TP53 status was retrospectively analyzed and its predictive and prognostic values were examined in 379 patients with advanced lung cancer (Stage IIIB-IV) harboring at least one classic NSCLC driver mutation. Staging of the primary lung tumor, lymph node status and metastasis were assessed based on the American Joint Committee on Cancer 7th edition Tumor, Node and Metastasis (TNM) staging system of NSCLC (29). Patients (female to male ratio, 1:1.3; median age, 56.5 years; range, 26–82 years) were treated at any of the nine participating centers between September 2015 and October 2016. The inclusion criteria were: i) Patients diagnosed with advanced-stage lung cancer (stage IIIB-stage IV) of any histology harboring at least one classic NSCLC driver mutation; and ii) the patient was treated at any of the nine participating centers between September 2015 and October 2016. The exclusion criteria was patients with early-stage lung cancer (stage IA-IIIA) of any histology. Written informed consent was obtained from each patient for participation in the present study. Capture-based targeted sequencing was performed on the plasma samples using a panel consisting of 168 lung cancer-associated genes, spanning 160 kb of the human genome. The present study was approved by the Ethics Committee of Jiangsu Province Hospital (Nanjing, China).

Next generation sequencing was performed using a commercial panel comprising 168 lung cancer-associated genes (Lung Plasma; Burning Rock Biotech) in a Clinical Laboratory Improvement Amendments-certified laboratory as previously described (30). Briefly, circulating cell-free DNA was acquired from 4–5 ml of plasma using the QIAamp Circulating Nucleic Acid kit (Qiagen China Co., Ltd.) according to the manufacturer's protocol. A minimum of 50 ng of DNA is required for next-generation sequencing library construction. A DNA library for the next-generation sequencing experiments were constructed. Fragments between 200 to 400 base pairs (bp) from the DNA were end-repaired, phosphorylated and ligated with adaptors (Agencourt AMPure XP kit; Beckman Coulter, CA, USA). Purified DNA with adaptors were then hybridized with capture probes baits, underwent hybrid selection with magnetic beads, and PCR amplified. The quality and the size of the fragments were assessed using a Qubit 2.0 fluorimeter (Thermo Fisher Scientific, Inc., Waltham, MA, USA) with a dsDNA high-sensitivity assay kit (Thermo Fisher Scientific, Inc.). Indexed samples were sequenced on a Nextseq500 sequencer (Illumina, Inc.) with pair-end reads. An average coverage of 11,816× was reached with a limit of detection of 0.2%.

Data were analyzed using optimized pipeline for somatic mutation calling as previously described (30). Briefly, the sequence data were mapped to the reference human genome (hg19) using Burrows-Wheeler Aligner (version 0.7.10) (31). Local alignment optimization and variant calling were performed using Genome Analysis Tool kit (version 3.2) (32,33) and VarScan (version 2.4.3) (34). Variants were filtered using the VarScan fpfilter pipeline; loci with depth <100 were filtered out. Base calling in plasma samples required ≥8 supporting reads for single nucleotide variations and 5 supporting reads for insertion-deletion variations. Variants with population frequency >0.1% in the ExAC (http://exac.broadinstitute.org/), 1,000 Genomes (35), dbSNP (https://www.ncbi.nlm.nih.gov/snp/) (36) or ESP6500SI–V2 (https://evs.gs.washington.edu/EVS/) databases were grouped as single nucleotide polymorphisms and excluded from further analysis. Remaining variants were annotated with ANNOVAR (2016-02-01 release) (37) and SnpEff (version 3.6) (38). Analysis of DNA translocation was performed using Factera (version 1.4.3) (39). Copy number variations (CNV) were analyzed based on the depth of coverage data of capture intervals using an in-house developed algorithm. The limit of detection for CNVs was 1.5 and 2.64 for deletions and amplifications, respectively.

Disruptive mutations, as described previously (15), were defined as any mutation leading to a stop codon or missense mutations occurring within the L2-L3 loop of the DNA-binding domain, leading to a substitution with an amino acid of a different polarity or charge group. All other mutations were defined as non-disruptive. Hotspot exons were defined as exons 5–8, as previously described (17).

Since the data were not equally distributed, data are presented as the median. All statistical tests were conducted in R version 3.3.3 (The R Foundation for Statistical Computing, Vienna, Austria; http://www.r-project.org) and R Studio version 1.1.383 software (40), and all tests were two-sided unless otherwise specified. Pearson's correlation test was used to assess correlation between two continuous variables. Fisher's exact test was used to assess the association between two categorical variables. Survival times were illustrated by Kaplan-Meier curves with the P-value determined by log-rank tests or Cox regression models when a co-variant was included. All survival analyses were adjusted for age, sex, smoking history, stage and histology. P<0.05 was considered to indicate a statistically significant difference.

The TP53 status of 379 patients with advanced lung cancer (Stage IIIB to IV) harboring at least one classic NSCLC driver mutation with various histological types was assessed in the present study. Among them, 294 patients had EGFR mutations, 24 had anaplastic lymphoma kinase (ALK) rearrangements, 14 had erb-b2 receptor tyrosine kinase 2 mutations, five had MET Proto-Oncogene (MET) mutations, two had B-raf proto-oncogene mutations, four had ROS proto-oncogene 1, receptor tyrosine kinase fusions, 11 had KRAS fusions and five had ret proto-oncogene fusions; the remaining 20 patients had dual driver mutations (Table I). In the examined cohort, 213 (56.2%) were female and 166 (43.8%) were male patients, and 156 had a history of smoking. The median age of this cohort was 56.5 years, ranging between 26 and 82 years. The cohort primarily consisted of adenocarcinoma (332/379), followed by squamous cell carcinoma (SqCC) (11/379) and small cell lung cancer (10/379). A total of 165 patients (43.5%) had not received tyrosine kinase inhibitors (TKI) as a treatment regimen and were considered TKI-naïve; among them, 84 were treatment-naïve and the remaining were previously treated with chemotherapy. A total of 214 patients (56.5%) were previously treated with TKI; among them, 184 patients were previously treated with one line of treatment. A total of 173 patients were treated with EGFR-TKI and the remaining 11 were treated with crizotinib, an ALK inhibitor (41). A total of 30 patients received two lines of treatment. Among them, 24 were treated with first and third generation EGFR-TKIs, two patients were treated with first generation EGFR-TKIs followed by a tyrosine-protein kinase MET inhibitor and the remaining three patients were treated with ALK inhibitors, Crizotinib followed by ceritinib (41). Table I summarized the detailed clinical characteristics of the cohort investigated.

A capture-based ultra-deep targeted sequencing analysis was performed as described in Materials and methods and (30) on the plasma samples obtained from 379 patients with advanced lung cancer to investigate their TP53 status and the prognostic value of the TP53 mutations. The prevalence of TP53 mutations in the cohort was 49.9% (189/379), which is comparable to its prevalence in the western population (22). Among them, 163 patients harbored mutations in hotspot exons: 48 Mutations were on exon 5, 31 on exon 6, 40 on exon 7 and 46 on exon 8 (Table I). Two patients had two TP53 mutations (data not shown). A total of 84 patients had disruptive mutations. The distribution of TP53 mutations is also presented in Fig. S1. No association between TP53 mutations and smoking history was observed when all TP53 mutations were taken into consideration (data not shown). The examined cohort primarily consisted of patients with adenocarcinoma, squamous cell carcinoma or small cell lung cancer. The percentages of TP53 mutations were comparable in patients with adenocarcinoma (48.3%) and small cell lung cancer (48.4%) (data not shown). All patients investigated in the present study with small cell lung cancer carried TP53 mutations, in line with previous studies (42), suggesting that a significant percentage of patients with small cell lung cancer have TP53 mutations. Next, the mutation spectra of patients with TP53 mutations were compared with patients without TP53 mutation (Fig. S2). The present results revealed that the most frequent mutation was EGFR in both groups. In addition, a larger number of RB1 mutations and MET amplifications were present in patients with TP53 mutation (Fig. S2).

The correlations between TP53 mutations, classified according to different systems, and clinical parameters, including but not limited to the TNM stage, smoking history and metastatic sites were further investigated. A positive correlation was identified between TP53 mutations and the TNM stage when all mutations were considered collectively. Patients with TP53 mutations were more likely to have advanced N (P=0.004, r=0.161) and M (P=0.004, r=0.151) stages of the disease (Fig. 1A and B). A positive correlation was also identified between liver metastasis and TP53 mutations when all mutations were considered collectively (P=0.001, r=0.187). In the cohort, 67.9% patients (55/81) with TP53 mutations had liver metastasis; in contrast, 45.5% (127/279) of patients without a TP53 mutation had liver metastasis (P=0.001) (Fig. 1C). These trends also existed when TP53 mutations were classified according to the location of the mutation or functional effects on the p53 protein. Hotspot exon and non-hotspot exon mutations demonstrated a significant correlation with N (P=0.036 and 0.012, respectively) and M (P=0.012 and 0.001, respectively) stage (data not shown). When TP53 mutations were classified as disruptive or non-disruptive mutations, TP53 disruptive mutations exhibited a non-significant correlation with N (P=0.08) and M (P=0.1) stages. A strong correlation was observed between liver metastasis and TP53 mutation, regardless of the classification system (P<0.001). Only TP53 hotspot exon mutations were significantly correlated with bone metastasis (P=0.032) (data not shown). Collectively, the present results suggested that only certain types of TP53 mutations were correlated with the clinical parameters analyzed, providing evidence for the hypothesis that not all TP53 mutations are equal.

Conflicting findings regarding the prognostic values of the TP53 status were reported, which can be partially attributed to the lack of a unifying classification system (21–24). The prognostic value of TP53 mutations was evaluated using the two aforementioned classification systems. Patients were grouped into three groups based on their treatment history: i) TKI-naïve (n=165); ii) previously treated with one line of TKI (n=184); and iii) treated with ≥2 TKI (n=30). In the TKI-naïve patients, 74 patients had TP53 mutations. Among them, 64 had mutations in hotspot exons (exons 5–8) (2,28) and the remaining 10 had mutations in other exons. A total of 33 patients had disruptive mutations and 45 had non-disruptive mutations. No association was observed between TP53 mutations and OS when all TP53 mutations were considered collectively or classified according to their location (hotspot exon vs. non-hotspot exon mutations) and functional effects on p53 protein (disruptive vs. non-disruptive; Fig. 2A-C). Notably, mutations occurring on exon 8 were found to be associated with OS (P=0.013) when controlling for age, sex, stage and histology. A total of 14 patients with mutations in exon 8 had a shorter median OS compared with the remaining 91 patients who had no mutations in exon 8 (Fig. 2D).

In patients previously treated with one line of TKI treatment (n=184), an analysis revealed that TP53 status, when all mutations were considered collectively, was found to be marginally associated with OS (P=0.05). Patients with TP53 mutations had a shorter OS compared with patients with WT TP53 (Fig. 3A). Such associations were significantly enhanced when only mutations occurring on exon 8 were considered (P=0.032; Fig. 3B). A total of 26 patients had mutations in exon 8, including 22 missense, three frameshift and one nonsense mutation (Fig. 3B). However, when all hot exon mutations were considered collectively, no association was observed (P=0.083; Fig. 3C). The same trend was observed in TKI-naïve patients, disruptive (P=0.081) and non-disruptive (P=0.106) mutations were not significantly associated with OS (Fig. 3D).

In the cohort, 99 patients were undergoing osimertinib treatment, a third-generation EGFR TKI (43). The impact of TP53 mutations on OS in osimertinib-treated patients was subsequently analyzed. In this cohort, 32 patients had WT TP53 and 67 patients had TP53 mutations (data not shown). No association was observed between TP53 status and OS, regardless of the classification system (data not shown). In the examined cohort of patients, 62 patients possessed EGFR 19 del and 37 possessed EGFR L858R (data not shown). No association was observed between TP53 status and OS, regardless of classification system, in patients harboring EGFR L858R (data not shown). In patients harboring EGFR 19 del concurrent to T790M treated with osimertinib, non-disruptive mutations (P=0.031) were found to be associated with OS (Fig. 4A). A total of 14 patients with non-disruptive TP53 mutations exhibited a significantly shorter OS compared with patients with WT TP53. All TP53 mutations (P=0.156), particularly disruptive mutations (P=0.690) as well as all hotspot exon mutations (P=0.128) including exon 8 (P=0.075) did not exhibit an association with OS (Fig. 4).

In patients treated with two lines of TKI, the analysis revealed that TP53 mutations, when considered collectively, were identified to be associated with OS (P=0.037; Fig. 5A). Mutations occurring on exon 8 (P=0.079) as well as all hot exon mutations (P=0.052) also exhibited an association with OS (Fig. 5B and C). In contrast, disruptive mutations (P=0.086) did not exhibit an association with OS, whereas non-disruptive mutations (P=0.048) exhibited a marginal association with OS (Fig. 5D). All analyses were controlled for smoking status. Collectively, the data provided evidence supporting the hypothesis that TP53 mutations are not equal. Furthermore, by comparing multiple TP53 mutation classification systems, it was identified that mutations in exon 8 may serve as prognostic biomarkers across all patients.