p53 R273 mutants are associated with more aggressive colorectal tumors relative to R175 mutants
Compared to most other cancers, in colorectal cancer (CRC) the relative representation of "hotspot" missense mutations among carriers of TP53 mutations is particularly high. Specifically, missense mutations in the four most commonly mutated p53 residues (R175, R248, R273 and R282) comprise approximately 37% of all TP53 mutations in this type of cancer (Fig. S1a). In contrast, mutations in these four residues encompass only 17% of all TP53 mutations in all other cancer types together. Although this might be simply due to the mutational signature of particular carcinogens, it might also suggest a more significant GOF effect of such missense mutations in CRC.
One obvious question is whether different hotspot mutations may exert different effects on disease features and patient outcome. To address this question, we set out to compare R175 structural mutations to R273 DNA contact mutations. Notably, these mutations together represent over 20% of all CRC tumors harboring TP53 mutations, as compared to only approximately 10% in all other cancers (Fig 1A). We analyzed clinical data from several patient cohorts, using the TCGA and ICGC open-source platforms as well as additional published datasets (16,17,18) (Supplementary Table 1). Remarkably, while R175 mutations are significantly more frequent than R273 mutations in early disease stages, the predominance of R175 mutations is abolished at later stages (Fig. 1b). This suggests that, relative to R175 mutations, R273 mutations might accelerate disease progression from early stages to advanced stages, involving cancer cell spreading to nearby lymph nodes (stage 3) and metastases to distant organs (stage 4).
Interestingly, when we analyzed the MSKCC CRC dataset, comprising 1134 cases of which ~90% were metastatic (19), we found that while both R175 and R273 mutants exhibited a similar percentage of liver, lung and lymph node first site metastases (data not shown), R273 mutants were significantly more associated with tumors that metastasize first to less common sites such as brain, bone, pelvis, peritoneum and gynecological sites (Fig. 1c). Importantly, unlike liver and lung metastases, metastatic lesions in these sites are usually considered unresectable, and thus incurable. Indeed, many studies have linked the presence of metastases at those sites to worse survival (20, 21, 22). Furthermore, R273 mutants were found to be significantly associated with multiple metastatic sites at the time of diagnosis of metastatic disease (Fig. 1d), further supporting the notion that R273 mutants selectively augment the metastatic capacity of CRC cancer cells. Importantly, R273 mutants were associated with significantly shorter disease-specific overall survival than R175 mutants (Fig 1e), regardless of patient age, sex, tumor location or presence of KRAS mutations (Fig 1f and Supplementary Table 2). Interestingly, while the impact of R273 mutations on overall survival was prominent in CRC patients presenting at stages 1-3 (Fig. S1b), it was not seen anymore when the patients presented with stage 4 disease (Fig. S1c); this is consistent with the notion that the main effect of R273 mutations is on the rate of progression from early stage CRC to advanced disease .
To explore the possibility that R273 mutant tumors might be associated with a particular mutational landscape, which may account for the observed clinical effects, we compared the co-occurrence of the most common gene mutations in CRC with either R175 or R273 mutations. Notably, other than SMAD4 mutations which showed a mild co-occurrence with R273 mutations (P=0.02), all other gene mutations were not differentially enriched in R273 mutated vs R175 mutated tumors (Fig S1d).
In sum, compared to R175 mutations, R273 mutations are preferentially associated with more advanced disease, higher rate of multiple and uncommon metastases, and shorter patient survival.
p53R273H orchestrates a distinct transcriptional signature
We next wished to elucidate the molecular mechanisms underpinning the differential impact of R273 vs R175 mutants in CRC, and to assess whether R273 mutations confer a true GOF. To that end, we utilized CRC-derived SW480 cells. SW480 is a microsatellite stable cell line, harboring APC and KRAS mutations; hence, it properly represents sporadic CRC. SW480 cells endogenously express two p53 mutants: p53R273H, and the less common p53P309S (23). SW480 cells depleted of their endogenous mutp53 by CRISPR/Cas9-mediated knockout (p53KO) were stably transduced with either p53R273H or p53R175H (Fig. 2a). Western blot analysis confirmed comparable overexpression of both mutants (Fig. 2b). As mutp53 GOF often involves changes in the cell transcriptome, we next subjected the different SW480 cell pools to RNA sequencing (RNA-seq) analysis, using the MARS-seq protocol (24). Clustering analysis revealed substantial differences between the transcriptome of the R273H cells and the parental p53KO cells (Fig. 2c). Surprisingly, overexpression of p53R175H had rather limited impact on the transcriptome of these cells (Fig. 2c). By comparing the observed transcriptional profiles, we generated a gene signature comprising 140 genes upregulated by p53R273H relative to both p53R175H and p53KO cells. This gene signature was defined as the “R273 signature” (Fig. 2d).
To further validate our conclusions, we adopted an alternative approach wherein SW480 cells were stably transduced with shRNA directed against the 3' UTR of the TP53 gene (shp53), followed by stable overexpression of shRNA-resistant p53R175H or p53R273H (Fig. 2e). The resultant cell pools were subjected to MARS-seq analysis as above. Clustering analysis of the data confirmed that, also by this approach, p53R273H had a stronger effect on the SW480 cell transcriptome than p53R175H (Fig. S2a). Importantly, the “R273 signature”, deduced from the reconstituted p53KO cells, was strongly correlated with the differences in gene expression between the R273H-reconstituted shp53 cells and the control (Fig. 2f) or R175H-reconstituted (Fig. 2g) cells, as determined by gene set enrichment analysis (GSEA).
Lastly, since the above RNA-seq analyses were done with ectopically overexpressed p53 mutants, we quantified the relative expression of representative R273 signature genes by RT-qPCR analysis in control (expressing endogenous mutp53) and p53KO SW480 cells (Western blot in Fig. S2b). As seen in Fig S2c, all tested genes were significantly downregulated in the knockout cells, consistent with their being positively regulated by p53R273H. Moreover, comparison by GSEA of our R273 signature to published RNA-seq data of SW480 cells before and after shRNA-mediated p53 knockdown (25) confirmed significantly higher expression of the R273 signature in the control cells, which harbor endogenous p53R273H (Fig. S2d). Thus, p53R273H drives a distinct transcriptional program in SW480 cells.
The R273 signature is upregulated in multiple CRC cell lines and tumors and is associated with poor survival
To assess the generality of the R273 signature we interrogated experimentally three additional CRC-derived cell lines, by expressing p53R273H and p53R175H ectopically in HCT116 and RKO cells depleted of their endogenous wtp53 (KO), and COLO-205 cells endogenously expressing truncated p53 (Fig. 3a and Fig. S3a,c). Reassuringly, RT-qPCR analysis of representative R273 signature genes confirmed that, in all three cell lines, p53R273H selectively upregulated these genes, albeit to varying extents (Fig. 3b, Fig. S3b,d). Moreover, using the cancer cell line encyclopedia (CCLE) database, we found that the R273 signature is significantly upregulated in CRC cell lines harboring R273 mutations, compared to CRC lines carrying protein-truncating TP53 mutations (Fig. 3c). The CCLE includes only three R175-mutated CRC lines; while their analysis indicated a similar trend as above, statistical significance could not be reached (p=0.067; data not shown).
We next wished to extend these findings to human CRC tumors. Importantly, GSEA analysis of the TCGA CRC cohort revealed that tumors harboring R273 mutations displayed significantly higher expression of the R273 signature than those with R175 mutations (Fig. 3d). Comparison of the R273-mutated tumors to tumors carrying truncating TP53 mutations yielded a similar trend, but the difference did not reach statistical significance (data not shown). Of note, the truncating mutations group is very heterogeneous, and not all cases may resemble a p53-null state. Yet, tumors with extremely low p53 mRNA levels, presumably owing to nonsense-mediated decay (3), are more likely to approximate true nulls. Indeed, when we included only truncating mutation cases displaying greatly reduced steady-state p53 mRNA, unequivocal association of R273-mutated tumors with the R273 signature was clearly evident (Fig. 3e). Interestingly, analysis of the entire set of CRC tumors revealed a remarkable degree of positive correlations between the expression levels of the genes comprising the R273 signature, which was not observed in three independent control signatures (Fig S3e,f). This suggests that many of the genes comprising the R273 signature may be subject to common transcriptional or post-transcriptional regulatory mechanisms.
Guinney et al. have recently employed comprehensive data analysis to define four consensus molecular subtypes (CMS) for colorectal cancer (26). Remarkably, when we compared our R273 signature with the cell-intrinsic transcriptional signatures of the four CMS subtypes, as determined by Sveen et al. (27), the R273 signature displayed a strong (R=0.66) and significant (p<2.2e-16) correlation with the CMS4 signature (Fig S4a). Furthermore, GSEA analysis confirmed that CRC tumors harboring R273 mutations are significantly associated with the CMS4 gene signature compared to tumors harboring R175 mutations or truncating mutation (Fig S4b). Interestingly, the GSEA analysis revealed that tumors harboring R175 mutations are significantly associated with the CMS2 gene signature, when compared to tumors harboring either R273 or truncating mutations (Fig S4c). Hence, R273 mutations and R175 mutations are differentially associated with distinct CRC molecular subtypes, implicating markedly different cancer-promoting biological processes (26).
Importantly, comparison of TCGA CRC tumors displaying high (upper quartile) R273 signature vs those with low (bottom quartile) signature revealed that high R273 signature was significantly associated with late-stage disease (Fig. 3f) and shorter patient survival (Fig. 3g). Furthermore, multivariate Cox regression analysis for overall survival, including age, sex, tumor location and the presence of KRAS mutations, demonstrated that high expression of the R273 signature is an independent prognostic factor (multivariate hazard ratio 2.314; 95% confidence interval 1.344–3.977; P=0.002; Supplementary Table 3).
In sum, the R273 gene signature is broadly enriched in CRC cells and tumors harboring R273 mutations, and is correlated with shorter patient survival. This further supports the hypothesis that the transcriptional output directed by R273 mutants endows CRC tumors with more aggressive features, which adversely affect patient outcome.
R273 mutants selectively promote cell spreading, migration and invasion
To elucidate oncogenic pathways that may contribute to the clinical impact of R273 mutations, we subjected the R273 signature to Gene Ontology analysis by METASCAPE (28). Interestingly, many observed pathways were directly or indirectly related to cytoskeleton dynamics (Fig. 4a), which is often associated with cancer-related properties such as cell adhesion, spreading, migration and invasion (29,30,31,32). Specifically, the Rho signaling pathway, ranking high in this analysis, can promote cancer by driving actin cytoskeleton remodeling and augmenting cell migration, survival, polarity, and more (33,34).
Phenotypically, the morphology of SW480 cells expressing p53R273H differed visibly from that of parental knockout cells or p53R175H expressors. This was evident as accelerated spreading, confirmed by time-lapse microscopy (Fig. 4b and Supplementary movies 1-3). Similar observations were made with RKO cells, depleted of their endogenous wtp53 and reconstituted with either p53R175H or p53R273H (Fig S5a). Importantly, RNA-seq analysis six hours after plating (Fig. S5b) showed that already at this early time point the R273 signature was upregulated in the p53R273H expressors to a similar extent as after 24 hours. This supports the notion that the inherent gene expression pattern dictated by p53R273H drives cell spreading, rather than being secondary to it.
Cell cycle analysis did not reveal differences between the effects of p53R273H and p53R175H (Fig. S5c). However, the p53R273H expressors displayed a significant increase in cell migration (Fig. 4c,d) and invasion (Fig. 4e,f), relative to p53R175H expressors or knockout cells. Moreover, while both p53R273H and p53R175H augmented the migration of p53-depleted RKO cells, the effect of p53R273H was significantly greater (Fig S5d,e). Thus, p53R273H preferentially promotes cell spreading, migration and invasion.
Rho signaling is one of the top enriched pathways in the R273 signature (Fig. 4a). In agreement, a Rho proteins GTPase activation assay confirmed that overexpression p53R273H of in SW480 cells augmented the activation of both Cdc42 and Rac1, relative to p53R175H overexpressors (Fig. 4g). Interestingly, RhoA activation was not differentially affected. Importantly, the migratory phenotype of p53R273H overexpressors was completely abolished by treatment with the Rac1/Cdc42 inhibitor MBQ-167 (Fig. 4h). Hence, p53R273H selectively drives Rac1/Cdc42-dependent cancer cell migration.
p53R273H preferentially promotes metastasis
We next wished to assess whether the differential impact of p53R273H in vitro is also reflected in a more aggressive phenotype in vivo. To that end, SW480 cells ectopically expressing either p53R175H or p53R273H were injected into the tail vein of NSG mice (Fig. 5a). Remarkably, 9 weeks after injection, the lungs of the mice injected with p53R273H -overexpressing cells displayed a significantly larger area of lung metastases than in mice injected with p53R175H overexpressors (Fig. 5b,c). Moreover, to better recapitulate CRC biology, we orthotopically injected SW480 cells harboring the two p53 mutants into the cecal wall of NSG mice (Fig 5d). Seven weeks later, mice were sacrificed and evaluated for distant organ metastases. Notably, four out of five mice in the R273H group developed both lung and liver metastases, while no metastases were observed in any of the mice injected with p53R175H overexpressors (Fig 5e,f). Thus, p53R273H not only confers increased migration and invasion in vitro, but also preferentially promotes metastatic behavior in vivo.
p53R273H is recruited to R273 signature genes and activates them via its transactivation domain
To explore the molecular mechanisms driving the transcriptional upregulation of R273 signature genes by p53R273H, we interrogated published p53 CHIP-seq data of SW480 cells (25), which express endogenous p53R273H (along with p53P309S). Remarkably, analysis of all mutp53 peaks using GREAT (35), revealed that the most significantly enriched cellular components associated with those peaks were related to cytoskeleton structure and function (Fig. 6a). Moreover, the mutp53 chromatin binding peaks were significantly correlated with the genes upregulated upon p53R273H overexpression in our SW480 RNA-seq (Fig. 6b), suggesting that regulation of their expression by p53R273H is mediated, at least in part, via the recruitment of p53R273H to the corresponding chromatin regions. To query experimentally this notion, we compared by ChIP-qPCR the binding of p53R273H and p53R175H to regulatory elements of representative R273 signature genes, in SW480 cells ectopically expressing either mutant. As seen in Fig. 6c, p53R273H indeed displayed significantly stronger binding than p53R175H to those regulatory regions.
Previous work has demonstrated that p53R273H can act as a potent transcriptional activator when recruited to DNA, e.g. as a GAL4 fusion protein (36,37,38). The N-terminal transactivation domain (TAD) is essential for this activity (36). In agreement, while transiently-transfected p53R273H augmented the expression of endogenous R273 signature genes in p53KO SW480 cells, this effect was lost when the cells were transfected with a TAD-mutated version of p53R273H (Fig. 6d and Fig. S6a), despite being expressed at comparable amounts in the transfected cells (Fig. 6e,f). In addition to p53R273H, the p53R273C mutation is also fairly common in human cancer, including CRC. As seen in Fig. S6b,c, p53R273C was also capable of transactivating endogenous R273 signature genes in transiently transfected p53KO SW480 cells.
Collectively, these observations support the notion that recruitment of R273-mutated p53 proteins to specific chromatin regions alters the expression of associated genes, in a TAD-dependent manner. These transcriptional alterations may underpin the observed biological effects of the R273 mutants, leading to enhanced tumor progression and worse patient outcome.