PPM1D Mosaic Truncating Variants in Ovarian Cancer Cases May Be Treatment-Related Somatic Mutations

Abstract

Mosaic truncating mutations in the protein phosphatase, Mg²⁺/Mn²⁺-dependent, 1D (PPM1D) gene have recently been reported with a statistically significantly greater frequency in lymphocyte DNA from ovarian cancer case patients compared with unaffected control patients. Using massively parallel sequencing (MPS) we identified truncating PPM1D mutations in 12 of 3236 epithelial ovarian cancer (EOC) case patients (0.37%) but in only one of 3431 unaffected control patients (0.03%) (P = .001). All statistical tests were two-sided. A combination of Sanger sequencing, pyrosequencing, and MPS data suggested that 12 of the 13 mutations were mosaic. All mutations were identified in post-chemotherapy treatment blood samples from case patients (n = 1827) (average 1234 days post-treatment in carriers) rather than from cases collected pretreatment (less than 14 days after diagnosis, n = 1384) (P = .002). These data suggest that PPM1D variants in EOC cases are primarily somatic mosaic mutations caused by treatment and are not associated with germline predisposition to EOC.

Two studies have recently reported truncating mutations in the protein phosphatase, Mg²⁺/Mn²⁺-dependent, 1D gene (PPM1D) (HGNC:9277) in lymphocyte DNA from ovarian cancer patients (frequency = 1.1%-1.5%) (1,2). Similar mutations were very rare in unaffected control patients (frequency = 0.017%-0.11%, respectively). Sequencing data indicated that these mutations may be because of somatic mosaicism rather than being present in the germline (1,2); in most instances, mutations were not present at a 1:1 ratio with the consensus copy of PPM1D, as expected in germline heterozygous patients., nor were mutations present in the ovarian tumor tissue from patients with lymphocyte DNA mutations. It has been suggested that because these mutations are present in lymphocyte DNA, they confer susceptibility to ovarian cancer and so may be important in the clinical management of women through genetic risk prediction and prevention strategies (1–3). However, as previous studies were retrospective it is possible that somatic mosaic mutations were the result of treatment in case patients rather than being present before diagnosis.

We have sequenced the 370bp region (exon 6) of PPM1D, where mutations are clustered, in 3374 epithelial ovarian cancer (EOC) case patients and 3487 unaffected control patients from the nine different, previously published case-control studies (Supplementary Table 1, available online) (4,5). Case patients were enriched for the serous histological subtype of EOC, which is the most common subtype, and/or a family history of ovarian cancer. Eighty-nine percent of the case patients were negative for BRCA1 or BRCA2 mutations, and one study was unscreened. All studies had ethics committee approval, and all patients provided informed consent (Supplementary Table 2, available online).

We used polymerase chain reaction (PCR) enrichment to amplify the target region of PPM1D using Fluidigm access array as previously described (6), followed by 100bp paired-end sequencing, using the Illumina HiSeq2000/HiScan (Supplementary Methods and Supplementary Table 3, available online). Burrows-Wheeler Aligner (BWA) (7) was used for sequencing read alignment against the human genome reference sequence (hg19). The Genome Analysis Toolkit (GATK) (8) was used for base quality-score recalibration, local insertion/deletion (indel) realignment, and substitution/indel discovery. After quality control filtering, 3236 case patients and 3431 control patients with greater than 80% of target bases being sequenced to a depth of 15X were included in the final analysis. The average coverage of the target region screened at 15X depth for the 6669 samples was 98.4%. To detect mosaic variants with a low frequency, a relatively high depth is required. In our data, 95% of target bases had greater than 50X depth and 86% of bases had greater than 100X depth (Supplementary Table 4, available online). All variants with an alternate allele frequency of 5% or higher were analyzed further. Where we found evidence of potentially deleterious truncating PPM1D mutations, we performed Sanger sequencing to confirm the presence of the mutation and the ratio of the normal and mutant alleles. All tests of statistical significance were two-sided, and the cutoff for statistical significance was a P value of less than .05.

Twelve truncating variants (0.37%) in PPM1D were identified in case patients with one truncating variant (0.03%) identified in a control (P = .001) (Table 1). Further investigation revealed the control had a prior history of breast cancer. All mutations in case patients were from patients for whom blood samples had been drawn after treatment (n = 1827) compared with case patients drawn pretreatment (n = 1384) (P = .002). The average time to blood draw in carriers was 1234 days postdiagnosis. We used 14 days after diagnosis as the threshold for being untreated as the majority of patients start platinum-based chemotherapy two to four weeks after diagnosis (9,10). Both massively parallel sequencing (MPS) and Sanger sequencing indicated that all but one (12 of 13) of the PPM1D mutations were mosaic; the alternate allele frequency in MPS was less for the mutant allele (average = 27.6%, range = 14.6–45.3). This was specific to PPM1D; for heterozygous variants identified in 18 other genes in the same samples, the average alternate allele frequency was 49.4% (P < .001 for difference with PPM1D variants) (Figure 1 and Table 1; Supplementary Figures 1 and 2, available online).

Table 1.

PPM1D protein truncating variants with time of blood collection for DNA

Summary			Mutation frequency, No. (%)
Control patients (n = 3431)			1 (0.03)
All case patients (n = 3236)			12 (0.37)†
Pretreatment case patients (n = 1384)‡			0 (0)||
Post-treatment case patients (n = 1827)§			12 (0.66)¶
All case patients		Alternate allele freq (MPS data)			Average 18 gene alternate allele freq#		Age at diagnosis, mean, y		Time to blood draw from diagnosis (pretreatment)‡, mean		Time to blood draw from diagnosis (post-treatment)§, mean
Carrier case patients (n = 12)		27.6			49.4		59.8		NA		1234 d
Noncarrier case patients (n = 3199)		NA			49.8		59.3		2 d		847 d
Individuals (study site)	Alternate allele freq (MPS data)			Average 18 gene alternate allele freq# (range)		Age at diagnosis, y		Histology††		Time to blood draw from diagnosis (post-treatment)§	DNA change	Predicted protein change
Case 1 (GRR)	37.8			51.7 (38.3–66.8)		55		Serous		3682 d	c.1280G>A	p.W427*
Case 2 (MAY)‡‡	45.3			49.2 (21.4–80.0)		86		HGS		65 d	c.1349delT	p.L450*
Case 3 (AOC)	40.3			49.2 (32.4–75.1)		60		HGS		576 d	c.1384C>T	p.Q462*
Case 4 (SEA)	31.8			50.8 (44.6–68.9)		61		LGS		314 d	c.1423G>T	p.E475*
Case 5 (SEA)	42.3			52.3 (45.7–66.9)		65		LGS		875 d	c.1535delA	p.N512fs
Case 6 (UKO)	25.8			49.4 (42.3–53.0)		55		HGS		1185 d	c.1608delG	p.R536fs
Case 7 (SEA)	19.4			52.9 (44.2–69.0)		67		HGS		847 d	c.1337delC	p.S446*
Case 8 (SEA)	17.3			45.0 (18.1–77.0)		66		LGS		540 d	c.1339_1342delGAGA	p.E447fs
Case 9 (GRR)	14.6			48.3 (27.0–68.4)		64		HGS		674 d	c.1403C>G	p.S468*
Case 10 (SEA)	17.8			50.5 (40.5–62.3)		60		LGS		849 d	c.1412delC	p.P471fs
Case 11 (GRR)	18.4			45.8 (30.0–55.1)		21		LGS		4625 d	c.1535delA	p.N512fs
Case 12 (SEA)	20.1			48.0 (25.0–75.0)		58		HGS		581 d	c.1579G>T	p.E527*
Control 1 (MAY)§§	14.7			46.7 (19.0–54.1)				Prior breast cancer		6 y after start Tamoxifen	c.1434C>A	p.C478*

* Indicates protein termination. Order of variants is sorted by those with alternate allele frequency higher than 25% and then those 25% or lower and within each group sorted by nucleotide position. AOC = Australian Ovarian Cancer Study; GRR = Gilda Radner Familial Ovarian Cancer Registry; HGS = high-grade serous; LGS = low-grade serous; MAY = Mayo Clinic Ovarian Cancer Study; MPS = massively parallel sequencing; SEA = SEARCH; UKO = United Kingdom Ovarian Cancer Population Study.

† Carrier frequency, all case patients vs control patients, P = .001.

‡ Case patients in which blood for genetic analyses was drawn at or no later up to 14 days after diagnosis (ie, pretreatment).

§ Case patients in which blood for genetic analyses was 14 or more days after diagnosis (ie, post-treatment); 25 case patients had unknown time between diagnosis and blood for DNA sample.

|| Carrier frequency, post-treatment case patients vs pretreatment case patients, P = .002.

¶ Carrier frequency, post-treatment case patients vs control patients, P < .001. Fisher’s exact test, two-tailed P values.

# Average alternate allele frequency for all other variants for each sample from MPS data and range.

†† Reported histology for case patients: serous (grade not reported), HGS (high-grade serous), LGS (low-grade serous).

‡‡ The variant in this individual did not show strong evidence of being mosaic by Sanger sequencing or pyrosequencing.

§§ This control patient had a prior history of breast cancer.

Figure 1.

Alternate allele frequency of PPM1D and other variants for the 13 samples with PPM1D truncating variants. A) Alternate allele frequency data from massively parallel sequencing. For the PPM1D variants, the frequency for each sample was graphed. For the other variants in 18 genes, the average of all identified variants in each individual, after homozygous variants were removed (alt freq >85%), was graphed. B) Alternate allele frequency data from pyro-sequencing for six cases with massively parallel sequencing alternate allele frequencies higher than 25%. For the PPM1D variants, the data from three to six replicates for each sample were averaged. Each of the six cases was heterozygous for either the single nucleotide polymorphism (SNP) BARD1 c.1134C>G or BRIP1 c.2637A>G. The data from three to six replicates of the control single-nucleotide polymorphism for each sample was averaged. Paired t tests were used to obtain two tailed P values.

Neither of these sequencing approaches is highly quantitative; our MPS analysis of other candidate genes in the 3236 case patients and 3431 control patients found multiple genetic variants that deviate from a 50% alternate allele frequency that showed no evidence of mosaicism by Sanger sequencing (Supplementary Figures 3 and 4, available online). Therefore, we also used pyrosequencing to more accurately determine the alternate allele frequencies of six of the PPM1D mutations with MPS alternate allele frequencies greater than 25% (Supplementary Methods, Supplementary Table 5, and Supplementary Figure 1, available online). Pyrosequencing indicated that PPM1D mutant alleles were present at statistically significant lower alternate allele frequencies than heterozygous variants in other genes for the same samples (P = .009) (Figure 1). Together, the three methods confirmed that five out of six PPM1D variants with MPS alternate allele frequencies of 25% to 45% also appear to be mosaic and not germline heterozygous mutations (Table 1; Supplementary Figures 1 and 2, available online). The variant in Case 2, which had the highest MPS alternate allele frequency of 45.3%, did not show strong evidence of being mosaic with any method.

In addition to the protein truncating variants, we identified eight different nonprotein truncating variants (6 missense and 2 synonymous changes) in 28 individuals and confirmed all variants by Sanger sequencing (Supplementary Table 6, available online). There was no evidence that any of these variants were mosaic from the MPS data or Sanger sequencing. Five of the variants were previously identified changes with assigned reference single-nucleotide polymorphism (rs) numbers.

The observation of mosaic protein truncating variants is consistent with the data presented in the two published studies that also suggest PPM1D mutations in EOC cases are mosaic (1,2). However, neither of these studies stratified mutation carriers by treatment status (pre- or post-treatment) and both publications suggested that these variants confer susceptibility to ovarian cancer. Ruark et al. estimated that the relative risk of ovarian cancer in PPM1D mutations carriers was 12 (95% confidence interval [CI] = 4.3 to 30) (1), which is similar to the relative risk of ovarian cancer associated with BRCA2 gene mutations (11,12). However, in our study, none of the PPM1D mutations were identified in case patients from whom blood (and lymphocyte DNA) was taken prior to disease treatment. This suggests that PPM1D mutations are caused by the treatment rather than being present before diagnosis and are therefore unlikely to be associated with susceptibility to ovarian cancer. This is supported by the findings of Akbari et al., who only identified PPM1D mutations in post-treatment case patients (2).

If PPMID mutations do not predispose to ovarian cancer, as our data suggest, then their clinical significance is less certain. Data from Akbari and colleagues suggest that PPMID mutations are associated with poorer outcome (2,13). However, the number of carriers in our study was too small to evaluate association with survival time and we are unable to confirm or refute this finding. If PPM1D mosaic mutations have a clinically significant affect on survival, this may explain the differences in mutation frequency between previous reports and our study (Supplementary Table 7, available online). Ascertainment of post-treatment cases to our studies was generally some time after diagnosis (median 18 months) and would have been biased to genotypes associated with better survival (Supplementary Figure 5, available online). Alternatively, we propose that chemotherapy damages the DNA and clones of cells with mutations may be more common immediately after treatment than some years later.

PPM1D-truncating mutations have been identified in 0.26% of breast cancer case patients (1). We identified one mosaic PPM1D mutation in a control patient who had a previous history of breast cancer and three years of Tamoxifen treatment. Tamoxifen is a selective estrogen receptor modulator and would be expected to have a different mechanism than platinum-based chemotherapy, which is standardly used in ovarian cancer. Our data would suggest that a similar investigation of PPM1D mutations and treatment in breast cancer is required.

One of the limitations of our study is the lack of detailed treatment data; however, the treatment for ovarian cancer at the time of these study collections was fairly standard. In all the case patients with clearly mosaic PPM1D variants, blood was drawn more than 300 days after diagnosis; therefore altering the threshold of greater than 14 days for treatment would not have affected the number of carriers. The small number of mutation carriers is another limitation of this study; a larger analysis of post-treatment case patients would be required to identify additional carriers to further examine survival associated with the presence of PPM1D mutations.

In summary, this study suggests that truncating variants in PPM1D in EOC case patients are mainly somatic mosaic mutations related to treatment and so are not likely to be associated with germline predisposition to EOC.

Funding

This work was funded by the Cancer Councils of New South Wales, Victoria, Queensland, South Australia and Tasmania, the Cancer Foundation of Western Australia, Cancer Research UK (C315/A2621, C490/A10119, C490/A10124, C490/A16561, C490/A6187, C1005/A12677, C1005/A6383, C1005/A7749), the Eve Appeal (The Oak Foundation), the National Institutes for Health (P30CA014089, P30CA016056, P30CA15083, P50CA136393, P50CA159981, R01CA122443, R01CA178535, R01CA61107, R01CA152990, and R01CA086381), the National Health & Medical Research Council of Australia (NHMRC; ID400413, ID400281), the Pomeranian Medical University, Queensland Cancer Fund, Roswell Park Cancer Institute Alliance Foundation, the Rudolf Bartling Foundation, the UK Department of Health, the UK National Institute for Health Research Biomedical Research Centres at the University of Cambridge and at the University College London Hospitals, and the US Army Medical Research and Materiel Command (DAMD17-01-1-0729).

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute or the National Institutes of Health.

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

We thank all the study participants who contributed to this study and all the researchers, clinicians, and technical and administrative staff who have made possible this work. In particular, we thank: the clinical and scientific collaborators listed at http://www.aocstudy.org/ (Australian Ovarian Cancer Study Group); E. Wozniak, A. Ryan, J. Ford, and N. Balogun (United Kingdom Ovarian Cancer Population Study—UKO study); Marie Mack, Craig Luccarini, Caroline Baynes, the SEARCH team, and Eastern Cancer Registration and Information Centre (SEARCH). Some of the sequencing was performed at the Source BioScience laboratories in the UK.

The authors would like to thank the National Heart, Lung, and Blood Institute Grand Opportunity (NHLBI GO) Exome Sequencing Project and its ongoing studies which produced and provided exome variant calls for comparison: the Lung GO Sequencing Project (HL-102923), the Women’s Health Initiative Sequencing Project (HL-102924), the Broad GO Sequencing Project (HL-102925), the Seattle GO Sequencing Project (HL-102926), and the Heart GO Sequencing Project (HL-103010).

The authors have no conflicts of interest to declare.