Diagnosis of Li-Fraumeni Syndrome: Differentiating TP53 germline mutations from clonal hematopoiesis

1 INTRODUCTION

Deleterious germline mutations in the TP53 gene (MIM# 191170) cause the Li-Fraumeni cancer predisposition syndrome (LFS1, MIM# 151623). LFS1 presents with a variety of tumor types and the TP53 gene is therefore covered by most diagnostic cancer gene panels. Due to the central role of the TP53 protein in tumor initiation, the classification of deleterious TP53 mutations identified in blood-derived DNA as disease-causing appears to be self-evident. The interpretation of TP53 variants identified in a germline diagnostic setting, however, remains challenging. Firstly, most pathogenic TP53 mutations are missense mutations (Bouaoun et al., 2016) and cannot be easily classified based on mutation type. Secondly, TP53 mutations may arise de novo, leading to somatic mosaicism (Forsberg, Gisselsson, & Dumanski, 2017). Next-generation sequencing (NGS) allows detecting genetic variants with a high read depth. While inherited, heterozygous germline variants usually show a variant fraction (VF) of approximately 50%, TP53 variants with a VF below 50% were described, suggesting de novo somatic mosaic variants (Weitzel et al., 2017). Using blood-derived DNA, Swisher et al. (2016) identified deleterious somatic mosaic TP53 variants in 10 out of 686 patients with ovarian cancer (OC). Paired neoplastic tissue was available for four women with TP53 mutations—in no case, the TP53 mutation was identified in the tumor-derived DNA. The evidence for deleterious TP53 mutations identified purely in blood-derived DNA but not in the tumor of the patient seeking advice requires further validation since it may have severe implications for genetic counseling.

2 MATERIALS AND METHODS

Genomic DNA was isolated from venous EDTA blood samples using standard methods. Using blood-derived DNA, we screened 523 unselected patients with primary diagnosis of OC (n = 281) or platinum-sensitive recurrent OC (n = 242) and 1,053 cancer-free female control individuals for deleterious variants in the TP53 (MIM# 191170) and PPM1D (MIM# 605100) genes by hybridization capture-based NGS (Agilent SureSelect XT protocol). The patient cohort was previously screened for pathogenic germline mutations in established cancer predisposition genes (observational AGO-TR1 study). Healthy controls were recruited by a study on civilization diseases (LIFE study, https://life.uni-leipzig.de/). The studies were approved by the local ethic committees. All participants gave their written informed consent. The study sample and the methodologies were previously described in detail (Harter et al., 2017). The AGO-TR1 study protocol was approved by the ethical committee of the Landesaerztekammer Nordrhein (Nr. 2014340) and registered (NCT02222883); all patients gave written informed consent prior to any study related procedure. The hybridization capture-based NGS method was suitable for the analysis of DNA derived from either blood- or formalin-fixed paraffin-embedded (FFPE) tumor samples (Agilent SureSelect XT protocol). For DNA isolation from FFPE tumor samples, hematoxylin and eosin-stained 3 μm tissue sections were analyzed by an experienced pathologist at the Institute of Pathology at the University Hospital Bonn, Germany. Tumor areas containing >80% tumor nuclei were chosen for DNA isolation using standard techniques. DNA quantification was performed using a Nanodrop® ND-1000 spectral photometer (NanoDrop Technologies, Wilmington, DE).

For the verification of TP53 variants identified in blood/ tumor-derived DNA by hybridization capture-based NGS, TP53-positive DNA samples were re-analyzed using an amplicon-based Fluidigm Access Array 48.48 system (Fluidigm, San Francisco, CA) for target enrichment, covering all coding TP53 exons and exon-flanking intronic sequences (TP53 reference transcript NM_000546.5). DNA libraries were sequenced in 150 base paired end mode with the Mid Output Kit v2. on a Nextseq 500 device (Illumina, San Diego, CA). For evaluation of the variants, BCL files were demultiplexed, converted into FASTQ format using bcl2fastq2 Conversion Software v2.19.1.403. Sequence reads were mapped to the human reference genome assembly GRCh37 including decoy sequences (hs37d5) using Burrows-Wheeler Aligner (BWA) v0.7.15 (Li & Durbin, 2009), and target-specific primer sequences were removed using BAMClipper v1.1.1 (Au, Ho, Kwong, Chan, & Ma, 2017). The Genome Analysis Toolkit (GATK) v3.8 (DePristo et al., 2011; McKenna et al., 2010) was used for realignment of insertions and deletions and quality recalibration. Variant calling was performed using Freebayes v1.1.0 (https://arxiv.org/abs/1207.3907; https://github.com/ekg/freebayes), claiming a minimum alternate allele fraction of 0.05 (via argument –min-alternate-fraction), and base and mapping qualities of at least 20 (via arguments –min-base-quality and –min-mapping-quality). All positions found by FreeBayes (Garrison E, Marth G. Haplotype-based variant detection from short-read sequencing. arXiv preprint arXiv:1207.3907 [q-bio.GN] 2012) with a minimum alternative allele fraction of 0.05 and a minimum read depth of 60 were considered as putative variants, irrespective of the predicted genotype. Read depths and base counts of positions with alternative allele fractions below 0.05 were obtained via the mpileup utility of samtools v1.9, with minimum mapping and base qualities set to 20 (via arguments -min-MQ and –min-BQ).

3 RESULTS

In our patient sample, potentially deleterious missense variants in the TP53 gene were identified in blood-derived DNA of six out of 523 patients with OC using hybridization capture-based NGS (Table 1). The VFs of three TP53 variants detected in three patients (#1–#3) were 55%, 50% and 49%, respectively, compatible with VFs usually observed for germline variants. In the remaining three patients (#4–#6), four TP53 variants with lower VFs of 34%, 26%, 17%, and 7%, respectively, were observed (Table 1). DNA samples derived from paired neoplastic tissue from all six patients were analyzed by NGS. TP53 variants with a VF of approximately 50% in blood-derived DNA (patients #1–#3) were also present in the corresponding tumor samples (Table 1). The VFs were elevated in the tumor samples of two patients (#1, #2), suggesting loss of the wild-type TP53 alleles. In contrast, the TP53 variants with lower VFs (patients #4–#6) were not or only barely detectable in the corresponding tumor. Of note, different somatic de novo TP53 variants were observed only in the tumors of the latter three patients (Table 1), which were classified nonfunctional in the IARC TP53 database (Bouaoun et al., 2016) and pathogenic according to the UMD TP53 database and Seshat TP53 variant classification tool (Soussi, Leroy, & Taschner, 2014). Two variants listed in the ClinVar database were classified likely pathogenic/pathogenic. In summary, these data suggest that deleterious TP53 variants identified in blood-derived DNA were not causal for the patients’ cancer in three out of six cases.

All TP53 variants identified in blood-derived DNA using hybridization capture-based NGS were independently verified using an amplicon-based assay for target enrichment prior to NGS, with similar VFs observed (Table 2). Thus, TP53 variants listed in Table 1 represent true positive NGS variant calls. No additional TP53 variants were observed in the verification analysis.

The occurrence of deleterious mutations with a low VF may be caused by chemotherapy-induced and/or age-related clonal hematopoiesis (CH), in which the deleterious mutations only affect the hematopoietic stem and progenitor cells in the bone marrow and no other compartments of the body (Genovese et al., 2014; Jaiswal et al., 2014; Swisher et al., 2016). To differentiate whether the occurrence of TP53 variants with low VFs may be chemotherapy-induced and/or age-related, we analyzed 1,053 cancer-free female control individuals for deleterious variants in the TP53 gene by hybridization capture-based NGS. In this large control sample with a mean age at blood draw of 59.3 years (range 19–80) similar to the mean age at blood draw in all 523 patients enrolled in the AGO-TR1 trial (59.9 years, range 18–93), no pathogenic TP53 variant and no other TP53 variant with a low VF was observed, suggesting that age-related CH affecting the TP53 gene represents a rare event.

At the time of the blood draw, patients with deleterious TP53 variants with a low VF in blood had completed first line taxane/platinum-based chemotherapy (Table 1). Consequently, we suggested that the low VF-variants observed in the TP53 gene were chemotherapy-induced rather than age-related. Mutations affecting the PPM1D gene were originally thought to represent mosaic events leading to predisposition to OC (Ruark et al., 2013). Subsequent studies have elucidated that such events are enriched in the peripheral blood of patients with prior chemotherapy (Coombs et al., 2017). In our study sample, 24 out of 523 patients (4.6%) carried truncating variants affecting the PPM1D gene (Table 3), with generally low VFs (≤40%) in blood-derived DNA which were not compatible with heterozygous germline alterations. In corresponding tumor-derived DNA samples, PPM1D variants were not or only barely detectable (Table 3). Traces of mutant alleles in the tumor may be explained by infiltration of the tumor tissue with blood cells. Of note, 18 out of 24 PPM1D-positive patients had completed first line platinum-based chemotherapy prior to blood draw. In five cases, blood was drawn during 1st line chemotherapy (Table 3). In the age-matched control sample (n = 1,053), PPM1D variants were extremely rare with only one 77-year-old woman carrying a nonsense variant with a low VF affecting the PPM1D gene (c.1654C>T, p.(Arg552*), VF 15%, 30 out of 198 reads).

In summary, 26 out of 523 (5.0%) patients enrolled in the AGO-TR1 trial carried TP53 and/or PPM1D variants with low VFs in blood-derived DNA (23× PPM1D only, 2× TP53 only, 1× PPM1D and TP53 [patient #6 in Tables 1 and 3]) versus one out of 1,053 (0.1%) in age-matched control individuals (1× PPM1D). Thus, the event of a CH is substantially enriched following standard chemo therapy and most likely account for the TP53 variants with low VFs observed in this study. Notably, the overall prevalence of pathogenic germline mutations in validated OC predisposition genes ATM (MIM# 607585), BRCA1 (MIM# 113705), BRCA2 (MIM# 600185), BRIP1 (MIM# 605882), MSH2 (MIM# 609309), MSH6 (MIM# 600678), RAD51C (MIM# 602774), RAD51D (MIM# 602954) according to Lilyquist et al. (2017) in the AGO-TR1 study was 25.2% (132 out of 523) (Harter et al., 2017). In the subgroup of patients showing variants with low VFs in the CH-associated genes TP53 and PPM1D, the proportion of germline mutation carriers was elevated (41.7%, 10/24; Table 3) compared with the overall patient sample, though not reaching levels of significance.

4 DISCUSSION

We demonstrate that deleterious TP53 variants identified in blood-derived DNA of patients with OC (AGO-TR1 trial, NCT02222883) were not causal for the patients’ cancer in three out of six TP53-positive cases. The paired analysis of blood/tumor-derived DNA of OC patients along with the analysis of 1,053 age-matched healthy female control individuals revealed that, in three out of six patients, TP53 mutations with low VFs arise from chemotherapy-induced CH. To avoid false-positive molecular genetic diagnoses of LFS1, we suggest that conspicuous TP53 test results in patients who received chemotherapy prior to blood draw should be complemented with additional tissue testing, excluding the hematopoietic compartment (e.g., tumor tissue). Following these analyses, we now consider the TP53 germline variants observed in the patients #1–#3 (Table 1) likely pathogenic.

Genovese et al. (2014) proposed that CH is associated with increased risks of hematologic cancer. In line with this suggestion patient #6, the only patient who was shown to carry both, TP53 and PPM1D variants, developed an acute myeloid leukemia. Whether chemotherapy-induced CH may be a risk factor for therapy-associated secondary hematologic malignancies needs to be clarified in larger prospective studies. In the subgroup of patients showing CH-associated alterations in the TP53 and PPM1D genes, however, the proportion of patients carrying germline mutations in validated OC predisposition genes was 1.7-fold higher than in the overall patient sample. Thus, we suggest that patients with a heterozygous inactivation of OC predispositions genes are prone to chemotherapy-induced CH. Whether germline cancer predisposition may be a risk factor for therapy-associated secondary hematologic events needs to be clarified.

CONFLICT OF INTEREST

Philipp Harter: Consulting or Advisory Role-AstraZeneca, Roche/Genentech, Tesaro, Clovis, Pharmamar Lilly, Sotio; Research Funding-AstraZeneca (Inst); Travel, Accommodations, Expenses-Medac; Stefan Kommoss: Honoraria-Astra Zeneca, Roche; Consulting or Advisory role-Roche, Tesaro, Astra-Zeneca; Travel, Accommodations, Expenses-Tesaro, PhermaMar, AstraZeneca; Katharina Prieske: Travel, Research funding-Medac Oncology; Honoraria-Astra Zeneca, Roche. Beyhan Ataseven: Advisory Role-Tesaro and Roche/Genetech; Honoraria-Roche/Genentech, AstraZeneca, Amgen; Travel Support-Roche/Genentech; Rita K. Schmutzler: Honoraria-AstraZeneca; Consulting or Advisory Role-AstraZeneca; Research Funding-AstraZeneca (Inst); Eric Hahnen: Honoraria-AstraZeneca, Consulting or Advisory Role-AstraZeneca, Research Funding-AstraZeneca (Inst); Konstantin Weber-Lassalle, Jan Hauke, Corinna Ernst, Frederik Marmé, Nana Weber-Lassalle, Dimo Dietrich, Julika Borde, Esther Pohl-Rescigno, Alexander Reuss, Christoph Engel, Julia Stingl, no relationships to disclose. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.