In cancer, missense mutations in the DNA-binding domain of TP53 are common. They abrogate canonical p53 activity and frequently confer gain-of-oncogenic function (GOF) through localization of transcriptionally active mutant p53 to noncanonical genes. We found that several recurring p53 mutations exhibit a sex difference in frequency in patients with glioblastoma (GBM). In vitro and in vivo analysis of three mutations, p53R172H, p53Y202C, and p53Y217C, revealed unique interactions between cellular sex and p53 GOF mutations that determined each mutation's ability to transform male versus female primary mouse astrocytes. These phenotypic differences were correlated with sex- and p53 mutation–specific patterns of genomic localization to the transcriptional start sites of upregulated genes belonging to core cancer pathways. The promoter regions of these genes exhibited a sex difference in enrichment for different transcription factor DNA-binding motifs. Together, our data establish a novel mechanism for sex-specific mutant p53 GOF activity in GBM with implications for all cancer.
Sex differences in cancer, including glioblastoma, have been observed in both incidence and outcome. We reveal that TP53, the most commonly mutated gene in cancer, contributes to sex differences through differential GOF activity. This discovery has critical implications for our understanding of p53 mutations and the importance of sex as a biological variable.
The tumor suppressor TP53 (p53), which regulates cell-cycle progression, DNA damage repair, apoptosis, and stem cell differentiation (1), is the most commonly mutated gene in cancer (2). It functions through direct protein–protein interactions and acting as a positive or negative regulator of transcription through its canonical DNA-binding site (3, 4). p53’s transcriptional activity is dependent on many interacting and context-specific factors such as the tissue or cell type, the mechanism of p53 activation, and the severity and longevity of DNA damage (5). A growing body of evidence indicates that sex has a profound influence on the role of p53 in development, disease, and aging. In mice and rats, loss of p53 in utero results in decreased female progeny compared with males – as a consequence of aberrant X-inactivation in females and altered X-gene dosage – which result in embryonically lethal neural tube defects (6, 7).
p53 also exhibits sex differences in the maintenance of neural progenitor cells (NPC) in the subventricular zone (8). Under normal conditions, post pubertal male mice NPCs are depleted faster than in female mice. Knocking out p53 in NPCs protects against depletion in males suggesting a male-specific proapoptotic function of p53 in NPCs (8).
Regardless of the geographic region, human females consistently exhibit greater longevity than males and other mammals (9). Work in Drosophila has shown that germline overexpression of wild-type (WT) p53 decreased lifespan in females and increased lifespan in males. Tissue-specific overexpression of WT p53 in the central nervous system had the opposite effect, increasing lifespan in females while decreasing lifespan in males (10). Together, these findings demonstrate that p53 activity is both sex and tissue dependent.
Sex differences in the effects of p53 mutations have also been observed in cancer. Overall, in nonreproductive tumors, p53 mutations occur more frequently in males, and are associated with worse survival (11). However, in individuals with Li-Fraumeni syndrome (LFS), a familial cancer predisposition syndrome associated with germline pathogenic variants in p53, females have an overall increased risk of developing cancer compared with males (12). For some other cancers, such as brain tumors, males with LFS exhibit higher risk. Mice harboring a germline mutant p53 and loss of the promyelocytic leukemia gene, exhibit sex differences in cancer type and survival. Male mice develop more soft tissue sarcomas and have shortened survival compared with females, who are more likely to develop osteosarcomas and exhibit longer survival (13). In a model of glioma, we previously observed that codeletion of p53 and the tumor suppressor neurofibromin 1 (Nf1) is sufficient for transformation of male, but not female astrocytes (14). Knocking out Nf1 and p53 in NPCs in utero also leads to faster glioma formation and disease progression in male mice than female mice (15).
The most common p53 alterations in cancer are missense mutations in the DNA-binding domain (DBD; ref. 16). In addition to the loss of canonical DNA binding and gene regulation, many p53 missense mutations exhibit gain-of-function (GOF) and are more oncogenic than TP53 deletion (17). Many of these neomorphic functions can be attributed to p53–DNA interactions at noncanonical binding sites that drive the aberrant expression of oncogenes and repression of tumor suppressors (18). The specific activity of different p53 mutations has been shown to be mediated by novel interactions with other proteins (such as NF-κB, ETS2, and NF-Y) and differential genomic localization (19). Despite efforts to specify GOF-mutant p53 phenotypes, characterizing the effects of mutant p53 GOF on transcription and tumorigenicity has been complicated by disparities in reported results, suggesting that, much like WTp53, mutant p53 activity, even for the same missense mutation, is diverse and context dependent.
Given the known sex differences in p53 function and cancer incidence and survival, it is important to consider whether p53 GOF mutations contribute to sex differences in cancer (20). Many of the pathways that contribute to sex differences including metabolism, cell-cycle progression, invasion and epithelial-to-mesenchymal transition, epigenetic dysregulation, and immunity (20), are associated with mutant p53 GOF activity. However, the influence of sex on mutant p53 GOF activity remains unknown. In glioblastoma (GBM), males have been shown to have both increased incidence and decreased overall survival compared with females (21). Notably, TP53 is among the most mutated genes in GBM with most mutations occurring as missense mutations in the DBD (22). Therefore, the intersection of sex differences in p53 function, sex difference in GBM incidence and outcome, and the high rate of p53 missense mutations in GBM make this an ideal system for interrogating the effects of sex on p53 GOF activity with potential applicability to p53 GOF mutations in all cancer.
Materials and Methods
Animal Studies Approval
Study was approved in accordance with an animal studies protocol (no. 2018205) approved by the Animal Studies Committee of Washington University School of Medicine per the recommendations of the NIH Guide for the Care and Use of Laboratory Animals.
Analysis of Missense Mutations Across Cancer
TP53 mutation data was aggregated from multiple studies (Supplementary Table S1). The frequency of missense mutations in the DBD was calculated as a fraction of the total p53 mutations in the dataset. Significance was calculated using Fisher exact test between each cancer type and the pan-cancer data set.
Meta-analysis of Mutation Incidence in Glioblastoma
TP53 mutation data was collected for 552 male and 442 female patients from The Cancer Genome Atlas (TCGA 2013), the Catalog of Somatic Mutations in Cancer (COSMIC– ALL CNS), and International Agency for Research on Cancer (IARC) datasets. The absolute number of mutations at each codon in the TP53 gene was quantified for each sex independently, and the sex ratio determine by dividing the number of male tumors with a mutation by female tumors with a mutation for each codon. Significant sex differences in mutation frequency were determined for each codon using Fisher exact test (P < 0.05).
Primary Mouse Astrocyte Isolation and Culture
Trp53f/− pups were produced by mating male Trp53−/− mice (B6.129S2-Trp53tm1Tyj/J – The Jackson Laboratory #002101) with Trp53f/f mice (B6.129P2-Trp53tm1Brn/J – The Jackson Laboratory #008462). Primary astrocytes were isolated from postnatal day 1 pups as previously described (14). The sex of the pups was determined by PCR of genomic DNA for Jarid1c/Jarid1d14.
Astrocytes from at least three male and three female pups were combined for each male and female.
Astrocytes were plated on poly-l-lysine (ScienCell) coated coverslips and fixed with 3.2% paraformaldehyde for 10 minutes at room temperature. Fixed cells were immunolabeled for GFAP and p53 (Methods, Table 1). Nuclei were counterstained with DAPI and mounted on microscope slides using Immu-Mount (Thermo Fisher Scientific). Images were taken using a fluorescent microscope (Olympus Bx60) and processed using Zeis Zen 3.1 software.
|Antibody .||Antigen .||Host .||Source .||Application .||Dilution .|
|2524||Mouse p53||Mouse||Cell Signaling Technology||Western blot||1:1,000|
|ChIP||5 μg antibody/25 μg DNA|
|SC-365062||GAPDH||Mouse||Santa Cruz||Western blot||1:200|
|E0431||Rabbit Immunoglobulin||Swine||Dako||IHC – Secondary/Biotin||1:750|
|926-32212||Mouse IgG||Donkey||LI-COR||Western blot – Secondary/IRDye 800CW||1:50,000|
|A11077||Rat IgG||Goat||Invitrogen||ICC – Secondary/Fluor 568||1:1,000|
|Antibody .||Antigen .||Host .||Source .||Application .||Dilution .|
|2524||Mouse p53||Mouse||Cell Signaling Technology||Western blot||1:1,000|
|ChIP||5 μg antibody/25 μg DNA|
|SC-365062||GAPDH||Mouse||Santa Cruz||Western blot||1:200|
|E0431||Rabbit Immunoglobulin||Swine||Dako||IHC – Secondary/Biotin||1:750|
|926-32212||Mouse IgG||Donkey||LI-COR||Western blot – Secondary/IRDye 800CW||1:50,000|
|A11077||Rat IgG||Goat||Invitrogen||ICC – Secondary/Fluor 568||1:1,000|
DNA Plasmids, Cloning, and Virus Production
Mouse Trp53 transcript variant 1 (NM_011640.3) was cloned into the retroviral vector pMSCV-IRES-GFP II (Addgene plasmid #52107). Trp53 point mutations were introduced using site-directed mutagenesis with the Q5 Site-Directed Mutagenesis Kit (New England Biolabs E0552S). Retrovirus was produced by plasmid transfection into Platinum-E (Plat-E) cells. Lentivirus to express Cre recombinase was produced by transfecting NIH HEK293T cells with Cre-IRES-PuroR plasmid (Addgene plasmid #30205), Δ8.9 and VSV-G. All transfections were performed with FuGENE 6 transfection agent (Promega E2691) in Opti-MEM Reduced Serum Medium (Gibco 31985070) for 5 hours, and the media replaced with standard DMEM/F12, 10% FBS for viral production. Virus was collected 48, 72, and 96 hours posttransfection and combined.
Primary Astrocyte Genomic Alteration
Trp53mutantp53/− astrocytes were produced by transduction with mutant p53-IRES-eGFP retrovirus followed by subsequent transduction with Cre-IRES-Puro lentivirus. Briefly, postnatal day 1 Trp53f/− astrocytes were transduced with a 1:1 ratio of viral media and growth media for 48 hours. Following stable expression of eGFP, cells were sorted by flow-sort for eGFP-positive cells. Sorted astrocytes were transduced with a 1:1 ratio of lentiviral-Cre-IRES-puro virus and growth media for 48 hours. Astrocytes were then selected for 1 week in 2.5 μg/mL puromycin (Sigma P8833). Trp53KO astrocytes was achieved by transduction with lentiviral-Cre-IRES-puro virus and selection for 1 week in 2.5 μg/mL puromycin. All primary astrocytes were transfected three passages postisolation. Cells were tested for Mycoplasma contamination by PCR (ATCC #30-1012K). All assays were performed between 4 and 15 passages after puromycin selection.
Western Blot Analysis
Total cell lysates were collected in RIPA buffer supplemented with proteinase and phosphatase inhibitors (Roche 11697498001, Roche 04906937001). Protein was separated by electrophoresis using 4%–12% gradient Bis-Tris NuPAGE gels (Invitrogen) and transferred to a nitrocellulose membrane. Membranes were blocked using Odyssey blocking buffer diluted 1:1 in 0.1% PBST. Membranes were then incubated in primary antibodies diluted in blocking buffer overnight at 4°C, followed by secondary antibodies diluted in blocking buffer for 1 hour at room temperature. Blots were imaged using the Bio-Rad ChemiDoc MP Imager and analyzed using the Bio-Rad Image Lab Software V6.1. Primary and secondary antibodies and their corresponding dilution factors can be found in Methods in Table 1.
In Vitro Growth Assay
3,000 cells were plated in triplicate in 10 mL media in 10-cm tissue culture plates. After incubation in normal tissue culture conditions for 8 days, the cells were washed with 1× PBS and fixed with 100% methanol at room temperature for 6 minutes. Plates were air dried at room temperature. Cells were stained for 1 hour at room temperature with Giemsa stain (Sigma GS500) diluted 1:20 in dH2O. Stained plates were washed three times in dH2O for 5 minutes at room temperature and air dried before imaging using the Imagescanner III (GE Healthcare). Growth was determined by percentage area of the plate covered by cells using ImageJ v1.52a. The global scale was set according to the width of a plate (10 cm), the image flattened, and the color channels split. Using the green channel only, the background was subtracted from each image, and the threshold set to 235 for each image. A circle of equal size was selected with each plate and the total percent area calculated using the “analyze particles” tool. Within each experiment, the percent area for each plate was normalized to the average of the Male KO plates. To test for interactions between sex and genotypes on the normalized growth rates, mixed effect linear regression models were used using the lmerTest R package, with variances controlled for the technical replicates within each plate. Regression terms included sex, the tested genotype, and their interaction.
In Vivo Tumorigenesis: Flank Implantations
Flank tumors were generated by implanting mutant p53 astrocytes subcutaneously into the left- and right-side flanks of NCr nude mice (Taconic). Male and female astrocytes were trypsinized and resuspended in DMEM/F12, and combined at a 1:1 ratio with Matrigel. One-hundred microliters of cell/Matrigel suspension equaling 750,000 male or female cells were injected into opposite flanks of 3 male and 3 female hosts. Flank tumors size was measured weekly using a digital caliper until the largest tumors reached a terminal volume of 1 cm3.
Total RNA was isolated using the RNeasy Mini Kit (Qiagen) from Trp53 KO or Trp53-mutant astrocytes. Samples were prepared according to library kit manufacturer's protocol, indexed, pooled, and sequenced on an Illumina HiSeq. Basecalls and demultiplexing were performed with Illumina's bcl2fastq software and a custom python demultiplexing program with a maximum of one mismatch allowed in the indexing read. RNA-sequencing (RNA-seq) reads were then aligned to the Ensembl release 76 primary assembly with STAR version 2.5.1a (23). Gene counts were derived from the number of uniquely aligned unambiguous reads by Subread:featureCount version 1.4.6-p5 (24). Isoform expression of known Ensembl transcripts were estimated with Salmon version 0.8.2 (25). Sequencing performance was assessed for the total number of aligned reads, total number of uniquely aligned reads, and features detected. The ribosomal fraction, known junction saturation, and read distribution over known gene models were quantified with RseQC version 2.6.2 (26).
All gene counts were then imported into the R/Bioconductor package EdgeR (27) and TMM normalization size factors were calculated to adjust for samples for differences in library size. Ribosomal genes and genes not expressed in the smallest group size minus one sample greater than one count-per-million were excluded from further analysis. The TMM size factors and the matrix of counts were then imported into the R/Bioconductor package Limma (28). Weighted likelihoods based on the observed mean-variance relationship of every gene and sample were then calculated for all samples with the voomWithQualityWeights (29). The performance of all genes was assessed with plots of the residual SD of every gene to their average log-count with a robustly fitted trend line of the residuals. Differential expression analysis was then performed to analyze for differences between conditions and the results were filtered for only those genes with Benjamini–Hochberg false-discovery rate adjusted P values less than or equal to 0.05. Interactions between sex and gene expressions were tested using the limma package.
Sparse Principal Component Analysis
Sparse principal component analysis (PCA) was performed on normalized gene counts using R/SparsePCA v0.1.2 (30) and variance plotted using the R package plot3D v1.3. Normalized gene counts for all genes with nonzero sparse loadings for the each of the first three principal components were used to generate heat maps. Heat maps were generated using R/heatmap3 v1.1.9.
GOF transcriptomes were defined as all genes that were differentially expressed between p53-mutant and p53 KO astrocytes (FDR < 0.05) within each sex. The GOF gene lists were then compared between males and females within each mutation to define the sex-specific and shared GOF transcriptomes. Pathway enrichment analysis was performed on shared and sex-specific differentially expressed gene lists using ShinyGO v0.65 (31). Enrichment analysis is based on hypergeometric distribution followed by FDR correction.
Whole genomic DNA was isolated from 106 cells using the QIAamp DNA Mini Kit (Qiagen 51304). gDNA was fragmented, indexed, and pool. Indexed pools were hybridized with Agilent SureSelectXT Mouse All Exon kit per manufacturer's instructions and sequenced on an Illumina NovaSeq-6000. Reads were analyzed using a DRAGEN Bio-IT processor (version 0 × 18101306) running software version 07.021.602.3.8.4. FASTQ files were mapped to mouse reference mm10 and output in CRAM format with duplicates marked. Hard filtered variants in the mouse exome target region were annotated with snpEff v4.3t.
Chromatin Immunoprecipitation Sequencing
10 million astrocytes were fixed with 1% formaldehyde in DMEM/F12 for 10 minutes are room temperature with gentle rocking. Fixation was quenched by adding glycine to a final concentration of 125 mmol/L. Cells were washed with cold PBS, scraped into 5 mL cold PBS supplemented with protease inhibitors (PI), phosphatase inhibitors (PPI), and phenylmethylsulfonylfluoride (PMSF), and transferred to 15 mL conical tubes. Cell suspension were pelleted by centrifugation at 4°C for 5 minutes at 1,000 × g and resuspended in 1 mL Farnham lysis buffer (plus PI, PPI, and PMSF) for 10 minutes with gentle vortexing every two minutes to isolate nuclei. Nuclei were pelleted by centrifugation at 4°C, for 5 minutes at 2,000 rpm. The nuclear pellet was resuspended in 1 mL RIPA buffer (plus PI, PPI, and PMSF). Sonication was performed on an Epishear probe-in sonicator (Active Motif) at 50% amplitude for 15 cycles of 10 seconds, with 20 seconds of rest between each cycle. Immunoprecipitation was performed as previously described, with an antibody recognizing mouse p53 (see Methods, Table 1).
Sequencing reads were generated using Illumina NovaSeq S4 (2 × 150bp) and processed using the ENCODE Transcription Factor and Histone ChIP-Seq processing pipeline v1.1.6 (http://github.com/ENCODE-DCC/chip-seq-pipeline2). The pipeline filtered and mapped reads to the Mus musculus genome (mm10), validated the quality of the data, and generated fold change signal tracks over the inputs using MACS2. Signal tracks of fold enrichment were visualized with the WashU Epigenome browser (https://epigenomegateway.wustl.edu). Motif search for peaks near gene transcript starting sites was conducted using Hypergeometric Optimization of Motif EnRichment (HOMER, v4.8.3; http://homer.ucsd+.edu/homer/index.html).
IHC of Mouse Tumors
Mouse flank tumors were dissected, drop fixed in 4% paraformaldehyde overnight at 4°C, and transferred to 30% sucrose. Tumors were imbedded in a paraffin block, sectioned, and mounted on microscope slides. Histologic features were determined by hematoxylin and eosin staining. IHC was performed against mouse p53. Antibodies and dilutions are listed in Methods in Table 1. IHC signal was developed with the ImmPACT DAB Peroxidase Kit (Vector Laboratories, SK-4105).
All in vitro experiments were carried out at least three times using cells from three independent thaws and platings. Comparisons between mutation frequencies were determined by Fisher exact test. Two-tailed Student t test was used for direct comparisons between mutants in the growth assay and tumor volumes. Differences in tumor growth rate were determined by two-way ANOVA. A Pe < 0.05 was considered statistically significant. Correlations between mutational burden and proliferation were calculated using Pearson correlation. Statistics were calculated using GraphPad Prism version 9.0.1.
Antibodies used in this study are listed in Table 1
Unique Materials Availability
Male and female astrocyte cultures null for p53 or expressing p53-mutant constructs are primary cultures of limited passage number. These will be freely shared upon request pending availability.
Data Accessibility Statement
All sequencing data are available through GEO via accession number GSE188710: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE188710
Sex Differences in the Incidence of p53 Mutations in GBM
Across cancer, TP53 mutations are most commonly missense mutations in the DBD compared with deletions, inactivating (truncating mutations), or missense mutations outside of the DBD. In patients with LFS, central nervous system (CNS) tumors in particular are significantly associated with germline mutations in the helix-loop-helix region of the DBD, suggesting that these mutations preferentially predispose LFS patients to brain tumors (32). To further investigate the association between mutant p53 in CNS tumors and glioblastoma (GBM), we asked whether this pattern was also present for somatic p53 mutations. TP53 mutation data were aggregated from multiple studies deposited in the cBioPortal database, including the pan-cancer MSK-IMPACT Clinical Sequencing Cohort and all available studies for glioblastoma, pan-CNS cancers, and the top five cancer types by incidence: breast, lung, skin, pancreatic, and colorectal (33). A complete list of studies and cancer subtypes included in this analysis can be found in Supplementary Table S1. We analyzed the frequency of missense mutations in the DBD as a percentage of all p53 mutations (Fig. 1A and B). Despite the DBD contributing 48% of the amino acids in the full-length p53 protein, missense mutations in the DBD account for 61.5% of p53 mutations in the pan-cancer dataset. In comparison with pan-cancer, missense mutations in the DBD are significantly more frequent in GBM (72.1%, P = 0.0001, Fisher exact test), CNS tumors (76.8%, P <0.0001), and colon cancer (66.4%, P = <0.0001), significantly less frequent in breast cancer (56.5%, P < 0.0001), and skin cancer (52.5%, P = 0.0015), and equally common in prostate (61.4%) and lung cancers (61.7%). In total, missense mutations in the DBD were observed in 19.7% of GBM cases, compared with inactivating mutations in 7.6% and complete deletion in 1%. Together, these data support an enhanced impact of missense mutations in the p53 DBD on GBM and CNS tumor development.
We next asked whether the frequency of individual mutations is influenced by patient sex. TP53 mutation data was combined and analyzed from GBM and CNS tumor specimens from The Cancer Genome Atlas (TCGA), the Catalogue of Somatic Mutations in Cancer (COSMIC), and the International Association for Research on Cancer (IARC) p53 database to determine the frequency of mutations at each codon in the DBD (34–36). In total, 554 male tumors and 442 female tumors were included in the analysis. Six codons, present in 10.8% of all cases were identified as having a sex difference in their mutation frequency by Fisher exact test P < 0.05 (Fig. 1C). Four codons were mutated more frequently in females: D184, Y205, V216, and V272, and two codons more frequently in males: Y220 and R282. Notably, both codons more commonly mutated in males are classified as hotspot mutations for their high mutation frequency across cancers (37). Codon R175 is among the best characterized hotspot mutations and has no sex difference in mutation frequency in GBM. The prevailing hypothesis for pan-cancer hotspot mutations is that these p53 GOF mutations confer a fitness advantage that drives their selection. Applying this same principle to our findings, we hypothesized that that p53 missense mutations have differential GOF effects in males and females that contribute to the selective advantage of certain mutations.
Modeling Mutant p53 GOF
To determine whether there are sex differences in p53 GOF activity, we developed a model of mutant p53 overexpression in mouse primary astrocytes. Briefly, astrocytes were harvested as previously described from the cortices of postnatal day one mouse pups harboring one p53 allele with flanking lox-P sites and one deleted p53 allele (14). The pups were sex-typed using the X- and Y-chromosome paralogs, Jarid1C/D. Astrocytes from a minimum of three pups were combined per sex to produce male and female Trp53f/− (p53 WT) astrocytes (Fig. 2A; ref. 14). Astrocyte purity was assessed using glial fibrillary acidic protein (GFAP) immunofluorescence (IF; Fig. 2B). To evaluate potential sex differences in p53 GOF, we selected three mutations: Trp53R172H, Trp53Y202C, and Trp53Y217C for evaluation. Trp53R172H (Hs TP53R175H) is among the most common mutations across cancer and has established GOF (17). It exhibits no sex difference in frequency in GBM. Trp53Y202C (Hs TP53Y205C) is the mutation with the most significant sex difference in frequency in female patients. Trp53Y217C (Hs TP53Y220C) has the most significant sex difference in frequency among the mutations more common in males. Male and female p53 WT astrocytes were transduced with retrovirus to stably overexpress mutant Trp53-IRES-eGFP for each missense mutation, and eGFP-positive cells were sorted by flow cytometry. A common cooccurring event in cancers expressing mutant p53 is the loss of the second wild-type p53 allele, known as loss-of-heterozygosity (LOH). LOH has been shown to be required for stabilized mutant expression and GOF activity in some mutants (38). To mirror these events in our in vitro cell model, WT and mutant p53 astrocytes were transduced with a lentiviral vector expressing Cre recombinase and Puromycin N-acetyl transferase (Pac). The cultures were selected for Cre/Pac-positive cells for one week with 2.5 μmol/L puromycin to produce p53 KO and p53mutant/− astrocytes (Fig. 2C). Figure 2D displays a representative Western blot measuring p53 protein levels in each cell line. Under low stress conditions, WTp53 is negatively regulated by the E3 ubiquitin ligase Mdm2. In response to DNA damage, p53 is phosphorylated inhibiting Mdm2 interactions and stabilizing the p53 protein. As a positive control, p53f/− astrocytes were treated with a combination of the DNA-damaging agent etoposide (10 μg/mL) and the Mdm2 inhibitor Nutlin-3a (10 μmol/L) or DMSO for 8 hours leading to stable induction of p53 protein. p53 KO astrocytes expressing Cre but lacking mutant p53 expression had no detectable p53. In contrast, high levels of overexpressed mutant p53 protein were observed in male and female astrocytes transduced with each of the retrovirus-encoded p53 mutations. To confirm the loss of the WT floxed p53 allele in Cre-positive cells, we performed a PCR with primers specific to the floxed and deleted p53 alleles (Fig. 2E). Introduction of Cre eliminated the floxed allele band indicating the complete loss of p53 at the WT locus. To ensure that WTp53 was completely depleted in the Cre-expressing astrocytes, WTp53 and p53KO astrocytes were treated with etoposide and Nutlin-3a or DMSO for 8 hours and p53 protein was detected by IF. Cotreatment induced high levels of nuclear localized p53 in WT astrocytes but resulted in no detectable p53 in the p53 KO astrocytes (Fig. 2F). IF for p53 in male and female astrocytes expressing mutant p53 confirmed that in the absence of treatment all three mutations expressed high levels p53 protein throughout the cell body (Fig. 2G).
Sex Differences in p53 GOF Phenotype
Accelerated proliferation is among the most frequently observed phenotypes attributed to mutant p53 GOF. In addition to the loss of WTp53 regulation of cell cycle, GOF-mutant p53 interacts with several transcription factors to increase the transcription of pro-proliferation genes (17). Thus, we measured the growth effects of each mutation in both sexes. 3,000 male or female p53 WT, p53 KO, p53R172H, p53Y202C, and p53Y217C astrocytes were plated in triplicate on 10-cm dishes, in three independent experiments, and allowed to expand for eight days before fixing and Giemsa staining nuclei. KO alone increased astrocyte proliferation compared with WT astrocytes but exhibited no detectable sex differences. All three p53 mutations exhibited a GOF growth phenotype in a sex-dependent manner. Expression of p53R172H in females and p53Y202C and p53Y217C in males significantly increased growth in comparison with p53 KO astrocytes and astrocytes expressing the same mutation in the opposite sex (two-way ANOVA, P < 0.05). While p53R172H in males, p53Y202C in females, and p53Y217C females were insufficient to increase proliferation over p53 KO (Fig. 3A–C), male p53R172H and female p53Y202C astrocytes formed denser colonies compared with KO, suggesting that these mutations may possess sex-dependent GOF activity (Fig. 3B). In female cells, p53Y217C expression appeared to decrease the growth phenotype compared with KO. To ensure that this effect was not a technical artifact of the assay, we plated a 5-fold titration of male and female p53 KO and p53Y217C astrocytes and incubated them for five days at which point the fastest growing cells reached overconfluency (Supplementary Fig. S1). This experiment confirmed the previous observation, with female p53Y217C growing slower than p53KO, p53KO exhibiting no sex difference, and male p53Y217C growing the fastest. Together, these data support that p53 GOF effects are dependent on an interaction between sex and mutation. Modeling the interaction between sex and mutation as a function of growth revealed a statistically significant interaction between sex and genotype for each mutation (mixed effect linear regression with variances controlled among technical replicates, p53R172HP = 7.78 × 10−5, p53Y202CP = 0.00413, p53Y217CP = 1.41 × 10−8).
Next, we asked whether the growth phenotypes were correlated with sex differences in tumorigenicity. Equal numbers of each male and female KO and p53-mutant astrocyte lines were injected into parallel flanks of three male and three female NCr nude mice (n = 6) as illustrated in Fig. 3D. Flank tumor volume was measured by digital caliper weekly until the first tumor reached terminal size. Male p53Y217, male p53Y202C, and female p53R172H tumors grew significantly faster than p53 KO tumors (two-way ANOVA, P < 0.05; Fig. 3E). When comparing between sexes, male p53Y217C and female p53R172H tumors grew significantly faster than the same mutation in the opposite sex, while the effects of neither p53Y202C nor p53 KO were significantly affected by sex (Fig. 3E inset, two-way ANOVA, P < 0.05). In total, 0/6 KO, 16.7% (1/6) of male and 83.3% (5/6) of female p53R172H, 66.6% (4/6) of male and 50% (3/6) of female p53Y202C, and 100% (6/6) of male and 16.6% (1/6) of female p53Y217C astrocyte injections formed tumors. In addition to the frequency of tumor formation, female p53R172H, male p53Y202C, and male p53Y217C tumors were significantly larger than tumors expressing the same mutation in the opposite sex (Fig. 3F, Supplementary Fig. S2, log-transformed Welch-corrected two-tailed t test, P < 0.05). When no tumor was recovered, the final tumor volume reflects the volume of the recovered Matrigel pellet. The sex of the host mouse had no observed effect on tumor size or frequency which is consistent with cell-intrinsic sex differences as we previously observed (14, 15). Resected tumors all had features indicative of gliomas including nuclear atypia and hypercellularity (Fig. 3G) and expressed high levels of mutant p53 protein (Fig. 3H). In comparison with p53 KO, all mutations exhibited a GOF in tumorigenicity compared with p53 loss regardless of sex. However, three distinct patterns of sex differences in GOF were observed. p53R172H mutants had the greatest female sex effect. p53Y202C had a trend toward greater tumorigenicity in males but had the smallest differences between sexes. p53Y217C GOF activity was greatest in males and was the only mutation which formed tumors from every injection supporting male-specific GOF for this mutation.
Mutant p53 and Genomic Instability
All mutations were introduced at the same astrocyte passage under equivalent conditions with WTp53 intact. Introduction of mutant p53 and loss of WT p53 function disrupts DNA repair and cell-cycle checkpoints leading to genomic instability and the acquisition of additional mutations (39). We questioned whether the observed sex differences in growth and tumorigenicity might be driven by differences in genomic instability or acquisition of sporadic oncogenic mutations. We performed whole-exome sequencing and measured mutational burden in male and female mutant p53–expressing cells and p53 KO cells. Loss of WT p53 function led to equivalently large increases in mutational burden in all eight cell lines. To determine whether overall mutational burden was impacting phenotype, the raw mutation count was plotted against the percent area from the growth assay in Fig. 2. There was no significant correlation between mutational burden and proliferation (Supplementary Fig. S3A). Most mutations identified were silent or intergenic. Further analysis of missense, frame-shift, and nonsense mutations predicted to have a moderate-to-high impact on gene expression or function, also exhibited similar frequencies between cell lines with no correlation between the number of mutations and growth (Supplementary Fig. S3B).
Finally, we asked whether there were any unique mutations that could be contributing to these phenotypic differences. Supplemental Table S2 lists every gene with a homozygous missense, frame-shift, or nonsense mutation in at least one cell line. Almost all mutated genes were altered across tumor-forming and non-tumor–forming cells. Three genes, Calcoco2, Rbmy, and Thap1 were identified as uniquely mutated only in male p53Y202C cells. None of these genes have been described as cancer-related. Together, these data suggest that genomic instability and the acquisition of sporadic mutations does not account for the observed sex differences.
P53 Mutants Drive Unique Transcriptional Programs in Males and Females
One of the primary mechanisms of mutant p53 GOF activity is the transcriptional regulation of genes in protumorigenic pathways. Previous studies have identified several oncogenic targets of both R175H and Y220C mutations (17). To determine whether mutant p53 exhibits GOF transcriptional activity, we performed mRNA-sequencing (RNA-seq) on RNA isolated from each cell line and compared the differentially expressed genes between each mutation and p53 KO within each sex. In all cases, introduction of mutant p53 led to significant changes in gene expression compared with p53 KO, confirming neomorphic transcriptional function (FDR < 0.05; Supplementary Fig. S4). We next compared the differentially expressed genes between male and female astrocytes within each mutation. p53 KO, p53R172H, p53Y202C, and p53Y217C yielded 7183, 9013, 10,902, and 10,748 significant differentially expressed genes between sexes, respectively (FDR < 0.05; Fig. 4A). Predictably, sex chromosome–linked genes including the X-linked genes Xist, Agtr2, and Eda, and the Y-linked genes Uty, Eif2s3y, Ddx3y, and Kdm5d are among the most differentially expressed genes across all mutations. However, these genes represent only a small fraction of the genes with a sex difference in expression.
To evaluate how gene expression varies as a function of p53 mutation and sex, we performed a sparse principal component analysis. The first three principal components account for 75.8% of total variance. Strikingly, the samples clustered primarily by their ability to form tumors with low-tumorigenic male and female KO, male p53R172H, and female p53Y217C clustering together, while high-tumorigenic female p53R172H, female p53Y202C, male p53Y202C, and male p53Y217C clustered independently (Fig. 4B; Supplementary Fig. S5). Among the cell lines with high tumorigenic potential, the first two principal components account for the majority of sex differences with male p53Y202C and p53Y217C separating in PC1 and female p53Y202C and p53R172H separating in PC2. Meanwhile, PC3 encompasses the majority of variation between mutants. The normalized expression for all genes contributing to each principal component are displayed in heat maps in Fig. 4C. All tumor-forming cell lines have distinct transcriptomes, suggesting that neither sex nor p53 mutation status alone explain the variation in gene expression. Rather, it is the interaction between mutation and sex that drives unique transcriptional phenotypes.
Sex and Mutant p53 Interact in the Regulation of Oncogenic Pathways
To characterize the sex-specific GOF transcriptome of each mutation, differentially expressed genes between mutant p53 and p53 KO were compared between sexes to define the sex specific (unique) and shared (overlapping) transcriptional impact of each mutation. We focused our analysis on those genes with an absolute log2 fold change greater than 0.5. KEGG pathway enrichment analysis was used to evaluate whether shared or sex-specific changes in gene expression could be contributing to the differing tumorigenic phenotypes. Concordant with differences in tumorigenesis, pathway analysis revealed three independent transcriptional patterns. p53R172H expression had a substantially greater impact in females than males with approximately 7.4× more female-specific differentially expressed genes (Fig. 5A). Pathway analysis revealed that shared differentially expressed genes were enriched for KEGG cancer pathways. In addition, shared genes were enriched for many cancer-associated pathways including cell cycle, DNA replication, p53 signaling, focal adhesion, senescence, and Rap1 signaling. The genes differentially expressed uniquely in males were not enriched for any cancer-associated pathways. Meanwhile, genes that were differentially expressed in females alone were also enriched for many KEGG cancer pathways and cancer-associated pathways including extracellular matrix receptor signaling, stem cell pluripotency, and PI3K–Akt signaling among others. This indicates that p53R172H has both female-specific and sex-independent GOF effects on gene expression in cancer-relevant pathways, and the combined effect is sufficient for tumorigenesis in females, while the shared GOF alone is insufficient in males.
The p53Y202C mutation had the smallest sex difference in GOF phenotype with both males and females forming large tumors in vivo, and the smallest differences in sex-specific genes of any mutation. For this mutation, the overlapping gene list was enriched for cancer pathways and cancer-associated pathways with very few enriched pathways in the sex-specific differentially expressed genes (Fig. 5B). Together, this reveals a second GOF transcriptional profile in which the shared GOF activity is sufficient for transformation in both sexes with very little sex-specific effect on transformation.
The p53Y217C mutation represents a third interaction between mutation and sex (Fig. 5C). Here, male-specific genes were the only subset enriched for KEGG cancer pathways, with very little pathway enrichment in the shared or female component. Male-specific genes were enriched for pathways in cancer, Rap1 signaling, proteoglycans in cancer, miRNAs in cancer, several tissue-specific cancer pathways, and cancer-associated metabolic pathways. These data support a model in which p53Y217C has male-specific GOF effects on transcription with little universal or female-specific activity in cancer-relevant pathways.
To further interrogate the interaction between sex and mutation, we modeled the changes in gene expression for all genes as a function of sex and mutation. We identified 9,287, 10,441, and 9,387 genes for p53R172H, p53Y202C, and p53Y217C, respectively with a significant interaction between sex and mutation (FDR < 0.05). Importantly, this analysis is highly concordant with the gene sets assembled by differential expression analysis above (Supplementary Table S3). Subsequent pathway enrichment analysis of the gene overlapping between both analyses revealed similar KEGG pathways to those identified in the differential expression analysis alone (Supplementary Tables S4–S6). Modeling the interaction of mutation and gene expression independent of sex also revealed a higher degree of concordance (99%) with the shared differentially expressed gene lists (Supplementary Table S7). Together, these analyses support that cell sex and p53 mutations interact to drive unique transcriptional programs.
Sex Differences in Mutant p53 GOF Activity Is Mediated Through Differential Genomic Localization
Having established that mutant p53 expression leads to unique GOF transcriptional profiles dependent on both mutation and sex, we next sought to determine the mechanism underlying mutant p53 transcriptional regulation. We performed chromatin immunoprecipitation with massively parallel DNA sequencing (ChIP-seq) to map mutant p53 localization. To understand the sex differences in mutant p53 function, we focused on the mutation with the greatest female-specific effect, p53R172H, and greatest male-specific effect, p53Y217C. Regions enriched for mutant p53 localization were then mapped relative to their distance from the transcriptional start sites (TSS) of the top 1000 upregulated genes for each mutation in males and female (Fig. 6A). In both males and females, p53R172H localization was enriched near the TSSs of a subset of upregulated genes indicating that mutant p53 directly regulates aberrant gene expression in both sexes. In p53Y217C cells, localization was enriched at TSSs of upregulated genes in males but displayed no pattern of localization in females.
Previous studies have established that mutant p53 lacks a consensus binding sequence, but instead relies on interactions with other transcription factors and DNA-binding proteins for localization to aberrant target genes. To identify potential mediators of mutant p53 activity, we performed a HOMER analysis on the enriched ChIP signals ± 5 kb of the TSS of the same 1,000 upregulated genes enriched for p53 localization. The top five significant known motifs identified in male and female p53R172H and male p53Y217C are presented in Fig. 6B. In all three analyses, mutant p53 was found to localize to the consensus sequence of a unique set of transcription factors, suggesting that the sex-specific GOF activity of each mutation may be driven by different sets of transcriptional mediators.
The transcription factor AP-2gamma (TFAP2C) is a known regulator of stem cell self-renewal and chemotherapy resistance (40). We found that in females but not males p53R172H localization is enriched at AP-2gamma DNA-binding motifs. WNT5A has been shown to be upregulated in glioma and is associated with poor overall survival (41). In our model, Wnt5a mRNA was significantly upregulated in female p53R172H and downregulated in male p53R172H. This difference in gene expression is consistent with greater ChIP-seq signal in females compared with males at the AP-2gamma motif near the TSS of Wnt5a (Fig. 6C). Similarly, p53Y217C localized to the androgen receptor-half site motif located upstream of the growth factor Fgf2 in males but not females correlating with sex differences in gene expression (Fig. 6D). These results suggest that sex differences in mutant p53 GOF gene regulation may be mediated through differential localization to the consensus sequences of known transcription factors upstream of oncogenes.
p53 is the most interrogated protein in cancer research. A PubMed search of p53, TP53, and Trp53 reveals an annual average of 6,000 p53 publications over past five years. Despite this herculean effort, our understanding of the most mutated gene in cancer remains incomplete. An obstacle to fully characterizing WT and mutant p53 activity is its context dependence in function. WT p53 is positioned at the center of a complex network of cellular pathways whose precise activation profile is determined by cell identity and the stimulus activating the p53 response (1). Similarly, p53 mutations result in a spectrum of cancer phenotypes dictated by the tumor type, cooccurring mutations, and specific p53 mutation (19). Mounting evidence in both normal development and disease support that sex is a key modulator of p53 function, and that a full understanding of WT and mutant p53 will require incorporating the scope and magnitude of these sex effects.
Most mutations in tumor suppressors prevent gene expression or protein production or result in an unstable protein that is quickly degraded. In contrast, p53 GOF activity selects for stable expression of mutant p53 protein over loss-of-function mutations or deletions. However, the importance p53 GOF mutations is rarely reflected in in vitro and mouse cancer models that more frequently rely on dominant negative p53 constructs, or p53 knockdown or deletion. In GBM, complete p53 loss is an uncommon event. Our analysis of p53-mutant tumors revealed that missense mutations in the DBD are particularly enriched in GBM and CNS cancers compared with other cancers. Together, with previous studies of LFS demonstrating an association between germline DBD mutations and a higher risk of brain tumor development, these findings support a CNS tissue–specific advantage of p53 GOF mutation in tumor development (32). This finding also suggests the fidelity of GBM models could be enhanced by inclusion of clinically relevant p53 GOF mutations.
In this study, we revealed a novel interaction between p53 mutation and cell sex. We found that while p53 deletion increased proliferation, it is insufficient to transform astrocytes and does not induce a sex difference in phenotype. In contrast, introduction of GOF mutant p53 into the same cells was sufficient for transformation, with the magnitude of tumorigenic potential dependent on the mutation and cell sex. The hotspot mutation p53Y217C exhibited the greatest GOF and greatest sex difference of all three tested mutations. Male p53Y217C cells were highly proliferative and tumorigenic while female p53Y217C cells exhibited no GOF phenotype compared with KO. Notably, this relationship is concordant with a greater mutation frequency of p53Y220C in human males. p53R172H exhibited the opposite effect with the GOF effect predominant in females with little impact on tumorigenesis in males. In humans, this mutation was found in equal prevalence in both sexes. Independent of p53 status, GBM is more common in males than females (incidence ratio of 1.6:1; ref. 42). If a mutation were to have equal GOF effects in both sexes, we would expect it to present at a similar incidence ratio to GBM overall. A potential explanation for the lack of sex difference in the incidence of p53R175H is that the female-specific GOF activity compensates for the other factors that would otherwise result in a greater incidence in males. The third mutation, p53Y202C, had the smallest sex difference in GOF. While males grew faster in vitro, both males and females formed tumors in vivo. In this instance, the cell-intrinsic sex effect is not concordant with the observed greater incidence in human females. The lack of concordance suggests additional factors involving such mechanisms as cell:cell interactions within the tumor microenvironment or systemic effects of immunity or metabolism, likely supersede the cell-intrinsic, sex-specific GOF effect of this mutation. In contrast, sex differences in p53Y217C GOF activity appear to be a dominant effector of sex differences in tumorigenicity. The differences in concordance between the laboratory and clinical phenotypes of these mutants may provide an important tool for investigating the mechanisms by which cell-intrinsic effects of p53 mutations interact with tissue- and systems-level factors to determine cancer phenotypes.
Mutant p53 promotes tumorigenesis through the aberrant regulation of oncogenic target genes. Using RNA-seq, we characterized a sex-specific and sex-independent GOF transcriptional profile – defined as the differential gene expression between mutant p53 and p53 KO – for each missense mutation. All three mutations resulted in sex- and mutation-specific gene expression patterns with thousands of differentially expressed genes. Comparing the GOF transcriptional profiles between sexes revealed three distinct gene expression patterns concordant with the tumorigenic phenotypes. In p53Y217C, the male GOF profile was enriched for many cancer pathways, with few pathways enriched in the shared or female-specific profiles. This male-specific GOF transcriptional profile is consistent with the male-specific GOF phenotype. p53R172H, which had the greatest female-specific GOF phenotype, exhibited enrichment for cancer-relevant pathways in both the female specific and shared differentially expressed genes, and no enrichment for cancer pathways in males. This points to a second model where p53R172H acts on a set of female-specific and sex-independent target genes whose effects interact to drive sex differences in tumor formation. In the third scenario, p53Y202C was only enriched for cancer pathways in the shared component indicating the GOF activity enacted by this mutation is sex independent, which is consistent with the similar rates of tumor formation that we observed in these cells. These findings indicate that mutant p53 can have both sex-dependent and sex-independent GOF activity, and that the impact of each on tumorigenesis is mutation specific.
Lopes-Ramos and colleagues recently demonstrated that in normal tissues, transcription factors that are not differentially expressed between males and females have different regulatory binding patterns in males and females (43). The differences in transcription factor binding thus contribute to sex-biased transcriptional programs. In our study, we demonstrated that sex differences in mutant p53 GOF are being driven by differential mutant p53 localization. In both male and female astrocytes, p53R172H localizes to the TSS of a subset of upregulated genes with greater overall enrichment in females. Whereas in p53Y217C, we only observed mutant p53 localization at upregulated genes in males with discernable pattern of p53 binding above input in females. This is consistent with a universal- and female-specific GOF by p53R172H and a male-specific GOF by p53Y217C.
Mutant p53 lacks a consensus response element and relies on interactions with other transcription factors for localization. By performing HOMER motif enrichment analysis for regions near the TSSs of the top upregulated genes, we identified several novel candidate transcription factors that may be contributing to sex differences in mutant p53 localization and gain of aberrant transcriptional output. Understanding the exact mechanisms of sex-specific mutant p53 localization and gene expression will require the interrogation of each potential binding partner. However, here we highlight two promising potential co-transactivators: female-specific p53R172H enrichment at the AP-2gamma motif upstream of Wnt5a and male-specific p53Y217C enrichment at the AR-halfsite motif upstream of Fgf2. AP-2gamma is an essential transcription factor for several developmental pathways including neural tube development and male gonad development. It also has been shown to act as an oncogene in breast cancer, melanoma, testicular cancer, neuroblastoma, ovarian cancer, and lung cancer (44). Expression of AP-2gamma is associated with stem cell renewal, proliferation, and therapy resistance (44). Wnt5a is upregulated in GBM and has been shown to drive proliferation and invasion in cancer stem cell (CSC) populations (45). Androgen receptor (AR) is known to positively regulate the expression Fgf2 in prostate cancer (46). In GBM, FGF2 promotes CSC maintenance through interactions with the receptor FGFR1 (47). AR is expressed in both male and female GBM, and androgen receptor antagonists are being evaluated as novel therapeutic strategies (48).
Despite the massive effort to dissect p53 biology, our understanding of this central tumor suppressor remains incomplete. Here, we demonstrated that the same mutations in p53 have unique GOF activity when expressed in otherwise isogenic male and female astrocytes. Our finding highlights sex as a critical determinant of mutant p53 GOF biology. This raises an important question: to what extent has the absence of a consideration of sex limited the approach to investigating p53 function, confounded the interpretation of results, and obscured our understanding of p53 biology? This possibility must be evaluated further as we move forward to improve our understanding of this complex pathway and increase the potential for developing treatments that target the unique effectors of p53 function in male versus female tumors.
M.V. Staller reports grants from NIGMS and grants from BWF during the conduct of the study. D. Baldridge reports grants from The Children's Discovery Institute of Washington University and St. Louis Children's Hospital during the conduct of the study. No disclosures were reported by the other authors.
N.C. Rockwell: Conceptualization, data curation, formal analysis, investigation, methodology, writing-original draft. W. Yang: Formal analysis, writing-review and editing. N.M. Warrington: Formal analysis, investigation, writing-review and editing. M.V. Staller: Resources, methodology, writing-review and editing. M. Griffith: Formal analysis, funding acquisition, writing-review and editing. O.L. Griffith: Formal analysis, funding acquisition, writing-review and editing. C.A. Gurnett: Funding acquisition, writing-review and editing. B.A. Cohen: Resources, formal analysis, funding acquisition, writing-review and editing. D. Baldridge: Resources, funding acquisition, writing-review and editing. J.B. Rubin: Conceptualization, resources, formal analysis, funding acquisition, methodology, writing-review and editing.
This work was supported by NIH grant RO1 CA174737-06 (to J.B. Rubin), the Children's Discovery Institute of Washington University and St. Louis Children's Hospital (to M. Griffith, O.L. Griffith, C.A. Gurnett, B.A. Cohen, D. Baldridge, J.B. Rubin), and Joshua's Great Things (to J.B. Rubin). We thank the Alvin J. Siteman Cancer Center at Washington University School of Medicine and Barnes-Jewish Hospital in St. Louis, MO, and the Institute of Clinical and Translational Sciences (ICTS) at Washington University in St. Louis, for the use of the Genome Technology Access Center, which provided RNA-, DNA-, and ChIP-seq services. The Siteman Cancer Center is supported in part by an NCI Cancer Center Support Grant #P30 CA091842 and the ICTS is funded by the NIH's NCATS Clinical and Translational Science Award (CTSA) program grant #UL1 TR002345. We also thank the Alvin J. Siteman Cancer Center at Washington University School of Medicine and Barnes-Jewish Hospital in St. Louis, MO, for the use of the Siteman Flow Cytometry, which provided flow sorting of eGFP-positive cells. We additionally thank Alexia Thia for her technical support including cell culture and Mycoplasma testing.