Measuring circulating neuroblastoma cells by quantitative reverse transcriptase–polymerase chain reaction analysis

Correlation with its paired bone marrow and standard disease markers




Histologic examination of bone marrow (BM) is an accepted clinical standard for the detection of metastatic neuroblastoma (NB). Circulating tumor cells in peripheral blood (PB) derive from depots other than BM, and its measurement may provide additional information in the management of patients with NB.


One hundred twenty patients with Stage 4 NB were evaluated for tumor cell content in PB by quantitative reverse transcriptase–polymerase chain reaction (qRT-PCR) analysis of GD2 synthase mRNA with a sensitivity of 1 NB cell in 106 normal cells. These findings were correlated with qRT-PCR analysis of their simultaneously sampled BM aspirates and 5 standard modalities of disease detection (histology, computed tomography/magnetic resonance imaging, bone scan, metaiodobenzylguanidine scan, and urinary homovanillic acid/vanillylmandelic acid levels).


Detection of GD2 synthase transcript was found in 62 patients: Eleven patients had positive (+) samples in their BM and PB (BM+PB+), 38 patients had BM+PB-negative (BM+PB−) specimens, and 13 patients had BM−PB+ samples. BM+PB+ paired samples had the highest transcript levels. When the extent of disease was scored (from 0 to 5) according to the number of positive disease detection modalities, the magnitude of the transcript level correlated with disease score. Ninety-one percent of patients with BM+PB+ samples had evidence of disease in ≥ 3 modalities, whereas 97% of patients with BM−PB− samples and 100% of patients with BM−PB+ samples had low disease scores ≤ 2. Marker positivity in BM correlated with disease score. Patients who had positive marker in BM or PB had higher rates of relapse and death compared with patients who had negative marker. Kaplan–Meier analysis demonstrated a significantly greater risk of death for patients who had BM+PB+ specimens compared with patients who had BM−PB− samples (P = 0.03).


BM monitoring should continue to be an integral part of disease follow-up for patients with Stage 4 NB. PB monitoring to complement tumor surveillance in the BM can be informative. Cancer 2004. © 2004 American Cancer Society.

Patients with high-risk neuroblastoma (NB) often succumb to relapse disease despite initial clinical remission. Bone marrow (BM) is a frequent site of tumor recurrence.1 According to criteria established by the International Neuroblastoma Staging System (INSS), an important component in the clinical management of NB is the monitoring of BM for the presence of NB cells.2 However, BM collection is an invasive procedure and is likely to be a traumatic experience for most children. In contrast, repeated sampling by venipuncture of peripheral blood (PB) is better tolerated and is likely to be more conducive to patient compliance. Circulating NB cells have been detected by immunocytology3, 4 and by reverse transcriptase–polymerase chain reaction (RT-PCR) analysis of tyrosine hydroxylase mRNA.4–6 Its detection correlated with BM involvement in 23 patients with NB at diagnosis and at relapse.5 More important, it has been demonstrated in large cooperative studies that the presence of NB cells in PB at diagnosis has an adverse effect on relapse-free survival among patients with Stage 4 NB who where age > 1 year at diagnosis.7, 8 In one report, BM and PB were collected concurrently, and NB cell detection was compared.9 Faulkner et al. analyzed 51 pairs of BM and PB samples from 35 patients with NB of different stages by immunocytochemistry and found that NB cell frequency was > 2 logs lower in blood. By detecting the expression of the molecular marker GAGE, a cancer testis antigen, paired samples from 18 patients with metastatic NB correlated strongly with active disease.10

In this study, we examined paired samples of simultaneously drawn BM and PB from a cohort of 120 patients with Stage 4 NB. These paired samples were assayed for tumor cell content by quantitative RT-PCR (qRT-PCR) of GD2 synthase mRNA. This surrogate molecular marker, which is the key enzyme required for GD2 synthesis, has proven to have clinical relevance in evaluating the efficacy of several adjuvant therapies for the treatment of metastatic NB in the minimal residual disease setting.11, 12 We correlated marker findings with five standard disease-detection modalities established by INSS, namely, histologic examination of BM biopsy and aspirates, computed tomography (CT) and/or magnetic resonance imaging (MRI), technetium 99m (99mTc) bone scan, 131I-metaiodobenzylguanidine (MIBG) or 123I-MIBG scan, and urinary catecholamine metabolites homovanillic acid/vanillylmandelic acid (HVA/VMA). By using the same criteria to compare GD2 synthase mRNA detection in these two tissue sources, we could determine whether PB monitoring may provide additional information in the management of high-risk patients either by complementing BM or may even serve as an acceptable and convenient alternative to BM in tumor surveillance.



One hundred twenty consecutive patients with Stage 4 NB who were age > 1 year at diagnosis were included in this study. They were in different phases of treatment at Memorial Sloan-Kettering Cancer Center (MSKCC). In accordance to International Neuroblastoma Response Criteria,2 an extent-of-disease work-up included 5 evaluation modalities: histologic examinations of BM aspirates and biopsy specimens, CT/MRI, 99mTc-bone scan, 131I-MIBG or 123I-MIBG scan, and urinary catecholamine metabolites. In the current study, 2 types of specimens were drawn: pooled BM aspirate samples (∼10.0 mL) obtained from 4 iliac crest sites and whole-blood samples (2.5 mL) collected in PAXgene blood RNA tubes (Qiagen, Valencia, CA). All samples were collected after the needle and/or plastic catheter was cleared of the initial draw to reduce the chance of contamination by skin, endothelial cells, or extraneous substances. PAXgene blood RNA tubes contain proprietary reagents to stabilize RNA from degradation by endogenous nucleases and prevent ex vivo gene induction during transport or storage at room temperature.13–15 The white blood cell (WBC) count was used to calculate the total nucleated cell number for each 2.5 mL sample of whole blood. Written informed consent was obtained from the patients and/or their guardians in accordance with the guidelines of the MSKCC Institutional Review Board.

RNA Extraction

Ten million mononuclear cells were isolated from BM aspirates by Ficoll centrifugation, and RNA was extracted as described previously.10 RNA from PB was extracted using the PAXgene blood RNA kit (Qiagen) according to the manufacturer's instructions. Based on the WBC count, 97% of patients had a minimum of 5 million nucleated blood cells in the 2.5-mL blood sample, and 69% had a minimum of 10 million nucleated blood cells. cDNA synthesis was performed as described previously.10, 16

Real-time qRT-PCR

Relative quantitation of GD2 synthase mRNA was achieved in a multiplex PCR using the ABI Prism 7700 Sequence Detection System (Applied Biosystems, Foster City, CA). Details of the procedure were reported previously.16 For each unknown test sample, the amount of GD2 synthase and its endogenous reference, glyceraldehyde-3-phosphate dehydrogenase (GAPDH), was determined from the respective standard curve. Dividing the GD2 synthase level by the GAPDH level resulted in a normalized GD2 synthase value. The sensitivity of the assay was 1 NB cell in 106 normal mononuclear cells. Based on the quantitation of a series of normal BM and PB mononuclear cells, a normalized GD2 synthase value ≤ 5 was defined as negative.

Statistical Analysis

Progression-free survival (PFS) and overall survival (OS) were measured in months from the sampling date. The probability of PFS and OS was estimated by the Kaplan–Meier method, and survival comparisons were made between groups with the log-rank test.


Detection of GD2 Synthase mRNA in Paired BM and PB Samples from 120 Consecutive Patients with Stage 4 NB

Of 120 paired BM/PB samples from 120 patients, 62 pairs had detectable GD2 synthase transcripts (Table 1). This marker was present in both BM and PB samples (BM+PB+) from 11 patients, whereas the remaining 51 patients had positive qRT-PCR results in BM samples only (BM+PB−; n = 38 patients) or in PB samples only (BM−PB+; n = 13 patients). Fifty-eight patients had negative qRT-PCR results in both BM and PB samples (BM−PB−). The percent agreement between BM and PB samples was only 58%. Paired samples that were BM+PB+ had the highest transcript levels.

Table 1. GD2 Synthase Transcript Levels in Paired Bone Marrow and Peripheral Blood Samples from 120 Patients with Stage 4 Neuroblastoma
Paired sampleNo. of patients% positiveMedian transcript level (range)a in units
  • BM: bone marrow; +: positive; PB: peripheral blood; −: negative.

  • a

    A GD2 synthase transcript level ≤5 units was defined as negative.

BM + PB+11929.1 (11.7–1076.4)22.5 (6.5–94.1)
BM + PB−383212.3 (5.2–186.1) 0.0 (0.0–3.3)
BM − PB+1311 1.1 (0.0–4.8) 6.8 (5.3–109.9)
BM − PB−5848 1.1 (0.0–4.9) 0.0 (0.0–4.8)

Correlation between Marker Status and Transcript Levels in BM versus PB and Clinical Outcome

Without pairing these samples, 49 of 120 patients (41%) had BM samples that were positive for GD2 synthase mRNA, and 24 of 120 patients (20%) had PB samples that were positive for GD2 synthase mRNA (Table 2). The median transcript level in positive BM samples was nearly twice as high as the median level in PB samples, although the variance of the transcript level was large. A transcript level > 10 was found in 69% of patients with BM+ samples and in 42% of patients with PB+ samples. Patients who had BM+ samples had higher relapse and death rates compared with patients who had BM− samples (relapse rate, 47% vs. 39%; mortality rate, 22% vs. 14%). Comparable rates were found in patients based on PB marker status (relapse rate, 46% vs. 42%; mortality rate, 25% vs. 16%).

Table 2. Relation between Marker Status and Transcript Levels in Bone Marrow versus Peripheral Blood, with Data on Clinical Outcome
Median transcript levela (units)No. of patients (%)No. of relapse (%)No. who DOD (%)
  • DOD: died of disease.

  • a

    A GD2 synthase transcript level ≤5 unit was defined as negative.

Bone marrow   
  ≤ 571 (59)28 (39)10 (14)
  14.6 (5.2–1076.4)49 (41)23 (47)11 (22)
  > 5–1015 (31)93
  > 10–10029 (59)106
  > 100–1000+ 5 (10)42
Peripheral blood   
  ≤ 596 (80)40 (42)15 (16)
  7.7 (5.3–109.9)24 (20)11 (46)6 (25)
  > 5–1014 (58)73
  > 10–100+10 (42)43

Correlation between the Magnitude of Positive Marker Status and Five Standard Modalities of Disease Detection

GD2 synthase positivity was correlated with the findings of five standard disease detection modalities at the time of sampling. These modalities included BM histology, CT/MRI, bone scan, MIBG scan, and urine catecholamine metabolites. Each positive modality was given a score of 1. Higher disease scores reflected more evidence of disease. Figure 1 shows that, among patients with marker-positive BM, 80% of patients who had transcript levels < 10 had low disease scores of 1–2, whereas 60% of patients who had transcript levels ranging from > 100 to 1000+ had high scores of 4–5. The difference between these 3 groups was statistically significant (P < 0.05).

Figure 1.

Disease score distribution among patients with positive marker in their bone marrow (BM) as stratified according to GD2 synthase transcript levels. Light gray bars: score 1–2; medium gray bars: score 3; dark bars: score 4–5.

Correlation between Marker Positivity and Histology Positivity with Disease Score

Among the patients who had paired samples that were positive for GD2 synthase mRNA (BM+PB+), 91% had evidence of disease in ≥ 3 modalities, whereas 97% of patients who had paired samples that were negative for GD2 synthase mRNA (BM−PB−) had scores ≤ 2 (Fig. 2). Patients with BM−PB+ status also exhibited less evidence of disease, all had scores ≤ 2; 9 of the 13 patients with BM−PB+ status (69%) had only 1 positive disease modality. In contrast, 12 of 38 patients (32%) with BM+PB− status had positive disease scores ≥ 3. Patients with marker + BM, regardless of PB status, correlated with more evidence of disease (Fig. 3A). A similar pattern was found among patients who had positive histology, although with lower detection sensitivity compared with qRT-PCR. The magnitude of transcript level correlated with histology-positive BM (Fig. 3B). Among 15 patients with BM transcript levels from > 5 to 10, 5 patients had positive histology, whereas all 5 patients who had transcript levels > 100 had positive histology (P < 0.01).

Figure 2.

Disease score distribution stratified according to paired sample status. BM: bone marrow; +: positive; PB: peripheral blood; −: negative.

Figure 3.

(A) Percentage of patients with positive findings on histologic examination versus quantitative reverse transcriptase–polymerase chain reaction (qRT-PCR) analysis according to disease score. (B) Percentage of patients with positive histologic findings according to transcript levels. BM: bone marrow; +: positive.

Prognostic Importance of Marker Positivity in BM and PB

Kaplan–Meier survival analyses were performed to evaluate patient outcomes based on BM and PB marker status. With a relatively short median follow-up of 15 months, no prognostic significance was achieved by either BM (PFS, P = 0.55; OS, P = 0.25) or PB (PFS, P = 0.34; OS, P = 0.16). Nevertheless, when paired samples were analyzed, there was a significantly greater risk of death for patients who had BM+PB+ marker status compared with patients who had BM−PB− marker status (P = 0.03) (Fig. 4).

Figure 4.

Kaplan–Meier analysis: prognostic significance (with respect to overall survival) of marker positivity in paired bone marrow (BM) and peripheral blood (PB) samples. +: positive BM+PB+ versus −; negative BM−PB−: P = 0.03; BM+PB+ versus BM+PB−: P = 0.17; BM+PB+ versus BM−PB+: P = 0.25.


The main objective of this study was to evaluate whether monitoring PB may provide additional information in the management of high-risk patients either by complementing BM or even serve as an acceptable and convenient alternative to BM in tumor surveillance. A large sample size (120 patients) of paired BM and PB (n = 240 samples) was studied. Only unique patients were included in this analysis to avoid potential statistical bias from repeated samplings from the same patient. In addition, only patients with Stage 4 NB who were age > 1 year at diagnosis were eligible for this study. Their samples were obtained during routine extent-of-disease evaluations at different phases of their cancer treatment at MSKCC. Comparisons between PB and the simultaneous BM were made in their detection of NB cells as measured by qRT-PCR of GD2 synthase mRNA. Marker positivity and the magnitude of transcript levels were correlated with concurrent standard disease-evaluation modalities.

Several observations were made. Twice as many BM samples were positive in the qRT-PCR analysis compared with PB samples, with a median transcript level in positive BM samples that was twice as high as the level in PB. In paired samples that were both positive in the qRT-PCR analysis, the transcript level in the BM was higher (2–64-fold) than the level in its corresponding PB in 7 of 11 patients. This was similar to the findings of Faulkner et al.9 Assaying their BM-PB pairs by immunocytochemistry, those authors reported that circulating (PB) tumor cell frequency was 100-fold less than the frequency of the paired BM sample. We demonstrated previously that, by using this qRT-PCR assay, the transcript level of GD2 synthase was highly correlated with the number of NB cells as measured by immunocytology.16 The lower positivity rate in PB could not be due to inadequate PB sampling. This was because the median number of nucleated cells per sample among the patients with PB+ specimens was 12 million, and it was 13.3 million among the patients with PB− specimens. In addition, GD2 synthase transcript in each sample was corrected for GAPDH, adjusting for RNA recovery and integrity. In our analysis of 240 samples, the level of GAPDH from PB, which was extracted using PAXgene RNA kit, was comparable to BM GAPDH. The use of PAXgene blood RNA tubes represented an attempt to circumvent a crucial factor that contributes to variability in target gene detection among different laboratories, namely, uncontrollable RNA degradation after phlebotomy prior to RNA extraction. The addition of other stabilizing reagents, such as phenol-isothiocyanate, to blood immediately after collection should be just as effective but, technically, would be more difficult.15 This blood collection procedure was followed by whole blood extraction, which allowed us to avoid some of the difficulties associated with methods that require red cell lysis and mononuclear cell extraction.17

The importance of BM assessment of microscopic disease was underscored by the fact that positive markers in BM, regardless of the PB RT-PCR status, correlated with potentially more evidence of disease, as determined by the disease score. The magnitude of transcript level also correlated with the disease score, suggesting that GD2 synthase mRNA is a qualitative measure of tumor load. Furthermore, one-third of all pairs exhibited marker positivity in BM and not in PB. One plausible explanation for the lack of concordance in the BM/PB pairs, with only 58% agreement, was that only a subset of tumor cells (e.g., stem lines) from the BM circulates in the blood. Thus, we interpret our findings not as a failure to detect tumor cells in PB compared with BM but as a reflection of their non-BM origin. Because of the heterogeneous treatment and sampling time of our cohort, the impact of marker status on patient outcome was not expected. Nevertheless, patients who had a positive marker in either BM or PB had higher rates of relapse and death compared with patients who had a negative marker. More important, the statistically significant difference in patient survival between concordant groups (BM+PB+ vs. BM−PB−) suggests that the addition of PB sampling in tumor surveillance may be informative.

It is also of great interest that, although only 13 patients had BM−PB+ samples with low disease scores (score 1 and 2) at the time of PB sampling, 6 patients subsequently relapsed, and 2 patients died of NB. Their relapse rate (46%) was comparable to the rate of patients with BM+PB− specimens who had the same disease scores (50%). It has been reported previously that other PB markers are detectable prior to patient relapse. For example, it has been demonstrated that tyrosine hydroxylase mRNA is a useful marker for circulating NB when measured by RT-PCR4–6 and qRT-PCR.18, 19 Its presence indicated an adverse clinical outcome at the time of diagnosis and at the time patients went off therapy.8 The expression of molecular marker, GAGE, was 80% concordant among 85 patients with melanoma who had their BM and PB pairs sampled immediately before undergoing surgical resection to render them disease free.20 Moreover, the detection of GAGE in PB was associated with poor survival in multivariate analyses.

Although the accessibility and tolerability of venipuncture have made PB sampling an attractive alternative to the invasive BM for high-risk patients with NB, formal comparison between BM and PB is crucial before the latter can be advocated as a substitute. The results of this study suggest that BM assessment should continue to be an integral part of disease follow-up for patients with Stage 4 NB. Our findings, however, also suggest that concurrent monitoring of PB can provide additional and potentially prognostic information on patient outcome. Since circulating tumor cells in PB can derive from depots other than BM, there may be a fundamental difference in the biology of tumor cells in these two compartments. The prognostic importance of circulating (PB) versus sequestered (BM) NB will require longer follow-up and further investigation. In addition, with the practice of PB stem cell transplantation, the detection and elimination of contaminating NB in harvested PB stem cells prior to autologous infusion will continue to be of great importance.


The authors thank Karen Danis for her assistance with data management.