Functional characterization of the p53 “mutome”

ABSTRACT

Phenotypic characterization of mutations in the tumor protein p53 (TP53) gene has so far focused on a handful of relatively frequent “hotspot” mutations, accounting for only ~ 30% of cases. We expanded the scope and quantitatively measured the impact of thousands of distinct TP53 mutations in vitro and in vivo, providing insights into the connections between structure, function, evolutionary conservation and clinical impact.

The tumor protein p53 (TP53) gene, encoding the p53 transcription factor, is the most frequently mutated gene in human cancer¹. Unlike other tumor suppressors, the majority of cancer-associated mutations in p53 are missense mutations residing in its DNA-binding domain (DBD)^2,3, leading to loss of tumor suppressive activity and, in some cases, also to gain of novel oncogenic functions (gain-of-function, GOF; reviewed in⁴). However, while thousands of distinct p53 mutations have been reported and variant-specific phenotypes have been observed, cancer-related research has so far focused almost exclusively on 6 relatively common hotspot mutations, which altogether account for only ~ 30% of the p53-mutated cases.

Despite the emerging understanding that p53 mutations are not all alike, personalized therapeutics lag behind and mutant-specific effects are widely overlooked, as highlighted by Sabapathy and Lane⁵. It is therefore important to determine the functional consequences of all distinct p53 mutations; this might be particularly valuable for the personalized treatment of cancer. We therefore sought to systematically study the impact of the entire spectrum of p53 DBD mutations, accounting for over 90% of tumor-associated mutations, in human cancer-derived cells. To quantitatively measure the anti-proliferative capacity of thousands of mutant p53 (mutp53) variants, we devised a massively-parallel mutational scan that allows to simultaneously measure the relative impact of a plethora of mutations. Using this assay, we measured the anti-proliferative functional capacity of a synthetically-designed library of ~ 10,000 distinct p53 variants (Figure 1)⁶.

Figure 1.

Comprehensive characterization of the functional outcomes of p53 mutations in vitro and in vivo. A synthetic library of 9,833 mutations in the tumor protein p53 (TP53) DNA-binding domain was expressed in human cells enabling quantitative measurements of the effect of distinct variants in cell culture and in mouse xenografts, providing a comprehensive phenotypic catalogue for p53 mutations. SNP – Single nucleotide polymorphism.

Specifically, we computationally designed a genomic library including nearly all p53 DBD mutations reported in tumor samples³, the majority of which were previously unstudied, as well as combinations of mutations with single-nucleotide polymorphisms (SNPs) found across human individuals (aimed to identify possible genetic interactions). Additionally, to unbiasedly characterize the effect of mutations across the entire DBD, we systematically generated all single-nucleotide alterations within the DBD, as well as single amino acid substitutions, premature termination codons and in-frame deletions. Technically, our approach employs a designed microarray-synthesized oligonucleotide library to generate a mutp53 lentiviral library. These viruses were then transduced into human cancer-derived cells, resulting in a mixed population of cells, each expressing a single mutp53 variant. Sampling of these cells at different time points post-infection allowed us to measure the dynamics of each variant along time and deduce a relative fitness score (RFS), indicative of each mutant’s functional outcome.

Using this library, we first examined how protein sequence variations affect wild-type p53 (wtp53)-like anti-proliferative capacity. Expectedly, regardless of position along the DBD, premature termination codons and frameshift mutations strongly disrupted p53 functionality, whereas synonymous mutations retained full wtp53 functionality. In contrast, the effects of substituting or deleting a single amino acid were strongly dependent on its position within the DBD, corresponding to well-defined structural motifs, and on the biochemical properties of the substituting amino acid. Interestingly, the majority of p53 mutations discretely segregated bimodally as either retaining wtp53 functionality or strongly disrupting it. Moreover, most residues dichotomized either as tolerant (retaining wtp53 functionality regardless of particular substitution) or highly intolerant (functionality is disrupted by nearly any mutation at that position).

We next compared the functional impact of protein sequence alterations to the relative evolutionary representation of each amino acid substitution (i.e. the percent of species in which that amino acid is present at a given position). This revealed that the mean relative representation of variants retaining wtp53 functionality is dramatically higher than that of non-functional variants. Thus, the functional effects measured in our assay faithfully reproduce the constraints that shape the DBD sequence along evolution. These associations between structure, function and conservation, underscore p53’s anti-proliferative capacity as a fundamental property under strong evolutionary selection.

We further observed high concordance between the functional outcome of each mutation and its prevalence in human tumors. This reinforces the notion that loss of anti-proliferative capacity is a key selective force in shaping the landscape of p53 mutations in cancer. Thus, tumor-associated mutations retaining wtp53-like anti-proliferative functionality are unlikely to be driver mutations. This was further demonstrated by comparing the age at tumor diagnosis in individuals harboring germline p53 mutations (Li-Fraumeni syndrome), when stratified by their specific mutation’s RFS.

Importantly, our in vitro measurements could not identify a growth advantage for commonly occurring “hotspot” mutations, compared to equally disruptive mutations observed less frequently in tumors. Therefore, we sought to pursue evidence for GOF under selective pressures operating in vivo, within a growing tumor. We injected library-expressing cells into immunocompromised mice and quantified the relative enrichment of each variant in the formed tumors. Interestingly, mutations that equally disrupted p53 functionality in vitro were found to span a broad phenotypic spectrum in the tumors; reassuringly, hotspot mutations now exhibited selective in vivo GOF.

Finally, we took advantage of our experimental platform to evaluate the significance of SNPs within the DBD. Thus, we measured the combinatorial effects of background SNPs with an additional acquired mutation. We found that, when presented on a 217M (Methionine at codon 217) background (rs35163653), missense mutations exhibited a more disruptive phenotype compared to the outcomes of the same mutations on a wild-type (217V) p53 background, highlighting the importance of experimental assessment of combinatorial genetic interactions.

Altogether, our data provides a comprehensive phenotypic catalogue of p53 mutations (Figure 1), quantifying the relative fitness effect of each mutation in distinct biological contexts and providing information regarding the effects of thousands of variants of unknown significance. This paves the way to studying the entire “p53 mutome” and supports the applicability of similar large-scale systematic scans to broaden our understanding of additional mutation-driven phenotypic landscapes. Our dataset might serve as a resource for subsequent clinical and in-depth mechanistic studies of the effects of specific p53 mutations. Furthermore, this library of p53 mutants may serve as a tool for further phenotypic studies in vitro and in vivo and for high-throughput drug screening.

Funding Statement

This work was supported in part by the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation and a Center of Excellence grant from the Israel Science Foundation (to M. Oren) and the European Research Council (to E. Segal).