All patients underwent treatment at the Princess Máxima Center for Pediatric Oncology between 2017 and 2023. Samples from patients were either prospectively obtained or retrospectively collected from biobanked material if informed consent from parents or guardians was obtained according to the declaration of Helsinki. BM-aspirates were collected from bilateral sites in EDTA tubes and processed within 24 hours after collection and stored in Trizol (Invitrogen, Carlsbad, CA) at -80°CC. PB samples were collected and processed the same way and stored in in PAXgene blood RNA tubes (QIAGEN, Venlo, the Netherlands) at -20°CC, or in Trizol (Invitrogen, Carlsbad, CA) at -80°CC. DNA and RNA isolation and cDNA synthesis was performed as previously described (30, 32, 40).

For the patients included in this report, tumor-derived mRNA was detected in the cellular fraction of bone marrow (BM) aspirates and/or peripheral blood (PB). For neuroblastoma, a multiplex (MPX) adrenergic marker panel was used which includes PHOX2B, TH and CHRNA3, and GAP43 and DBH for BM and PB (28, 29). For rhabdomyosarcoma, the MPX panel used for BM includes MYOD1, MYOG, PDLIM3, ACTC1, ZIC1 and PAX3/7-FOXO1 and for PB, SNAI2, CDH11, THEM47 and MEGF10 (30). Expression of the markers was normalized to GUSB expression and positivity was scored according to thresholds published by Van Zogchel et al. and Lak et al. (28–30) To calculate the level of infiltration in the neuroblastoma samples, expression levels of the mRNA markers in BM or PB were related to the expression in neuroblastoma cell line IMR32.

Detection of RASSF1A-M in ctDNA was performed using the method described by Van Zogchel et al. (32, 41) This method subjects every sample to two ddPCR reactions after cfDNA extraction, once with methylation-sensitive restriction enzymes (MRSE) added and once without. This allows for accurate discrimination between methylated and unmethylated RASSF1A. As an input control, two assays for the reference gene ACTB are used of which one amplicon is digested by the enzymes and one is not.

For the patients in this report, a MYCN primer-probe set and an NAGK (normal diploid reference gene) primer-probe set were added in a duplex PCR reaction. MYCN copy number was then determined by calculating the ratio of MYCN to the reference gene (42, 43).

Methylation analysis of cfDNA using cell-free reduced representation bisulphite sequencing (cfRRBS) was performed as previously described (39). The diagnosis is established based on the highest estimated tumor fraction using reference-based deconvolution as described by Moss et al. (39).

See Supplementals for a detailed description of patients and methods.

A 3-year-old girl presented with lethargy, abdominal pain, vomiting and a palpable abdominal mass ( Table 1 ). An MRI-scan revealed a tumor of the left kidney with several pulmonary metastases and skeletal lesions, suspicious for metastases ( Figure 2A ). A fine needle tissue biopsy performed to differentiate between clear cell sarcoma (CCSK) and Wilms tumor showed extensive necrosis and was of no diagnostic value. An MRI of the brain showed no abnormalities ( Figure 2B ) and bone marrow testing showed no metastases. Chemotherapy was started empirically according to the SIOP-RTSG UMBRELLA trial CCSK regimen. At nephrectomy, the tumor was identified as a metastatic regressive type-intermediate risk Wilms tumor. During post-operative chemotherapy, the patient developed progressive bradycardia, vomiting, eye movement disorders and epileptic seizures. Imaging (by MRI and CT-scan) of the brain showed new leptomeningeal lesions ( Figures 2C, D ) that could not be biopsied. Analysis of cerebrospinal fluid showed atypical cells of unclear origin. Again, a definitive diagnosis could not be made. Shortly thereafter, the patient died. Postmortem analysis of the intracerebral lesions showed a small blue round cell tumor which was negative for Wilms tumor markers Keratin7, PAX8 and WT1. Methylation analysis (cfRRBS) of cfDNA isolated from the bio-banked plasma sample taken at diagnosis estimated a second primary tumor, an embryonal rhabdomyosarcoma (81% estimated tumor fraction). This was later confirmed by methylation analysis by EPIC array and immunohistochemical analysis of the postmortem biopsy. Prompted by the simultaneous occurrence of two primary tumors, whole-exome sequencing (WES) was performed for additional germline analysis. No aberrations were detected.

An 11-year-old boy presented with fever, vomiting, abdominal pain and severe back pain ( Table 1 ). MRI ( Figure 3A ) showed a retroperitoneal tumor extending into the right suprarenal area, with locoregional lymph node involvement and a potential skeletal metastasis in the S1 vertebra. The differential diagnosis included rhabdoid tumor, neuroblastoma and soft-tissue sarcoma. On [¹²³I]mIBG imaging ( Figure 3B ) there was heterogeneous [¹²³I]mIBG uptake of the primary tumor, with large [¹²³I]mIBG negative parts of the primary tumor and locoregional lymph node metastases. In contrast to the MRI, multiple skeletal metastases with faint [¹²³I]mIBG uptake (SIOPEN score of 4) were detected. Because of the heterogeneous [¹²³I]mIBG uptake, an [¹⁸F]FDG-PET CT was performed ( Figure 3C ), confirming all lesions on [¹²³I]mIBG imaging plus identifying additional skeletal lesions, as suspected on the MRI. Analysis of the urinary catecholamines showed normal homovanillic acid (HVA) and vanillylmandelic acid (VMA) and slightly elevated levels of dopamine, 3-methoxytyramine (3MT) and norepinephrine. Histologically, the primary tumor biopsy did not express common neuroblastoma markers (negative for PHOX2B and TH, but positive for Chromogranin A). SNP-array showed a gain of chromosome 17q and next-generation sequencing detected a SMARCA4 mutation [c.4574T>C; p.(Leu1525Pro)]. Based on DNA-methylation (using the institutional classifier of Heidelberg), the tumor was classified as a rhabdoid tumor (score 0.73). Methylation analysis of cfDNA from the blood plasma was performed to aid in the diagnosis of this case. Deconvolution of cfRRBS data estimated both rhabdoid tumor (44.2% tumor fraction) and neuroblastoma (22.8% tumor fraction). In addition, RT-qPCR was performed on diagnostic BM and PB samples using the neuroblastoma MPX-panel, which came back positive for mRNA markers CHRNA3 and GAP43 (Figure 3D). At tumor resection following induction chemotherapy, tumor cells stained positive for chromogranin A and synaptophysin and focal ganglion cells, suggestive of neuroblastoma. The diagnosis was amended and the patient received postoperative chemotherapy according to the GPOH NBL2009 high-risk neuroblastoma regimen.

A 2-year-old boy presented with progressive episodes of coughing, dyspnea and a palpable mass on the chest ( Table 1 ). A CT-scan showed multiple masses of the rib, vertebrae and pleura, and a filled right hemithorax (pleural effusion) with a mediastinal shift and deviation of the trachea and left main bronchus ( Figure 4 ). The subsequent risk of respiratory failure during anesthesia precluded tumor biopsy at diagnosis. Methylation analysis using cfRRBS on the cfDNA at diagnosis estimated a fraction of Ewing sarcoma (28% tumor fraction). After the first chemotherapy course the tumor responded and a biopsy was feasible, and the diagnosis of an Ewing sarcoma was confirmed.

A 6-week-old girl presented with abdominal distention and a palpable abdominal mass ( Table 1 ). MRI showed a small tumor in the right adrenal gland with massive liver infiltration, resulting in a very large liver that extended into the pelvis ( Figure 5A ). The progressive hepatomegaly caused feeding problems, respiratory distress and edema of the lower abdomen and legs. Analysis of urinary catecholamines showed elevated VMA, HVA, dopamine, 3MT, norepinephrine and normetanephrine levels. As the patient’s clinical condition did not allow for an [¹²³I]mIBG SPECT scan or tumor biopsy to be performed, chemotherapy was immediately started. After chemotherapy, only a very small (8x10 mm) calcified primary tumor lesion remained ( Figure 5B ). RT-qPCR of BM and PB samples at diagnosis showed detection of all neuroblastoma mRNA-markers, including the neuroblastoma-specific marker PHOX2B ( Figure 5C ). To be able to discriminate between high- and non-high-risk disease, a ddPCR to detect MYCN status was performed on cfDNA isolated from plasma at diagnosis. No amplification was detected ( Figure 5C ), and a non-high risk neuroblastoma was diagnosed. The patient was treated with chemotherapy according to a medium-risk protocol. The patient responded very well to treatment and is still in complete remission, indicating that the clinical course is in line with the risk stratification.

A 5-year-old girl presented with a painless swelling in the left forearm ( Table 1 ). Imaging by [¹⁸F]FDG-PET showed a soft tissue tumor without metastases ( Figures 6A, B ). Tumor biopsy showed an embryonal rhabdomyosarcoma. MyoD1 staining (heterogeneously expressed in embryonal rhabdomyosarcoma) was considered ambiguous in the bone marrow aspirate ( Figure 6C ). A RT-qPCR using a rhabdomyosarcoma mRNA-panel (30) was then performed on the archived diagnostic BM sample. No tumor invasion was detected ( Figure 6D ), and the patient was diagnosed with non-metastatic rhabdomyosarcoma. This was later confirmed when the immunohistochemistry staining was reevaluated. The patient was treated according to a localized treatment protocol (RMS2005).

A 6-year-old boy, treated for a high-risk neuroblastoma, suffered from a bone marrow recurrence two years after diagnosis ( Table 1 ). During relapse treatment, [¹²³I]mIBG imaging showed progressive disease for which he was treated with [¹³¹I]mIBG therapy ( Figure 7A ). After [¹³¹I]mIBG therapy ( Figure 7B ), he suffered from pancytopenia despite autologous hematopoietic stem cell reinfusion and severe hemorrhagic cystitis for which he received continuous bladder irrigation through a urinary catheter. When further progression of the disease was suspected, an [¹⁸F]FDG-PET scan was performed, as [¹²³I]mIBG scans are less reliable post-[¹³¹I]mIBG therapy. This demonstrated diffuse bone marrow activation, reduced uptake in regions with known skeletal metastases and an enhanced liver uptake ( Figure 7C ). As a similar image can be caused by radionuclide therapy imaging and bone marrow activation, these results could not confirm our clinical suspicions. Urinary catecholamines could not be measured due to the continuous bladder irrigation. As an alternative, RT-qPCR was performed on a PB sample using our neuroblastoma mRNA marker-panel (28, 29), revealing exceedingly high expression of all markers. BM was not tested due to the patient’s clinical condition. In the same sample, a high percentage of RASSF1A-M was found in plasma using the RASSF1A-M ddPCR. On retrospective analysis, all sequentially collected liquid biopsies of this patient reflected the tumor burden and predicted his recurrence and progressive disease months before the actual clinical diagnosis. As shown in Figure 7D , RASSF1A-M stayed positive throughout and BM RT-qPCR became positive at immunotherapy, followed by the PB RT-qPCR. The liquid biopsy analyses, both the RASSF1A-M ddPCR and BM mRNA RT-qPCR, clearly indicated progressive disease. This resulted in initiating palliative therapy and the patient died a week after the analyses were performed.

In cases 1, 2 and 3, we describe the use of cfDNA for diagnostic purposes. Tumor histology is known to correlate with the tumor’s DNA methylation pattern (70, 71). Clinically, methylation analysis is performed using an EPIC array, which is costly and requires a relatively large amount of DNA. In cases where tumor biopsy cannot be performed, such as with patient 2 and 3, methylation analysis of cfDNA could be used as a substitute diagnostic tool. cfRRBS on cfDNA is cost-effective and feasible even with small pediatric sample volumes (39). In cases 1 and 2, retrospective analysis of the diagnostic cfDNA samples resulted in a clear diagnosis. Particularly for patient 1, an earlier diagnosis of a second primary tumor might have resulted in a more patient-tailored treatment, although the rarity of the case precludes any certainty. With case 2, the uncertainty of the primary diagnosis is reflected in the result of the methylation analysis, which clearly shows that the SMARCA4 loss detected in this patient correlates with a specific methylation pattern detected in cfDNA. However, unlike conventional tissue analysis, the cfRRBS methylation pattern also indicated neuroblastoma. Previous research has shown that loss of SMARCA4 in neuroblastoma does occur and correlates with poor prognosis (72). In this case, we demonstrated that the methylation pattern of this tumor had a distinct signature which indicated both a rhabdoid tumor and a neuroblastoma. Although the outcome of cfRRBS did not directly impact clinical decision making in this case, the effect of the SMARCA4 mutation on the methylation pattern indicates opportunities for further research on the diagnostic and prognostic value of this technique. Furthermore, case 2 illustrates how mRNA positivity of BM and PB for neuroblastoma-specific markers supported the diagnosis of metastatic neuroblastoma. Currently, several tumor-specific mRNA panels are being validated for clinical application in neuroblastoma and rhabdomyosarcoma tumors. For neuroblastoma, it has been shown repeatedly that tumor-derived mRNA can be sensitively detected in bone marrow and blood and that both correlate with clinical outcome (26, 44, 45, 73, 74). Newly developed MPX mRNA panels combining these markers decrease the required volume of blood or BM, the turnaround time and the (financial) resources required for the procedure (40).

The prognostic use of cfDNA is shown in case 4, a patient in whom no amplification of MYCN, nor methylated RASSF1A was detected, and who therefore was treated according to a non-high-risk protocol. cfDNA-based detection of MYCN amplification has been described using qPCR (75, 76) and more recently ddPCR (42, 43). Hypermethylation of the tumor suppressor gene RASSF1A was shown to correlate with prognosis in neuroblastoma, rhabdomyosarcoma and Wilms tumors (32, 77–80). Our group has previously shown RASSF1A-M to correlate with outcome in neuroblastoma when it is detected in both the primary tumor and metastasized tumor cells in the BM (78). Of note, tumors with MYCN amplification had significantly higher RASSF1A-M levels than tumors without amplified MYCN (78). Detection of RASSF1A-M by ddPCR in plasma of patients with high-risk neuroblastoma clearly showed a significantly poorer event-free survival when hypermethylated RASSF1A exceeded 27,681 copies/mL at diagnosis (32). In rhabdomyosarcoma, detectable levels of RASSF1A-M in diagnostic plasma samples clearly correlate with poor outcome, especially in patients with metastatic disease (80). RASSF1A-M detection could therefore be complementary to the rhabdomyosarcoma RT-qPCR-panel (80).

In case 5, the rhabdomyosarcoma-specific mRNA panel was negative. A cohort study performed by our group showed that this panel is more sensitive than conventional diagnostic techniques to detect metastatic rhabdomyosarcoma (30). Its negativity was therefore sufficient to confirm that no BM infiltration had occurred. Case 5 demonstrates how panels such as these could assist to establish metastatic status and subsequent risk profile.

In case 6, neuroblastoma-specific mRNA levels significantly increased prior to the first event and failed to decrease during second line therapy, which was in line with the clinical course of the disease. The extremely elevated mRNA-markers in PB confirmed progressive disease rather than toxic bone marrow depression. Furthermore, case 6 illustrates that RASSF1A-M detection enables sensitive monitoring of response to therapy (32, 41, 78). For this patient with poor outcome, RASSF1A-M analyzed by ddPPCR never became undetectable, but fluctuated between 10 (3)-10 (6) copies RASSF1A-M/mL plasma ( Figure 7D ). It should be noted that for disease monitoring, evidence indicates that combining cfDNA and mRNA analysis appears to increase sensitivity (41). A combination of ddPCR and RT-qPCR enables a more accurate quantitation of tumor markers (32).