Purpose:

We evaluated the predictive and prognostic value of circulating tumor DNA (ctDNA) in patients with Ewing sarcoma (EWS) treated in the EWING2008 trial.

Experimental Design:

Plasma samples from 102 patients with EWS enrolled in the EWING2008 trial were obtained before and during induction chemotherapy. Genomic EWSR1 fusion sequence spanning primers and probes were used for highly specific and sensitive quantification of the levels of ctDNA by digital droplet PCR. ctDNA levels were correlated to established clinical risk factors and outcome parameters.

Results:

Pretreatment ctDNA copy numbers were correlated with event-free and overall survival. The reduction in ctDNA levels below the detection limit was observed in most cases after only two blocks of vincristine, ifosfamide, doxorubicin, and etoposide (VIDE) induction chemotherapy. The persistence of ctDNA after two VIDE blocks was a strong predictor of poor outcomes. ctDNA levels correlated well with most established clinical risk factors; an inverse correlation was found only for the histologic response to induction therapy. ctDNA levels did not provide simple representations of tumor volume, but integrated information from various tumor characteristics represented an independent EWS tumor marker with predictive and prognostic value.

Conclusions:

ctDNA copy number in the plasma of patients with EWS is a quantifiable parameter for early risk stratification and can be used as a dynamic noninvasive biomarker for early prediction of treatment response and outcome of patients.

Translational Relevance

In Ewing sarcoma (EWS), poor prognosis and the lack of new therapeutic options have created an urgent clinical need for more differentiated therapy stratification. In this respect, EWS is representative for many sarcomas and other translocation-positive malignancies; the methodology used in this study is directly applicable to several additional entities for which blood-based biomarkers are not available. In this study, the clinical value of circulating tumor DNA copy numbers in the plasma was evaluated in the largest series of patients with EWS treated uniformly in the EWING2008 trial. The methodology used here for highly sensitive detection of DNA fusion sequences is feasible in a clinical setting and comparable with the well-established quantification of minimal residual disease in, for example, lymphocytic leukemia. Establishing circulating tumor DNA as a biomarker in translocation-positive solid tumors has great potential for improving the understanding of the efficacy of individual therapeutic elements and for refined risk-adapted therapy stratification.

The detection and quantification of circulating cell-free tumor-specific DNA (ctDNA) have gained increasing importance for use as a serum marker for noninvasive diagnostics, therapy response monitoring, risk stratification, and detecting recurrent disease in recent years (1–3). For many solid tumors in adults, commercial assays have been developed to target recurrent hotspot mutations in oncogenes or suppressor genes present in a substantial proportion of cases (4–8).

Such recurrent point mutations are rare in pediatric tumors. Sarcomas, specifically Ewing sarcoma (EWS), are characterized by a particularly low mutation rate and do not share common disease-associated hotspot mutations (9–12). The genetic features of EWS are dominated by the presence of EWSR1 rearrangements with a member of ETS family genes (13). EWSR1-FLI1 and EWSR1-ERG are the two most common fusion genes, occurring in 85% to 90% and approximately 10%, respectively (14, 15). Although treatment of EWS is clinically challenging because of its multifaceted clinical manifestations, poor prognosis (particularly in advanced or metastatic disease), and intense multimodal treatment, few methods are available for close therapy monitoring and dynamic risk stratification. In contrast to other pediatric solid tumors, such as neuroblastoma, hepatoblastoma, or germ cell tumors, validated blood-based tumor markers for EWS are lacking. Imaging and histologic evaluation of tumor biopsies are the only techniques for monitoring the response to treatment and early risk stratification in these patients.

In most current treatment approaches, all patients with EWS receive identical, highly intense induction chemotherapy associated with considerable toxicity and the risk of overtreatment in some patients. Compared with other childhood cancers, the incidence of secondary neoplastic malignancies is relatively high at 5% to 10% in patients with EWS, with a median latency of 8 years (16). Because of the lack of promising new agents for individualized targeted therapy, systemic therapy continues to be based on multiagent cytostatic therapy. Serial minimally invasive blood sampling for ctDNA quantification may support improved risk stratification and reduce radiation exposure and anesthesia risks for pediatric patients by minimizing repeated imaging scans. In addition, evaluation of ctDNA levels may improve the understanding of tumor activity and viability in the therapeutic setting. Previous studies, including our own, have shown that quantification of fusion genes in plasma samples of patients with EWS is feasible and correlates with tumor volume and activity (17–23).

This study was conducted to evaluate the diagnostic and prognostic impact of ctDNA quantification in patients with EWS. We analyzed ctDNA copy numbers at diagnosis and during induction therapy in 102 children, adolescents, and young adults treated in the EWING2008 trial (24, 25). Figure 1 shows a graphical summary illustrating our study design. We showed that ctDNA quantification enables early risk stratification as well as initial assessment of treatment responses in EWS.

Figure 1.

Graphical summary of the study design. Plasma samples from 102 patients with EWS from the EWING2008 trial were collected before treatment start and during induction therapy at the first, second, third, and sixth VIDE cycle, as well as after the completion of induction therapy (OP/RT). To quantify ctDNA, the patient-specific genomic EWSR1-FLI1 and EWSR1-ERG fusion sequences were used. These genes were first identified from tumor biopsy samples by multiplex PCR assays or next-generation sequencing. After designing breakpoint-spanning primers and probe sets, tumor-specific sequences were quantified in the plasma samples by digital droplet PCR. Finally, ctDNA levels were correlated with clinical established risk factors and outcomes.

Figure 1.

Graphical summary of the study design. Plasma samples from 102 patients with EWS from the EWING2008 trial were collected before treatment start and during induction therapy at the first, second, third, and sixth VIDE cycle, as well as after the completion of induction therapy (OP/RT). To quantify ctDNA, the patient-specific genomic EWSR1-FLI1 and EWSR1-ERG fusion sequences were used. These genes were first identified from tumor biopsy samples by multiplex PCR assays or next-generation sequencing. After designing breakpoint-spanning primers and probe sets, tumor-specific sequences were quantified in the plasma samples by digital droplet PCR. Finally, ctDNA levels were correlated with clinical established risk factors and outcomes.

Close modal

Patient characteristics

We included 489 plasma samples from 102 patients with EWS enrolled in the EWING2008 trial (NCT00987636) in our multicenter study, EFACT. The EFACT study was amended to the EWING2008 trial in April 2014 as an accompanying research study in conjunction with the infrastructure for sample processing in the PROVABES (PROspective VAlidation of Biomarkers in Ewing Sarcoma) network. Patient plasma samples were prospectively collected in selected trial centers until the end of recruitment for the EWING2008 trial in March 2019 and centrally analyzed in batches at the Erlangen study center.

The median age of the study cohort was 14.9 years (range: 2–38 years). At the date of diagnosis, 31/98 (31.6%) patients had metastases, the primary tumor was localized in the pelvis in 25/99 (25.3%) patients, the tumor volumes of 38/96 (39.6%) patients were greater than 200 mL, and 44/56 (78.6%) displayed good histologic responses as confirmed by histology from resected tumor tissue after induction therapy. Thirty-eight patients (37%) had relapsed or died by the time of data closure in the EWING2008 trial. Patients were monitored for a median follow-up time of 2.5 years (range: 0–7.9 years). The characteristics of patients in the EFACT study are listed in Table 1.

Table 1.

Overview of the patient cohort (EWING2008; N = 1,421).

Patient characteristics
EFACT N = 102Non-EFACT N = 1,319
Number of patientsPercentage (%)Number of patientsPercentage (%)P value
Sex 
 Male 65 63.7 759 57.5  
 Female 37 36.6 560 42.5 0.25 
Stage at diagnosisa 
 Localized disease 67 68.4 884 69  
 Metastases 31 31.6 398 31 0.91 
Localization of primary tumora 
 Pelvis 25 25.3 302 23.5  
 Other 74 74.7 982 76.5 0.71 
Tumor volumea 
 <200 mL 58 60.4 735 59.3  
 ≥200 mL 38 39.6 504 40.7 0.91 
Histological responsea 
 Good 44 78.6 521 70.1  
 Poor 12 21.4 222 29.9 0.22 
Relapsea 
 No 64 62.7 842 64.4  
 Yes 38 37.3 465 35.6 0.75 
Age at diagnosis (median, range) 14.9 years (2.3–38.5) 15.1 years (0.1–72) <0.01b 
Patient characteristics
EFACT N = 102Non-EFACT N = 1,319
Number of patientsPercentage (%)Number of patientsPercentage (%)P value
Sex 
 Male 65 63.7 759 57.5  
 Female 37 36.6 560 42.5 0.25 
Stage at diagnosisa 
 Localized disease 67 68.4 884 69  
 Metastases 31 31.6 398 31 0.91 
Localization of primary tumora 
 Pelvis 25 25.3 302 23.5  
 Other 74 74.7 982 76.5 0.71 
Tumor volumea 
 <200 mL 58 60.4 735 59.3  
 ≥200 mL 38 39.6 504 40.7 0.91 
Histological responsea 
 Good 44 78.6 521 70.1  
 Poor 12 21.4 222 29.9 0.22 
Relapsea 
 No 64 62.7 842 64.4  
 Yes 38 37.3 465 35.6 0.75 
Age at diagnosis (median, range) 14.9 years (2.3–38.5) 15.1 years (0.1–72) <0.01b 

aMissing values excluded.

bClinically not relevant.

All patients were treated in the EWING2008 trial. Briefly, induction therapy for all patients consists of six multiagent chemotherapy elements, including vincristine, ifosfamide, doxorubicin, and etoposide (VIDE). Following local therapy, tumor resection, radiation, or a combination of both was performed upon assessment by a multidisciplinary review board. Depending on individual risk stratification, the patients were allocated to the respective therapy arm R1 [standard-risk arm for patients with localized disease and good histologic response (<10% viable tumor cells)], R2 (high-risk arm for poor responders with localized disease or patients with large tumor volumes that were noneligible for assessment of histologic response and patients with primary pulmonary metastases), and R3 (very high–risk arm for patients with disseminated disease; refs. 26, 27). During consolidation therapy, R1 patients were treated with eight chemotherapy elements including vincristine, actinomycin D, and ifosfamide (VAI; male patients) or cyclophosphamide (VAC; female patients) combined with zoledronic acid upon randomization (25). Consolidation treatment in R2 patients consisted of eight cycles of VAI or high-dose busulfan-melphalan with autologous stem cell reinfusion (27). R3 patients were treated with eight cycles of VAC or high-dose chemotherapy using treosulfan-melphalan followed by autologous stem cell reinfusion to eight cycles of standard adjuvant chemotherapy VAC (24).

Response assessment was carried out by MRI or CT and optional PET/CT after the second and fifth VIDE block within the induction therapy. In some cases, additional imaging studies were performed for specific clinical indications.

Sampling and storage of blood specimens

Blood samples were collected into EDTA tubes at the time of diagnosis; at the first, second, third, and sixth VIDE block; and after operation/radiotherapy for ctDNA quantification. To prevent the degradation of cell-free DNA and minimize contamination with genomic DNA from leukocytes, blood samples were immediately centrifuged at 1,200 × g for 10 minutes. The plasma was separated from peripheral blood cells, aliquoted into microtubes, and frozen at −80°C. All participating clinical centers followed a standard operational procedure, including preparation of blood samples within a maximum of 2 hours after the blood was drawn. Plasma samples were shipped on dry ice to the central laboratory for further analysis.

The leading ethics committee “Ärztekammer Westfalen-Lippe und der Westfälischen Wilhelms-Universität Münster” (2008–391-f-A; EudraCT 2008–003658–13) and human research ethics committees at each hospital approved the study. All patients or their legal guardians gave written informed consent for the collection of blood samples and analysis of ctDNA in accordance with the Declaration of Helsinki.

Identification of genomic fusion sites

Genomic fusion sequences were identified from fresh-frozen cryopreserved biopsies or paraffin-embedded tissues if sufficient material was available for multiplex PCR (18). Next-generation sequencing (NGS) was performed for cases in which minute paraffin-embedded tissue biopsies with only minimal DNA quantity were available and for sequencing of the genomic fusion sequences directly from plasma samples from patients with no accessible tumor. Highly fragmented DNA from paraffin-embedded tissues and short fragmented ctDNA were analyzed using customized libraries targeting intronic and exonic regions of EWSR1, FLI1, and ERG commonly involved in chromosomal translocation using the Thermo Fisher Scientific Ion S5 System or whole-genome sequencing libraries using an Illumina HiSeq 4000 instrument or Novaseq S4 instrument (Illumina). Libraries were generated according to manufacturers' instructions using the maximum amount of cell-free DNA (16 μL eluate). Potential EWSR1–FLI1 and EWSR1–ERG fusions were identified using socrates v1.13.1 (28) after data preparation using samtools v 0.1.19 (29) and mapping to the human genome (Genome Reference Consortium Human Build 37, hg19) using bowtie2 (30) with the setting “sensitive-local.” The most promising breakpoint candidates (rated by the number of identified breakpoint copies) were confirmed by PCR and breakpoint-spanning primers.

For whole-genome sequencing, base calls were converted into BAM files using Illumina2bam (https://github.com/wtsi-npg/illumina2bam) and demultiplexed using BamIndexDecoder from the same package. Reads were mapped to hg38 using the BWA-MEM software (31) with default settings. Aligned BAM files were loaded into the Integrative Genomics Viewer (32) and relevant genomic regions (EWSR1, FLI1, and ERG) were manually screened for a cluster of discordant reads indicating a translocation to one of the other genes (33).

Identification of the patient-specific fusion sequence from plasma samples by whole-genome sequencing is independent of tumor biopsy but requires a ctDNA ratio above 10% with the sequencing coverage chosen in this study.

Digital droplet PCR for quantification of cell-free circulating DNA

Cell-free circulating DNA was isolated using the QIAsymphony Circulating DNA Kit with the QIAsymphony SP (Qiagen) instrument or QIAamp MinElute ccfDNA Kit (Qiagen) for manual isolation according to the manufacturer's instructions. We used 0.6 to 3 mL plasma samples for isolation of cell-free DNA depending on the patient's age and weight.

Tumor-specific cell-free DNA (ctDNA) was analyzed by quantifying patient-specific fusion sequences using breakpoint-spanning primer and probe sets. Because of the high fragmentation of cell-free DNA (fragments ∼135–170 bp; refs. 8, 23), personalized digital droplet PCR (ddPCR) assays were designed to generate small amplicon lengths. Using the same criteria, an assay for quantifying single-copy gene albumin (ALB) was generated and used as the reference gene.

ctDNA Quantification was performed on a QX200 reader system (Bio-Rad) using probe‐based quantification assays with 2× ddPCR Supermix for Probes (no dUTP) and double-quenched FAM-ZEN-IBFQ probes. A regular QX200 reaction contained 7 μL cell-free DNA for high-sensitivity detection of patient-specific ctDNA fragments and 1 μL cell-free DNA for quantification of the reference gene ALB. A no-template control and at least two wild‐type negative controls were run on each plate with adjusted amounts of DNA to ensure that no false-positive data were generated. All assays were conducted in duplicate. ddPCR assays were performed according to the manufacturer's instructions and analyzed using QuantaSoft Analysis Pro (Version 1.0.596; Bio-Rad).

Samples were regarded as positive when at least three positive droplets were detected. Samples with one or two positive droplets were counted as non-quantifiable. The absolute numbers of fusion sequence copies were calculated by normalization of fusion‐specific probe signals to the signal of the single-copy human ALB. ddPCR assays for detecting ctDNA copies achieved a sensitivity of approximately 0.1% ctDNA copies/ALB copies (∼3 ctDNA copies/mL plasma) detectable.

Statistical analyses

For statistical analyses, we used the validated data set of the EWING2008 trial after data closure (24, 25). Statistical analyses and illustrations were conducted using SPSS Statistics 27 software (SPSS, Inc.), SAS 9.4, and PRISM 8 software packages (GraphPad, Inc.). Event-free survival (EFS) and overall survival (OS) were calculated using the Kaplan–Meier method from trial registration to the first event (death) or last follow-up. Events were defined as relapse, second malignancy, or death. Univariate comparisons were estimated using the log-rank test. Multivariate analyses were performed by Cox's proportional hazard method. Chi-square or Fisher test was used to test proportions. The significance level was set at P < 0.05 (two-sided). No alpha corrections were conducted because of multiple testing.

EFACT patient cohort

Participants involved in this study were treated in the EWING2008 trial. The analyzed cohort of 102 individuals is representative of the entire trial cohort (Table 1). Sequencing of the genomic breakpoint identified a EWSR1–FLI1 fusion in 95 patients, EWSR1–ERG fusion in 6 patients, and EWSR1–CREM in 1 patient.

ctDNA levels at induction chemotherapy

The absolute numbers of ctDNA copies were calculated as the ratio of fusion‐specific probe signals and single-copy human ALB signal (fusion gene/ALB copies). Pretreatment ctDNA was detected in 93/102 individuals (91%). In the nine cases without detectable baseline ctDNA before induction therapy, tumors had been removed in an initial surgery in 5 patients. One of these patients, with initial partial resection of the tumor, had detectable ctDNA after the first VIDE block and was still positive at VIDE2. The other 4 patients remained negative during treatment. Two additional cases without detectable pretreatment ctDNA levels had very small osseous tumors of 3 and 11 mL, respectively, in peripheral extremity segments. In the last mentioned case, ctDNA was detected after the first chemotherapeutic block in response to therapy; remaining time points were again negative. No suitable primer probe assay could be established, or the quality of plasma samples was insufficient in one case each.

At the time of diagnosis, ctDNA levels ranged from 0.0005 to 0.86 (median of 0.097). Higher ctDNA levels can be quantified in tumors with larger volumes compared with tumors with smaller volumes (r = 0.56, P < 0.001; Fig. 2A).

Figure 2.

ctDNA levels (fusion gene/ALB copies) at initial diagnosis and during induction chemotherapy. A, Quantified pretreatment ctDNA levels correlated with initial tumor volumes. Unfilled circles represent nonmetastatic tumors (n = 51); black filled circles represent metastatic tumors (n = 21). B, ctDNA levels before treatment start (VIDE1A) and during therapy after the first, second, third, and sixth VIDE cycle (VIDE1E, VIDE2, VIDE3, and VIDE6), as well as after the completion of induction therapy, before starting consolidation therapy (OP/RT). Blue circles represent plasma samples from patients without event; red circles represent plasma samples from patients with event. Gray bars represent the median. All non-shown P values are >0.05.

Figure 2.

ctDNA levels (fusion gene/ALB copies) at initial diagnosis and during induction chemotherapy. A, Quantified pretreatment ctDNA levels correlated with initial tumor volumes. Unfilled circles represent nonmetastatic tumors (n = 51); black filled circles represent metastatic tumors (n = 21). B, ctDNA levels before treatment start (VIDE1A) and during therapy after the first, second, third, and sixth VIDE cycle (VIDE1E, VIDE2, VIDE3, and VIDE6), as well as after the completion of induction therapy, before starting consolidation therapy (OP/RT). Blue circles represent plasma samples from patients without event; red circles represent plasma samples from patients with event. Gray bars represent the median. All non-shown P values are >0.05.

Close modal

Eighty-three percent (43/52) of patients with no events and all patients (39/39) with events displayed detectable pretreatment ctDNA levels (VIDE1A) with a median of 0.0469 (range: 0.0005–0.576) and 0.165 (range: 0.002–0.857), respectively. A rapid reduction in ctDNA levels was observed after the start of induction chemotherapy. However, the group of patients who had relapsed exhibited longer-lasting higher ctDNA levels during induction therapy (Fig. 2B).

In the group of patients who developed no events, 70% (30/43) of patients displayed detectable ctDNA copy numbers after the first chemotherapeutic block (VIDE1E) with a median of 0.040 (range: 0.0001–0.485) and 33% (17/51) at the second chemotherapeutic block (VIDE2, median: 0.001; range: 0.0001–0.11). All patients were negative at VIDE3, VIDE6, or before consolidation chemotherapy began.

In the group of patients with events, ctDNA copies were detectable in 97% (32/33) of patients after the first VIDE cycle (VIDE1E) with a median of 0.056 (range: 0.001–0.68), 57% (21/37) at VIDE2 (median: 0.006; range: 0.0002–0.123), and 26% (10/38) at VIDE3 (median: 0.04; range: 0.0006–0.072). We found that 21% of patients had detectable ctDNA copies at VIDE6 or before consolidation chemotherapy (8/38, median: 0.005; range: 0.001–0.13). To date, all patients with detectable ctDNA copy numbers at VIDE3, VIDE6, or consolidation chemotherapy had relapsed, of which 11 of 16 patients (69%) had died from the disease.

Correlation of ctDNA quantification with clinical risk factors

We evaluated the correlation of ctDNA quantification for therapy response monitoring and early risk stratification. ctDNA levels over the course of induction chemotherapy were correlated with established adverse risk factors, specifically the tumor volume >200 mL, presence of metastases, localization in the pelvis, osseous localization, and age at diagnosis ≥15 years, as well as a poor histological response after induction chemotherapy (Fig. 3A).

Figure 3.

A, Quantified mean ctDNA levels (fusion gene/ALB copies) before (VIDE1A) and after (VIDE1E) the first chemotherapy block separated, according to different clinical risk criteria. All non-shown P values are >0.05. B, Quantified mean ctDNA levels (fusion gene/ALB copies) at different risk groups. All non-shown P values are >0.05. C, Mean tumor volumes in the study groups: pelvic tumors versus nonpelvic tumors; metastatic versus nonmetastatic tumors; and osseous versus nonosseous. All non-shown P values are >0.05. D, Results of multivariate analyses. F, F value; ηp2, partial eta squared.

Figure 3.

A, Quantified mean ctDNA levels (fusion gene/ALB copies) before (VIDE1A) and after (VIDE1E) the first chemotherapy block separated, according to different clinical risk criteria. All non-shown P values are >0.05. B, Quantified mean ctDNA levels (fusion gene/ALB copies) at different risk groups. All non-shown P values are >0.05. C, Mean tumor volumes in the study groups: pelvic tumors versus nonpelvic tumors; metastatic versus nonmetastatic tumors; and osseous versus nonosseous. All non-shown P values are >0.05. D, Results of multivariate analyses. F, F value; ηp2, partial eta squared.

Close modal

In all risk factor subgroups, ctDNA plasma levels were higher before the first VIDE cycle (VIDE1A) than after chemotherapy (VIDE1E). Male and female patients showed similar ctDNA levels at both time points. In the group of patients aged ≥15 years at diagnosis, a small increase in ctDNA copies was detectable after chemotherapy (1.7-fold). Patients with pelvic tumor localization showed 2.4-fold higher ctDNA levels at VIDE1A (P < 0.01) and 2.1-fold higher ctDNA levels at VIDE1E. Similarly, patients with an initial tumor volume larger than 200 mL exhibited 2.5-fold higher ctDNA levels at VIDE1A (P < 0.01) and 4.1-fold higher levels at VIDE1E (P < 0.01) than patients with smaller tumor volumes. In patients with known metastases at diagnosis, 2-fold increased ctDNA levels were observed before (P = 0.01) and after the first VIDE cycle. ctDNA levels in osseous tumors were 2.9-fold higher at VIDE1A (P < 0.01) and 2.2-fold higher at VIDE1E compared with non-osseous tumors. In contrast to previously mentioned risk factors, patients with poor histological responses showed lower ctDNA numbers compared with patients with good histologic responses (0,5-fold, VIDE1A; 0,45-fold, VIDE1E; Fig. 3A). Tumor volume, metastatic status, and tumor location were equally distributed in the good and poor response groups (Supplementary Fig. S1). In addition, different ctDNA values before treatment began to be correlated with the classification of the current strata R1, R2 localized, R2 pulmonary, and R3 with increasing ctDNA copy numbers (Fig. 3B).

The level of detectable ctDNA copies at the beginning of therapy was mainly determined by the tumor volume, pelvis tumor localization, presence of metastases, and osseous tumors. This effect was also recognizable after the first VIDE cycle (VIDE1E) but only showed significant differences for tumor volumes smaller or larger than 200 mL (P < 0.01; Fig. 3A). Test groups were split according to tumor size to examine whether the different ctDNA levels were mainly attributable to an unbalanced distribution of tumor size in the investigated groups (pelvis tumor localization, presence of metastases, and osseous tumors). Tumor volume distribution was not related to the various ctDNA levels in all three test groups (Fig. 3C). Multivariate analyses also indicated the independence of variables related to tumor volume and confirmed the independence of pelvis tumor localization and presence of metastases (Fig. 3D).

Correlation of ctDNA copy numbers with EFS and OS

Follow-up data were available for all 102 patients. Significant differences were observed for pelvic versus nonpelvic tumors and for metastatic versus nonmetastatic disease. Patients with primary tumor origin in the pelvis and the presence of metastasis at the date of diagnosis had inferior EFS and OS. The EFS estimates since registration in patients with pelvic and nonpelvic tumors were ∼25% and ∼60%, respectively (P = 0.002). The OS data were ∼50% and ∼80%, respectively (P = 0.049). Patients with known metastases at the date of diagnosis displayed an EFS of ∼35% compared with ∼60% in patients without metastases (P = 0.017). The OS rates were ∼50% and ∼85% in metastatic versus nonmetastatic disease, respectively (P = 0.015). Thus, the outcome of the EFACT cohort is consistent with previous treatment results after stratification with clinical prognostic parameters (Supplementary Fig. S2).

To determine whether ctDNA at diagnosis is correlated with outcomes, we analyzed the EFS and OS for three groups of individuals with high (ctDNA: 0.1448–0.8569), medium (ctDNA: 0.0207–0.1447), and low (ctDNA: 0–0.0206) levels. The calculated EFS rates were ∼25%, ∼50%, and ∼70% (P = 0.008), whereas the calculated OS rates were ∼50%, ∼75%, and ∼90% (P = 0.047), respectively (Fig. 4A). This effect was particularly pronounced in metastatic disease (Supplementary Fig. S3), whereas, in localized disease, the ctDNA levels only separated the high-risk group in EFS (Supplementary Fig. S3).

Figure 4.

EFS and OS by quantified pretreatment ctDNA levels (A) and by detectable ctDNA levels at the second chemotherapy block VIDE2 (B).

Figure 4.

EFS and OS by quantified pretreatment ctDNA levels (A) and by detectable ctDNA levels at the second chemotherapy block VIDE2 (B).

Close modal

The most significant prediction of EFS is the absence or presence of ctDNA copies at VIDE2. Patients with ctDNA copy numbers present in VIDE2 displayed EFS rates of ∼40% versus ∼70% and OS rates of ∼65% compared with ∼85% in those without ctDNA in VIDE2 (P = 0.033 and 0.220), respectively (Fig. 4B). Thus, quantification of ctDNA at the time of the first and second VIDE cycle provides a reliable parameter for early risk stratification of patients with EWS.

EWS presents with a markedly heterogeneous spectrum, ranging from limited extraosseous manifestation to large tumors of the pelvis or spine and disseminated disease with distant metastases to the lung, bone marrow, and distant parts of the skeletal system. Patients with EWS with metastatic or recurrent disease have a very poor prognosis. Optimal risk stratification is therefore critical for improving treatment outcomes, given that innovative drugs with the potential to profoundly affect survival results are currently unavailable (34).

In many adult solid tumor diseases, detection and quantification of ctDNA has become increasingly relevant as a noninvasive biomarker, as it allows for close and repetitive assessment of therapy responses and prognostic predictions at specific time points during therapy (35, 36). Recurrent hotspot mutations in oncogenes or tumor-suppressor genes are commonly used as tumor-specific targets. The most frequent mutations in EWS, affecting STAG2, CDKN2A, and TP53, account for less than 20% of patients overall at diagnosis (10, 34). Fusion genes are causative for EWS and, in contrast to point mutations, stable during the clonal evolution of tumor cells and therefore also capture metastases and relapse lesions (9, 11). However, using fusion genes as targets for ctDNA quantification requires the design of an individual primer-probe-set for each patient. The regular turn-around time for ctDNA quantification at the first time point (at diagnosis) takes about two weeks and is composed of DNA preparation, identification of genomic fusions sequence, primer/probe design and synthesis, and ctDNA quantification using ddPCR. Once the patient-specific assay is established, follow-up samples can be measured within four days. This approach is standard in the therapy monitoring of lymphoblastic leukemia by minimal residual disease diagnostics utilizing clone-specific immunoglobulin and T-cell receptor rearrangements. Researchers, including our own research group, previously showed that fusion genes could be detected in plasma samples of EWS and used to quantify ctDNA for the evaluation of tumor load and tumor activity (17, 18, 20–22). To validate the clinical prognostic relevance of ctDNA copy numbers in EWS, larger cohorts of patients with uniform treatment regimens must be studied.

Shulman and colleagues analyzed EWS-specific ctDNA in a retrospective study in a cohort of 94 patients enrolled on the COG Ewing sarcoma biology study AEWS07B. The pretreatment levels of ctDNA were detected in 53% of banked plasma specimens using a NGS hybrid capture assay and an ultra-low-pass whole-genome sequencing assay. The outcomes of patients with detectable ctDNA copies at diagnosis were inferior compared with those in patients without detectable ctDNA copies. In patients with localized disease, EFS and OS were 48.6% and 79.8%, respectively, for patients with detectable ctDNA. Patients without detectable ctDNA had EFS and OS of 82.1% and 92.6%, respectively. In patients with metastatic disease, the EFS was 34.1% with detectable ctDNA and 85.7% without detectable ctDNA (19). In our study, the vast majority of EWS patients (93/102; 91%) had detectable ctDNA before starting treatment. Therefore, we classified the patients into high, medium, and low ctDNA levels for EFS and OS analysis (Fig. 4A). For the entire EFACT cohort, we calculated EFS of ∼25%, ∼50%, and ∼70% and OS of ∼50%, ∼75%, and ∼90%. Separate analysis of patients with localized or metastatic tumors shows that the less treatable patients with metastatic EWS in particular can benefit from a risk classification based on ctDNA quantification (Supplementary Fig. S3).

The higher detection rate of ctDNA in our study is likely mainly attributable to technical differences. Plasma specimens in our study were prospectively collected and processed on-site within 2 hours according to a standardized procedure to ensure optimal sample quality. Furthermore, in our study, the genomic fusion sequence was determined from tumor biopsies in most cases and quantified by a highly sensitive ddPCR assay. The background signal of non-tumor DNA was significantly lower in tumor biopsies compared with in the patient's plasma samples, and the overall DNA content was much higher, facilitating the detection of fusion sequences. Sole evaluation of blood at the time of diagnosis has logistical advantages; nevertheless, omission of an initial tumor biopsy is currently not desirable because of its considerable benefit for examining established diagnostic and prognostic markers, so that genetic characterization for marker generation does not require additional effort.

In our study, all patients were participants in a clinical trial with uniform induction therapy. Patients were investigated at the time of diagnosis and consecutively several times after sequential chemotherapy blocks. By separating study participants into three groups with high, intermediate, and low ctDNA levels at diagnosis, we observed an association between the prognosis and ctDNA level in patients with EWS (Fig. 4A). Our data revealed a strong association of pretreatment ctDNA levels with the tumor volume, presence of metastases, pelvic tumors, and osseous tumor localization.

The novelty of our study is the quantitative monitoring of ctDNA levels in response to each chemotherapy block during the course of induction therapy to assess the dynamics of treatment response following uniform chemotherapy. Detectable ctDNA at the second VIDE cycle independently indicates poor prognosis. Furthermore, patients with positive ctDNA levels at the third VIDE cycle or later form a very high–risk group. All 16 patients tested positive after VIDE2 had relapsed, and 14/16 had died of tumor disease by the end of the study. Notably, 4/16 patients were initially classified as low risk (R1) based on the tumor volume, localization, and absence of metastases.

Interestingly, all clinical risk factors, such as tumor size, pelvic location, and metastasis, showed the expected association with elevated ctDNA levels despite the presence of the risk factor of "poor histologic response" after the completion of induction therapy, which showed the tendency towards an inverse correlation with low ctDNA levels before and after the first VIDE block. This difference is most evident in metastatic disease and less pronounced in localized disease (Supplementary Fig. S4). We interpreted the higher ctDNA levels in good responders to have resulted from decaying tumor cells with insufficient clearance of ctDNA, as apoptotic and necrotic tumor cells that have been insufficiently cleared by phagocytes may be the main source of ctDNA (37–39). The effect of an unbalanced distribution of patients in the investigated groups, for example, owing to a particularly large number of patients with large tumor volumes or metastatic tumors in the group with a good histological response, can be excluded. In patients with good histologic responses, tumors may appear as more fragile and are therefore more sensitive to therapy. Rapid cellular turnover of tumor cells and regions of hypoxia increase apoptosis and necrosis, which, in combination with inadequate clearance, lead to the accumulation of ctDNA and its release into the blood system (40). To validate this hypothesis in vivo, representative tumor samples must be obtained during the course of induction therapy. However, this approach is not feasible in practice, particularly for tumors with poorly accessible localization.

At the time of study initiation, insufficient data on the variation of ctDNA levels under treatment were available to define an explicit statistical approach. We therefore decided to investigate a representative sample of the EWING2008 study cohort (Table 1) to reflect the heterogeneity of the disease and to obtain information on ctDNA dynamics in all subgroups. Limitations of this, therefore, are that the size of the study population was determined by completion of sample collection with data closure of the EWING2008 trial and that no statistically significant results could be achieved in different subgroups.

In summary, our data show that ctDNA levels represent valuable biomarkers for early risk stratification and monitoring of treatment responses in patients with EWS. Notably, ctDNA was most informative in the first three chemotherapy blocks of induction therapy. The combination of ctDNA quantification with advanced imaging techniques and functional imaging using metabolic features will optimize therapy assessment (22). The use of a serum marker, such as ctDNA, will allow for closer monitoring of therapy success and thus provide opportunities for early, personalized, risk-adapted therapy.

P. Peneder reports grants from Austrian National Bank and Kapsch Group during the conduct of the study. E.M. Tomazou reports grants from Austrian National Bank's Jubiläumsfonds and charitable donation of Kapsch Group during the conduct of the study. U. Dirksen reports grants from German Cancer Aid and Mayer Foundation outside the submitted work. No disclosures were reported by the other authors.

M. Krumbholz: Conceptualization, data curation, investigation, visualization, methodology, writing–original draft, writing–review and editing. J. Eiblwieser: Formal analysis, investigation, visualization, methodology, writing–review and editing. A. Ranft: Data curation, software, validation, visualization, writing–review and editing. J. Zierk: Data curation, software, formal analysis, writing–review and editing. C. Schmidkonz: Investigation, writing–review and editing. A.M. Stütz: Formal analysis, investigation, writing–review and editing. P. Peneder: Software, formal analysis, investigation, writing–review and editing. E.M. Tomazou: Formal analysis, funding acquisition, writing–review and editing. A. Agaimy: Writing–review and editing, material support. T. Bäuerle: Writing–review and editing, administrative support. W. Hartmann: Writing–review and editing, material support. U. Dirksen: Conceptualization, data curation, funding acquisition, project administration, writing–review and editing. M. Metzler: Conceptualization, data curation, funding acquisition, writing–original draft, project administration, writing–review and editing.

This work was supported by the EURO EWING Consortium (EEC)—international clinical trials to improve survival from Ewing sarcoma (grant agreement number 602856) to M. Metzler and U. Dirksen, German Cancer Aid to U. Dirksen (DKH 108128, 70113419, 70112018), EraNet consortium PROspective VAlidation of Biomarkers in Ewing Sarcoma [PROVABES ERA-Net-TRANSCAN (01KT1310)], Trettner-Stiftung (T0355/31554/2018/sm) to M. Metzler and U. Dirksen, by grants of the Madeleine Schickedanz Kinderkrebs-Stiftung, “Schornsteinfeger helfen krebskranken Kindern”, and the Gerd and Susanne Mayer Foundation (to U. Dirksen). E.M. Tomazou was supported by a fellowship of the Austrian Science Fund (FWF, Elise Richter Fellowship 730 V506-B28), by the Austrian National Bank's Jubiläumsfonds (OeNB Project Number: 17876), and by an institutional grant financed by a charitable donation of Kapsch Group. The EWING2008 trial was sponsored by the University Hospital Muenster. We thank all participants for their assistance with the study; particularly, we thank Susanne Jabar, Christiane Schaefer, and Dagmar Clemens from the EWING2008 study group. The authors thank Sabine Semper, Perdita Weller, Tatjana Flamann, and Ursula Jacobs for providing technical support.

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.

1.
Russano
M
,
Napolitano
A
,
Ribelli
G
,
Iuliani
M
,
Simonetti
S
,
Citarella
F
, et al
Liquid biopsy and tumor heterogeneity in metastatic solid tumors: the potentiality of blood samples
.
J Exp Clin Cancer Res
2020
;
39
:
95
.
2.
Corcoran
RB
,
Chabner
BA
. 
Application of cell-free DNA analysis to cancer treatment
.
N Engl J Med
2018
;
379
:
1754
65
.
3.
Stewart
CM
,
Tsui
DWY
. 
Circulating cell-free DNA for non-invasive cancer management
.
Cancer Genet
2018
;
228–229
:
169
79
.
4.
Kruger
S
,
Heinemann
V
,
Ross
C
,
Diehl
F
,
Nagel
D
,
Ormanns
S
, et al
Repeated mutKRAS ctDNA measurements represent a novel and promising tool for early response prediction and therapy monitoring in advanced pancreatic cancer
.
Ann Oncol
2018
;
29
:
2348
55
.
5.
Tie
J
,
Wang
Y
,
Tomasetti
C
,
Li
L
,
Springer
S
,
Kinde
I
, et al
Circulating tumor DNA analysis detects minimal residual disease and predicts recurrence in patients with stage II colon cancer
.
Sci Transl Med
2016
;
8
:
346ra92
.
6.
Moss
J
,
Zick
A
,
Grinshpun
A
,
Carmon
E
,
Maoz
M
,
Ochana
BL
, et al
Circulating breast-derived DNA allows universal detection and monitoring of localized breast cancer
.
Ann Oncol
2020
;
31
:
395
403
.
7.
Choudhury
AD
,
Werner
L
,
Francini
E
,
Wei
XX
,
Ha
G
,
Freeman
SS
, et al
Tumor fraction in cell-free DNA as a biomarker in prostate cancer
.
JCI Insight
2018
;
3
:
e122109
.
8.
Heidrich
I
,
Ackar
L
,
Mossahebi Mohammadi
P
,
Pantel
K
. 
Liquid biopsies: potential and challenges
.
Int J Cancer
2021
;
148
:
528
45
.
9.
Crompton
BD
,
Stewart
C
,
Taylor-Weiner
A
,
Alexe
G
,
Kurek
KC
,
Calicchio
ML
, et al
The genomic landscape of pediatric Ewing sarcoma
.
Cancer Discov
2014
;
4
:
1326
41
.
10.
Tirode
F
,
Surdez
D
,
Ma
X
,
Parker
M
,
Le Deley
MC
,
Bahrami
A
, et al
Genomic landscape of Ewing sarcoma defines an aggressive subtype with co-association of STAG2 and TP53 mutations
.
Cancer Discov
2014
;
4
:
1342
53
.
11.
Anderson
ND
,
de Borja
R
,
Young
MD
,
Fuligni
F
,
Rosic
A
,
Roberts
ND
, et al
Rearrangement bursts generate canonical gene fusions in bone and soft tissue tumors
.
Science
2018
;
361
:
eaam8419
.
12.
Brohl
AS
,
Solomon
DA
,
Chang
W
,
Wang
J
,
Song
Y
,
Sindiri
S
, et al
The genomic landscape of the Ewing Sarcoma family of tumors reveals recurrent STAG2 mutation
.
PLoS Genet
2014
;
10
:
e1004475
.
13.
Riggi
N
,
Suva
ML
,
Stamenkovic
I
. 
Ewing's sarcoma
.
N Engl J Med
2021
;
384
:
154
64
.
14.
Delattre
O
,
Zucman
J
,
Plougastel
B
,
Desmaze
C
,
Melot
T
,
Peter
M
, et al
Gene fusion with an ETS DNA-binding domain caused by chromosome translocation in human tumours
.
Nature
1992
;
359
:
162
5
.
15.
Sankar
S
,
Lessnick
SL
. 
Promiscuous partnerships in Ewing's sarcoma
.
Cancer Genet
2011
;
204
:
351
65
.
16.
Caruso
J
,
Shulman
DS
,
DuBois
SG
. 
Second malignancies in patients treated for Ewing sarcoma: a systematic review
.
Pediatr Blood Cancer
2019
;
66
:
e27938
.
17.
Shukla
NN
,
Patel
JA
,
Magnan
H
,
Zehir
A
,
You
D
,
Tang
J
, et al
Plasma DNA-based molecular diagnosis, prognostication, and monitoring of patients with EWSR1 fusion-positive sarcomas
.
JCO Precis Oncol
2017
;
2017
:
PO.16.00028
.
18.
Krumbholz
M
,
Hellberg
J
,
Steif
B
,
Bauerle
T
,
Gillmann
C
,
Fritscher
T
, et al
Genomic EWSR1 fusion sequence as highly sensitive and dynamic plasma tumor marker in Ewing sarcoma
.
Clin Cancer Res
2016
;
22
:
4356
65
.
19.
Shulman
DS
,
Klega
K
,
Imamovic-Tuco
A
,
Clapp
A
,
Nag
A
,
Thorner
AR
, et al
Detection of circulating tumour DNA is associated with inferior outcomes in Ewing sarcoma and osteosarcoma: a report from the Children's Oncology Group
.
Br J Cancer
2018
;
119
:
615
21
.
20.
Hayashi
M
,
Chu
D
,
Meyer
CF
,
Llosa
NJ
,
McCarty
G
,
Morris
CD
, et al
Highly personalized detection of minimal Ewing sarcoma disease burden from plasma tumor DNA
.
Cancer
2016
;
122
:
3015
23
.
21.
Klega
K
,
Imamovic-Tuco
A
,
Ha
G
,
Clapp
AN
,
Meyer
S
,
Ward
A
, et al
Detection of somatic structural variants enables quantification and characterization of circulating tumor DNA in children with solid tumors
.
JCO Precis Oncol
2018
;
2018
:
PO.17.00285
.
22.
Schmidkonz
C
,
Krumbholz
M
,
Atzinger
A
,
Cordes
M
,
Goetz
TI
,
Prante
O
, et al
Assessment of treatment responses in children and adolescents with Ewing sarcoma with metabolic tumor parameters derived from (18)F-FDG-PET/CT and circulating tumor DNA
.
Eur J Nucl Med Mol Imaging
2020
;
47
:
1564
75
.
23.
Abbou
SD
,
Shulman
DS
,
DuBois
SG
,
Crompton
BD
. 
Assessment of circulating tumor DNA in pediatric solid tumors: the promise of liquid biopsies
.
Pediatr Blood Cancer
2019
;
66
:
e27595
.
24.
Dirksen
U
,
Bhadri
V
,
Brichard
B
,
Butterfass-Bahloul
T
,
Cyprova
S
,
Faldum
A
, et al
Efficacy of add-on treosulfan and melphalan high-dose therapy in patients with high-risk metastatic Ewing sarcoma: Report from the International Ewing 2008R3 trial
.
J Clin Oncol
2020
;
38
:
abstract 11501
.
25.
Dirksen
U
,
Koch
R
,
Bhadri
V
,
Brichard
B
,
Butterfass-Bahloul
T
,
Cyprova
S
, et al
Efficacy of maintenance therapy with zoledronic acid in patients with localized Ewing sarcoma: Report from the International Ewing 2008 trial
.
J Clin Oncol
2020
;
38
:
abstract 11523
.
26.
Whelan
J
,
Le Deley
MC
,
Dirksen
U
,
Le Teuff
G
,
Brennan
B
,
Gaspar
N
, et al
High-dose chemotherapy and blood autologous stem-cell rescue compared with standard chemotherapy in localized high-risk Ewing sarcoma: results of Euro-E.W.I.N.G.99 and Ewing-2008
.
J Clin Oncol
2018
;
36
:
JCO2018782516
.
27.
Dirksen
U
,
Brennan
B
,
Le Deley
MC
,
Cozic
N
,
van den Berg
H
,
Bhadri
V
, et al
High-dose chemotherapy compared with standard chemotherapy and lung radiation in Ewing sarcoma with pulmonary metastases: results of the European ewing tumour working initiative of national groups, 99 trial and EWING 2008
.
J Clin Oncol
2019
;
37
:
3192
202
.
28.
Schroder
J
,
Hsu
A
,
Boyle
SE
,
Macintyre
G
,
Cmero
M
,
Tothill
RW
, et al
Socrates: identification of genomic rearrangements in tumour genomes by re-aligning soft clipped reads
.
Bioinformatics
2014
;
30
:
1064
72
.
29.
Li
H
,
Handsaker
B
,
Wysoker
A
,
Fennell
T
,
Ruan
J
,
Homer
N
, et al
The sequence alignment/map format and SAMtools
.
Bioinformatics
2009
;
25
:
2078
9
.
30.
Langmead
B
,
Salzberg
SL
. 
Fast gapped-read alignment with Bowtie 2
.
Nat Methods
2012
;
9
:
357
9
.
31.
Li
H
. 
Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM
. Available from: https://arxiv.org/abs/1303.3997.
32.
Robinson
JT
,
Thorvaldsdottir
H
,
Winckler
W
,
Guttman
M
,
Lander
ES
,
Getz
G
, et al
Integrative genomics viewer
.
Nat Biotechnol
2011
;
29
:
24
6
.
33.
Peneder
P
,
Stütz
AM
,
Surdez
D
,
Krumbholz
M
,
Semper
S
,
Chicard
M
, et al
Accurate detection and classification of pediatric sarcomas based on cell-free DNA fragmentation patterns
.
Nat Commun
2021
;
12
:
3230
.
34.
Grunewald
TGP
,
Cidre-Aranaz
F
,
Surdez
D
,
Tomazou
EM
,
de Alava
E
,
Kovar
H
, et al
Ewing sarcoma
.
Nat Rev Dis Primers
2018
;
4
:
5
.
35.
Siravegna
G
,
Mussolin
B
,
Venesio
T
,
Marsoni
S
,
Seoane
J
,
Dive
C
, et al
How liquid biopsies can change clinical practice in oncology
.
Ann Oncol
2019
;
30
:
1580
90
.
36.
Dudley
JC
,
Diehn
M
. 
Detection and diagnostic utilization of cellular and cell-free tumor DNA
.
Annu Rev Pathol
2021
;
16
:
199
222
.
37.
Thierry
AR
,
El Messaoudi
S
,
Gahan
PB
,
Anker
P
,
Stroun
M
. 
Origins, structures, and functions of circulating DNA in oncology
.
Cancer Metastasis Rev
2016
;
35
:
347
76
.
38.
Diaz
LA
 Jr
,
Bardelli
A
. 
Liquid biopsies: genotyping circulating tumor DNA
.
J Clin Oncol
2014
;
32
:
579
86
.
39.
Kustanovich
A
,
Schwartz
R
,
Peretz
T
,
Grinshpun
A
. 
Life and death of circulating cell-free DNA
.
Cancer Biol Ther
2019
;
20
:
1057
67
.
40.
Takai
E
,
Yachida
S
. 
Circulating tumor DNA as a liquid biopsy target for detection of pancreatic cancer
.
World J Gastroenterol
2016
;
22
:
8480
8
.