Abstract
Lung cancer, the leading cause of cancer-related mortality in the United States, occurs primarily due to prolonged exposure to an array of carcinogenic compounds in cigarette smoke. These carcinogens create bulky DNA adducts, inducing alterations including missense mutations in the tumor suppressor gene TP53. TP53 is the most commonly mutated gene in many human cancers, and a specific set of these variants are enriched in lung cancer (at amino acid residues V157, R158, and A159). This perspective postulates that lung-enriched mutations can be explained, in part, by biological selection for oncogenic gain-of-function (GOF) mutant p53 alleles at V157, R158, and A159. This hypothesis explaining tissue-specific TP53 mutations is further supported by mouse model studies of the canonical TP53 hotspots showing that tumor spectra and GOF activities are altered with mutation type. Therefore, although smoking-related lung cancer unequivocally arises due to the mutagenic environment induced by tobacco carcinogens, this perspective provides a rationale for the preferential selection of lung-enriched V157, R158, and A159 mutant p53.
Introduction
Although the p53 tumor suppressor, as the “guardian of the genome,” has been studied extensively, the availability of more complete sequencing data across the entire genome has allowed the current era of human genetic studies to be more comprehensive than ever (1, 2). One of the most active areas of research is in mutant p53 gain-of-oncogenic function (GOF), whereby single nucleotide missense mutations exhibit not only loss of tumor suppressor function but also acquisition of novel activities such as increased invasion, chemoresistance, and metabolic reprogramming (3, 4). Identification of novel tumorigenic functions may open new avenues for p53-targeted therapies.
Somatic alterations in the TP53 gene occur primarily at one of six major canonical hotspots in the DNA-binding domain: R175, G245, R248, R249, R273, and R282. Specifically, the eight most common mutants individually range in frequency from 2.9% to 7% and together account for 27.7% of all TP53 mutations in a database of 11,145 sequenced tumors (5). Notably, when lung cancer cases are reviewed in isolation, the frequency of mutations at amino acid residues V157 and R158 increases five- to six-fold (0.9%–4.6% and 1.1%–5.9%, respectively) and surpasses that of many of the traditional hotspots. This phenomenon has been reported extensively and attributed to preferential DNA adduct formation at these codons by carcinogenic polycyclic aromatic hydrocarbons (PAH) found in cigarette smoke (6–10).
The PAH benzo[a]pyrene, which is metabolically activated to benzo[a]pyrene-diol-epoxide (BPDE), preferentially binds the N2 position of guanine at TP53 codons 157, 248, and 273 in normal human bronchial epithelial cells and normal human fibroblasts (6). These bulky DNA adducts generate a COSMIC smoking signature described as G:C > T:A transversion mutations with a bias for the nontranscribed DNA strand, where DNA is less efficiently repaired (11). Moreover, DNA adduct formation by BPDE is enhanced at methylated CpG sites in both lung-enriched V157F and R158 codons and the canonical hotspots G245, R248, and R273 (6). Although the precise mechanism is not yet understood, it has been proposed that cytosine methylation increases the electron density of the amine group of the paired guanine. This enhances its affinity with BPDE and acrolein, a mutagenic aldehyde also found in cigarette smoke (12–14). Alternative hypotheses have been put forth as well, with the methyl group of 5-methylcytosine allowing increased intercalation of BPDE with subsequent increase in covalent interaction; or increased probability of base flipping of the paired guanine into the major groove, allowing it to interact with BPDE (12, 13). A variety of in vitro and in vivo studies to examine mutations induced by exposure of p53 cDNA, bronchial epithelial cells, lung fibroblasts, and human TP53 knock-in mouse embryo fibroblasts to BPDE have confirmed a high frequency of G>T transversions at codons V157 and R158 in addition to the canonical p53 hotspots (12–18).
TP53 mutational spectra in lung tumors
To investigate the clinical significance of V157, R158, and A159 mutations in lung cancer, we evaluated lung adenocarcinoma and lung squamous cell lung cancer sequencing data from the Pan-Lung Cancer Study of The Cancer Genome Atlas (TCGA; ref. 19). As noted above, the frequency of R158 and V157 mutations in lung cancers is markedly elevated. This finding is not observed in head and neck squamous cell carcinomas (HNSCC), for which cigarette smoking is also the primary risk factor (Fig. 1). In the TCGA database which includes 528 patient samples, TP53 is altered in 367 cases of HNSCC (69.5%), and 253 (68.9%) of these are missense mutations (20). These frequencies are nearly identical to the TCGA Pan-Lung Cancer Study, where 1,144 patient samples exhibit TP53 alterations in 776 cases (67.8%), and 505 (65.1%) of these are missense mutations (19). However, in sharp contrast to lung cancer, only five samples harbor V157 mutations (5/5 V157F, 2.0% of total), three have R158 mutations (2/3 R158L, 1.2% of total), and two have A159 mutations (1/2 A159P, 0.8%). In bladder urothelial carcinoma where cigarette smoking is also a major risk factor, there is a much lower rate of TP53 mutations (21). Among 412 patient samples, 74 (18.0%) exhibit alterations in TP53, and 57 (77.0%) of these are missense mutations. No V157 mutations and only one R158 mutations (R158H) exist in this cohort, and although there are four A159 mutations, the overall lower frequency of TP53 mutations suggests that changes in the p53 tumor suppressor do not play a major role in bladder tumorigenesis. These data show that the V157, R158, and A159 TP53 mutations occur preferentially in lung cancer over HNSCC and bladder cancer despite the critical role that cigarette smoke exposure plays in all three tumor types. Hence, it is possible that these TP53 mutations are not only induced by tobacco carcinogens but exhibit a relative abundance in the lung driven by selective biological pressure which may be tissue-specific.
Patterns in TP53 mutations vary by tumor type. Number of somatic missense mutations are plotted for the most frequently mutated amino acid residues in the TP53 gene, by solid tumor type. Data are from The Cancer Genome Atlas (TCGA; ref. 19). Total numbers of samples with TP53 missense mutations are: lung adenocarcinoma and squamous cell carcinoma, 505; breast adenocarcinoma, 179; head and neck squamous cell carcinoma (HNSCC), 253; gastric adenocarcinoma, 131; ovarian serous cystadenocarcinoma, 179; esophageal adenocarcinoma, 110; hepatocellular carcinoma (HCC), 71; colorectal adenocarcinoma, 83; pancreatic adenocarcinoma, 72.
Our analysis of the larger and more recently curated AACR Project GENIE cohort revealed a similar enrichment for mutations at V157, R158, and A159 among non–small cell lung cancer samples (Table 1). Specifically, in lung adenocarcinoma (n = 5,060 samples), the overall prevalence of TP53 alterations was 46.1%, or 2,333 samples, and R158 was the third most frequent amino acid residue at which mutations occurred (increasing four-fold in frequency to 4.2%, from 1.0% when compared with non-lung cancers). Alterations in TP53 in general were more common in squamous cell lung cancer (n = 559 samples) than in adenocarcinomas, with an overall prevalence of 78.4%, or 438 samples. Mutations at the traditional hotspot R248 and the lung-enriched hotspot R158 occurred most commonly, each with a frequency of 7.4%, or 22 of 297 missense mutations. The lung-enriched hotspots V157 and A159 also occurred with higher frequency than in non-lung cancers. In small cell lung cancer (SCLC, n = 236 samples), TP53 alterations occurred in 78.0%, or 184 samples. This is lower than in other published cohorts, where alterations were seen in 98% of samples (22). Although the general distribution of mutations is starkly different from that in lung adenocarcinoma and squamous cell lung cancers, with mutations seen at a variety of amino acid residues including H179, Y220, and P278, the lung-enriched loci V157 and R158 continue to be hotspots in SCLC. Together, these findings show that although minor differences exist among histologic groups—primarily, high R158 frequencies in squamous cell lung cancer in particular—mutations at the lung-enriched hotspots have a high relative abundance in all subtypes of lung cancer.
TP53 mutations in AACR Project GENIE registry v3.0.0
Recent studies have defined subsets of lung adenocarcinoma based upon frequently co-occurring mutations, namely KRAS and TP53 or STK11, with impacts on treatment response and survival (23–25). We sought to determine whether comutational events differ among lung cancers harboring the lung-enriched mutations in the Project GENIE cohort. Similar to all TP53-altered lung adenocarcinomas, the most frequently co-occurring mutations in V157-mutated, R158-mutated, and A159-mutated lung adenocarcinomas were found in KRAS and EGFR (which are mutually exclusive). Co-mutated TP53 and KRAS were slightly more common among lung-enriched p53-mutated samples (35.6%, or 52 of 146 samples vs. 26.0%, or 434 of 1,667 samples; Fig. 2A and Table 2A). Co-occurrence of KRAS and STK11 in p53-mutant tumors was rare in both groups (2.7%, or 4 of 146 samples vs. 3.1%, or 52 of 1,667 samples), consistent with existing literature (26). Among lung squamous cell cancers, the most commonly comutated genes with TP53 were KMT2D and CDKN2A (Table 2B). In the lung-enriched mutant p53 cohort, the frequency of co-occurring CDKN2A mutations was higher (31.4%, or 11 of 35 samples) than among squamous cell lung cancers with any TP53 mutation (20.6%, or 59 of 287 samples; Fig. 2B). Together, these findings show that common co-occurring mutations (KRAS in lung adenocarcinoma and CDKN2A in squamous cell lung cancer) are even more frequent in tumors harboring lung-enriched p53 mutations.
Co-occurring mutations with TP53 in lung adenocarcinoma and lung squamous cell samples from AACR Project GENIE v3.0.0. Prevalence of common co-occurring mutations in A. lung adenocarcinoma, and (B) squamous cell lung cancers is shown for tumors harboring any mutation in TP53 vs. the cohort of lung-enriched mutations.
AACR GENIE lung adenocarcinoma co-occurring mutations
AACR GENIE squamous cell lung cancer co-occurring mutations
Although a comparison of lung tumors occurring in current or former smokers versus never smokers would better elucidate the relationship between tobacco carcinogens and the lung-enriched p53 mutations, 64 of the 65 samples with missense mutations at V157, R158, or A159 in the TCGA Pan-Lung Cancer study are from patients with a history of cigarette smoke exposure (19). However, this delineation has been addressed in older datasets, with the conclusion that the p53 mutation spectra are highly different between smokers and nonsmokers (7, 9). Newer datasets such as the AACR Project GENIE, which is not yet fully clinically annotated, may provide more detailed knowledge as smoking rates continue to decline in the United States (27).
In addition to the enriched loci at which TP53 mutations occur in lung cancer, the frequency of specific nucleotide changes is also altered in these tumors. Smoking-related lung cancer arises in a highly mutagenic environment and is associated with a specific G>T mutational signature. This G>T signature is found frequently in p53 mutations in lung cancer – in contrast to the G>A transitions that predominate p53 mutations in non-lung cancers – and supports the important role of smoking in oncogenic p53 mutations in the pathogenesis of lung cancer. Given this, one would expect that if G>T transversions at the canonical hotspots are possible and sufficient for lung tumorigenesis, the presence of non-classical nucleotide changes at noncanonical codons might be unnecessary. Notably, although G>T transversions are also the most common nucleotide substitution in V157 and R158 mutations, a minority of tumors show a G>C transversion (19). Furthermore, TCGA data reveals that A159, despite including a guanine, exclusively participates in G>C transversions (A159P) or C>T transitions (A159V) rather than the canonical G>T smoking-related signature. Finally, it has been noted that the G>T smoking signature dominates the nucleotide changes at the canonical hotspots R273 (51.6% of R273 mutations in the Pan-Lung Cancer Study), G245 (84%), and R248 (32%) which still occur with moderate frequency, albeit reduced. In contrast, at R175 there are no G>T transversions in the Pan-Lung Cancer Study but rather an increase in C>G transversions. Furthermore, there is an overall decrease in frequency of mutations at this traditional hotspot, which is usually among the most commonly altered amino acid residues in other solid tumors (19). This diversity of nucleotide changes suggests that factors beyond cigarette smoke exposure may play a role in selection for p53 mutation and cancer progression in the lung.
Additional support for preferential biological selection of the V157 mutant is illustrated by a case report of a 22-year-old patient with Li Fraumeni-like syndrome due to germline V157D p53 and PMS2 mutations who developed lung cancer with loss of heterozygosity (LOH) restricted to lung tumor cells (28). Notably, the proband's father also carried a germline p53 V157D mutation with heterozygous mutation noted in a sample of the colon cancer from which he died at age 31. Although one interpretation would be that selective pressure exists specifically in the lung for mutations at V157F p53, it is also possible that LOH in the lung adenocarcinoma resulted from its inability to tolerate the presence of even a single allele of wild-type p53 tumor suppressive capacity. Larger scale studies of Li Fraumeni patients harboring different p53 mutations have demonstrated a pattern of variable tumor type and age at onset depending on the mutant allele (29, 30).
Together, this evidence supports our hypothesis that while mutations at codons V157, R158, and A159 may occur as a result of tobacco carcinogen exposure, they may be preferentially selected in lung cancer for additional reasons such as GOF activities with inherent benefit to lung tumors.
Biochemical and molecular rationale for selection of p53 mutations
It is plausible that the lung-enriched mutations are selected for structural alterations resulting in a highly debilitated mutant protein. Fersht and colleagues characterized V157F as a severely destabilizing mutation, with a large number of neighboring side chains affected by the substitution of a larger, hydrophobic phenylalanine for valine, which eliminates function at physiological temperature (31). Located in the β-sandwich strand S4, V157F mutations cause global effects on the DNA-binding core domain and loop-sheet helix motif, which directly contacts DNA (32). Additional structural studies, however, have reported that the most common p53 mutant proteins exhibit a wide range of structural debility, suggesting that p53 mutations are selected for more than just a disrupted structure (5, 33).
Early in vivo studies of GOF effects by mutant p53 demonstrated allele-specific differences in tumor type and invasiveness (34–36). Specifically, Olive and colleagues found, in genetically engineered murine models, that p53R270H/− and p53R172H/− mice exhibited similar survival to that of p53−/− mice, but the heterozygous mutant p53 mice exhibited an increased incidence of epithelial tumors with evidence of invasion and/or metastasis whereas none of the null mice in the study developed carcinomas (34). Similarly, p53R270H/+ mice had an increased incidence of carcinomas, most commonly lung adenocarcinomas. In contrast, p53R172H/+ mice were found to develop osteosarcomas at a rate twice that of p53R270H/+ mice. These results together demonstrated GOF effects (invasion and metastasis) due to structural (R175H) and conformational (R273H) mutations in p53 and the emergence of distinct tumor spectra depending on the specific allele present. More recently, Hanel and colleagues demonstrated differential GOF in tumorigenesis, with accelerated tumor onset of all tumor types and shortened survival in humanized p53 knock-in (HUPKI) p53hupkiR248Q/− mice, in contrast to p53hupkiG245S/− mice which were similar to p53 null mice in tumor latency and survival (37). To examine the effect of p53 alterations on K-ras-induced lung cancer, Jackson and colleagues generated a compound conditional mouse harboring the LSL-K-rasG12D allele and combinations of p53LSL.R270H, p53LSL.R172H, and p53Flox alleles (38). This work demonstrated that mice with endogenous expression of the contact mutant p53R270H develop significantly greater tumor burden compared with mice expressing the structural mutant p53R172H. The authors postulated that this difference in tumorigenic potential may explain the relative underrepresentation of mutations at the conventional hotspot codon R175 in human non–small cell lung cancers. Using a similar model, Turrell and colleagues found genotype-specific changes in transcriptional GOF, with increased dependency on mevalonate pathway gene expression in R270H but not R172H lung tumor cells (39). Furthermore, p53R270H mutant KrasG12D-driven lung tumors exhibited a robust decrease in cell proliferation and survival upon mevalonate pathway inhibition by statin treatment, whereas p53-null and R172H lung tumors displayed no antiproliferative or apoptotic effect with statin therapy. Collectively, these data demonstrate that mutant p53 proteins exhibit GOF activities during tumor progression in vivo, and also that different hotspot mutations in p53 manifest GOF properties of varying magnitudes and tumor spectra. As yet, no murine studies have examined the lung-enriched p53 mutations at V157, R158, and A159.
Conclusions
We have proposed a paradigm for tissue-specificity in the selection of p53 mutations, which translates into varying frequencies of hotspot mutations unique to specific tumor types. Additionally, we hypothesize that while the unique pattern of mutations at V157, R158, and A159 in lung cancer is in part due to the mutagenic effect of tobacco carcinogens, these specific codons may also impart an oncogenic GOF leading to selection for V157, R158, or A159 mutations. Notably, the clustered nature of the lung-enriched loci suggests structurally or functionally related mutant p53 proteins, although the innate hypersensitivity of this genomic region to DNA adducts cannot be ruled out. From a translational standpoint, elucidating the biological activities associated with the lung-enriched p53 mutations will add to the growing understanding of mutant p53 GOF and may allow for specific therapies targeted against GOF pathways. Existing treatment strategies against mutant p53 include restoration of wild-type p53 function or degradation of mutant p53 proteins. The GOF pathways induced or repressed by lung-enriched mutant p53 may identify actionable targets for precision therapy in certain subsets of lung cancer. As an example, statins hold promise in targeting the mevalonate pathway, which is responsible for disordered and invasive morphologic GOF exhibited by p53 R273H mutant breast cancers (40). Many additional questions remain, including the impact of tissue-environment and microenvironment context on GOF activities and whether different mutant p53 alleles function through different molecular mechanisms.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: J.A. Barta, S.B. McMahon
Development of methodology: J.A. Barta
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.A. Barta
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.A. Barta, S.B. McMahon
Writing, review, and/or revision of the manuscript: J.A. Barta, S.B. McMahon
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.A. Barta
Study supervision: S.B. McMahon
Acknowledgments
This study was supported by NCI R01 CA164834 (to S.B. McMahon) and American Cancer Society 1300042-IRG-16-244-10 (to J.A. Barta).
- Received April 20, 2018.
- Revision received July 25, 2018.
- Accepted August 22, 2018.
- Published first September 17, 2018.
- ©2018 American Association for Cancer Research.