar across the cell lines evaluated (Fig. 2D). Overall, these data reveal a potential GOF activity of p53(A347D) in eliciting a glycolytic switch.

In order to obtain a more global picture of mutant dimeric p53 occupancy across the genomic landscape and assess whether p53(A347D) may bind novel sequences, we performed ChIP of endogenously expressed p53 followed by next-generation sequencing (ChIP-seq) in both U2OS cells (Supplementary Data S4 and S5) and LFS osteoblasts (Supplementary Data S6). ChIP-seq analysis revealed significant genome-wide overlap in binding (389/520) between p53^+/+ and p53^+/AD cells. This suggested that p53(A347D) may not have dominant negative activity in preventing WT p53 from binding its transcriptional targets. Strikingly, p53(A347D) in p53^AD/AD cells exhibited a starkly different binding landscape relative to WT p53, with only 9 of 571 peaks in common and 105 of 571 unique peaks, thereby indicating that p53(A347D) has not only lost WT DNA binding ability but also gained the capacity to associate with novel genomic loci (Fig. 3A). Corroborating this finding, p53 in LFS osteoblasts retained the ability to associate with regions bound by both WT p53 and dimeric mutant p53 in U2OS cells (Fig. 3B).

The genomic distribution of p53 binding sites in p53^AD/AD cells was found to be proportionally distinct from both p53^+/+ and p53^+/AD cells with over three quarters (90/117) detected at promoters, defined as regions within ±1,000 bp of transcription start sites as compared with roughly 6% (30/468) in cells bearing only WT p53 (Supplementary Fig. S5A). De novo motif enrichment analysis revealed that, as expected, p53 motifs were the top enriched in p53^+/+ and p53^+/AD cells, suggesting that p53 binding at canonical elements is permitted even in the presence of mutant p53 (Supplementary Fig. S5B). However, only cells expressing p53(A347D) exhibited enrichment of ETS transcription factor family motifs, albeit at much lower confidence than p53 motifs in WT p53 cells, which may support a potential chromatin binding GOF by p53(A347D) (Supplementary Fig. S5B). Although these data are intriguing and seemingly in concordance with prior studies demonstrating that certain cancer-related mutant p53 proteins associate with ETS sites and ETS family members (11, 38, 39), such canonical p53 mutant proteins harbor mutations within the DNA binding domain, whereas the p53(A347D) mutant has a WT DNA binding domain.

In fact, when ectopically expressed in H1299 cells, only p53(R175H) and not p53(A347D) could be coimmunoprecipitated with ETS1 (Supplementary Fig. S5C) or ETS2 (Supplementary Fig. S5D), suggesting that any association of p53(A347D) with ETS motifs may be through direct binding or at the very least independent of ETS proteins. A related study by Gencel-Augusto and colleagues has identified cooperative binding of p53(A347D) and PPARα at PPAR response elements to promote a PPAR transcriptional program in mice harboring the same OD mutation (40). Although we were unable to detect PPAR motif enrichment in U2OS cells bearing mutant p53, we found that p53(A347D) could be coimmunoprecipitated with PPARα to a greater degree than could WT p53 (Supplementary Fig. S5E). Furthermore, in a similar assay, p53(A347D) was found to associate with PPARγ, although this binding interaction necessitated the presence of PPARα (Supplementary Fig. S5F).

We next asked whether etoposide treatment might either increase or alter the distribution of genomic binding of dimeric mutant p53. As expected, although WT p53 chromatin association significantly increased following etoposide treatment (Supplementary Fig. S5G), binding to mutant-specific peaks did not increase after etoposide in U2OS p53^AD/AD cells (Fig. 3C). We then performed functional enrichment analysis on genomic regions occupied by WT and/or mutant p53 (Fig. 3D). Those ontologies that were highly enriched in p53^+/+ versus p53^AD/AD cells were not enriched in p53^+/AD cells. This indicated that p53 in the heterozygous mutants may retain WT DNA binding ability. However, several gene ontologies, including those related to mitochondrial respiration, were identified to be highly enriched in the p53^AD/AD cells exclusively, which was in concordance with pathway analysis of differentially expressed genes in U2OS p53^AD/AD cells (Fig. 3D). On the other hand, glycolysis was not seen to be enriched in the p53^AD/AD ChIP-seq gene set, which suggests that the glycolytic switch may not be directly transcriptionally induced by dimeric mutant p53. Furthermore, by intersecting RNA-seq expression data with ChIP-seq peaks, we were able to identify significant differential regulation of genes bound uniquely by p53(A347D) relative to cells with either WT p53 or no p53 (Fig. 3E). GSEA of osteoblast expression data also revealed that ChIP targets identified in p53^+/AD and p53^AD/AD U2OS cells were enriched in LFS osteoblasts compared with WT osteoblasts, ultimately indicating that patient-derived cells may express transcriptional targets of p53(A347D) (Supplementary Fig. S5H). Consistency between two cellular contexts not only strengthens the case for novel chromatin binding by p53(A347D) but also suggests the relevance of studying LFS-derived mutant p53 with either CRISPR-generated or patient-derived cells. Although there are far fewer genes potentially regulated by p53(A347D) than WT p53, these results implicate a neomorphic role of dimeric mutant p53 in transcriptional regulation that may result in mitochondrial alterations.

Because the results from both RNA-seq and ChIP-seq analyses revealed that the expression of mitochondrial electron transport proteins may be altered in cells expressing p53(A347D), we used confocal microscopy to assess mitochondrial morphology in the U2OS series of cells by staining with MitoTracker Red. This revealed dramatic mitochondrial network aberration in p53^AD/AD U2OS cells characterized by greater mitochondrial area and greater percentage of cells with altered mitochondria when compared with p53^+/+, p53 KO, or p53^+/AD cells (Fig. 4A and B; Supplementary Fig. S6A and S6B). Knockdown of mutant p53 significantly reduced the aberrant morphology of mitochondria. The morphologic change we observed was similar to that seen in cells with WT p53 that were treated with etoposide (Fig. 4C; Supplementary Fig. S6A). Because p53^+/AD cells did not exhibit altered mitochondrial morphology, this suggests that WT p53 may be haplosufficient in maintaining normal mitochondrial networks. These results implicate an extranuclear function of p53(A347D) in potentially inducing mitochondrial stress. Although p53 knockdown was able to revert the stressed mitochondrial phenotype basally in p53^AD/AD cells, etoposide-mediated changes in mitochondrial morphology were unable to be so rescued (Supplementary Fig. S6A). It remained possible that the altered mitochondrial morphology seen in the untreated p53^AD/AD cells evolved during the generation of these CRISPR cells as part of the selection process. Thus, we sought to confirm this key observation using an orthogonal method—introducing mutant p53 into U2OS KO cells by retroviral expression. Importantly, the ectopically expressed mutant p53(A347D) phenocopied the aberrant mitochondrial morphology observed in U2OS p53^AD/AD cells (Fig. 4D).

Mitochondrial fragmentation reflects a shift in the dynamic mitochondrial equilibrium toward mitochondrial fission and is ultimately associated with alterations in metabolism and mitochondrial dysfunction (41–43). Although a causal link between mitochondrial fragmentation and apoptosis is not yet clear, evidence suggests that a shift toward mitochondrial fission sensitizes cells to apoptosis and vice versa (44). Accordingly, we sought to assess whether dimeric mutant p53–dependent mitochondrial network changes may drive sensitization to apoptosis both basally and under stress. In addition to its well-characterized role as an activator of transcription, p53 possesses transactivation-independent, cytosolic effects, including the inhibition of autophagy and induction of apoptosis (45). Cytoplasmic p53 rapidly localizes to mitochondria under cell death–inducing conditions to drive mitochondrial outer-membrane permeabilization and subsequent cytochrome c release independent of its transcriptional activity, which is supported by the observation that transactivation-deficient mutant p53 retains the ability to induce apoptosis (46–49). Along with our findings on aberrant mitochondrial morphology, we hypothesized that mutant p53(A347D)-bearing cells may similarly retain proap­optotic function(s) and thereby be more vulnerable to genotoxic stress. Indeed, p53^AD/AD U2OS cells displayed markedly attenuated viability following treatment with etoposide as compared with p53^+/+, p53 KO, and p53^+/AD cells (Fig. 5A; Supplementary Fig. S7A). Decreased cell viability upon etoposide treatment was rescuable by inhibiting caspase-3, with zVAD-FMK, confirming that p53^AD/AD cells preferentially undergo apoptosis. Further, the viability of cells with ectopically expressed p53(A347D) was also significantly decreased upon etoposide treatment (Fig. 5B), indicating that the increased death seen upon treatment with this chemotherapeutic agent was not an indirect outcome of selecting the mutant p53–expressing cells. Interestingly as well, expression of other dimer-forming mutants, including E343K, L344A, and A347S (13), rendered cells somewhat more sensitive to etoposide treatment, whereas the monomeric A347P mutation (13) did not elicit such sensitivity (Fig. 5B). Enhanced PARP and caspase-3 cleavage was detected in p53^AD/AD cells upon etoposide treatment, corroborating the finding that etoposide induces apoptosis (Fig. 5C). Additionally, MSCs derived from LFS patient iPSCs expressing heterozygous p53(A347D) mutant protein exhibited increased PARP and caspase-3 cleavage following etoposide treatment (Fig. 5D). Although p53^+/AD U2OS cells were similarly sensitive to etoposide as both WT p53 and p53 KO cells, cell-cycle profiling revealed that both p53^+/AD and p53^AD/AD U2OS cells arrested to a greater degree after etoposide treatment (Supplementary Fig. S7B). Notably, our results indicating severe transcriptional impairment of mutant p53(A347D) suggested that cell death caused by etoposide was independent of transcription. This was confirmed when we treated cells with actinomycin D and found that the mutant p53(A347D) cells still underwent cell death upon etoposide treatment (Fig. 5E). Verifying that etoposide sensitivity was not due to an off-target effect, we observed propensity for apoptotic cell death in p53^AD/AD cells after treatment with daunorubicin, another inhibitor of topoisomerase II (Supplementary Fig. S7C).

We went on to assess whether the mutant p53(A347D) cells were also sensitive to other known agents that ­produce WT p53-dependent cell death. When we tested the impact of treatment with the ferroptosis inducer IKE (50), only cells with WT p53 (p53^+/+ and p53^+/AD) underwent ferroptosis under these conditions, which was confirmed via rescue with the ferroptosis inhibitor ferrostatin-1 (ref. 51; Fig. 5F; Supplementary Fig. S7D). Further, treatment with the topoisomerase I inhibitor camptothecin did not kill mutant p53(A347D) cells (Fig. 5G).

Given the well-defined apoptogenic role cytoplasmic p53 plays at mitochondria and the hyperstability of p53(A347D), we posited that p53(A347D) drives basal levels of stress, which leads to enhanced apoptotic cell death after topoisomerase II inhibition (46). In order to test this hypothesis, we assessed the association of p53 protein from the different U2OS allelic series with mitochondria. Cell fractionation analysis revealed that p53 protein from U2OS p53^+/+ and p53^+/AD cells was associated with the mitochondrial fraction to a much lesser extent than p53 protein from p53^AD/AD cells, with p53^+/AD cells exhibiting intermediate mitochondrial p53 levels (Fig. 6A). The elevated levels of dimeric mutant p53 at mitochondria may be due to the greater amounts of this hyperstable protein or may reflect preferential localization or binding to mitochondria. The amount of the p53^AD/AD variant associated with mitochondria was increased to an even greater extent when cells were treated with etoposide, suggesting that protein levels alone may not explain increased levels of mitochondrial mutant p53 (Fig. 6A and B).

We utilized pifithrin-μ (PFTμ), a small-molecule inhibitor reported to disrupt p53 binding to mitochondrial antiap­optotic proteins and prevent p53-mediated apoptosis (52). Remarkably, treatment with this compound dramatically reduced the level of mutant p53 within the mitochondrial fraction (Fig. 6B; Supplementary Fig. S8A and S8B). In accordance, we found that p53(A347D) could be found associated with the mitochondrial antiapoptotic proteins, Bcl-2 and Bcl-xL, to a significantly greater extent than does WT p53 (Fig. 6C). Relevantly, p53 association with Bcl-2 and Bcl-xL inhibits their ability to block the proapoptotic activities of BH3-only proteins (53), implicating the mechanism by which p53(A347D) promotes apoptosis. These observations prompted us to determine whether the interaction of mutant p53 with mitochondria might be involved in its ability to render cells hypersensitive to etoposide treatment. At an etoposide dose that did not affect significantly the viability of p53^+/+ or p53 KO cells, cotreatment of PFTμ and etoposide dramatically rescued the decreased cell viability in p53^AD/AD cells observed upon etoposide treatment alone (Fig. 6D; Supplementary Fig. S8C). Finally, etoposide treatment attenuated reduced (GSH) to oxidized (GSSG) glutathione ratios in p53(A347D)-bearing cells relative to both untreated cells and p53^+/+ and p53 KO cells (Supplementary Fig. S8D). This suggests that etoposide induces oxidative stress to a greater extent in mutant p53–bearing cells, thereby corroborating the direct apoptogenic role of p53(A347D) at mitochondria.

Taken together, our findings indicate that dimeric mutant p53 may inhibit mitochondrial antiapoptotic proteins to induce cell death to a greater extent than WT p53 under induction of specific apoptotic stimuli. Ultimately, we discovered that though mutant dimeric p53 has lost the ability to induce the expression of proapoptotic genes, it retains the license to kill through transactivation-independent effects at mitochondria, thereby driving elevated sensitivity to certain DNA-damaging agents.

We next examined functional consequences of the LFS-associated A347D mutation that might be more directly associated with certain aspects of tumorigenesis using different assays. First, we examined the growth characteristics of the U2OS cell line series and found that p53^AD/AD and p53^+/AD U2OS cells exhibited higher in vitro growth rate in monolayer cell culture compared with WT p53 and p53 KO cells (Fig. 7A). We also assessed the ability of the U2OS allelic series of cells to engraft and proliferate in vivo using xenograft tumor assays and observed that whereas U2OS cells harboring WT p53 (p53^+/+ or p53^+/AD) did not form tumors over the time course of the experiments, both U2OS KO and p53^AD/AD cells formed tumors at equivalent rates (Fig. 7B).

To examine possible early stages of the metastatic process, we used an invasion assay described previously by Guzman and colleagues (54), which involves generating cell spheroids under low adhesion conditions and then embedding them within a 3D biopolymer matrix, followed by monitoring the spontaneous invasion of cells into the matrix. Strikingly, LFS patient-derived p53^+/AD fibroblasts exhibited significantly greater invasion into the surrounding matrix compared with WT p53 fibroblasts (Fig. 7C). As it had been predetermined that U2OS cells are not amenable to this assay, we did not pursue the invasion potential of the A347D mutation in our U2OS allelic series of cells.

Finally, we examined the ability of LFS iPSC-derived osteoblasts to engraft and grow in nude mice. We found that 6/10 and 5/8 LFS (p53^+/AD) osteoblasts successfully engrafted and formed tumors, whereas none of the WT p53 (p53^+/+) osteoblasts produced palpable tumors in two independent experiments (Fig. 7D). Although monolayer growth suggests p53(A347D) GOF in driving enhanced proliferation, these results overall indicate that increased tumorigenicity (in a xenograft model) may be an LOF phenotype. It is also possible that although p53(A347D) may promote greater proliferation than both WT p53 and p53 KO cells, the ability to induce mitochondrial stress and promote cell death imparts a selective disadvantage. Although the precise mechanism by which p53(A347D) may promote tumorigenesis through either GOF or LOF effects is not yet fully elucidated, it is clear from these data and the observation that individuals bearing the p53(A347D) in the germline suffer from cancers that cells bearing p53(A347D) demonstrate greater oncogenic potential and malignant properties over cells with WT p53. Ultimately, we describe p53(A347D) as a hyperstable dimeric protein that has lost canonical transcriptional activity, which facilitates tumorigenesis, while gaining the ability to induce mitochondrial changes and apoptosis, thereby retaining certain tumor-suppressive characteristics (Fig. 7E).

We have characterized key activities of the LFS mutant p53(A347D) that preferentially forms dimers and not tetramers and is unable to form hetero-oligomers with WT p53 when coexpressed ectopically. Although we identified p53(A347D) to have significantly lost canonical transcriptional activity and WT DNA binding ability, selective p53-dependent transactivation may still occur, given our finding of a set of upregulated and downregulated genes whose loci are occupied by dimeric mutant p53. We also observed the enrichment of glycolysis and mitochondrial electron transport chain–related transcripts in both U2OS p53^AD/AD cells and LFS osteoblasts, which translated to elevated glycolysis in mutant p53–bearing U2OS cells. Furthermore, dimeric mutant p53 drives mitochondrial network aberrations, which implicates glycolytic reprogramming as a compensatory response. Surprisingly, we found that despite an impaired potential to activate many proapoptotic targets, p53(A347D) retains the ability to preferentially stimulate apoptosis through the direct mitochondrial cell death pathway in a transactivation-independent manner. Finally, we discovered that cells bearing dimeric mutant p53 are more proliferative, invasive, and tumorigenic than WT p53 cells.

A growing body of work has examined myriad aspects of the impact of mutant p53 GOF activities on key protumorigenic processes such as metabolic reprogramming, invasion, and stemness as well as the mechanistic underpinnings and clinical implications of mutant p53 GOF (11, 12, 55). However, these insights have been near exclusively associated with DNA binding domain mutants, and little is known about the functions of tumor-associated p53 OD variants. To our knowledge, this is the first study to perform a deep dive into the molecular functions of oligomerization-defective mutant p53 that may ultimately possess therapeutic implications. Here, we studied the p53(A347D) mutation expressed in a diverse set of endogenous contexts (U2OS cells, patient-acquired primary fibroblasts, patient-derived MSCs and osteoblasts) and utilized orthogonal experimental approaches to characterize the GOF and LOF activities of this dimeric p53 variant, which are summarized in Supplementary Tables S1 and S2. Given striking similarities between derivatives from LFS patient–derived cells and CRISPR-generated mutants, we posit the relevance of both systems to study other LFS-derived p53 variants.

Tetramerization of p53 has been well demonstrated to be necessary for efficient DNA binding, and certain mutations within the p53 OD were shown to abolish both tetramer formation and transcriptional activity to varying degrees (13, 23). Although prior findings have demonstrated that p53(A347D) may have lost selective transcriptional activity, our observations demonstrate more globally in a physiologically relevant context that endogenously expressed p53(A347D) has a significantly impaired ability to activate a canonical p53 transcriptional program. This loss in transactivation potential most likely leads to the accumulation and hyperstability of p53(A347D) due to the inability of this p53 variant to induce its negative regulator MDM2. Although we have demonstrated previously that dimeric p53, in both WT and tumor-derived variants, is preferentially bound and degraded by MDM2, such an enhanced degradation may only occur in contexts where MDM2 is present in adequate quantities (19). Ultimately, these results implicate impaired canonical transcriptional activity as a clear LOF phenotype of p53(A347D) that may promote tumorigenesis. Yet, p53(A347D) gains a novel chromatin binding signature char­acterized by the enrichment of ETS binding motifs at predicted mutant p53 binding sites. Although this finding implies cooperativity with ETS family members, especially ETS1 and ETS2, which have previously been demonstrated to interact with p53 DNA binding domain mutants to drive transcription (38, 39, 56) the lack of such an interaction by p53(A347D) suggests that this dimer-forming variant may independently bind DNA, albeit less efficiently than WT p53. This is a distinct possibility given that p53(A347D) possesses a conformation that is not fully denatured, implicated by inaccessibility to PAb240, yet not completely well folded. We cannot rule out the possibility that p53(A347D) either interacts with other ETS family members to regulate transcription or coordinates with PPAR proteins, although we did not observe evidence for the upregulation of a PPAR-dependent transcriptional program.

Transcriptomics revealed both glycolysis and mitochondrial respiration–related genes to be enriched in dimeric mutant p53–bearing cells, yet only mitochondrial electron transport genes were predicted to be bound by p53(A347D). This seeming discrepancy suggests that mitochondrial changes may be regulated, in part, transcriptionally by mutant p53, whereas glycolytic reprogramming may be an indirect compensatory outcome of mitochondrial network alterations that occur. We additionally report that dimeric mutant p53 is found associated with mitochondria at higher levels both basally and after etoposide treatment, which may also contribute significantly to mitochondrial network changes. Although difficult to delineate whether these changes are mediated through gain of transcriptional or transactivation-independent function, it is clear that p53(A347D) drives mitochondrial aberrations that are characteristic of stress.

The regulation of mitochondrial fission and fusion dynamics is critical in sustaining mitochondrial health and function under metabolically or environmentally stressful conditions (43, 57). Excessive mitochondrial fission leads to fragmentation and is both a sensitizer to and a requisite step in apoptosis (57–59). Furthermore, mitochondrial fragmentation has been linked to glycolytic reprogramming in cancer cells and cancer-associated myofibroblasts (60, 61). WT p53 has been shown to regulate fission–fusion dynamics by transcriptionally inducing mitofusins (Mfn1 and Mfn2), which are essential for mitochondrial fusion, and by promoting the mitochondrial translocation of dynamin-related protein 1 (Drp1), which induces fission (46, 62). Though we cannot definitively label mitochondrial aberrations driven by p53(A347D) as either mitochondrial fission or fusion, when placed into the context of apoptotic sensitivity and increased glycolysis, unopposed fission may be the more likely explanation for this phenotype. Further investigation into the precise mechanism by which p53(A347D) may tilt the mitochondrial dynamic balance either transcriptionally or through direct protein–protein interactions is warranted, especially given that (i) we found elevated levels of mutant dimeric p53 associated with mitochondria, (ii) WT p53 already functions in this capacity, and (iii) dysregulated mitochondrial dynamics may be a targetable vulnerability of cancers (46, 63). However, of note is that though p53^AD/AD cells may be primed to die, they do not undergo apoptosis under basal conditions, implying that mitochondrial network alteration may be necessary, but not sufficient, for cell death.

Mitochondrial fission may additionally be tumor-supportive in certain contexts. Higher levels of mitochondrial fission have been associated with increased proliferative capacity and invasion in some cancer cell lines, whereas mitochondrial fusion opposes growth (64). Mitochondrial fission may thereby serve as a potential mechanism of tumorigenesis in cells bearing p53(A347D). Furthermore, mitochondrial fragmentation has been linked to glycolytic reprogramming in cancer cells and cancer-associated myofibroblasts (60, 61). Elevated glycolysis is a well-demonstrated and frequent protumorigenic alteration that supports the increased anabolic demands of rapidly proliferating cancers and helps maintain cellular redox potential (65). Therefore, we can hypothesize that mitochondrial fission and associated metabolic reprogramming may be mechanisms by which p53(A347D) mediates the observed enhanced proliferative, invasive, and tumorigenic capacities of cells bearing this mutation.

In addition to its proapoptotic functions as a transcription factor, p53 possesses transactivation-independent apoptogenic roles. WT p53 rapidly translocates to mitochondria in response to stress signals, where it can interact with and induce the oligomerization of both proapoptotic BH3-only proteins Bax and Bak and VDAC, leading to pore formation in the outer mitochondrial membrane and subsequent release of apoptogenic factors from mitochondria (66, 67). Additionally, p53 physically interacts with and inhibits antiapoptotic proteins Bcl-2, Bcl-xL, and MCL1, thereby releasing proapoptotic tBid, Bax, and Bak from the Bcl-2/Bcl-xL inhibitory complex and leading to the induction of apoptosis (46, 53). The DNA ­binding domain of p53 was determined to be the binding interface for Bcl-2, and mutations within the DNA binding domain abolish this binding interaction (68, 69). Consequently, several tumor-derived DNA binding domain missense mutants were found to be unable to promote cytochrome c release, whereas a C-terminally deleted p53 protein preserved the ability to induce apoptosis (70, 71). Yet some recent evidence suggests that p53(R181E) preserves the ability to induce apoptotic cell death, suggesting mutation specificity in terms of apoptogenic potential (72). We demonstrate that p53(A347D), which has a WT DNA binding domain, binds both Bcl-2 and Bcl-xL to an even greater extent than WT p53 does, suggesting that higher levels of the mutant protein basally and after induction alone do not account for mutant p53 cells’ propensity to apoptose. The direct mitochondrial apoptogenic role for p53(A347D) is supported by our finding that PFTμ, which has been shown to attenuate p53 association with mitochondria and preclude p53 binding to Bcl-2 and Bcl-xL, both decreases the amount of p53 found in the mitochondrial fraction and rescues p53^AD/AD cells from etoposide-mediated cell death (52, 73). Not only p53(A347D) but other dimeric p53 mutants may similarly possess transactivation-independent proapoptotic activity, given that ectopically expressed dimer but not monomer mutants were able to sensitize cells to etoposide. This proapoptotic ability may not be a universal feature of dimeric p53, however, as the dimer-forming K351N mutant demonstrates defective mitochondrial trafficking in ovarian cancer cells and attenuated p53-dependent mitochondrial fragmentation in yeast (74). Furthermore, our observation that not all inducers of genotoxic stress can promote heightened cell death in mutant p53 cells indicates specificity in terms of what can induce p53(A347D) to kill. Ultimately, these data describe p53(A347D) as a mutant tumor suppressor that retains the license to kill but only under specific orders.

Germline mutations in the p53 OD have been found to have variable impacts on clinical outcomes dependent on the quaternary structure the specific p53 mutant may form (18). Although clinical data stratifying survival by monomeric and dimeric p53 status are limited, the association with cancer development is clear. Limited in vitro evidence has demonstrated increased proliferation and colony formation in both monomeric and dimeric p53 mutants expressed ectopically in H1299 cells relative to both WT p53 and tetrameric p53 variants (18). We expand on this finding to observe that endogenously expressed mutant p53 may facilitate augmented proliferation in 2D culture, invasion into an experimentally derived extracellular environment in a 3D setting, and tumor engraftment and growth in ectopic xenograft models. Despite the tumorigenic characteristics imparted by p53(A347D), this mutant still preserves apoptogenic ability, thereby suggesting that chemotherapies may retain efficacy in cancers bearing such a mutant. Although we suspect that the GOF activities of p53(A347D) may be applicable to other transcriptionally inert, dimeric p53 mutants that accumulate to intracellular high levels, whether other OD mutants demonstrate this phenotype remains to be shown. Ultimately, our results serve to delineate the LOF and GOF abilities of the dimer-forming p53(A347D) mutant, which not only hints at the complexities inherent in different p53 mutants but also implicates the efficacy of specific proapoptotic therapeutic strategies.

Dimethyl sulfoxide (Sigma), nutlin-3a (Sigma), etoposide (Sigma), zVAD-FMK (AdooQ Biosciences), daunorubicin (Sigma), camptothecin (Sigma), imidazole ketone erastin (Selleck Chemicals), ferrostatin-1 (Selleck Chem), actinomycin D (Calbiochem), and PFTμ (Calbiochem) were utilized at indicated concentrations and times.

The following commercial primary antibodies, supplemented with 0.02% sodium azide, were used: anti-MDM2 (1:1,000; Cell Signaling Technology, D1V2Z), anti-p53 (1:1,000; Bio-Rad, VMA00019), anti-p53 (1:500; Santa Cruz Biotechnology, sc-126), anti-p53 PAb1620 (1:1,000; EMD Millipore, OP33), anti-p53 PAb240 (1:500; Santa Cruz Biotechnology, sc-99), anti–β-actin (1:5,000; Sigma, A2228), anti-p21 (1:2,000; Cell Signaling Technology, 12D1), anti–Myc-tag (1:1,000; Cell Signaling Technology, 9B11), anti–α-Tubulin (1:1,000; Cell Signaling Technology, 2144S), anti-PARP (1:1,000; Cell Signaling Technology, 9542S), anti–caspase-3 (1:1,000; Cell Signaling Technology, 9662S), anti–COX-IV (1:1,000; Cell Signaling Technology, 4844S), CD24 (BD Biosciences, 555427), CD73 (BD Biosciences, 560847), CD105 (eBioscience, 12-1057), OCT4 (Santa Cruz Biotechnology, sc-9081), and SSEA4 (R&D Systems, FAB1435P). In-house mAb anti-p53 (1:10; mixture of DO1/1801 hybridoma supernatant) was also used. Goat anti-mouse and goat anti-rabbit horseradish peroxidase–conjugated secondary antibodies (1:5,000; Sigma) were utilized. Mouse TruBlot ULTRA: anti-mouse Ig HRP (ROCKLAND, eB144) secondary antibody was also used.

The HA-p53 and FLAG-p53 plasmids were described previously (19). The p53 mutations L330A, E343K, L344A, A347S, A347P, and A347D were introduced into the HA-p53 construct as previously reported using the QuikChange II site-directed mutagenesis kit (Agilent Technologies; ref. 19). Myc-DDK-tagged BCL2 (RC204498), Myc-DDK-tagged BCL2L1 (RC201314), Myc-DDK-tagged PPARα (RC216237), and Myc-DDK-tagged PPARγ (RC201538) were purchased from OriGene Technologies. Myc-ETS1 and Myc-ETS2 were kindly gifted by Dr. Luis A. Martinez (Stony Brook Cancer Center, Department of Pathology, Renaissance School of Medicine, Stony Brook University, Stony Brook, NY).

This study was approved by the National Heart, Lung, and Blood Institute (NHLBI)–NIH Internal Review Board (ClinicalTrials.gov identifiers NCT00406445 and NCT01143454) and the University of Utah Institutional Review Board (IRB #41211, Cancer Genetics Study). The study was conducted in accordance with the U.S. Common Rule and after the subjects provided written informed consent for their participation.

U2OS osteosarcoma cell lines varying in p53 status (U2OS parental; U2OS CRISPR p53 KO, p53^+/AD, p53^AD/AD), primary dermal fibroblasts with either WT p53 (L53-WT) or p53^+/AD (L53-M1, L53-M2), the p53-null lung carcinoma H1299 cell line, and HEK293T cells were cultured in Dulbecco's modified Eagle medium (DMEM) containing 10% fetal bovine serum (FBS) at 37°C in a humidified 5% CO₂ environment. Dermal fibroblasts were obtained from an LFS family in which two siblings harbored the germline p53(A347D) mutation (L53-M1 and L53-M2) and a third sibling harbored WT p53 (L53-WT). Human fibroblasts (L53-M1) from one of the LFS siblings were prepared at the NHLBI–NIH, and the fibroblasts (L53-WT and L53-M2) from the two additional siblings were prepared at the University of Utah using standard protocols. Either WT p53 dermal fibroblasts or L53-M1 fibroblasts were utilized to generate iPSCs, which were then differentiated into MSCs and osteoblasts according to the methods provided below. Single clones were isolated, and patient fibroblast derivatives with mutant p53 are termed LFS-1, LFS-2, LFS-3, LFS-4, and LFS-5. WT p53 derivatives are termed WT-1, WT-2, and WT-3. Cell lines were routinely assessed for Mycoplasma infection using the Lookout Mycoplasma PCR Detection Kit (Sigma).

U2OS p53(A347D) mutant cells were generated using a modification of CRISPR–Cas9 genome-editing technology (75). A 100-base-pair single-stranded oligodeoxynucleotide (ssODN) flanking the cut site and containing the mutation of interest was used as a donor template. U2OS cells (2 × 10⁶) were electroporated with 200 pmol/L of ssODN (IDT), 100 pmol/L of EnGen Cas9 NLS (New England Biolabs), and 120 pmol/L of single-guide RNA (Synthego) using an Amaxa Nucleofector II device. Single-cell clones were selected via limiting dilution, and mutant clones were confirmed by genomic DNA and cDNA sequencing. Three single-cell clones of p53^+/AD cells, two single-cell clones of p53^AD/AD cells, and two single-cell clones of p53 KO cells were pooled after confirming similar growth rates and were used for all experiments with U2OS cells unless otherwise indicated. U2OS p53 KO cells were generated as described previously (19).

Mutant p53 constructs (E343K, L344A, A347S, A347P, and A347D) were cloned into the pMXs-IRES-Puro retroviral expression vector (Cell Biolabs) and cotransfected with packaging plasmids into HEK293T cells to produce viruses. Each virus with either empty vector or mutant p53 constructs was infected into U2OS p53 KO cells and selected by 1 μg/mL of puromycin for at least 1 week.

Somatic cell reprogramming with Sendai viruses (SeV) was described previously (76–79). WT and LFS (L53-M1) fibroblasts were reprogrammed by transducing SeVs expressing Yamanaka four factors (OCT4, SOX2, KLF4, and MYC) using the CytoTune-iPS 2.0 reprogramming kit (Invitrogen). The reprogrammed iPSCs were maintained in human embryonic stem cell (hESC) media (DMEM/F12; Corning) containing 20% (v/v) KnockOut Serum Replacement (Invitrogen), β-mercaptoethanol, nonessential amino acids, L-glutamine, penicillin–streptomycin antibiotics, and 10 ng/mL bFGF for 3 weeks. The iPSC clones with hESC morphology and positive expression of the hESC surface marker SSEA4 were picked and expanded in StemMACS iPSC-Brew XF medium (Miltenyi Biotec). Expression of hESC pluripotency factor OCT4, hESC surface marker SSEA4, and alkaline phosphatase was confirmed by immunofluorescent staining (Supplementary Fig. S9A).

In vitro differentiation of iPSCs to MSCs was performed by an SB-431542/7.5% CO₂-based MSC differentiation method described previously (78, 80). iPSCs were cultured in StemMACS iPSC-Brew XF medium supplemented with 10 μmol/L TGFβ inhibitor SB-431542 and 7.5% CO₂. After culturing for 1 month, the majority of cells acquired an MSC-like spindle-shaped morphology. The differentiated cells were then maintained and expanded in the modified MSC differentiation medium [KnockOut DMEM (Gibco) supplemented with 10% KnockOut serum replacement, β-mercaptoethanol, nonessential amino acids, penicillin–streptomycin, glutamine, 20 ng/mL EGF, and 20 ng/mL bFGF] under regular culture conditions. The expression of MSC markers CD73 and CD105, but not pluripotent stem cell marker CD24, was verified by immunofluorescent staining (Supplementary Fig. S9B). For osteogenic differentiation, 2 × 10⁴ CD73⁺/CD105⁺/CD24⁻ MSCs were plated and cultured in osteogenic differentiation medium (α-MEM supplemented with ­10% FBS, 10 mmol/L β-glycerol phosphate, 200 μmol/L ascorbic acid, and 0.1 μmol/L dexamethasone) for 15 days (76–79).

Cells were lysed with TEB Lysis Buffer (10 mmol/L Tris-HCl pH 7.5, 137 mmol/L NaCl, 10% glycerol, and 1% Nonidet P-40 with 50 nmol/L PMSF and inhibitor cocktail containing 100 μmol/L Benzamidine, 300 μg/μL Leupeptin, 100 mg/mL Bacitracin, and 1 mg/mL a2-macroglobulin). Lysates were cleared by 10,000-rpm centrifugation at 4°C, and protein concentration was measured using the Bradford assay (Bio-Rad). Equivalent lysates were separated by SDS-PAGE and electrotransferred onto nitrocellulose membranes (Bio-Rad) and blocked with phosphate buffer (PBS) containing 0.1% Tween 20 (Sigma) and 5% nonfat dry milk for 30 minutes. The above antibodies were applied. Pierce ECL Western Blotting Substrate (Thermo Fisher) or Immobilon Western Chemiluminescent HRP Substrate (EMD Millipore) was used for protein visualization.

U2OS p53 KO cells were transfected with HA-p53, HA-p53(A347D), HA-p53(R175H), or empty vector pcDNA3 using Lipofectamine 2000 (Thermo-Fisher) according to indicated combinations. Cells were lysed with immunoprecipitation (IP) buffer [150 mmol/L NaCl, 50 mmol/L Tris-HCl (pH 8.0), 1 mmol/L EDTA, 1% NP40, 10% glycerol, protease inhibitor]. Protein lysates and Protein G Sepharose 4 Fast Flow beads (Cytiva) were incubated with anti-p53 PAb1620, PAb240, or DO-1 (Santa Cruz Biotechnology) overnight at 4°C. Samples were washed with IP buffer three times and subjected to immunoblotting with anti-p53 (DO-1, Santa Cruz Biotechnology), anti–Myc-tag, or anti–β-actin antibodies.

H1299 cells were cotransfected with HA-p53, HA-p53(A347D), HA-p53(R175H), or the empty vector pcDNA3 and/or Myc-ETS1 and Myc-ETS2 using Lipofectamine 2000 (Thermo Fisher) according to indicated combinations. Cells were lysed with IP buffer [150 mmol/L NaCl, 50 mmol/L Tris-HCl (pH 8.0 or 7.4), 1 mmol/L EDTA, 1% NP40, 10% glycerol, protease inhibitor]. Protein lysates were incubated with Pierce Anti-HA Agarose (Fisher Scientific) overnight at 4°C. Samples were washed with IP buffer three times and subjected to immunoblotting with anti-Myc-tag, anti-p53 (DO-1), or anti–β-actin antibodies.

U2OS p53 KO cells were cotransfected with indicated plasmids and combinations using Lipofectamine 2000 (Thermo Fisher) and treated with 10 μmol/L etoposide for 6 hours. Cells were lysed with IP buffer [150 mmol/L NaCl, 50 mmol/L Tris-HCl (pH 7.4), 1 mmol/L EDTA, 1% NP40, 10% glycerol, protease inhibitor], and protein lysates were incubated with Pierce Anti-HA Agarose (Fisher Scientific) overnight at 4°C. Samples were washed with IP buffer three times and subjected to immunoblotting with anti–Myc-tag, anti-p53 (DO-1), or anti–β-actin antibodies.

Following harvesting cells by scraping with PBS, cells were resuspended in phosphate buffer (PBS + 10% glycerol + 10 mmol/L EDTA + 0.5% NP-40, 0.1 mol/L KCl, protease inhibitor mixture), incubated on a rolling shaker at room temperature for 20 minutes in the presence or absence of glutaraldehyde at indicated concentrations (Sigma). Samples were then boiled with protein sample buffer containing SDS and subjected to immunoblot analysis as described previously.

Cells were treated with 100 μg/mL cycloheximide (Sigma) and harvested at the indicated time points. Cell lysates were subjected to immunoblotting with the indicated antibodies. Signal intensity of p53 and actin was measured using ImageJ software (version 1.53K). The p53 signal was normalized to the actin signal.

RNA extraction was performed using the RNeasy Mini Kit (Qiagen) according to the manufacturer's protocol, and concentration was measured using the NanoDrop 2000 (Thermo Fisher). QuantiTect Reverse Transcription Kit (Qiagen) was used to generate cDNA. Power SYBR Green Master Mix (Thermo Fisher) was mixed with cDNA and following real-time PCR primers. The RPL32 gene was used as a housekeeping control. Quantitative PCR was performed with the StepOnePlus Real-Time PCR System (Applied Biosystems) and quantified by double delta Ct analysis.

The following qRT-PCR primers were used: RPL32_F 5′-TTCCTGGTC CACAACGTCAAG-3′, RPL32_R 5′-TGTGAGCGATCTCGGCAC-3′, MDM2_F 5′-TTGGCGTGCCAAGCTTCTCT-3′, MDM2_R 5′-TACCTGAGTCCGATGATTCC-3′, CDKN1A_F 5′-GGCGGCAGACCAGCATGACA-3′, CDKN1A_R 5′-GCAGGGGGCGGCCAGGGTAT-3′, TIGAR_F 5′-CCAGGCTCGCAGCTTCA-3′, TIGAR_R 5′-GGTTTCGACTCCAGGTGCAA-3′.

Bulk RNA was extracted from cells grown in 15-cm plates using TRIzol (Invitrogen) and the RNeasy Mini Kit (Qiagen). Cells grown on separate 15-cm plates were considered biological replicates. RNA-seq was performed on RNA libraries prepared from two biologically independent experiments in U2OS cells. The first prep contained U2OS parental and U2OS CRISPR p53 KO, p53^+/AD, and p53^AD/AD cells treated with either DMSO or 20 μmol/L etoposide for 24 hours with four biological replicates per condition. The second prep contained U2OS parental and U2OS CRISPR p53 KO, and p53^AD/AD cells treated with either DMSO or 20 μmol/L etoposide for 24 hours with two biological replicates per condition. Additionally, bulk RNA was prepared from three clones of LFS osteoblasts and two clones of WT osteoblasts grown in 10-cm plates. U2OS RNA samples from the first prep were sent in TRIzol to BGI Genomics for further sample preparation, quality control (QC), and sequencing. U2OS RNA samples from the second prep and osteoblast RNA samples were sent to the Cancer Research & Prevention Institute of Texas (CPRIT) UTHealth Cancer Genomics Center for further library preparation, QC, and sequencing.

Resulting .fastq files were aligned to human genome reference hg38 using the kallisto pseudoalignment and quantification algorithm, as described by Bray and colleagues (81). Quantification of RNA read counts was performed at the gene level for downstream analysis. No outliers were observed with respect to total sequencing depth or gene coverage. In order to perform an integrated analysis of expression data from two separate library preps performed in U2OS cells, ComBat-Seq was utilized to correct for batch effect (82). Using the DESeq2 analysis package (83), the combined read-count matrix of all samples annotated by treatment and genotype was converted to a variance-stabilized gene expression matrix, and projected into 2D visualization by uniform manifold projection. On this projection, replicates of the same genotype and treatment condition can be observed to cocluster, without concerning outliers. The major axis of separation between datapoints was found to be etoposide treatment, though within each treatment arm, cosegregation of gene expression patterns is seen corresponding to genotype.

Within each treatment arm, pairwise differential gene expression was computed by the DESeq2 negative binomial test for p53^AD/AD versus p53^+/+ genotype, p53 KO versus p53^+/+ genotype, and p53^AD/AD versus p53 KO genotype, with adjustment of P values for multiple hypothesis testing by Benjamini–Hochberg correction. Differential gene expression for LFS OB versus WT OB was also computed similarly. The full data tables ranked by log-fold-change are provided in the Supplementary Data files.

In order to highlight biological pathways in addition to individual genes of interest, pathway enrichment was performed by GSEA for each gene set described above on the WebGestalt online platform (84) using the entire set of biological pathways defined in the Human Wikipathways database (85). Enriched pathways were ranked by enrichment score, and the weighted set cover was used to reduce redundancies. Pathways that were found to be positively or negatively enriched in indicated datasets were plotted as a dot plot, with relative dot size indicative of enrichment score and relative color representing the FDR. All analyses were performed in the R version 4.1.2 statistical computing environment. Additionally, GSEAs for the set of differentially expressed genes in LFS osteoblasts versus WT osteoblasts were performed with the GSEA software (86) using the p53 signaling pathway defined in the KEGG Pathway Database (87).

ChIP experiments were performed as previously described (88). U2OS cells with either p53^+/+ or p53^AD/AD grown under indicated conditions were harvested and lysed with RIPA buffer [150 mmol/L NaCl, 1% NP-40, 0.5% deoxycholate, 0.1% SDS, 50 mmol/L Tris-HCl (pH 8.0), 5 mmol/L EDTA, and protease inhibitors] and incubated on ice for 20 minutes. Lysates were sonicated to generate DNA fragments with an average length of ∼500 to 1,000 bp. Following sonication, samples were incubated overnight with indicated antibodies and protein A/G beads were preblocked with 1 mg/mL BSA and 0.3 mg/mL salmon sperm DNA. The following antibodies were utilized for ChIP: anti-p53 (DO-1, Santa Cruz Biotechnology, sc-126) and rabbit IgG (Diagenode, C15410206). DNA was purified using the QIAquick PCR Purification Kit (Qiagen). Quantitative ChIP was performed on a StepOnePlus Real-Time PCR system (Thermo Fisher) using Power SYBR Green Master Mix (Thermo Fisher) versus genomic standard DNA and input DNA. ChIP primers designed with Primer Express (Applied Biosystems) were derived from the UCSC Human Genome Browser hg19 assembly, and sequence specificity was confirmed with the UCSC Human Genome Browser in silico PCR tool. Primer sequences are provided below.

The following ChIP-qPCR primers were utilized: CDKN1A_F 5′-GTGGCTCTGATTGGCTTTCTG-3′, CDKN1A_R 5′-CTGAAAACAGGCAGCCCAAG-3′, PUMA_F 5′-GAACGCCCGTCGGTCGGTCTGT-3′, PUMA_R 5′-CAAGTCAGGACTTGCAGGCGC-3′.

Two biologically independent ChIP-seq experiments were performed in U2OS cells. The first contained two biological replicates of U2OS p53^+/+, p53^+/AD, and p53^AD/AD cells and one replicate of U2OS p53 KO cells as a control. The second experiment contained two biological replicates of U2OS p53^+/+ and p53^AD/AD cells treated with either DMSO or 20 μmol/L etoposide for 24 hours and one replicate of U2OS p53 KO cells treated under the same conditions as a control. Additionally, ChIP-seq was performed on two clones of WT osteoblasts (WT-1 and WT-2) and two clones of LFS osteoblasts (LFS-1 and LFS-2).

p53 ChIP-seq was performed using modified previous methods (78). Cells were fixed in 1% formal­dehyde at room temperature for 10 minutes and quenched by 125 mmol/L glycine. Collected cells were sonicated in ChIP lysis buffer [50 mmol/L HEPES-KOH (pH 7.5), 140 mmol/L NaCl, 1 mmol/L EDTA, 0.1% Na-Deoxycholate, 0.1% SDS, 1% Triton X-100, protease inhibitor] on ice. Anti-p53 antibody (Santa Cruz Biotechnology, sc-126) was added to lysate supernatants for overnight incubation at 4°C. The cell lysates and antibody mixtures were diluted by the same amount of ChIP dilution buffer [50 mmol/L HEPES-KOH (pH 7.5), 140 mmol/L NaCl, 1 mmol/L EDTA, 1% Triton X-100] and incubated with magnetic protein G Dynabeads (Thermo Fisher Scientific, 10004D) for 3 hours at 4°C. Immunoprecipitates on beads were washed with ChIP high-salt buffer [20 mmol/L Tris-HCl (pH 8.0), 500 mmol/L NaCl, 2 mmol/L EDTA, 0.1% SDS, 1% Triton X-100] and eluted in TE buffer. DNA isolated from the immunoprecipitates was used for library preparation and sequencing at the CPRIT UTHealth Cancer Genomics Center.

ChIP-sequencing data analyses were conducted at Galaxy (usegalaxy.org) following the tutorials provided by Galaxy Training! In brief, .fastq files were trimmed by Trim Galore! and mapped to human reference genome (hg19) by Bowtie2. PCR duplicates were removed by SAMtools RmDup before MACS2 was used for calling peaks. Unique peaks were identified by bedtools intersect. Heat maps were generated using deepTools (bamCoverage, computeMatrix, and plotHeatmap). Peak annotation and motif prediction were performed using HOMER software (89). Identification of enriched biological processes (Fig. 3D) was performed using EnrichR. GSEA for Supplementary Fig. S3B and S3C was performed using default KEGG pathway gene sets with the metric for ranking genes as “Signal2Noise” and the enrichment statistic as “classic.” GSEA for Supplementary Fig. S5H was performed using default U2OS p53^+/+, p53^+/AD, and p53^AD/AD ChIP targets as gene sets with the metric for ranking genes as “Diff_of_Classes” and the enrichment statistic as “classic.” The analyzed results were considered significant when the FDR q value was less than 0.25 and the nominal P value was less than 0.05.

U2OS cells (3 × 10⁵) varying in p53 status were seeded in 6-well plates, allowed to adhere overnight, and transfected with SMARTpool sip53 (Horizon Discovery) and DharmaFECT1 Transfection Reagent for 48 hours. Cells were then split 1:2 into 6-well plates containing coverslips, let to adhere overnight, and treated with either DMSO or 10 μmol/L etoposide for 24 hours. Treated cells were washed twice with 1× PBS, and MitoTracker Red CMXRos was added in a final concentration of 100 nmol/L. Cells were incubated at 37°C for 30 minutes, and later fixed with 4% paraformaldehyde (Thermo Fisher) for 1 hour, washed twice with 1× PBS, and permeabilized with PBS/0.5% Triton X-100 (Sigma) for 15 minutes, followed by blocking with 2% goat serum (Sigma) for 30 minutes at room temperature. Immunofluorescence staining was performed by adding 100 μL of diluted (1:1,000) primary antibody [mouse anti-p53 (DO-1)] solution to coverslips. After 1-hour incubation at room temperature, coverslips were washed with 1× PBS and then incubated for 1 hour with 100 μL of diluted (1:100) secondary antibody (Alexa Fluor 488 goat anti-rabbit or Alexa Fluor 594 goat anti-mouse antibodies; Life Technologies). Coverslips were washed three times with 1× PBS after each of the above steps and mounted on a microscope slide with Vectashield (Mounting Medium with DAPI; Vector Laboratories), and images were analyzed by confocal laser scanning microscopy (Zeiss LSM 800). Quantification of the percentage of cells with aberrant mitochondria was performed by a trained observer using blinded and randomized files. Quantification of mitochondrial area/cell was performed in ImageJ. Mitochondrial network visualization was also performed for U2OS p53 KO cells stably expressing either vector or p53(A347D) with U2OS p53^AD/AD cells in the same experiment as a comparison according to the above protocol.

Mitochondrial and cytosolic fractions were isolated with the Pierce Mitochondria Isolation Kit for Culture Cells (Thermo Fisher) according to the manufacturer's protocol. Signal intensity of p53 and COX-IV was measured using ImageJ software (version 1.53K). The p53 signal was normalized to the COX-IV signal.

U2OS parental and U2OS CRISPR cells (p53 KO, p53^+/AD, and p53^AD/AD) were plated at a density of 10,000 cells/well in Seahorse XFp cartridges the day prior to analysis. Seahorse XFp Glycolytic Rate Assay and ATP Production Assay were performed according to the manufacturer's protocol. To assess GSH:GSSG ratios, cells varying in p53 status were plated in triplicate at 1 × 10⁴ cells/well in 96-well plates. After cells adhered to the plate overnight, media containing 0.1% FBS and either 20 μmol/L etoposide or no treatment were added for 24 hours. Reduced and oxidized glutathione levels were assessed using the GSH/GSSG-Glo Assay (Promega) according to the manufacturer's protocol. Lysates were transferred to a white-walled 96-well plate prior to luminescence detection.

Cells (5 × 10³) were plated in a 96-well plate (Corning) in an appropriate growth medium and allowed to adhere overnight. Cells were then treated with indicated drugs for 48 hours. The Neutral Red Cell Viability Assay (BioVision) was utilized to measure cell viability according to the manufacturer's protocol.

U2OS parental and U2OS CRISPR cells (p53 KO, p53^+/AD, and p53^AD/AD) were seeded in 6-well plates in triplicate at an appropriate density to yield subconfluent plates at harvest and let adhere to plate overnight. Cells were treated with either DMSO or 20 μmol/L etoposide for 24 hours, collected with trypsin, and counted. ∼5 × 10⁵ cells were resuspended in 200 μL cold PBS, after which 200 μL cold 100% ethanol was added to fix and permeabilize cells. Cells were incubated for 30 minutes on ice, washed once with cold PBS, and resuspended in 400 μL propidium iodide staining solution (50 μg/mL ribonuclease A and 62.5 μg/mL propidium iodide (Sigma) in PBS). Cells were transferred to FACS tubes and incubated at 37°C for 15 minutes protected from light. Samples were immediately analyzed with the FACSCelesta flow cytometer (Becton Dickinson) using a 488-nm laser and collecting fluorescence emission with a 610/20 bandpass filter. FlowJo software (Becton Dickinson) was used to determine the percentage of cells at each stage of the cell cycle.

U2OS parental and U2OS CRISPR cells (p53 KO, p53^+/AD, p53^AD/AD) were seeded at an initial density of 2.5 × 10⁵ cells/well in a 6-well plate with three technical replicates/cell line. Cells were trypsinized and counted with the Countess 2 FL Automated Cell Counter.

Accutase was purchased from MP Biomedicals. Pepsin-treated (PT) bovine type I collagen was obtained from Advanced BioMatrix as a 5.9 to 6.1 mg/mL solution. Growth ­factor–reduced, Phenol Red-free BME/Matrigel was obtained as an 8.9 to 10 mg/mL solution from BD Biosciences. DMEM solution (10 ×), NaOH (1 N), and sodium bicarbonate solution (7.5%) were purchased from Sigma-Aldrich and sterile-filtered before use. Gibco 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) buffer (1 M) was obtained from Invitrogen and sterile-filtered before use. CellMask membrane dye was obtained from Molecular Probes by Thermo Fisher Scientific.

Spheroids were formed from 2,000 cells per spheroid in an appropriate growth medium using a centrifugation method described previously (90) using ultralow attachment 96-well U-bottom Nunclon Sphera plates (Thermo-Fisher Scientific). Spheroids and organoids were allowed to form for 24 hours at 37°C under 5% carbon dioxide.

Collagen solution (1 mg/mL) was prepared by diluting the high-concentration PT collagen monomer stock solution. Appropriate amounts of collagen stock solution were prepared with 10% (v/v) 10 × DMEM, 2.5% (v/v) HEPES buffer, 2.5% (v/v) sodium bicarbonate, and distilled deionized H₂O. To prevent the self-assembly of collagen monomers, all solutions were held and mixed at 4°C. NaOH was added to adjust the pH to 7.4.

To prepare gels loaded with a single spheroid, 1 mg/mL collagen solution was prepared as described above at 4°C to prevent gelation. Neutralized collagen (200 μL) was added to a chamber consisting of a 5-mm glass cylinder glued to a coverslip-bottom cell culture dish. A nylon mesh was placed on the inner circumference of the cylinder to anchor the collagen gel. The spheroid was added to the liquid collagen in 5 μL growth medium. The chamber was then transferred immediately to a 37°C incubator. The collagen gels were overlaid with 50 μL growth medium after 1 hour and surrounded by additional liquid to prevent drying during extended incubation periods. To prepare collagen gels loaded with dispersed cells, the 1 mg/mL collagen solution was prepared omitting 50% of the water and neutralized at 4°C. The water volume was substituted with ice-cold growth medium containing the desired number of cells (20,000–40,000 cells/mL). Subsequently, 100 μL of the cell-loaded collagen or collagen/BME was added to the chamber and gelled at 37°C as described above for spheroid-loaded collagen gels. After 1 hour, gels were overlaid with 30 μL growth medium. In all cell experiments, time points are described considering time zero (t = 0 minutes) to be the time at which the cells or spheroids were placed into the collagen or collagen/BME solution.

For invasion analysis, spheroids were generated from cells fluorescently labeled with CellTracker Orange (7 μmol/L dye for 40 minutes at 37°C) and imaged at 10× magnification using transmitted light microscopy at 1 hour and fluorescent confocal imaging at 24 hours after implantation (z-stacks from 80–120 μm total with step size of 10 μm). We used these magnified fluorescence images to determine the invaded area, defined as the difference between the area of the spheroid at t = 1 hour and an ellipse that circumscribes 90% of the invasive cells at t = 24 hours. Maximum intensity projections were thresholded and used to generate masks from which the noninvading spherical core of the spheroid, which is well discernible under these matrix conditions, was removed. The remaining structures, termed invasive structures, were analyzed regarding their area that reflects how far the cells have invaded. The total area of invasive structures per spheroid was used to assess invasion efficiency.

All mouse experiments were reviewed and approved by the UTHealth Houston Animal Welfare Committee following Institutional Animal Care and Use Committee guidelines. Tumor xenografts were performed as described previously (77, 78, 91). WT and LFS iPSC-derived osteoblasts (1 × 10⁷) were injected subcutaneously into the right and left flanks of 7- to 8-week-old NU/NU mice. After 3-month inoculation, mice were sacrificed, and tumors were extracted and weighed. For U2OS cells, 1 × 10⁷ parental and U2OS CRISPR pools (p53 KO, p53^+/AD, and p53^AD/AD) were injected subcutaneously into the right and left flanks of 7- to 8-week-old NU/NU mice. Mice were sacrificed, and tumors were extracted and weighed at the indicated time.

Statistical analyses were conducted in Excel and GraphPad Prism 7. Results, unless indicated otherwise, are expressed as mean ± SEM. A P value of <0.05 determined via two-tailed Student t test of unknown variance was considered statistically significant for hypothesis testing. The Benjamini–Hochberg procedure was utilized to correct for multiple hypotheses where appropriate. The following format was used to denote significance based on P value: P < 0.05 (*), P < 0.01 (**), P < 0.001 (***).

The SuperSeries containing all four sequencing experiments as well as each individual sequencing experiment have been deposited to the Gene Expression Omnibus (GEO). The GEO accession number for the SuperSeries is GSE226407.

T. Kawase reports personal fees from Astellas Pharma Inc. outside the submitted work. J.D. Schiffman reports other support from ItRunsInMyFamily.com and Peel Therapeutics, Inc. outside the submitted work. C. Prives reports grants from the NIH during the conduct of the study. No disclosures were reported by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

J.H. Choe: Conceptualization, data curation, formal analysis, supervision, investigation, methodology, writing–original draft, writing–review and editing. T. Kawase: Formal analysis, validation, investigation, visualization, methodology, writing–review and editing. A. Xu: Data curation, formal analysis, validation, investigation, visualization. A. Guzman: Formal analysis, investigation. A.Z. Obradovic: Data curation, formal analysis, visualization. A.M. Low-Calle: Investi­gation. B. Alaghebandan: Investigation. A. Raghavan: Investigation. K. Long: Investigation. P.M. Hwang: Resources. J.D. Schiffman: Resources. Y. Zhu: Resources, investigation. R. Zhao: Formal analysis, supervision, investigation, methodology. D.-F. Lee: Resources, supervision, funding acquisition, methodology, writing–review and editing. C. Katz: Conceptualization, formal analysis, supervision, investigation, visualization. C. Prives: Conceptualization, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

We are grateful to David R. Tong for his invaluable generation of the U2OS CRISPR cell lines used in this study. David passed away in 2022 and is deeply missed for his contributions to this study and to the Prives laboratory in general. We are grateful to the members of the Prives and Lee laboratories for helpful discussions. Ella Freulich, Ruoyu Wang, and Ping Xu are thanked for expert technical assistance. Ping-Yuan Wang is thanked for help with key protocols. J.D. Schiffman is the Helen Clise Presidential Endowed Chair in Li–Fraumeni Syndrome Research at Huntsman Cancer Institute at the University of Utah. This work was supported by NIH/NCI grant R35CA220526 to C. Prives, NIH/NCI grant R01CA246130 to D.-F. Lee and R. Zhao, and NIH/NCI grant R15CA213426 to Y. Zhu. A. Xu was a CPRIT Postdoctoral Fellow in the Biomedical Informatics, Genomics and Translational Cancer Research Training Program (BIG-TCR; CPRIT grant RP210045). The UTHealth Cancer Genomics Core is supported by CPRIT (CPRIT RP180734). Research reported in this publication utilized the Genetic Counseling and Biorepository and Molecular Pathology Shared Resources at Huntsman Cancer Institute at the University of Utah and was supported by NIH/NCI grant P30CA042014.

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.