Abstract

Introduction

Despite advances in treatment and survival, mortality for pediatric patients with solid tumors that suffer of metastatic or relapsed disease remains high (1–7). During the course of the disease, children face many invasive procedures to acquire tumor material as well as imaging under general anesthesia to determine disease dissemination and response evaluation. Sampling of blood or other liquids produced by the human body, e.g. ‘liquid biopsies’ form a potential source of biomarkers that can be collected in a less invasive manner which could reduce the number of stressful procedures. Moreover, liquid biopsies contain material from both the primary tumor and metastatic lesions, thereby offering a more comprehensive view of the disease and could assist in clinical decision making (8–10). An important challenge is the correct choice of the marker. Studies focusing on the detection of tumor-derived mRNA from blood and bone marrow using a tumor-specific RNA panel have shown promising results for improving risk stratification at diagnosis, as seen in neuroblastoma and rhabdomyosarcoma (10–12). However, these methods still require the use of bone marrow, and the potential for response monitoring with this approach has not been shown yet for rhabdomyosarcoma (10). Cell-free DNA (cfDNA) from plasma holds great potential for diagnostic and prognostic purposes in pediatric solid tumors (13–16). We have previously described hypermethylated RASSF1A as a marker for cell free tumor DNA in several pediatric tumors. However, the level of methylation of RASSF1A differs in different types of pediatric tumors, which limits its use (16). In contrast to adult malignancies, mutations in pediatric tumors are scarce. Often, they have copy number alterations (CNAs) or translocations resulting in fusion genes which are considered early tumor-driving events and remain present during the entire course of the disease (17–19). In rhabdomyosarcoma, the fusion gene between PAX3 or PAX7 and FOXO1 is an important characteristic within the alveolar subtype (3, 20–22). In Ewing sarcoma, EWSR1 pairs with several fusion partners from the ETS family of transcription factors (5, 17). Neuroblastoma tumors often have amplification of MYCN, loss of heterozygosity of chromosome 1p and 11p and gain of 17q (4). Most of these CNAs result in a unique chromosomal fusion, however it is mostly unknown to which chromosome. These genetic events are formed by DNA sequences which are exclusive to a patient, thereby forming a perfect target to detect tumor-derived DNA since these sequences are not present in the background of healthy cell-free DNA, which is always present in blood. For pediatric patients with a solid tumor, fluorescent in situ hybridization (FISH), shallow whole genome sequencing (sWGS) or single nucleotide polymorphism (SNP) array has become available for routine diagnostics to identify clinically relevant fusion genes, as well as genomic deletions or amplifications. As these genomic aberrations are independent of gene activity, their presence could potentially be used to detect and quantify tumor burden. Historically, the identification of the exact breakpoint sequence has been time- and resource consuming, as WGS followed by Sanger sequencing validation was necessary (23). However, this procedure can be sped up by using a targeted approach for genomic breakpoint sequencing, like targeted locus amplification/capture (TLA/TLC). TLA/TLC is a technique that uses crosslinking of physically proximal sequences to selectively amplify and sequence regions of >100 kb surrounding specific primer or probe binding sites without prior detailed locus information. TLA can be applied to cells, while TLC is optimized for formalin fixed paraffin embedded (FFPE) material (24–27). The breakpoint sequence revealed by the TLA/TLC technique can be used to design an assay that targets the patient-specific breakpoint. We use droplet digital PCR (ddPCR) to detect these targets in small volumes of plasma from patients with pediatric solid tumors, since ddPCR allows for absolute quantification combined with high sensitivity. In this report, we investigate the possibility of designing patient-specific assays for cell free tumor DNA detection in patients with neuroblastoma, rhabdomyosarcoma and Ewing sarcoma, using TLA/TLC – based breakpoint sequences. Furthermore, we study whether the presence of these specific breakpoints correlates to residual and recurrent disease and, thus, its potential as marker for treatment response.

Methods

For a graphical overview of the methods, see Figure 1 .

Open in a separate window

Figure 1

Workflow for the development of a patient-specific assay. At primary diagnosis, tumor material is collected through biopsy or resection. The tissue is then analyzed in the regular diagnostic pipeline. This means copy number analysis through SNP array for neuroblastoma tumors, and fusion gene detection through immunohistochemistry or RT-qPCR for rhabdomyosarcoma and Ewing sarcoma. Based on the identified altered regions/copy number aberrations and fusion partner, for targeted locus amplification (TLA) or targeted locus capture (TLC) is performed on cellsor FFPE material, respectively. The breakpoint sequence(s) are then used for a patient-specific ddPCR design which is then measured on cell- free DNA from EDTA blood.

Results

Results of ddPCR assay in single and multiple breakpoints

Patient-specific ddPCR assays were designed for 15 breakpoints identified in 13 cases. An illustrative example of a ddPCR assay with one breakpoint is shown in Figure 2 for the cfDNA from diagnostic plasma and genomic DNA from the primary tumor from patient NB2049. In case more than one tumor-specific ddPCR could be designed, we aimed to combine these in a multiplex assay ( Figure 3 ). For NB2050, two breakpoints, both chromosome 2-2 breakpoints in the amplified MYCN locus, were massively amplified relative to the reference gene, resulting in overloading of the droplets and failure to quantify the cfDNA targets accurately in undiluted cfDNA from diagnostic plasma. cfDNA diluted 500 times enabled correct quantification of the different targets. Both Figures 2 , 3 illustrate the range that can be covered by ddPCR and the possibilities of absolute quantification.

Open in a separate window

Figure 2

2D plot from the ddPCR assay for NB2049 with (A). cfDNA from the diagnostic plasma sample (total cfDNA input 25.6 ng/well) and (B). positive control with DNA from FFPE material from the primary tumor (total cfDNA input 3.3 ng/well). Blue dots; droplets positive for patient-specific breakpoint (FAM channel) Green dots; droplets positive for ACTB(HEX channel) Grey dots; droplets negative for both targets.

Open in a separate window

Figure 3

2D plot from the ddPCR assay for NB2050 with 2 patient-specific breakpoints (A). cfDNA from the diagnostic plasma sample (total cfDNA input can not be determined due to overload of the droplets) (B). positive control with DNA from FFPE material from the primary tumor (total cfDNA input 1.7 ng/well). (C) Dilution of diagnostic plasma 50 times and (D). 500 times Blue dots; droplets positive for both patient-specific breakpoint (FAM channel) Green dots; droplets positive for Actin Beta (HEX channel) Pink dots; droplets positive for breakpoint Chr 2;2 nr 1 Purple dots; droplets positive for breakpoint Chr 2;2 nr 2 (with 450 nM and 125 nM primer and probe concentrations, respectively) Orange dots; droplets positive for both breakpoints and Actin Beta Black dots; droplets positive for breakpoint nr 1 and Actin Beta Salmon-colored dots; droplets positive for breakpoint nr 2 and Actin Beta Grey dot; droplets negative for both targets.

Presence of patient-specific targets correlates with disease stage

Sequential cfDNA samples obtained during the clinical disease course were measured by ddPCR for the patient-specific breakpoint and by the RASSF1A-M assay ( Figure 4 for patients with neuroblastoma, Figure 5 for patients with rhabdomyosarcoma and Ewing sarcoma). In all plasma samples taken at initial diagnosis in patients with neuroblastoma and rhabdomyosarcoma, the patient-specific targets were present. In neuroblastoma, presence of tumor-derived targets followed the clinical course, decreasing after start of treatment and reappearing before or at relapse. In patients NBL2061 and NBL2101 the tumor-specific target is clearly detectable in the cfDNA before a relapse is detected by imaging or standard bone marrow evaluation. In the two patients with rhabdomyosarcoma, the targets in the tumor-derived cfDNA disappeared quickly after start of therapy and did not reappear during therapy for relapse (RMS108) or progressive disease (RMS006). In both cases, no samples were drawn at diagnosis of relapse or progressive disease. For the patient with Ewing sarcoma, the specific breakpoint target was not detected in two cfDNA samples taken during therapy for relapse, even though design of a patient-specific breakpoint was successful, as determined in the positive control. Unfortunately, no sample taken at initial diagnosis was available for this patient.

Open in a separate window

Figure 4

Levels of patient-specific targets, reference gene ACTIN beta (ACTB) and methylated RASSF1A (RASSF1A-M) in cell-free DNA (cfDNA) from 10 neuroblastoma patients at diagnosis and during the course of the disease. Dx, diagnosis; Dx 4S, diagnosis INSS stage 4S; Dx 4, diagnosis INSS stage 4; nCt 1L, after n courses in first line therapy; nCt 2L, after courses in second line therapy; 3L, third line therapy; 1EV 3L, first evaluation third line therapy; 2EV 3L, second evaluation third line therapy; IT-0, before GD-2 immunotherapy IT-3 after 3 cycles of GD-2 immunotherapy; IT-6, after 6 cycles of GD-2 immunotherapy; MAT, myeloablative therapy and autologous stem cell transplantation; RT 2L, after radiotherapy during second line therapy; MT, maintenance treatment. PD, progressive disease; R, relapse; RD, refractory disease. Green blocks indicate first line treatment, orange blocks indicate added treatment for refractory disease, red blocks indicate treatment for progressive or relapsed disease.

Open in a separate window

Figure 5

Levels of patient-specific targets, reference gene ACTIN beta (ACTB) and methylated RASSF1A (RASSF1A-M) in cell-free DNA (cfDNA) in 2 patients with rhabdomyosarcoma (RMS026 and RMS092) and 1 patient with Ewing sarcoma (ES010) at diagnosis and during the course of the disease. Dx, diagnosis; nCt 1L, after n courses in first line therapy; M, maintenance; FUP-follow up; EVn 2L, evaluation number n during second line therapy; EOT, end of treatment; R, relapse; Start 2L, start second line therapy; R, relapse; PD, progressive disease. Green blocks indicate first line treatment, red blocks indicate treatment for progressive or relapsed disease.

For NBL2053, NBL2061, NBL2086, NBL2100, RMS006 and RMS108, the tumoroid that was used to identify the patient-specific breakpoint for TLA was grown from a tumor sample taken at relapse. However, we could detect the exact same breakpoints in plasma taken at initial diagnosis. This illustrates clearly that the targeted chromosomal breakpoints in neuroblastoma and the PAX3-FOXO1 fusion gene in the rhabdomyosarcoma patient remain stable during the course of the disease.

We show that patient-specific targets identified in tumor material by TLA can be detected in cfDNA from diagnostic plasma, furthermore, the presence of these targets track clinical course in neuroblastoma.

Discussion

In this study we demonstrate the feasibility to identify a patient-specific target, based on chromosomal structural variants, and design a patient-specific assay for use in liquid biopsies in different pediatric solid tumors. Moreover, we show that the presence of these targets in plasma at initial diagnosis for neuroblastoma and rhabdomyosarcoma, and that its presence during the course of the disease, corresponds to detectable or minimal residual disease status in neuroblastoma.

In patients with neuroblastoma, we observed that the level of tumor-derived cfDNA, as measured by the patient-specific targets, already increased before the clinical diagnosis of relapse or progressive disease was made. This finding suggests a potential for monitoring treatment response in neuroblastoma by detecting tumor-derived cfDNA. This is in line with data from our previous study on hypermethylated RASSF1A (16), but is also shown by others. Su et al. reported that the total amount of cfDNA increases before the recurrence of high risk neuroblastoma (29), which can be explained as the majority of the present cfDNA at relapse is tumor derived (16). More recently, Lodrini et al., showed the applicability of detecting tumor-derived cfDNA MYCN and ALK copy number alterations and ALK hotspot mutations in longitudinal plasma samples from patients with neuroblastoma (30). The study of Bosse et al., that predominantly included neuroblastoma patients with an event (91%), showed at least one pathogenic genomic alteration detected in 56% of the samples (31). However, only 20% and 10% of neuroblastoma tumors harbor MYCN amplification or ALK mutation, respectively (4). With the development of TLA/TLC, patient-specific targets for use in liquid biopsies can be detected for any CNAs, as illustrated in our study, which significantly increases the number of patients eligible for monitoring of disease with tumor-derived cfDNA.

In our study, we did not observe re-appearance of the patient-specific breakpoint in samples from patients with rhabdomyosarcoma and Ewing sarcoma. This might be due to a lack of well-timed samples, especially for the patient with Ewing sarcoma. Re-appearance of the patient-specific breakpoint has been described by Eguchi-Ishimae in a patient with fusion-positive rhabdomyosarcoma that suffered from relapse (32). Recently, Ruhen et al. published analysis of cfDNA from plasma in a cohort of 18 patients with rhabdomyosarcoma, which showed a rapid decrease of cfDNA targets after initiation of therapy and an increase at relapse (33). This rapid decrease of tumor-derived cfDNA was also observed by Klega et al. in patients with Ewing sarcoma and fusion-positive rhabdomyosarcoma, often becoming undetectable at the start of the second cycle of chemotherapy (34). Moreover, they observed that in patients with Ewing sarcoma the detection of tumor-derived cfDNA after start of treatment was related to the level of tumor necrosis (34). The relation to tumor burden and monitoring of relapse in Ewing sarcoma was also demonstrated in the recent study by Shulman et al. (35) They also designed a patient-specific assay for the fusion genes in 6 patients with Ewing sarcoma, using data from WGS of the tumor material. In 2 patients that remained in complete remission, the fusion breakpoint disappeared after initiation of treatment. In 4 patients that suffered from relapse, cfDNA levels of the breakpoint reflected presence of relapse and response to therapy (35). These findings from other reports underline the potential of a patient-specific target as a treatment response marker and early relapse detection. But the timing of blood sampling is crucial. In the 2 patients with rhabdomyosarcoma in our study, we did not have samples taken right before or at diagnosis of relapse, only samples taken after start of relapse therapy. Standardized and uniform sampling in a larger cohort of patients with rhabdomyosarcoma and Ewing sarcoma is essential to validate these markers for clinical use. Clinical trials are now being conducted, with liquid biopsy sampling being implemented in the current EpSSG FaR RMS trial for pediatric and adult rhabdomyosarcoma in Europe and in the US a focused trial in adults into liquid biopsies for solid tumors (36, 37).

The use of a patient-specific molecular target as marker for minimal residual disease has been implemented firmly in leukemia (38, 39). In solid tumors, our group has previously described the feasibility to design patient-specific DNA markers from aberrations detected by WGS, and successful marker detection in BM of patients with neuroblastoma (23), but this has not reached clinical practice yet. Some important factors contributing to this lack of translation should be considered. One challenge is defining a target. At initial diagnosis, the search for potential targets can be guided by clinical information and tumor histology, focusing on early oncogenic events that remain present throughout the course of the disease, for example the PAX3-FOXO1 fusion in rhabdomyosarcoma. This fusion gene is considered the tumor-driving event in this tumor entity (3, 20, 40). Also for neuroblastoma, amplification of the MYCN gene, gain of 17q and loss of heterozygosity of chromosome 1 have been found to be recurring events, occurring extremely early in tumor development and remaining present during the course of the disease (41–43). Clonal evolution can affect the suitability of targets. Studies of paired primary and relapse neuroblastoma tumor have shown that mutations detected at relapse represent outgrowth of clones already present at diagnosis or de novo events, but most structural events remain present in the relapse sample (41, 44). Combining a panel of targets from diagnosis and then updating this panel again at relapse, using fresh genetic data from the relapsed tumors, might maintain sensitivity of the patient-specific ddPCR assays. In other types of pediatric solid tumors, it might be more challenging to identify patient-specific targets that remain stable throughout the course of the disease. For example, in osteosarcoma, many structural variations have been reported throughout the genome in primary tumors (45, 46), but extensive studies on the stability of these regions in recurrent and progressive disease is lacking. Nonetheless, as many pediatric tumors harbor any structural variant (insertions, deletions, translocations) (18), this approach could benefit cfDNA research in other pediatric tumors as well (42, 45). In our center, regions involved in translocations and copy numbers were previously evaluated by FISH, RT-qPCR and SNP arrays as part of regular clinical investigations and this information can direct the investigations into patient-specific targets. Considering these recurrent regions with copy number aberrations, it would be interesting to explore a multiplex approach of TLA/TLC for neuroblastoma, targeting several regions often carrying amplifications. This approach has shown its potential for detection of translocations in acute leukemia (26).

Another important challenge is the time and effort necessary to identify a patient-specific target. The procedure for FFPE-TLC takes 2 to 3 weeks, shorter than the process based on WGS as described by Subhash et al. (47) Furthermore, FFPE-TLC opens up the possibility to analyze archived samples of patients presenting with late relapse. If no FFPE material is available and not enough cells are available directly after biopsy or surgery, then the time depends on growth of the tumoroids. This can differ significantly. TLA-based approaches to determine the patient-specific breakpoint also preclude the known objections to WGS, with the risk of unsolicited findings and their impact on patients’ lives (48, 49).

In this study, we used a ddPCR-based approach for the detection of the patient-specific targets in cfDNA. Other reports have used hybrid capture sequencing (e.g., TranSS-Seq by Klega et al.) (34). All approaches correlate well with each other (34, 50, 51). The choice of a platform for cfDNA detection depends on the availability at a specific center, the costs and whether multiplexing is necessary, in tumors with multiple targets. A next generation sequencing platform can offer a wider range of targets to be tested, but on the other hand it can be time-consuming to validate, is less flexible and more costly. The possibility of multiplexing targets in cfDNA on the ddPCR is more limited but not impossible, as reported here and in previous publications (52, 53). ddPCR thereby offers a rapid testing modality, also very suited for monitoring of residual disease during treatment and follow-up in a clinical setting.

In some cases, it might be impossible to design a patient-specific assay, which might be due to absence of an appropriate chromosomal region or presence of the sequence in the normal genome (as illustrated by case NB2100). Alternative liquid biopsy-based platforms could be explored, as we have demonstrated previously. We have developed and validated RNA panels for the detection of circulating tumor cells in the cellular compartment of blood and bone marrow for patients with neuroblastoma and rhabdomyosarcoma (10–12). Furthermore, as we also applied in the current study, an enzyme-based ddPCR for methylated RASSF1A in cfDNA is also suited for the detection of tumor-derived cfDNA. Since hypermethylation of RASSF1A has been found in many types of tumors (54–60), this assay offers another approach for liquid biopsy-based disease monitoring. The combination of both RNA and DNA-based platforms for the analysis of liquid biopsies could be complementary, as we have showed previously in a cohort of patients with rhabdomyosarcoma (61).

Conclusion

In this study, we demonstrate that patient-specific targets can be identified using targeted locus amplification in different pediatric solid tumors. Furthermore, we show that these patient-specific targets can be detected in cfDNA from plasma and their presence may correlate to (minimal) residual or recurrent disease. This approach holds promise for use in daily clinical practice.

Data availability statement

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding author.

Ethics statement

The studies involving human participants were reviewed and approved by medical ethics committees of Academic Medical Center Amsterdam and University Medical Center Utrecht. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.

Author contributions

LZ, NL, HF, MM, JSt, ESp and GT contributed to conception and design of the study. Acquisition and analysis of results was performed by LZ, NL, NG, IS, ESt, JSw, HF, ESp, MB, MK, WB, MTM, WK, FH, JM and WL. Interpretation of data was done by LZ, NL, IS, ESt, HF, ESp and GT. LZ and NL wrote the first draft of the manuscript. IS, ESt, JSw, HF, ESp wrote sections of the manuscript. All authors contributed to the article and approved the submitted version.

Acknowledgments

Funding Statement

LZ was supported by Liquidhope, a TranScan-2 project by Koningin Wilhelmina Fund, KWF Kankerbestrijding TRANSCAN 8352/TRS-2018- 00000715. NL and JSt were supported by Children Cancer Free (KiKa) grant number 312. MTM received financial support from the Deutsche Forschungsgemeinschaft (#408083583). The work conducted by Cergentis has received funding from the European Union’s Horizon 2020 SME instruments program SEQURE under grant agreement No. 806446.

References