Interrogating the genomics of circulating tumor DNA (ctDNA; the liquid biopsy) has advantages in patients in whom tissue biopsy is difficult. However, the reported concordance between genomic analysis of tissue DNA and ctDNA is variable among studies. Herein, we characterized the clinical implications of the relationship between mutations in TP53 genes in tissue DNA versus ctDNA. The molecular profiles of both liquid (Guardant Health) and tissue (Foundation Medicine) biopsies from 433 patients were analyzed (pan-cancer setting). In 71/433 (16%) cases, all same TP53 mutations were detected in both tissue DNA and ctDNA; in 18/433 (4%), same mutation plus additional mutation/mutations; and in 27/433 (6%), different TP53 mutations were detected. In 99/433 (23%) cases, TP53 mutations were detected only in tissue DNA; in 43/433 (10%), only in ctDNA; and in 175/433 (40%), no TP53 mutations were detected in either test. When TP53 mutations were identical in tissue and ctDNA, the alterations were enriched for nonsense mutations, and survival was significantly shorter in multivariate analysis (as compared with different mutations in ctDNA vs. tissue or no mutations); this finding was independent of tumor type, time interval between tests, and the %ctDNA for TP53 mutations. In summary, in 16% of 433 patients with diverse cancers, TP53 mutations were identical in tissue DNA and ctDNA. In these individuals, the alterations were enriched for stop–gain (nonsense) mutations (results in a premature termination codon). Though unknown confounders cannot be ruled out, these patients fared significantly worse than those whose ctDNA and tissue DNA harbored different TP53 mutation portfolios or no TP53 mutations.
Cancer is caused by a sequence of acquired somatic genomic aberrations (1). Recent advances in cancer genome sequencing led to large-scale projects characterizing the genomic landscape of many tumor types (2, 3). These projects contributed to the identification of cancer genomic involvement in specific tumor types (4), and, in addition, it was realized that, often, each patient's tumor harbors a unique mutational profile (5, 6). In the last decade, developments in sequencing technology made tumor genomic analysis feasible for clinical application. In parallel, many new drugs targeting specific cancer genes have been developed and approved. These coupled endeavors hold promise for the application of a personalized medicine approach in cancer treatment (7–9). This approach already led to major clinical achievements such as targeting BRAF V600E mutations in melanoma (10) and EGFR mutations in lung cancer (11), among others. It was also shown, in a meta-analysis of clinical trials, that a personalized (biomarker-based, especially genomic biomarkers) approach is more beneficial compared with chemotherapy or targeted therapy of unselected patients (12).
Traditionally, genomic analysis of tumors is performed on DNA from tumor specimens obtained by surgery or biopsy. Circulating tumor DNA (ctDNA) represents tumor-derived DNA molecules that circulate in the bloodstream. These molecules can be isolated and analyzed for genomic alterations (“liquid biopsy”; ref. 13). ctDNA has the advantage of requiring only a small tube of blood and, hence, being less invasive than tissue biopsy. However, tissue biopsy still remains standard in most cases for the initial diagnosis and molecular profiling. The advantages of liquid biopsy are especially exploitable for cases in which tumor material is not sufficient for molecular profiling, to monitor response to therapy, and to identify the emergence of mutations causing resistance and/or that can be targetable by other drugs (14–17).
The concordance level between genomic analysis of tumor DNA and ctDNA when taken simultaneously can be as high as 80%–90% (18–21), though other investigators have shown lower concordance (15). It should also be kept in mind that there may be biologic reasons that underlie differences between tissue and ctDNA results: (i) ctDNA can be suppressed by therapy; (ii) tissue DNA tests the genomics in a small piece of tissue, whereas ctDNA may reflect shed DNA from multiple metastatic sites (hence, better representing tumor heterogeneity; ref. 15); and (iii) shedding of DNA into the bloodstream may be limited from some sites. To our knowledge, the clinical implications of the relationship between tumor DNA and ctDNA concordance for TP53 are not established.
Somatic mutations in the tumor suppressor TP53 gene are the most frequent alterations in human cancers. More than 80% of somatic and germline TP53 alterations are missense mutations leading to the synthesis of a stable mutant protein that accumulates in the nucleus of tumor cells (22). Analysis of the data generated by the various cancer genome projects highlights the high frequency of TP53 mutations and reveals that several TP53 hot spot mutants are the most common oncoprotein variants expressed in several types of tumors (22). Hence, focusing on mutations in TP53 seems an optimal starting point to investigate the clinical importance of comparison between tumor tissue DNA and ctDNA taken at different time points of the disease course.
In the current study, we hypothesized that patients with greater concordance between tissue and blood TP53 alterations might have poorer survival because the concordance might reflect higher variant allele fraction of TP53 in tumor. We analyze a unique clinical database of 433 patients with different tumor types. All patients underwent molecular profiling of tumor DNA and ctDNA. We focus our investigation on the relationship between both tests for TP53 mutation status and the clinical correlates of this relationship.
Materials and Methods
The molecular profiles of both liquid and tissue biopsies from 433 consecutive eligible patients with different cancer types, seen at the University of California San Diego Moores Cancer Center starting in June 2014 were analyzed. We included patients who had no prior treatment as well as those with prior systemic therapies. The last ctDNA specimen was obtained on 09/2017. Eligibility implied adequate follow-up and demographic data availability and patient meeting UCSD IRB guidelines for consent or waiver. Demographics of each of these patients were provided by chart review, including, but not limited to age, gender, cancer diagnosis, tumor origin, date of biopsy report or blood test, date of diagnosis, and survival time (Fig. 1). The study was performed in accordance with the UCSD internal review board-approved protocol (NCT02478931) and for any investigational therapies for which the patient gave written consent.
The ctDNA molecular profiles were provided by Guardant Health Inc. (http://www.guardanthealth.com/); tissue testing was performed by Foundation Medicine (https://www.foundationmedicine.com/genomic-testing#how-does-it-work). Both laboratories are Clinical Laboratory Improvement Amendment (CLIA)-accredited and Foundation Medicine test is now FDA approved for reporting pathogenic variants. Detailed Materials and Methods are previously published (23, 24).
As reported in Lanman and colleagues (24), 5–30 ng of ctDNA was isolated from plasma (two 10 mL Streck tubes drawn for each patient) and sequencing libraries were made with custom in-line barcode molecular tagging and complete sequencing at 15,000 × read depth. The panels use hybrid capture followed by next-generation sequencing (NGS) of the crucial exons in a panel of 54–73 genes and report all four major types of genomic alterations (indels, fusions, point mutations, and copy-number amplifications). Post-sequencing bioinformatics matches the complementary strands of each barcoded DNA fragment to remove false-positive results (24). The variant allele fraction (%ctDNA) is calculated as the number of mutated DNA molecules divided by the total number (mutated plus wild type) of DNA fragments at that allele. We used the maximum %ctDNA if a patient had two different TP53 alterations, unless we referred to a specific TP53 alteration. The majority of cell-free DNA is wild type; hence, the median %ctDNA of somatic alterations is <0.5%. The analytic sensitivity reaches detection of one to two single-mutant fragments from a 10-mL blood sample (0.1% limit of detection), and the analytic specificity is greater than 99.9999% (24).
Tissue NGS was performed at Foundation Medicine (CLIA laboratories) with assay panels of 236 or 315 genes, with detailed Materials and Methods as previously described (www.foundationmedicine.com; ref. 23). Average depth of sequencing was greater than 250x, with 100x at >99% of exons. This method of sequencing allows for detection of copy-number alterations, gene rearrangements, and somatic mutations with 99% specificity and >99% sensitivity for base substitutions at ≥5 mutant allele frequency and >95% sensitivity for copy-number alterations. A threshold of ≥8 copies for gene amplification was used. All tissue samples (Formalin-fixed, paraffin-embedded) were reviewed by a pathologist to ensure a sample volume of ≥1 mm3, nucleated cellularity ≥80% or ≥30,000 cells and that ≥20% of the nuclei in the sample were derived from the tumor.
Variants of unknown significance
Synonymous alterations and other variants of unknown significance (VUS) were excluded and only characterized alterations were included in the analysis (23).
Characterized alterations refer to pathogenic alterations that are not VUSs.
Definition of TP53 mutations concordance type
For each patient, the specific TP53 alterations (or their absence) were compared between ctDNA and tissue DNA. We classified the concordance patterns into six types (Supplementary Table S1; Fig. 2): (i) Exactly the same TP53 mutation in tissue DNA and ctDNA; (ii) different TP53 mutations in tissue DNA and ctDNA; (iii) TP53 mutation/s only in tissue DNA, (iv) TP53 mutation/s only in ctDNA; (v) no TP53 mutation in both tissue DNA and ctDNA; and (vi) exactly the same + others TP53 mutations in tissue DNA and ctDNA.
Statistical analysis was performed using the R programming language version 3.5.1 (25). Cox proportional hazards regression model and multivariable analysis was performed using the survival (26) R package. Pairwise comparisons for the continuous time interval variable were performed by pairwise Wilcoxon test with FDR correction for multiple comparisons. Survival was examined by Cox regression; patients still alive at the last follow-up were censored at that time. Survival time was calculated from time of diagnosis and the defined event was death. Quartile thresholds were created using R ntile command; in case of quartile borders value overlap, the algorithm assigns the first samples to the lower quartiles. Mutations type annotation and description of their landscape of TP53 gene were performed using MutationMapper tool in cBioPortal (https://www.cbioportal.org/mutation_mapper; ref. 27) and Seshat (https://p53.fr/tp53-database/seshat; ref. 28). We defined T125M, Q331R, and Q331H as splice mutations despite the fact that cBioPortal defined them as missense mutations because despite the fact that they appear in exons, they influence splicing (29). (However, even if defined as missense mutations, the statistical relationships do not change). Visualization was performed by the R packages of ggplot2 (29), forest plots by the R package of forestmodel (30). Balloon plots representing the residuals of the χ2 test were performed using the R package of corrplot (31).
Dataset is available as Supplementary Table S2.
Our cohort included 433 individuals with several different tumor types (Fig. 1). All but 28 patients had advanced disease that was surgically unresectable and/or metastatic. For each patient, genomic analysis was carried out twice: (i) for DNA obtained from tissue biopsy (FoundationOne test) and (ii) for ctDNA (Guardant360 test). The median time between the tissue biopsy date and the blood draw for ctDNA was 3.7 months (range, −14.6 to 241.4). The median age at the time of ctDNA liquid biopsy was 62 years (range, 19–93 years) and the median age at diagnosis was 58.6 years (range 2.2 to 90.5 years). Approximately 55% of patients were women (Supplementary Table S3). Mutations in TP53 gene were found in 215 (49.7%) tissue biopsies and in 159 (36.7%) ctDNA analysis. The landscape of the mutations in both tissue and ctDNA are given in Supplementary Fig. S1; Supplementary Table S4. In both tissue DNA and ctDNA, most of the mutations are located in the active domain of TP53.
Relationship between tissue and ctDNA TP53 mutations
The type of TP53 mutation correlated with whether or not it was found in both tissue and ctDNA
Characterization of the relation of specific TP53 mutations in tissue DNA as compared with ctDNA resulted in six mutation concordance categories (see Materials and Methods; Supplementary Table S1; Fig. 2). There was an association between mutation type and the mutation concordance categories. Thus, testing the three major mutations concordance categories of: (i) “all same TP53 mutations,” (ii) “only tissue DNA TP53 mutations,” and (iii) “only ctDNA TP53 mutations” concordance categories—there was strong association (P = 2.6 e−10, χ2 test) between the mutation type and mutation concordance category. There were more (1) frameshift, (2) splice site, (3) in-frame deletion, and (4) "others" mutations in "only tissue DNA TP53 mutations" category whereas there were more nonsense mutations (“stop–gain”) in “all same TP53 mutations” category (Fig. 3; Supplementary Table S5).
Same TP53 mutations were more likely to be found in tissue and ctDNA when there were shorter time intervals between the two tests
There was a significant association between mutation concordance category and the absolute time interval between the tests (P = 0.038, χ2 test). As shown in Fig. 4A; Supplementary Table S6—“all same TP53 mutations” category is associated with the shorter time interval of <2 months and “only ctDNA TP53 mutations” category is associated with the longer time interval of >6 months between tissue biopsy and blood draw. A time-dependent linear analysis (instead of using the 2 and 6 month cutoffs between ctDNA and tissue sampling) showed no substantial change in the conclusions (see Supplementary Table S7).
Colorectal cancer was more likely to have the same TP53 mutations in tissue DNA and ctDNA than other tumor types
There was a significant association between tumor types and mutations concordance categories (P = 4.534e−07, χ2 test). As shown in Fig. 4B; Supplementary Table S8—“all same TP53 mutations” category was positively associated with colorectal tumors and negatively associated with brain tumors. “Only tissue DNA TP53 mutations” category was associated with brain tumors. In addition—“Same + other TP53 mutations” category was overrepresented in head and neck tumor type.
Relationship between tissue and ctDNA TP53 mutations and survival
Detecting the same TP53 mutations in blood and tissue was associated with shorter survival
We tested for the association between mutation concordance category and survival from diagnosis (Fig. 5). Testing for all six mutation concordance categories was significant (P = 0.014, cox regression test). The shortest survival was for patients in “all TP53 same mutations” category. This result was significant after multivariate cox regression controlling for age, sex, %ctDNA, number of mutations in ctDNA, tumor type and time interval between the two genomic tests (OR, 1.57; P = 0.03, cox regression test). In contrast, patients with the other mutation concordance categories had longer survival, which was comparable with patients with “no TP53 mutations in both” (OR range, 0.85–1.14, not significant). We also analyzed survival from ctDNA blood test and found that all same TP53 mutations versus other mutation categories (i.e., discordant ctDNA vs. tissue DNA TP53 mutation status) had significantly shorter survival: HR, 2.05; 95% confidence interval, 7–15.7months; P = 3 e−05.
Higher TP53 %ctDNA was associated with shorter survival from diagnosis
There was a significant association between higher %ctDNA in the blood for TP53 mutations and shorter survival from diagnosis (P = 0.0003, cox regression test; Fig. 6A). This factor was not significant in the multivariate analysis controlling for mutation concordance category, age, sex, tumor type, and time interval between the genomic testing (Fig. 5B).
Higher number of characterized mutations (pathogenic alterations that are not VUS—in TP53 together with other genes) was associated with shorter survival from diagnosis
There was a significant association between mutation number in tissue DNA (P = 0.0004, cox regression test) or in ctDNA (P = 1.1e−6, cox regression test) and survival (Fig. 6B and C, respectively). This was also confirmed for ctDNA in multivariate analysis controlling for mutation concordance category, age, sex, tumor type and time interval between the genomic testing (Fig. 5B).
In this study, we analyzed a dataset of 433 patients with cancer with various tumor types. The patients underwent genomic testing of tumor DNA and ctDNA at different points of time in their disease course. Mutations in TP53 gene were found in 49.7% tissue biopsies, and in 36.7% ctDNA analysis. The lower frequency of TP53 mutations in ctDNA is expected because the yield of ctDNA is lower compared with tissue DNA. We focused our analysis on characterized mutations in the TP53 gene (those that are no VUS). Specifically, we investigated the relationship between TP53 mutation state in the tumor DNA and ctDNA. Accordingly, we defined 6 mutational categories describing this relationship: (i) exactly the same TP53 mutation in tissue DNA and ctDNA; (ii) different TP53 mutations in tissue DNA and ctDNA; (iii) TP53 mutation/s only in tissue DNA, (iv) TP53 mutation/s only in ctDNA; (v) no TP53 mutation in both tissue DNA and ctDNA; and (vi) exactly the same + others TP53 mutations in tissue DNA and ctDNA.
We noted that there were several genomic and clinical associations with mutation concordance categories. First, mutation concordance category was associated with the mutations type. Specifically, there were more frameshift, splice site, in-frame deletion, and "others" mutations in "only tissue DNA TP53 mutations" category whereas there were more nonsense mutations (“stop–gain”) in “all same TP53 mutations” category. This may reflect increased selective advantage of nonsense mutations compared with missense mutations (32) that cause them to have higher variant allele frequency, and thus higher detection rates in ctDNA (33). An alternative explanation might be the possible presence of nonsense mutations in more aggressive parts of the tumor that more easily release DNA to the blood. Nevertheless, it is unclear why frameshift mutation are more abundant in “only tissue DNA TP53 mutations” category as they should have comparable selective advantage with nonsense mutations.
Second, as expected, the same mutation in TP53 appeared more frequently in both tissue DNA and ctDNA in cases for which the time interval between the tests was shorter, possibly reflecting bigger genomic distance at greater time intervals (34) due to tumor evolution. By contrast, there was higher frequency of “only ctDNA TP53 mutations” in patients for which there were more than 6 months between tissue DNA and ctDNA tests. This could be due to tumor evolution (including possible treatment-related selective pressure) that is more marked in the group with largest time interval between the tissue DNA and ctDNA tests.
The third association was with tumor type. The yield of ctDNA testing in brain tumors is among the lowest among tumor types (35), probably due to lower exposure of brain tumor to the circulation due to the blood–brain barrier (14). By contrast, higher rates of “all same TP53 mutations” categories were observed for colorectal cancer, perhaps reflecting higher frequency of ctDNA detection in this tumor type (14). Interestingly, we previously found colorectal cancers (vs. other malignancies) to also have higher concordance between blood and tissue results for KRAS alterations (36).
The fourth finding was the association of mutation concordance category with survival. Specifically, “all same TP53 mutations” category was associated with lower overall survival (OS), whereas the other mutations concordance categories had OS comparable with the category of “no TP53 mutation in both.” In multivariate analysis, this finding was independent of other potentially confounding factors, including the time interval between tests and the %ctDNA (which is believed to reflect tumor burden). The reason for this finding is unclear, but could be hypothesized to be related to a higher variant allelic fraction of TP53 mutation in the tumor (either primary or metastatic), which merits investigation in further studies where such data are available
There are several limitations in this study. The cohort is heterogonous and included diverse tumor types; future analyses of larger more homogenous cohorts are required to validate our findings. On the other hand, our observations may speak to the generalizability of the results across tumor types. In addition, the time intervals between the tissue DNA test and the ctDNA were variable. Nevertheless, multivariable analysis showed that the association between mutation concordance category and OS was independent of tumor type and time interval between tests. Also, it is conceivable that some of the different mutations in ctDNA represent clonal hematopoiesis (37). Another issue is that, though we found that when TP53 mutations were the same in tissue and ctDNA, survival was significantly shorter in multivariable analysis, statistical significance does not necessarily imply clinical significance, as some of the actual survival times were not dramatically different; furthermore, it is conceivable that there were unknown confounders in the analysis. Finally, we did not consider the effect of therapy on detection of ctDNA alterations, and we limited our analysis to the most common gene—TP53; future analyses might consider, including other genes, though they occur at significantly lower frequencies, and might interrogate the relationship with therapy. Notably, this is a real-world cohort, rather than a prospective analysis with pre-specified analyses. The above mentioned heterogeneity of the cohort represents real-world application of tissue DNA and ctDNA analysis and enables interesting insights into the clinical meaning of the relation of genomic findings in tissue DNA and ctDNA. However, verifying the true clinical significance of testing for TP53 or any other component of ctDNA will require prospective studies with more focused clinical questions.
To summarize, in this study we characterized the relationship between genomic testing of TP53 in tissue DNA and ctDNA. Patients with the same TP53 mutation portfolio in tissue and liquid biopsy had tumors enriched for premature termination codons (nonsense mutations). Furthermore, we have shown that the relationship between TP53 mutations in tissue DNA and ctDNA has clinical meaning. Specifically, patients in whom specific characterized TP53 mutations appear unchanged between tissue DNA and ctDNA had shorter survival.
Disclosure of Potential Conflicts of Interest
S. Rosenberg reports grants from Hadassah France during the conduct of the study, grants from Trudy Mandel Louis Charitable Trust and Munsa Fund, and personal fees from Barcode Diagnostics (consultant) outside the submitted work. S. Kato reports personal fees from Roche (speakers' fee), personal fees from Foundation Medicine (consulting) during the conduct of the study, and research grants from ACT Genomics, Sysmex, Konica Minolta, and OmniSeq. R. Kurzrock reports research funding from Genentech, Merck Serono, Pfizer, Boehringer Ingelheim, TopAlliance, Takeda, Incyte, Debiopharm, Medimmune, Sequenom, Foundation Medicine, Konica Minolta, Grifols, Omniseq, and Guardant; reports consultant and/or speaker fees and/or advisory board for X-Biotech, Neomed, Pfizer, Actuate Therapeutics, Roche, Turning Point Therapeutics, and TD2/Volastra; has an equity interest in IDbyDNA and CureMatch Inc.; serves on the Board of CureMatch and CureMetrix; and is a co-founder of CureMatch. No disclosures were reported by the other authors.
S. Rosenberg: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing-original draft. R. Okamura: Data curation, validation, writing-review and editing. S. Kato: Data curation, validation, writing-review and editing. T. Soussi: Formal analysis, validation, investigation, methodology, writing-review and editing. R. Kurzrock: Conceptualization, resources, supervision, funding acquisition, validation, investigation, methodology, writing-original draft, writing-review and editing.
The work was funded in part by National Cancer Institute grant P30 CA023100 and the Joan and Irwin Jacobs Fund philanthropic fund. Hadassah France.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.