Mutation profiles in circulating cell-free DNA predict acquired resistance to olaparib in high-grade serous ovarian carcinoma.

Although resistance to poly(ADP-ribose) polymerase inhibitors (PARPi) has gradually become a major challenge in the maintenance therapy for high-grade serous ovarian carcinoma (HGSOC), there are no universal indicators for resistance monitoring in patients. A key resistance mechanism to PARPi is the restoration of homologous recombination repair (HRR), including BRCA reversion mutations and changes in DNA damage repair proteins. To explore mutation profiles associated with PARPi resistance, we undertook targeted 42-gene deep sequencing of circulating cell-free DNA (cfDNA) extracted from HGSOC patients pre- and post-treatment with olaparib maintenance therapy. We found that pathogenic germline mutations in the HRR pathway, including BRCA1/2, were strongly associated with improved clinical outcomes, and newly acquired MRE11A mutations significantly shortened the progression-free survival (PFS) of patients. Furthermore, dynamic fluctuations of somatic mutation sites in CHEK2:p.K373E and CHEK2:p.R406H can be used for evaluating the therapeutic efficacy of patients. MRE11A:p.K464R might be a vital driving factor of olaparib resistance, as patients with newly acquired MRE11A:p.K464R in post-treatment cfDNA had significantly shorter PFS than those without it. These findings provide potential noninvasive biomarkers for efficacy evaluation and resistance monitoring of olaparib treatment, and lay the foundation for developing combination treatment after olaparib resistance.

cell‐free DNA

circular ligation amplification and sequencing

DNA damage repair

germline BRCA1/2

genomic DNA

high‐grade serous ovarian carcinoma

homologous recombination repair

mutant allele frequency

maximum mutant allele frequency

next‐generation sequencing

poly(ADP‐ribose) polymerase

poly(ADP‐ribose) polymerase inhibitor

peripheral blood

progression‐free survival

white blood cell

INTRODUCTION

High‐grade serous ovarian carcinoma is the most common pathological type of ovarian epithelial tumor. Its tumorigenesis is generally asymptomatic and early diagnosis is difficult, resulting in approximately 75% of patients being diagnosed at advanced stage. In addition, HGSOC patients are more likely to relapse and develop drug resistance after treatment; the 5‐year survival rate of patients with advanced disease is less than 30%. Therefore, clinical experts have been looking for ways to reduce relapse and improve patients’ prognosis.

Poly(ADP‐ribose) polymerase inhibitors open a new chapter in maintenance therapy of HGSOC patients. When the HRR pathway is disrupted, such as BRCA1/2 mutations, tumor cells will produce homologous recombination deficiency and form a synthetic lethal effect with PARPi. , Olaparib, as a representative of PARPi, has been mainly used in maintenance therapy of platinum‐sensitive ovarian cancer patients, , and it has provided significant long‐term survival benefit in clinical studies at home and abroad. However, with the widespread application of olaparib, its drug resistance is becoming increasingly prominent. Mounting evidence has suggested that PARPi resistance is closely related to HRR restoration, including BRCA1/2 reversion mutations and changes in DDR proteins (e.g., ATM, 53BP1, RAD51, and CDK12). , , , Moreover, our previous study showed that acquired KRAS mutation or mutations of other components of the MAPK pathway lead to primary and acquired PARPi resistance. Therefore, exploring a viable method for monitoring molecular changes of HGSOC patients will contribute to solving the clinical dilemma of olaparib resistance.

Cell‐free DNA analysis can achieve continuous and repeatable monitoring of tumor molecular characterizations at different stages of diagnosis and treatment in patients. , Cell‐free DNA analysis has gradually highlighted its technical advantages in clinical practice, such as risk stratification, response assessment, and resistance monitoring. For example, it can predict the risk of postoperative recurrence of colorectal cancer, guide the application of targeted therapy in non‐small‐cell lung cancer, , and screen prostate cancer patients who benefit from PARPi treatment. Similarly, BRCA reversion mutations detected by cfDNA can predict primary or acquired resistance to rucaparib in HGSOC patients, suggesting that cfDNA analysis could be an effective means for efficacy evaluation and resistance monitoring of olaparib.

In this study, we investigated the development of genetic resistance to olaparib in a retrospective study using sequential cfDNA sampling in platinum‐sensitive HGSOC patients. We identified the cfDNA mutation profiles, determined the correlations between cfDNA mutation profiles and patient prognosis, and screened out mutation biomarkers and potential drug targets in HGSOC patients with olaparib resistance.

MATERIALS AND METHODS

Collection of samples and clinical information

Peripheral blood samples (10 ml) were collected using LBgard Blood Tubes (Biomatrica) prior to olaparib treatment and at later follow‐ups (every 3 months until 12 months after treatment) in patients. Healthy subjects were recruited to collect 10 ml PB samples as the baseline for cfDNA analysis. Clinical information from patients was collected for subsequent analysis.

Peripheral blood sample processing

All PB samples were processed within 3–4 days, including sample collection, transportation, and separation of blood components. The separation process was carried out as follows. First, unqualified PB samples with coagulation or hemolysis were excluded. Next, PB samples were centrifuged at 1600 g for 10 min at 4°C to separate plasma and WBC. Plasma was transferred into a new 2 ml microcentrifuge tube, followed by centrifugation at 16,000 g for 10 min at 4°C to remove any remaining cellular components. Plasma was then transferred into a new 2 ml microcentrifuge tube for cfDNA extraction. White blood cells were collected using a new 2 ml microcentrifuge tube for isolating gDNA. All plasma and WBC were immediately stored at −80°C until DNA extraction. Sample processing time did not exceed 1 h.

Genomic DNA extraction

Genomic DNA was extracted from 100 μl WBC using the TIANamp Blood DNA Kit (Tiangen) following the manufacturer’s instructions and was quantified by a Nanodrop 2000 spectrophotometer (Thermo Fisher Scientific). Quality control was as follows: concentration >1 ng/μl and no protein or RNA contamination. Extracted gDNA was stored at −80°C until sequencing.

Cell‐free DNA extraction

Cell‐free DNA was extracted from 4 ml plasma using the QIAamp Circulating Nucleic Acid Kit (Qiagen) in accordance with the manufacturer’s instructions. Extracted cfDNA was quantified by a Qubit 3.0 fluorometer (Life Technologies) and 2100 Electrophoresis Bioanalyzer Instrument (Agilent Technologies). Quality control was as follows: (i) a typical peak around 167 bp, (ii) no obvious gDNA contamination, and (iii) total content >10 ng. Extracted cfDNA was stored at −80°C until sequencing.

Library construction and sequencing

The gDNA was sonicated into short fragments using a Covaris ultrasonicator, with the peak around 200 bp. The linear cfDNA molecules were denatured and circularized to form templates for rolling circle amplification reaction. Next, 100 ng fragmented gDNA and amplified cfDNA were used for library construction using the KAPA sequencing library construction kit (Kapa Biosystems) following the manufacturer’s instructions. Sequencing libraries were captured by a 42‐gene panel, followed by sequencing on an Illumina HiSeq 2500. The sequencing scheme of 250 paired ends was used to obtain the full sequence information. Unique sequencing reads were determined by the AccuraGen proprietary algorithm. The average coverage depth for all probes was approximately 690× in gDNA and 68,014× in cfDNA. Cell‐free DNA samples from healthy subjects were detected following the procedure described above.

Sequencing data processing

The variants of gDNA and cfDNA were called using AccuraGen’s NGS pipeline (AccuraGen Biotechnology Co., Ltd.). Briefly, sequencing reads were aligned to the hg19/GRCh37 reference genome by the Burrows–Wheeler Alignment tool (http://bio‐bwa.sourceforge.net/bwa.shtml). Duplicate removal were carried out using SAMBLASTER followed by single nucleotide polymorphism and indel calling using FreeBayes. Variants in low‐complexity regions were removed and annotated using GEMINI. In addition, background noise introduced by random NGS error was removed by the AccuraGen proprietary algorithm and cfDNA sequencing data from healthy subjects was used as the baseline for cfDNA analysis. Somatic mutations were determined by cross‐checking cfDNA sequencing data with germline and clonal hematopoietic variants from gDNA. Mutant allele frequency was calculated by comparing the variant reads to total sequencing reads of the variant site.

Statistical analysis

Pearson correlation analysis was used to assess correlations between different parameters. Unpaired or paired t‐test was used for comparison between two groups, and ordinary one‐way ANOVA was used for comparison among three groups. The differences between the proportions of two or three groups were analyzed by χ2‐test and Kruskal–Wallis test with Dunn’s multiple comparisons test, respectively. Log–rank (Mantel–Cox) test was utilized to determine the differences in survival curves. The above analyses were undertaken using GraphPad Prism 7 (GraphPad Software). Results were considered statistically significant when p < 0.05.

RESULTS

Procedure for cfDNA analysis in HGSOC patients

This work was carried out retrospectively on prospectively collected samples from platinum‐sensitive HGSOC patients with olaparib maintenance therapy. Peripheral blood samples from 25 HGSOC patients and 37 healthy subjects were collected for cfDNA analysis and baseline detection, respectively (Figure 1). The clinical characteristics of patients are described in Table 1. Cell‐free DNA was extracted from plasma samples collected from patients (Figure 1A, B0–B4) and healthy subjects. A personalized panel containing 42 genes was customized to detect germline and somatic mutations (Table S1). Circular ligation amplification and sequencing was used for mutation detection (Figure 1B,C), in which gDNA of B0 samples were used to identify germline mutations, and 96 cfDNA (35 healthy subject‐derived and 61 patient‐derived) samples were detected for analyzing somatic mutations in patients (Figure 1B). For data processing, we estimated tumor load in cfDNA using Max MAF. , Variant‐supporting reads >2 and without germline mutations were defined as the criteria for somatic mutations.

FIGURE 1

Study schema and method for cell‐free DNA (cfDNA) analysis in patients with high‐grade serous ovarian carcinoma. (A) Diagram showing the schema of peripheral blood (PB) sample collection. PB samples are collected at the following time points: pretreatment (B0), then every 3 months after olaparib treatment (B1, B2, B3, and B4). The duration of PB collection is determined according to the follow‐up of patients. (B) Flow diagram showing the processing, detection, and analysis of cfDNA. C, Schema of the method of cfDNA detection. CLAmp‐seq, circular ligation amplification and sequencing; gDNA, genomic DNA; RCA, rolling circle amplification; WBC, white blood cell

TABLE 1

Demographics of our cohort of patients with high‐grade serous ovarian carcinoma treated with olaparib

	All (n = 25)	Unrelapsed	Relapsed
		(n = 6)	(n = 19)
Age, years	53.24 (41–66)	51.83 (45–65)	53.68 (41–66)
Stage
II	3	0	3
III	16	3	13
IV	6	3	3
NACT
Y	3	1	2
N	22	5	17
Number of chemotherapy regimens
2	15	4	11
≥2	10	2	8
Metastasis (enrollment)
Y	14	3	11
N	11	3	8
Cytoreductive surgery (enrollment)
Y	11	3	8
N	14	3	11
Response of chemotherapy regimens (enrollment)
CR	11	3	8
PR	14	3	11
Pathogenic germline mutations of BRCA
Y	8	3	5
N	17	3	14
Somatic mutations of BRCA
Y	4	1	3
N	21	5	16
Benign mutations of BRCA
≤3	7	2	5
3 <No. ≤5	7	1	6
>5	11	3	8
Somatic mutations of TP53
Y	11	2	9
N	14	4	10

Abbreviations: BRCA, including BRCA1 and BRCA2; CR, complete response; N, No; NACT, new adjuvant chemotherapy treatment; PR, partial response; Y, Yes.

Dynamic changes of Max MAF are associated with prognosis of olaparib maintenance therapy

By integrating the follow‐up information of patients, we plotted the distribution maps of PFS, cfDNA yield, and Max MAF in pretreatment patients (Figure 2A), and analyzed the correlations between them. We found that Max MAF was significantly positively correlated with PFS relative to cfDNA yield (Figure 2B,C). Moreover, Max MAF was significantly increased in post‐treatment patients, suggesting that it could play an important role in affecting the prognosis of olaparib treatment, but cfDNA yield did not (Figure 2D,E). Next, we plotted the dynamic changes of cfDNA or Max MAF with treatment progress under different outcomes, and observed that the worse the patients’ prognosis, the higher the proportion of patients with increased Max MAF, in which the proportion of “Relapsed < 12 m” even reached 100% (Figure 2F). We then compared the fluctuations of cfDNA or Max MAF and CA125 levels during olaparib treatment, respectively, and found that cfDNA yield did not show a consistent trend with CA125 (Figure S1A–C). For Max MAF, although the changes of Max MAF and CA125 were irregular in the “Unrelapsed” group (Figure S1D), their fluctuations tended to be consistent in the relapsed patients, especially in the “Relapsed < 12 m” group (Figure S1E,F). Furthermore, there was a significant difference in the trend of proportional changes in increasing cfDNA yield and Max MAF (Figure 2G). These results suggested that increased Max MAF in post‐treatment could predict the patients’ relapse.

FIGURE 2

Associations between cell‐free DNA (cfDNA) or maximum mutant allele frequency (Max MAF) and the prognosis of patients with high‐grade serous ovarian carcinoma. (A) Distribution maps of progression‐free survival (PFS), cfDNA yield, and Max MAF in pretreatment (B0). (B, C) Pearson correlation analysis between PFS and cfDNA yield (B, p = 0.1484) or Max MAF (C, p = 0.0065) at baseline (B0). (D, E) Differences of cfDNA yield (D, p = 0.9069) or Max MAF (E, p = 0.0081) in patients before treatment (Pre‐T) and after treatment (Post‐T), paired t‐test. (F) Dynamic changes of cfDNA yield or Max MAF with treatment process in different outcomes. (G) Difference of the tendency of the proportion of increased cfDNA yield or Max MAF in different outcomes; χ2‐test, p < 0.0001. (H, I) Survival analyses of the changes of cfDNA yield (H, p = 0.5577) and Max MAF (I, p = 0.0043) in patients. HR, hazard ratio; mPFS, median PFS; PARPi, poly(ADP‐ribose) polymerase inhibitor

To validate the effect of Max MAF on patients’ prognosis, we plotted the survival curves and found increased Max MAF was significantly associated with poor prognosis of patients (Figure 2I), but cfDNA yield was not (Figure 2H). It confirmed the feasibility of monitoring olaparib resistance by Max MAF detection.

Pathogenic germline mutations of HRR pathway influence the prognosis of olaparib maintenance therapy

The defects of the HRR pathway can form a synthetic lethal effect with PARPi in ovarian cancer. Therefore, we plotted a germline mutation map and screened out pathogenic or likely pathogenic mutations (pathogenic mutations) in patients (Figure 3A,B). Furthermore, we found that the whole mutation load had no effect on patients’ outcomes (Figure 3C,D), but patients with pathogenic mutations had significant survival benefits compared to those without them (Figure 3E,F), highlighting the effectiveness of synthetic lethal effect in clinical practice.

FIGURE 3

Pathogenic germline mutations of homologous recombination repair pathway improve the survival benefits of patients with high‐grade serous ovarian carcinoma (HGSOC). (A) Germline mutation map of HGSOC patients at baseline (B0). (B) Distribution map of pathogenic germline mutations in patients at baseline (B0). (C) Differences in variant counts among different outcomes; ordinary one‐way ANOVA, p = 0.1015. (D) Pearson correlation analysis between progression‐free survival (PFS) and germline mutation load; p = 0.9127. (E) Differences in PFS between carriers and noncarriers of pathogenic mutations; unpaired t‐test, p = 0.0211. (F) Survival analysis in patients with pathogenic germline mutations; p = 0.0229. (G) Differences in PFS between carriers and noncarriers of germline BRCA1/2 (gBRCA) mutations; unpaired t‐test, p = 0.0281. (H) Survival analysis in patients with gBRCA mutations (H, p = 0.1090). (C, E, G) Data are presented as mean ± SEM. HR, hazard ratio; mPFS, median PFS; PARPi, poly(ADP‐ribose) polymerase inhibitor

Considering the high proportion of pathogenic gBRCA mutations, we analyzed and found that patients with gBRCA mutations had a degree of survival advantage (Figure 3G,H). These results suggested that HRR‐related pathogenic germline mutations in HGSOC patients affected the efficacy of olaparib maintenance therapy and were closely related to patients’ prognosis.

Somatic mutations under pathogenic germline mutations are associated with prognosis of olaparib maintenance therapy

We further plotted the distribution maps of somatic mutations in patients with pathogenic germline mutations (Figure S2A–C) and observed that high‐frequency (≥3) mutations of CHEK2, TP53, MRE11A, and ATM in pretreatment (Figure S2B) and post‐treatment (Figure S2C). Next, we explored their effect on patients’ PFS by survival analysis. The results showed that patients with new MRE11A mutations in post‐treatment had an extremely poor prognosis, with significantly shorter PFS compared to noncarriers (Figure S2I). However, no differences in PFS were observed for other high‐frequency mutations or new mutations, including TP53, MRE11A, and ATM in pretreatment (Figure S2E–G) and CHEK2 (Figure S2H), ATM (Figure S2J), and TP53 (Figure S2K) in post‐treatment.

Furthermore, we observed that high‐frequency (≥3) new mutations were mainly concentrated in MRE11A:p.K464R (Figure S2D). Therefore, we explored its role in patients’ prognosis and found that newly acquired MRE11A:p.K464R was strongly associated with poor prognosis of patients (Figure S2L). It revealed that MRE11A:p.K464R could be a reliable indicator for resistance monitoring of olaparib in HGSOC patients with pathogenic germline mutations.

Somatic mutations in cfDNA influence the prognosis of olaparib maintenance therapy

Based on the above analyses, we realized that somatic mutations could be a more universal indicator for efficacy evaluation or resistance monitoring of olaparib maintenance therapy. Therefore, we plotted somatic mutation maps in cfDNA. As shown in Figure 4A,B, 80% (20/25) of patients had an increased mutation load, and 96% (24/25) of patients developed new mutations. Next, we compared the differences of somatic mutation load in patients before and after olaparib treatment among different outcomes, and found that increased somatic mutation load in post‐treatment predicted a poor prognosis (Figure 4C) and has a negative correlation trend with patients’ PFS (Figure 4D). In addition, the higher ratio of new somatic mutations in total mutation load, the worse patients’ prognosis (Figure 4E), indicating that it might be a crucial factor for olaparib resistance in HGSOC patients.

FIGURE 4

Associations between somatic mutation profiles and the prognosis of olaparib maintenance therapy in patients with high‐grade serous ovarian carcinoma. (A, B) Distribution maps of somatic mutations in pretreatment (Pre‐T) (A) and post‐treatment (Post‐T) (B) patients. (C) Changes of mutation load in patients before and after olaparib treatment. Paired t‐test: Unrelapsed, p = 0.6109; Relapsed > 12 months, p = 0.8893; Relapsed < 12 months, p = 0.0004. (D) Pearson correlation analysis between progression‐free survival (PFS) and increased mutation load in post‐treatment; p = 0.1817. (E) Differences in the ratio of new mutations among different outcomes. Kruskal–Wallis test with Dunn’s multiple comparisons test: Unrelapsed versus Relapsed > 12 months, p = 0.2162; Unrelapsed versus Relapsed < 12 months, p = 0.0192. (F–H) Survival analyses of somatic mutations (mut) in (F, p = 0.2762), (G, p = 0.0267), and (H, p = 0.0759) in Pre‐T patients. (I–K) Survival analyses of new somatic mutations in (I, p = 0.0005), (J, p = 0.0728 ), and (K, p = 0.1787) in Post‐T patients. (L–N) Survival analyses of combined somatic mutations of pre‐ + post‐ (L, p = 0.0305), pre‐ + post‐ (M, p = 0.0057), and pre‐ + post‐ (N, p = 0.0006) in patients. (C, E) Data are presented as mean ± SEM. HR, hazard ratio; mPFS, median PFS; PARPi, poly(ADP‐ribose) polymerase inhibitor

Next, we summarized all high‐frequency (≥3) somatic mutations in cfDNA (Table S2), and analyzed their impact on patients’ prognosis. Due to the extremely high frequency of CHEK2 in pretreatment (96%) and post‐treatment (100%) (Figure 4A,B), we only analyzed the role of new CHEK2 mutations in patients’ prognosis, but no meaningful result was observed (Figure S3B). For other somatic mutations, compared with TP53 (Figure 4F) and PMS2 (Figure S3A), MRE11A mutations could significantly prolong patients’ PFS in pretreatment (Figure 4G). Similarly, ATM mutations also tended to improve patients’ prognosis (Figure 4H). In post‐treatment, we unexpectedly observed the opposite effect of new MRE11A mutations compared with pretreatment, which significantly shortened the PFS of patients, indicating a strong correlation with olaparib resistance (Figure 4I). We also found that new TP53 and ATM mutations had opposite trends in their impacts on patients’ prognosis (Figure 4J,K). However, new mutations of EGFR, PIK3CA, NF1, and ERBB2 did not affect patients’ PFS (Figure S3C–F).

Furthermore, we evaluated the combined effect of somatic mutations on patients’ prognosis and observed that the combination of pre‐MRE11A + post‐TP53 significantly differentiated patients’ prognosis (Figure 4L). Pre‐ATM combined with post‐MRE11A or post‐TP53 also shortened patients’ PFS (Figure 4M,N). However, we found that only the combination of pre‐ATM + post‐TP53 greatly improved the efficiency of prognostic assessment (Figure 4N). In addition, although the combination of MRE11A mutations with other somatic mutations distinguished patients’ PFS well, the efficiency was significantly reduced compared with MRE11A alone (Figures 4L,M and S3G–I), revealing the specificity and reliability of MRE11A mutations as an indicator for efficacy evaluation and resistance monitoring of olaparib.

Somatic mutation sites in cfDNA are associated with the prognosis of olaparib maintenance therapy

We have confirmed that somatic mutations in cfDNA were closely related to patients’ prognosis. To better elucidate their role in disease progression, we further summarized the somatic mutation sites in cfDNA (Table S3) and analyzed the relationships between high‐frequency (≥3) mutation sites and patients’ prognosis. We first plotted the distribution maps of mutation sites, and found that CHEK2:p.K373E and CHEK2:p.R406H had no new mutations in post‐treatment (Figure 5A,B). In contrast, the proportion of newly acquired MRE11A:p.K464R was up to 52% (13/25), and contributed 92.9% (13/14) of new mutation load in MRE11A (Figure 5B). Furthermore, we observed CHEK2:p.K373E and MRE11A:p.K464R in pretreatment significantly prolonged the PFS of patients (Figure 5C,E), whereas CHEK2:p.R406H did not (Figure 5D).

FIGURE 5

Associations between somatic mutation sites and the prognosis of olaparib maintenance therapy in patients with high‐grade serous ovarian carcinoma. (A, B) Distribution maps of somatic mutation sites in pretreatment (A) and post‐treatment (B) patients. (C–E) Survival analyses of (C, p = 0.0219), (D, p = 0.1319), and (E, p = 0.0267) in pretreatment. (F, G) Mutation (mut) tracking of (F) and (G) in cell‐free DNA (cfDNA) detection during olaparib treatment. (H, I) Survival analyses of higher (H, p = 0.0091) and higher (I, p = 0.0002) in post‐treatment. (J) Mutation tracking of in cfDNA detection during olaparib treatment. (K) Longitudinal representation of from all patients with a timepoint available. (L) Survival analysis of newly acquired (G, p = 0.0005) in post‐treatment. (M, N) Survival analyses of combined somatic mutation sites of post‐ :p.  post‐ :p. (M, p = 0.0120) and post‐ :p.  post‐ :p. (N, p < 0.0001) in patients. HR, hazard ratio; mPFS, median PFS; PARPi, poly(ADP‐ribose) polymerase inhibitor

In post‐treatment, we drew the dynamic curves of CHEK2:p.K373E and CHEK2:p.R406H during the sampling period, and observed that the content of them increased in relapsed patients (Figure 5F,G). We also plotted the relationships between their fluctuations and disease progression in all patients. The results showed that the proportion of disease progression in patients with increased CHEK2:p.K373E and CHEK2:p.R406H (“higher”) was as high as 66.7% (10/15) and 100% (9/9), respectively (Figure S4A,B). Survival analysis also confirmed that increased CHEK2:p.K373E and CHEK2:p.R406H predicted a poor prognosis (Figure 5H,I). In addition, their combined effect was only observed to distinguish the prognosis of patients in post‐treatment, and the efficiency was not improved relative to CHEK2:p.R406H alone (Figure S4C,D).

For MRE11A:p.K464R, its fluctuation in relapsed patients came from newly acquired mutations (Figure 5B,J). By analyzing the dynamic changes of MRE11A:p.K464R in all patients, we were surprised to find that once patients acquired the MRE11A:p.K464R, it often predicted disease progression or drug resistance, with the proportion as high as 92.3% (12/13). In contrast, only 25% (3/12) of patients with MRE11A:p.K464R in pre‐existing or negative groups developed disease progression (Figure 5K). As expected, newly acquired MRE11A:p.K464R greatly shortened patients’ PFS, suggesting that it plays a vital role in olaparib resistance (Figure 5L). Furthermore, we analyzed the role of CHEK2:p.K373E + MRE11A:p.K464R and MRE11A:p.K464R + CHEK2:p.R406H in post‐treatment in patients’ PFS. The results showed they significantly reduced the survival benefits of patients (Figure 5M,N, Table S4). Furthermore, the combined three mutation sites also highlighted the negative impact on patients’ prognosis (Figures S4E and S5, Table S4). In addition, we analyzed the effects of mutations or newly acquired mutations in MRE11A:p.K464R, CHEK2:p.R474C, PIK3CA:p.E707K, ATM:p.I1407T, and CHEK2:p.Y390C on patients’ prognosis (Figure S6A–G). However, no meaningful results were observed.

Compared with single‐mutation sites, we were surprised to find that, except for MRE11A:p.K464R + CHEK2:p.R406H, combined mutation sites did not improve the efficiency of prognostic assessment in patients, especially when compared with MRE11A:p.K464R or CHEK2:p.R406H (Figures 5M,N, S4C–E, and S6H,I). By analyzing the characteristics of mutation sites, we found that variations of CHEK2:p.K373E and CHEK2:p.R406H came from dynamic fluctuations of content during olaparib treatment. Increased CHEK2:p.K373E and CHEK2:p.R406H often predicted a poor prognosis, suggesting that they had great potential as monitoring indicators for olaparib maintenance therapy. Interestingly, newly acquired MRE11A:p.K464R was often accompanied by relapse (6/13) or subsequent short‐term disease progression (6/13), indicating that it could be a vital driving factor for olaparib resistance in patients (Figure 5K) and is expected to be a reliable indicator and intervention target for olaparib resistance. Encouragingly, compared with CA125, it was further confirmed that the fluctuations of CHEK2:p.K373E, CHEK2:p.R406H, and MRE11A:p.K464R were in good agreement with CA125, and had a good potential for clinical transformation (Figure S7).

In conclusion, we found that mutation profiles in cfDNA can be used for efficacy evaluation and resistance monitoring in HGSOC patients with olaparib maintenance therapy. Our results will contribute to constructing clinical prediction models of olaparib treatment, guide the formulation of combination therapy after olaparib resistance, and optimize the clinical application of PARPi. In addition, the discovery of MRE11A:p.K464R will provide a focus for developing targeted drugs and will promote the progress of diagnosis and treatment of ovarian cancer.

DISCUSSION

In our study, we examined the possibility of mutation profiles in cfDNA as a predictive biomarker for responses and outcomes of olaparib maintenance therapy. Platinum‐sensitive HGSOC patients who had a relapse and were scheduled to receive olaparib maintenance therapy were enrolled.

In recent years, cfDNA analysis as a noninvasive detection technology has been widely used in clinical practice, such as noninvasive prenatal testing, organ transplantation monitoring, and cancer fluid biopsy. Genomic or epigenomic alterations of cells can be analyzed by cfDNA detection, including point mutations, copy number alterations, and methylation changes. In addition, cfDNA can also be used to detect exosomal DNA, chromosomal rearrangements, and viral or bacterial DNA fragments in plasma. , For ovarian cancer studies, cfDNA analysis can be used to undertake a comprehensive molecular analysis of tumors, , , such as BRCA reversion mutations and copy‐number alterations. This evidence indicates that it is theoretically feasible to solve the clinical dilemma of olaparib resistance by cfDNA analysis. However, previous cfDNA studies mainly focused on tumor characteristics and BRCA variations, and explored their associations with patients’ prognosis. , , , Previous studies ignored the critical role of whole mutation profiles in PARPi treatment and lacked the specific and universal indicators of PARPi resistance. To our knowledge, our work represents the most comprehensive study of cfDNA mutation profiles before and after olaparib treatment in relapsed HGSOC patients, identifying the associations between somatic mutation sites and olaparib resistance.

Here, we undertook personalized cfDNA detection using CLAmp‐seq to determine mutation profiles of each patient. We found that increased Max MAF in post‐treatment predicted a poor prognosis (Figure 2) and patients with pathogenic germline mutations in the HRR pathway had significantly superior PFS compared with negative ones (Figure 3). In addition, we also observed that MRE11A:p.K464R under pathogenic germline mutations was associated with shortened PFS, which has not been reported in relevant ovarian cancer studies (Figure S2L). If validated by further studies, this finding could profoundly influence the olaparib treatment in HGSOC patients with pathogenic gBRCA mutations.

Recent studies have shown that somatic mutations in cfDNA are closely related to drug resistance in patients. , , We speculate that somatic mutations will be more promising to screen the universal indicators for olaparib resistance. In our results, new somatic mutations in post‐treatment predict a poor prognosis, and the larger proportion of new mutation load, the worse the patients’ prognosis (Figure 4E). Furthermore, we unexpectedly found that MRE11A mutations in pretreatment and post‐treatment could both effectively differentiate the prognosis of patients, but their effects on prognosis were completely opposite. In the subsequent analyses, we found that MRE11A mutations mainly came from MRE11A:p.K464R (17/18). This is an interesting phenomenon worthy of further study. MRE11A, as an important component of the MRE11A–RAD50–NBS1 complex, plays a crucial role in DDR. , In addition, MRE11A is involved in cellular bioenergetic regulation, protecting the mitochondria and preventing T cell pyroptosis. Therefore, we speculate that the balance of biological functions of MRE11A is disrupted in olaparib treatment, leading to the opposite prognostic effects. In brief, patients with MRE11A:p.K464R in pretreatment have more stable cellular bioenergetic regulation and more vigorous immune microenvironment, which improve the survival benefits of patients. After olaparib treatment, newly acquired MRE11A:p.K464R combined with accumulated DNA damage highlight the DDR function of MRE11A, leading to olaparib resistance. Of course, further research is needed to verify our speculation. In addition, we observed that CHEK2:p.K373E and CHEK2:p.R406H were strongly correlated with patients’ PFS, which could be used for efficacy evaluation of olaparib.

For the system of cfDNA analysis, CLAmp‐seq has the following advantages. Our detection was based on high‐throughput sequencing with extremely high efficiency compared with droplet digital PCR or BEAMing (beads, emulsion, amplification, magnetics digital PCR). Moreover, due to low content of cfDNA, low‐frequency mutations. and rare mutations, the sensitivity of detection technology was greatly challenged, , , , such as NGS or cancer personalized profiling by deep sequencing (CAPP‐Seq). The CLAmp‐seq method can achieve accurate and specific amplification of cfDNA by applying the strategy of “single chain loopization and concatemer error correction,” with higher accuracy and reliability. We also filtered the sequencing data using clonal hematopoietic mutations of WBC , , to avoid the false positives before undertaking the cfDNA mutation analysis.

Inevitably, our cfDNA analysis did have certain limitations, including the inability to define the evolution profiles of existing mutations in the course of treatment, to detect new second primary cancers or identify cancer metastasis. Moreover, due to the short length and low content of tumor‐derived DNA fragments, it is easy to lead to false negative results. Therefore, cfDNA detection techniques need to be further optimized to continuously improve the threshold of variant detection. However, new therapeutic strategies based on our findings could be developed in future clinical studies. For example, olaparib can be actively given to patients with MRE11A:p.K464R in pretreatment. Noncarriers can dynamically monitor MRE11A:p.K464R by cfDNA detection, providing early warning for olaparib resistance. Further studies in larger cohorts are warranted to validate the role of MRE11A:p.K464R in olaparib treatment due to the limitation of sample size in this study.

In summary, dynamic analyses of mutation profiles in cfDNA could facilitate the timely adjustment of treatment strategies and improve the survival benefits of HGSOC patients with olaparib maintenance therapy. The discovery of MRE11A:p.K464R will not only provide a complementary or alternative indicator for diagnosis and treatment of ovarian cancer but also provide a focus for targeted drugs, and lay the theoretical basis for the development of combination therapy after olaparib resistance.

CONFLICT OF INTEREST

The authors have no conflict of interest.

ETHICS STATEMENT

The study was approved by the Ethics Committee of Tongji Hospital, Tongji Medical College, Huangzhong University of Science and Technology. Study subjects provided informed consent to allow the collection of blood and analysis of clinical and genetic data for this work.

Figure S1

Click here for additional data file.

Figure S2

Click here for additional data file.

Figure S3

Click here for additional data file.

Figure S4

Click here for additional data file.

Figure S5

Click here for additional data file.

Figure S6

Click here for additional data file.

Figure S7

Click here for additional data file.

Table S1

Click here for additional data file.

Table S2

Click here for additional data file.

Table S3

Click here for additional data file.

Table S4

Click here for additional data file.

Supinfo S1

Click here for additional data file.