Detection of Circulating and Disseminated Neuroblastoma Cells Using the ImageStream Flow Cytometer for Use as Predictive and Pharmacodynamic Biomarkers

Neuroblastoma (NB) is a heterogeneous tumor occurring in 10.2 per million children. It has one of the lowest survival rates of all childhood cancers, with only 67% of patients surviving to 5 years (1). Treatment of neuroblastoma depends on the International Neuroblastoma Risk Group (INRG) classification, which is determined by patient age, stage, and genetics including MYCN gene amplification, classifying patients as low, intermediate, or high-risk (2). For high-risk patients defined in Europe as metastatic disease over 1 year of age or MYCN amplified, intensive multimodality approaches are used including induction chemotherapy, surgical resection of the primary tumor, consolidation with myeloablative therapy and autologous stem cell rescue and local radiotherapy, followed by immunotherapy and differentiation therapy. Despite an improved initial response to treatment, survival remains poor (<50% at 5 years) due to eventual drug resistant relapse (3).

To better understand tumor evolution and drug resistance, it is important to consider intratumoral heterogeneity (4), but performing multiple biopsies from different tumor areas is not feasible in most patients. Studying circulating tumor cells (CTCs) from blood may inform intratumor heterogeneity and tumor evolution (5, 6), and such studies are gaining importance in the clinic as their detection is a noninvasive procedure involving only the collection of peripheral blood (7). However, as CTCs are very rare, present in around one in 10⁶ leucocytes in patients with prostate cancer (8), it is important to increase sensitivity in detection assays by including an enrichment step.

Technologies recently developed for CTC analysis include microfluidic devices with antibody-coated microspots (CTC chip), high-throughput microfluidic mixing devices [Herringbone- Chip; (9), or ultrasound-based isolation in microfluidic devices (10)]. A dielectrophoretic array method has been used to isolate disseminated tumor cells (DTCs) from bone marrow (BM) from patients with neuroblastoma for tumor genetic analysis (11), and magnetic bead-based enrichment has been developed to isolate DTCs from bone marrow samples using anti-GD₂ and NCAM antibodies (12). The Amnis Image Stream Imaging flow Cytometer (ISx) combines the features of classical flow cytometry, including an impartial analysis of a large number of cells in a short period of time with essential features of fluorescence microscopy allowing multiparameter cell analysis (13).

Detection of circulating free DNA and circulating tumor DNA (ctDNA) in plasma or serum of cancer patient blood is another type of liquid biopsy with higher levels detected in patients with metastatic disease (14). Recent studies have shown the feasibility of detecting ctDNA in cell-free DNA (cfDNA) extracted from plasma to determine genetic aberrations including in neuroblastoma (15–19).

In this study, we detected CTCs and DTCs from blood and BM samples in patients with neuroblastoma identified by GD₂ expression and the absence of CD45 expression using the ISx. We show that higher numbers of CTCs in newly diagnosed patients with high-risk neuroblastoma are associated with a failure to achieve a complete BM metastatic response after first-line induction therapy and that CTCs and DTCs can be used as pharmacodynamic biomarkers of novel targeted treatments, for example, MDM2 inhibitors in early-phase clinical trials. To our knowledge, this is the first study to report use of the ISx to detect neuroblastoma CTCs and DTCs in neuroblastoma clinical samples.

A panel of human neuroblastoma cell lines, N-type [SHSY5Y (20–22), NGP (23), and NBLW-N (24, 25)] and S-type [SHEP, SKNAS, and NBLW-S; (20, 21)] were used to develop a protocol for detection of neuroblastoma cells using the ISx. Cell lines were routinely maintained in RPMI medium supplemented with 10% v/v FBS (Gibco) and 1% v/v penicillin–streptomycin (Sigma) in 5% CO₂ in air in a humidified incubator at 37°C, regularly tested, and found to be free from Mycoplasma using MycoAlert (Lonza). Authentication of the NBLW cell line was performed as described previously (26). NGP and SHSY5Y cell lines were authenticated using short tandem repeat genotyping (New Gene Limited, International Centre for Life). SKNAS, SHEP, and SKNBE(2c) (Be2C) cells were authenticated using cytogenetic analysis (27). All cell lines were used within six passages or within a period of 6 months.

A GD₂ antibody conjugated with PerCP Cy5.5 (Peridinin chlorophyll protein, BD Pharmingen, 563438, 14.G2a) and anti-neural cell adhesion molecule (NCAM) conjugated with phycoerythrin (PE) CF594 antibodies (BD Pharmingen, 562328, B159) were used as neuroblastoma markers in different cell lines alongside DAPI (BD Pharmingen, 564907) nuclear staining. Imagestream data were analyzed using IDEAS image stream analysis software using methods described previously (28).

This study was undertaken in accordance with the ethical principles of the Declaration of Helsinki. Following institutional review board approval (ethics reference number 14/NW/0154), local institutional approval and written informed consent from the patient or carer, clinical samples (4–8 mL blood and 1–3 mL BM) were obtained from patients with newly diagnosed or relapsed neuroblastoma from five UK Paediatric Oncology Principal Treatment Centers (Supplementary Fig. S1B and S1C). Samples were collected in Cell Save (Veridex, Menarine Diagnostics) tubes, sent by post at room temperature within 72 hours of collection, and processed within 96 hours of collection. In the clinical study, 40 patients with neuroblastoma were recruited (32 high, six low, and two intermediate risk), 24 were studied solely at diagnosis, 14 solely at relapse, and two at both diagnosis and relapse (Supplementary Fig. S1A and S1D). In total, 42 blood samples were studied for CTCs (one fail) and 35 paired BM samples. In seven cases, only a blood sample was obtained (Supplementary Fig. S1C). International Neuroblastoma Risk Group (INRG) and other clinical characteristics of patients are shown in Table 1 and Supplementary Table T2. For ex vivo Nutlin-3 treatment, four blood and two BM samples were collected in EDTA tubes from four patients. Clinical information is correct up to May 31, 2018. All newly diagnosed high-risk patients studied were treated on the European High Risk Neuroblastoma trial [HR-NBL1; (29)].

Samples (blood and BM) were blocked to prevent nonspecific antigen binding, red cells lysed, and enriched for nonhematopoietic cells using previously reported methods (30, 31). Cells were then permeabilized in perm-wash buffer (BD Pharmingen) and stained with immunofluorescent antibodies including GD₂-PerCp, NCAM-PE, and CD45- PE-Cy 7 (BioLegend, 560915, H130) and nuclei stained with DAPI. Following incubation for 1 hour, stained cells were washed, resuspended in PBS, and processed on the ISx according to the manufacturer's protocol, and the presence of CTCs and DTCs detected using IDEAS software (See Supplementary Methods- and Supplementary Fig. S2).

CTCs and DTCs were detected on the basis of brightfield morphology, GD₂ expression, a nuclear signal, and the absence of CD45 expression. Potential CTCs were gated in GD₂ and CD45 scatter plots and visually confirmed from immunofluorescence images, tagged, and counted manually. The numbers of DTCs were counted using IDEAS software as there were too many to count manually. The diameter of CTCs and DTCs in neuroblastoma patient blood and BM samples was calculated as described previously (28). The mean diameter of CTCs/DTCs was compared with the mean diameter of white blood cells (WBCs; n = 5000). The number of CTCs used to calculate diameter ranged from four to 100 and for DTCs from four to 2,000. DNA ploidy was determined in all blood and BM samples with ≥4 cells based on WBC DNA and CTC/DTC DNA ratio from a DNA histogram using IDEAS software using the formula CTC DNA Ploidy = Mean CTC DNA/Mean WBC DNA. If the ratio of CTC DNA was >1.25× WBC DNA ploidy, then it was considered hyperdiploid, if not diploid (32).

Plasma was separated from blood (n = 3 cases) and BM (n = 3 cases) samples in Cell Save tubes by two sequential centrifugations (2,000 × g, 10 minutes) and stored at -80°C in 1-mL aliquots. cfDNA was isolated from aliquots of double spun plasma using the QIAamp DNA Blood Maxi Kit (Qiagen). Following isolation, the cfDNA yield was quantified using the Qubit Fluorometer (Thermo Fisher Scientific) as per the manufacturer's instructions. cfDNA fragment size was determined using 1% agarose gel electrophoresis.

Two neuroblastoma cell lines, p53 wt (SHSY5Y) and p53 mutant [Be2C; (27)], and blood and BM samples were exposed to Nutlin-3, an MDM2 inhibitor, and upregulation of p53 and p21 detected by immunofluorescence using the ISx and compared with DMSO (dimethyl sulfoxide) controls (Sigma-Aldrich, UK, D2650). Nutlin-3 (Selleckchem) was dissolved in DMSO to a stock concentration of 100 μmol/L. Cells were treated with 10 μmol/L Nutlin-3 or an equal volume of DMSO for 24 hours prior to fixation, followed by incubation with GD₂ PerCp (BD Pharmingen), p53 Alexa Fluor-647 (Cell Signaling Technology, 2533S, 1C12) at 1:50, p21 Alexa Fluor-488 (Cell Signaling Technology, 5487, 12D1, at 1:50) and DAPI (BD Pharmingen) and run on the ISx as described above. Cell-cycle analysis was performed using IDEAS software with round single in focused cells gated on scatter plots. Using the intensity of the DAPI histogram, mitotic cells and cells in G₁, S, and G₂–M were gated to observe the effect of Nutlin-3 treatment and DMSO in neuroblastoma cell lines, healthy volunteer blood samples, and neuroblastoma patient blood and BM samples. (see Supplementary Methods for more details).

Using SNP arrays (n = 11 cases), array comparative genomic hybridization (CGH; n = 21 cases) and multiplex ligation PCR-dependent amplification (MLPA; n = 6 cases), cytogenetic analysis was undertaken on primary tumors or BM metastases from all except 4 cases (Supplementary Table T2). For SNP arrays, DNA samples were hybridized to Infinium CytoSNP-850K v1.1 BeadChip (Illumina, Inc) according to the manufacturer's instructions. Illumina IDAT files were analyzed using BlueFuse Multi software. Using targeted next-generation sequencing (26) p53 gene mutational status was determined for the cases treated with Nutlin 3 (Supplementary Table T3). Affymetrix Oncoscan arrays (OncoScan FFPE CNV) performed by Eurofins Genomics (Ebersberg) were used for detecting copy number abnormalities from cfDNA from plasma with CEL files analyzed using Nexus (Biodiscovery) software.

Statistical tests were performed using GraphPad Prism (version 6.04) and IDEAS software (Amnis-Imagestream Imaging Flow cytometer). The number of cases with BM involvement and the presence or absence of CTCs and DTCs was assessed using a Fisher exact test. A Mann–Whitney U test was used to calculate the relationship between numbers of CTCs and DTCs with BM involvement and BM response to induction chemotherapy. To determine the correlation between numbers of CTCs/DTCs detected by qRT-PCR, for the neuroblastoma mRNAs tyrosine hydroxylase (TH) and PHOX2B (33), a Spearman rank correlation test was used (SPSS). All P values reported were two-tailed and considered significant if P ≤ 0.05. Data is presented as mean ± SEM.

A panel of neuroblastoma cell lines (n = 6) comprising neuronal (N) and substrate-adherent (S) types were used to determine expression of the neuroblastoma cell surface markers (GD₂ and NCAM) using the ISx. N-type cell lines (NGP, SHSY5Y, and NBLW-N) were found to express a higher percentage of GD₂ and NCAM⁺ cells than S-type cell lines (SHEP, NBLW-S, and SKNAS). NGP cells were the most strongly positive N-type cell line for GD₂ and NCAM (Supplementary Table T1).

CTCs and DTCs were detected from a scatter plot of cells that were GD₂⁺/CD45⁻ with an intact nucleus (Fig. 1A and E). Potential CTCs/DTCs were gated on the basis of intensity scatter plots of GD₂ versus CD45 for one patient sample and the template saved as a standard and used for all remaining clinical samples using IDEAS software. Dots in scatter plots were linked to corresponding cell imagery and visualized to help define gating boundaries (Fig. 1A–F). Once a gate was defined, cell imagery of that population could be inspected and CTCs/DTCs were visually confirmed on the basis of a round single cell from the brightfield image (BF), with positive GD₂ expression, negative CD45 expression, and an intact DAPI-stained nucleus (Fig. 1A–C). Confirmed CTCs/DTCs were then tagged and saved for further analysis of cell diameter and DNA ploidy.

CTCs were detected in 26 of 42 patient blood samples (mean = 40/mL, range, 1–264/mL at diagnosis; mean = 6/mL, range, 1–39/mL at relapse, Fig. 1G). DTCs were detected in 25 of 35 BM samples (mean = 30,342/mL, range, 57–291, 635/mL at diagnosis; mean = 2,124/mL, range, 112–15,688/mL at relapse, Fig. 1H). Table 1 shows a summary of cases studied in relation to clinical risk group and MYCN status with 32 of 40 (80%) cases high-risk and 26/40 (65%) studied at diagnosis (Supplementary Fig. S1A, S1D, and S1E).

BM involvement with neuroblastoma on either one or two aspirates and/or trephines as reported by a consultant hematologist based on morphology was present in 24 of 41 cases (including 1 case at diagnosis and relapse). In patients with reported BM involvement, CTCs were detected in 19 of 24 cases and DTCs in 20 of 21 cases (in three cases DTCs not studied; Supplementary Fig S1E). In patients without reported BM involvement, CTCs were detected in six of 17 and DTCs in five of 17 cases (Table 1). The presence of CTCs or DTCs was associated with BM involvement (P < 0.01 for CTCs and P < 0.0001 for DTCs, Fisher exact test). Similarly, there was an association between the number of CTCs and DTCs and BM involvement (P < 0.0001 Mann–Whitney test, Fig. 1I and J). CTCs were detected in two of 42 cases in the absence of DTCs and BM involvement (one low-risk and one high-risk), and in one case in the absence of DTCs and the presence of BM involvement (Supplementary Fig. 1E). In three of 35 cases, DTCs were detected in the absence of CTCs and presence of BM involvement, and in two cases in the absence of CTCs and absence of BM involvement. There was no association between the numbers of CTCs or DTCs and the presence of MYCN amplification in the primary tumor or other metastatic site biopsied. CTC and DTC numbers from 16 and 21 patients, respectively with untreated high-risk neuroblastoma treated on the SIOPEN HR-NBL-1 trial were compared with neuroblastoma -specific mRNA detected by qRT-PCR for TH and PHOX2B in blood and neuroblastoma. There was a weak correlation between CTC numbers and level of PHOX2B mRNA expression in BM only (r = 0.45, P< 0.05; data not shown).

NCAM was expressed in only three of 11 GD₂⁺/CD45⁻ blood samples initially examined, whereas in GD₂⁺/CD45⁻ cells from BM aspirates weak NCAM⁺ DTCs were observed in nine of 11 samples. From these early observations, it appears that there are differences in NCAM expression in neuroblastoma cell lines compared with NB CTCs and DTCs. It has been reported that polysialylated NCAM and nonpolysialylated NCAM expression differ between in vitro and in vivo conditions (34). Hence NCAM expression was not considered further to detect and confirm CTCs/DTCs in this study.

To determine whether the number of CTCs or DTCs at diagnosis were associated with BM response to chemotherapy in high-risk neuroblastoma patients with BM involvement at diagnosis (n = 21), numbers of CTCs/DTCs were plotted against BM response after first-line induction therapy (COJEC or N7; Fig. 1K and L). BM response after induction chemotherapy was determined by the presence of neuroblastoma cell infiltration in either BM aspirate or trephine biopsy according to the International Neuroblastoma Response Criteria Bone Marrow Working Group classification (35). Absence of morphologic evidence of neuroblastoma on two aspirates and two trephines was considered a complete response (CR), that is, BM in CR and a positive BM aspirate or trephine at one or more sites was considered an incomplete response, that is, BM not in CR (35, 36). Patients with higher numbers of CTCs at diagnosis were found to have an incomplete BM response after first-line induction therapy (Fig. 1K, P < 0.01, Mann–Whitney U test). In contrast, there was no association between numbers of DTCs at diagnosis and BM response to first-line induction chemotherapy (Fig. 1L, P = 0.38).

The DNA content of WBCs and CTCs/DTCs in the CD45-depleted blood cell population was determined from a DNA histogram plotted using IDEAS software. On the DNA histogram of CD45-depleted cells, visually confirmed CTCs/DTCs were overlaid to determine DNA content of CTCs/DTCs, from which ploidy could be extrapolated (Fig. 2A and C). The ploidy status determined in this way from 24 of 42 CTC samples and 25 of 35 DTC samples is shown in Supplementary Table T2. For CTCs, 19 of 24 samples were diploid and five of 24 hyperdiploid and for DTCs 17 of 25 diploid and eight of 25 hyperdiploid.

The DNA ploidy of CTCs in five cases determined using the ISx was compared with ploidy from a corresponding primary tumor or BM aspirate determined using high-density SNP arrays. Case-No-15 CTCs were found to be diploid using the ISx (Fig. 2A) and the corresponding primary tumor SNP array showed diploidy with multiple segmental chromosomal abnormalities (SCA) without MYCN amplification (Fig. 2B). Case-No-33 CTCs detected at first relapse were hyperdiploid using the ISx (Fig. 2C) and a SNP array performed 15 months later at further relapse on a BM aspirate showed near-triploidy (Fig. 2D). MYCN and ALK amplification were detected in the SNP array profile from this case as well as 1p and 6q loss as shown in Fig. 2D. MYCN and ALK amplification were the only SCAs detected in the SNP array from a lymph node metastasis from the same patient at diagnosis when diploidy was present. The concordance of ploidy results from five of five cases with primary tumor or BM aspirate SNP arrays with CTC ploidy results from the same patients suggests that measurement of ploidy status using the ISx is accurate and reliable (Supplementary Table T2). In addition, the ploidy from CTCs/DTCs of three cases determined using the ISx was compared with the ploidy from SNP arrays performed on cfDNA collected at the same time and found to be concordant (Supplementary Table T2). However, the low frequency of hyperploidy indicates that ploidy could only be used to distinguish a diploid WBC from a hyperploid CTC/DTC in a minority of cases.

Cell diameter was also investigated as an additional feature to differentiate CTCs/DTCs from WBCs. The diameter of all CTCs and DTCs in samples with ≥4 CTCs/DTCs was measured using IDEAS software in 21 and 25 samples, respectively (Supplementary Fig. S3A and S3B). A significant difference was observed between the diameter of WBCs in blood or BM samples versus CTCs or DTCs, respectively (Wilcoxon signed rank test, P < 0.0001, Supplementary Fig. S3G and S3H). However, after plotting the values for each patient, the intrapatient variability between CTC/DTC diameter- and WBC diameter–suggested cell size would not be a useful parameter to isolate neuroblastoma CTCs/DTCs from residual WBCs.

cfDNA collected in Cell Save tubes (n = 6 samples) was analyzed by Oncoscan FFPE arrays and copy number variations (CNV) successfully detected in five of six samples. In one case, blood and BM plasma were collected at the same time and the CNV found to be identical in both samples (Fig. 3A and B). In three cases, MYCN amplification was detected by SNP arrays in cfDNA (Fig. 3C and D). In two cases, the numbers of SCAs detected in the blood plasma at relapse increased from diagnosis illustrating temporal heterogeneity (Fig. 3C and D).

MDM2 inhibitors prevent binding of MDM2 to p53 so stabilising and activating p53 leading to increased transcription of target genes including p21, which mediates a G₁ cell-cycle arrest. p53-mutant Be2C cells and p53 wt SHSY5Y cells were treated with the MDM2 inhibitor Nutlin-3 for 24 hours or DMSO control, immunostained with GD₂, p53, p21, and DAPI and run on the ISx. p53 and p21 expression was determined by the percentage of p53- and p21-positive cells from intensity histograms of p53 and p21 (Fig. 4A; Supplementary Fig. S4A and S4B). Mutant Be2C cells expressed a higher percentage of p53-positive cells at baseline compared with SHSY5Y cells with no increase following Nutlin-3 treatment, whereas in SHSY5Y cells there was a statistically significant increase in p53 expression following Nutlin -3 treatment (Fig. 4A; Supplementary Fig. S4A and S4B; Fig. 5G). In p53-mutant Be2C cells, there was low baseline expression of p21 and no increase following Nutlin-3 treatment (Supplementary Fig. S4B; Fig. 5H), resulting in a virtually unchanged cell-cycle profile (Fig. 5F), whereas in p53 wt SHSY5Y cells, there was an almost 10-fold increase in p21 expression (Fig. 4A; Supplementary Fig. S4A, Fig. 5F–H) resulting in a strong G₁–S cell-cycle arrest and increase in G₁–S ratio from 6.4 to 33 (Supplementary Fig. S4C; Fig. 5F).

To validate the panel of antibodies with GD₂-PerCp, PE Cy7 CD45, p53-Alexa Fluor-647, p21 Alexa Fluor-488, and DAPI and to compensate data with all fluorochromes, healthy volunteer blood samples (n = 3) were spiked with SHSY5Y cells or Be2c cells, treated with Nutlin-3 for 24 hours or DMSO control and p53 and p21 expression measured. After 24 hours, there was increased expression of p53 and p21 in GD₂⁺/CD45- SHSY5Y neuroblastoma cells and residual GD₂⁻/CD45⁺ WBCs remaining after WBC depletion (Figs. 4B and 5G and H). The fold increase in p21 in CD45⁺ cells, although statistically significant, was not as great as for SHSY5Y cells (Fig. 5H), resulting in an increased G₁ population but unchanged G₁–S ratio after Nutlin-3 treatment (Fig. 5F). In contrast, there was no change in p53, p21, or cell-cycle profile after Nutlin-3 treatment of spiked p53-mutant Be2c cells compared with DMSO control cells (Fig. 5F–H).

We next treated neuroblastoma patient blood (n = 4) and BM samples (n = 2) with Nutlin-3 ex- vivo for 24 hours or DMSO control and measured p53 and p21 expression (Supplementary Table T3). In 4 of 6 samples, GD₂⁺ cells were present, but a BM and blood sample from a patient with low-risk neuroblastoma at diagnosis had no detectable GD₂⁺/CD45⁻ CTCs or DTCs. In the four Nutlin-3–treated samples (three blood and one BM) with CTCs/DTCs present, increased p53 and p21 expression was seen in GD₂⁺/CD45⁻ CTCs and DTCs compared with DMSO control (Figs. 4E and F and 5A, B, G, and H). In Case-No-12, a high number of CTCs (n = 264/mL) and DTCs (n = 6,383/mL) were detected (Supplementary Table S3), and a histogram was generated for cell-cycle changes in DTCs and p53 and p21 expression in CTCs and DTCs, before and after 10 μmol/L Nutlin-3 treatment (Figs. 4E and F and 5A, B, and E–H). It was not possible to generate histograms from CTCs from the two blood samples taken at relapse due to low numbers of CTCs (Supplementary Table S3). In CTCs and DTCs, baseline nuclear and cytoplasmic p53 was detectable with nuclear accumulation of p53 following Nutlin-3 treatment (Figs. 4E and F and 5A and G). Consistent with p53 activation, there was increased expression of p21 in CTCs and DTCs following Nutlin-3 treatment (Figs. 4E and F and 5B and H), with a 12-fold increase in p21 in DTCs (Fig. 5B). This led to an increase in the G₁ population of Case-No-12 DTCs (56% for DMSO treated and 70%-Nutlin-3 treated), but no reduction in S-phase in DTCs (Fig. 5E and F), so an unchanged G₁–S ratio (9.8-DMSO and 9.2-Nutlin-3).

To determine the effect of Nutlin-3 on WBCs, four patient samples and three spiked healthy volunteer blood samples were treated with Nutlin-3 or DMSO control and p53, p21, and cell-cycle arrest measured. Compared with neuroblastoma cell lines, CTCs and DTCs, there was very low baseline expression of p53 in WBCs, but in all cases there was nuclear p53 accumulation following Nutlin-3 treatment (Figs. 4C and D and 5C and G; Supplementary Fig. S4E), and increased p21 expression compared with DMSO controls (Figs. 4C and D and 5D and H; Supplementary Fig. S4F). This led to an increase in G₁ population after Nutlin-3 treatment but no change in G₁–S phase ratio, that is, 9.1 after Nutlin-3 treatment compared with DMSO control (8.7) in WBCs from Case-No-12 BM (Fig. 5F; Supplementary Fig. S4D).

The aim of this study was to detect neuroblastoma CTCs from blood and DTCs from BM using the high-resolution ISx to use as biomarkers in neuroblastoma. Recently, we reported CTC detection and characterization from patients with esophageal, hepatocellular, thyroid, and ovarian cancers using the ISx (28, 37). Various techniques have been used to detect CTCs such as DEPArray and CTC-iChip. DEPArray has been previously used for neuroblastoma cell lines and patient BM samples (11) but this, to our knowledge, is the first to detect neuroblastoma CTCs and DTCs using the ISx. Previously, we reported between 0–118 CTCs/mL in thyroid cancer and 0–20 CTCs/mL in hepatocellular carcinoma (28) using similar methods on the ISx. This compares with 0–264 NB CTCs/mL and a mean of 12 CTCs/mL in this study.

Seeger and colleagues reported >10⁴ neuroblastoma cells per 10⁵ nucleated cells in BM samples from 103 of 267 patients and >10⁴ NB cells per 10⁵ nucleated cells in two of 174 patient blood samples from patients with high-risk metastatic neuroblastoma at diagnosis using anti-GD₂ immunocytology. They concluded that quantifying neuroblastoma cells in BM and peripheral blood at diagnosis and during induction therapy provides an important poor prognostic marker for patients with stage IV neuroblastoma (38). In our study, higher number of CTCs were detected at diagnosis (1–264/mL) compared with relapse (1–39/mL) due to clearance of CTCs from the blood by chemotherapy and likely earlier detection of relapsed disease. It is not possible to compare the sensitivity of our method for detection of CTCs with other published studies as CD45⁺ cells were depleted prior to analysis. In five of 17 cases, where DTCs were detected without BM involvement, this may have been due to sampling variation, the assessment of BM on the basis of morphology alone or neuroblastoma cells passing through the BM rather than homing there. Although GD₂ is expressed in >90% of neuroblastoma BM metastases at diagnosis and relapse (39), it is possible that using this technique we are missing a small proportion of tumor cells that do not express GD₂. A future study should compare the sensitivity of the ISx to detect DTCs with BM examination using immunocytology and also compare with neuroblastoma-specific mRNA expression by qRT-PCR. In patients with untreated high-risk neuroblastoma, there was a weak correlation between numbers of CTCs with the level of the neuroblastoma-specific mRNA PHOX2B detected by qRT-PCR in BM, but only small numbers of patients were studied (n = 22).

Higher numbers of CTCs were detected in untreated patients with high-risk neuroblastoma who did not achieve a BM CR after first-line induction therapy versus those who did, suggesting that CTC enumeration may prove useful to guide the length of induction chemotherapy in patients with high-risk neuroblastoma in the future. However, this was not the case for DTCs, which could be due to much higher numbers of DTCs which were not visually confirmed. These observations now need to be extended to a larger, prospective study of high-risk neuroblastoma. Because of our sample size of 40 patients including 16 relapse cases, it was not possible to evaluate the prognostic significance of CTC/DTC numbers. DNA ploidy of tumor cells was not useful for distinguishing CTCs from WBCs due to the presence of diploidy in the majority of neuroblastoma CTCs studied, but ploidy of CTCs/DTCs did reflect the ploidy status of the primary tumor.

We also evaluated the use of cfDNA for detection of circulating nucleic acids from blood and BM plasma collected for CTC/DTC studies in Cell Save tubes. In neuroblastoma, the genomic profile of the tumor is necessary for treatment stratification. Various methodologies have been reported for detecting neuroblastoma genomic profiles with SNP arrays now frequently used in national reference laboratories (40). Chicard and colleagues reported ctDNA copy number analysis using Oncoscan arrays in 66 of 70 patients with copy number profiles obtained in 74% of patients (19).

In five of six blood or BM plasma samples collected at the same time as CTCs/DTCs, copy number abnormalities (CNAs) were detected in ctDNA using Oncoscan FFPE arrays. In one case, the paired blood and BM plasma showed the same CNA confirming the usefulness of BM plasma for detecting CNA in neuroblastoma as reported previously (12), and in another case, the CNA from plasma cfDNA were identical to those in a concurrently biopsied metastatic disease site. cfDNA extracted from blood and BM plasma isolated from samples collected in Cell Save tubes for CTCs/DTCs could be used to detect metastatic neuroblastoma tumor-specific genomic alterations and should be evaluated prospectively in future clinical trials.

We are developing MDM2 inhibitors as a novel therapy for neuroblastoma (41–43) and MDM2 inhibitors are currently being evaluated in adult and pediatric early-phase clinical trials; therefore, we sought to establish whether CTCs could be used as a circulating pharmacodynamic biomarker. The optimum pharmacodynamic biomarker for MDM2 inhibitor activity is activation of the p53 pathway in tumor cells detected by increased expression of a p53-induced gene. Macrophage inhibitory cytokine (MIC1) has been used as a surrogate marker in plasma samples (44), but elevation of MIC1 levels are not specific for tumor cells highlighting the importance of developing less invasive, but tumor-specific pharmacodynamic proof-of-mechanism biomarkers. In this study, detection of increased p53 and p21 expression in SHSY5Y cells following Nutlin-3 treatment, but not mutant p53 Be2C cells is consistent with our previous studies of MDM2 inhibitors in these cell lines (42, 45).

In healthy volunteer blood samples exposed to MDM2 inhibitors ex vivo, blood spiking studies with neuroblastoma cells and then patient blood and BM samples, we demonstrated increased p53 and p21 protein expression and an increase in G₁ population in CTCs, DTCs, and WBCs after Nutlin-3 consistent with our previous studies testing p53 function in diagnostic neuroblastoma biopsies showing induction of the p53 pathway following ex vivo irradiation (46). This study demonstrates proof-of-concept to use p21 as a sensitive and specific pharmacodynamic biomarker of MDM2 inhibitor activity for CTCs and DTCs detected by the ISx, but its usefulness for CTCs may be limited by small numbers of CTCs present in relapsed blood. p21 expression and the G₁ population also increased in WBCs following Nutlin-3 highlighting the importance of studying tumor cells rather than surrogate WBC to study proof-of-concept pharmacodynamic biomarkers of MDM2 inhibitor activity.

In conclusion, this is the first study to demonstrate the clinical utility of neuroblastoma CTCs detected by the ISx as noninvasive pharmacodynamic biomarkers of novel therapies in early-phase clinical trials. A future, larger study is needed to investigate whether they are predictive biomarkers of response and also to determine whether they are potential prognostic biomarkers to further refine risk stratification in high-risk neuroblastoma.

No potential conflicts of interest were disclosed.

Conception and design: S. Merugu, E. Gavens, D. Jamieson, D.A. Tweddle

Development of methodology: S. Merugu, L. Chen, E. Gavens, M.L. Robinson, D. Jamieson, D.A. Tweddle

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Merugu, L. Chen, E. Gavens, M. Brougham, G. Makin, A. Ng, D. Murphy, M.L. Robinson, J.H. Wright, S.A. Burchill, A. Humphreys, N. Bown, D.A. Tweddle

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Merugu, L. Chen, E. Gavens, A. Ng, A.S. Gabriel, M.L. Robinson, J.H. Wright, S.A. Burchill, A. Humphreys, D. Jamieson, D.A. Tweddle

Writing, review, and/or revision of the manuscript: S. Merugu, L. Chen, E. Gavens, H. Gabra, M. Brougham, G. Makin, A. Ng, D. Murphy, A.S. Gabriel, S.A. Burchill, D. Jamieson, D.A. Tweddle

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Merugu, L. Chen, A. Ng, N. Bown, D.A. Tweddle

Study supervision: L. Chen, A. Ng, D. Jamieson, D.A. Tweddle

We are very grateful to all patients and families for donating samples, to research nurses at all study sites [Great North Children's Hospital Newcastle (GNCH), Royal Manchester Children's Hospital, Royal Hospitals for Sick Children in Edinburgh, Glasgow and Bristol] for collecting samples and to Geoff Bell (GNCH) for coordinating this national, multicenter study. We are thankful to the following for providing cell lines: Sue Cohn (NBLW), Barbara Spengler (Be2C), Rogier Versteeg (NGP), Jean Bernard (SKNAS), and Penny Lovat (SHSY5Y, SHEP). We are also grateful to Lina Hamadeh for statistical support and the Newcastle University Flow Cytometry Core Facility. This work was funded by a Newcastle University Overseas Research Studentship Award (to S. Merugu), the Children's Cancer & Leukaemia Group (CCLG) & Little Princess Trust (CCLGA 2016 08; to S. Merugu, D. Tweddle, L. Chen, D. Jamieson), Children with Cancer UK (to S. Merugu), the Newcastle upon Tyne Hospitals NHS Charity (E. Gavens, D. Tweddle, L. Chen), the DUBOIS Childhood Cancer Fund, SPARKS (11NCL05; to L. Chen, D. Tweddle), Neuroblastoma UK & Niamh's Next Step (to L. Chen, D. Tweddle), the North of England Children's Cancer Research Fund (D. Tweddle), Santander Universities & Association of Physicians intercalated scholarship (to J. Wright), Mrs Anna Upjohn (to M. Robinson), Cancer Research UK (CRUK-C9380/A25138; to D. Jamieson, E. Gavens), Action Medical Research/Great Ormond Street Hospital Children's Charity (GN2390; to A. Gabriel, D. Tweddle), Charity Soul, and the Sir Bobby Robson Foundation (for purchase of the Image Stream flow cytometer). The authors acknowledge the support of the UK National Institute for Health Research (NIHR) Clinical Research Network: Cancer (UKCRNID16957) and the Experimental Cancer Medicines Centre (ECMC) Paediatric network (to G. Makin, D. Murphy, A. Ng, D. Tweddle).

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.