Neuroblastoma is a tumor of pluripotent neural crest cells of the sympathetic nervous system (1–3). Neural crest or neuroblastic tumors reflect various stages of differentiation and are classified as either neuroblastoma, ganglioneuroblastoma, or ganglioneuroma (3, 4). Neuroblastoma occasionally regresses or matures to ganglioneuroblastomas or ganglioneuromas, either spontaneously or posttreatment. It is the most common extracranial solid tumor of childhood, accounting for 7–10% of all pediatric malignancies (one-half of the neonatal malignancies and one-third of infant neoplasms) and causing a disproportionately high number of pediatric cancer deaths (15%). Approximately one-half of the cases are diagnosed by age 2 and more than two-thirds of the cases are diagnosed by age 5.
Morphologically, neuroblastoma tumors are small, round, blue cell pediatric tumors. Differential diagnostic entities for consideration include non-Hodgkin's lymphoma, acute leukemia, rhabdomyosarcoma, Ewing's sarcoma, retinoblastoma, medulloblastoma, some small-cell carcinomas, Wilm's tumor, and peripheral neuroepithelioma (PNET). The diagnostic evaluation requires clinical information such as anemia and “blueberry muffin” skin; morphologic findings such as Homer Wright rosettes and neuropil in the marrow; positive immunohistochemical (IHC) staining with neuron-specific enolase (NSE), chromogranin A, synaptophysin, and more recently, CD56; karyotypic abnormalities, including deletion of the short arm of chromosome 1 and repeating sequences of extrachromosomal double minutes (DM); homogeneous staining regions (HSR); and fluorescence in situ hybridization (FISH) analysis showing N-myc amplification. Other helpful clinical tests include increased urinary catecholamines and metabolites such as homovanillic acid (HVA), vanillylmandelic acid (VMA), dopamine, norepinephrine, epinephrine, and metanephrine. Electron microscopy shows distinctive dendritic processes and neurosecretory granules. Neuroblastoma can originate anywhere in the sympathetic nervous system. The most common primary site is located intra-abdominally, with 40% of tumors located in the adrenal and 25% in a paraspinal ganglion. The majority of children older than 1 year of age present with disseminated disease, most commonly involving lymph nodes, liver, the central nervous system (CNS), and especially the bone marrow. Neuroblastoma is second only to acute lymphoblastic leukemia in frequency of malignant infiltration of the marrow in children.
The International Staging System for Neuroblastoma (ISSN) incorporates elements of both the Evan's and Pediatric Oncology Group staging systems (5). The most important prognostic factors are disease stage and patient age. Infants younger than 1 year of age generally do well with some tumors reportedly regressing spontaneously. Among infants, DNA ploidy is an important prognostic factor (6). Treatment for neuroblastoma includes surgical resection, chemotherapy, radiation therapy, high-dose chemotherapy with purged or unpurged stem cell rescue, retinoic acid, and biologic response modifiers. In contrast to other pediatric malignancies, there has been only gradual improvement in the survival of neuroblastoma patients.
Ten years ago, while evaluating a bone marrow sample for possible leukemia, we noted that the marrow sample contained an extremely bright (the fourth decade on a four-decade scale) CD56 NCAM antigen cell population that was also CD45−. The tumor was diagnosed subsequently as a neuroblastoma. We evaluated other cases of neuroblastoma using a CD45−CD56+ dual-color phenotypic analysis and demonstrated similar results. This information, to our knowledge, had not been previously described in the flow cytometric literature (7). Using the approach of selecting at least two positive markers with one negative marker to analyze “rare events,” we added to the CD45−CD56very bright+ two-color assay another marker associated with neuroblastoma, i.e., NSE. NSE is not a specific enzyme, however, for neuroblastoma. We refined the assay by using a more specific gangliosided2 (GD2) antigen and, most recently, included a fourth color using a CD81 cell surface marker.
MATERIALS AND METHODS
Tissue samples received in transport media (RPMI 1640 with 16% fetal calf serum [FCS], 99 U/ml penicillin, 99 μg/ml streptomycin, and 41 U/ml heparin) were manually disaggreated by scraping with glass slides and washing two times in buffer (phosphate buffered saline [PBS], pH 7.4, with 2% FCS and 0.1% sodium azide). The cells were enumerated with a Sysmex K-1000 (TOA Medical Electronics, Kobe, Japan). The cell number and viability were determined manually by a trypan blue dye exclusion test in a hemocytometer chamber. Cytospins were prepared for morphologic review by the pathologists. Bone marrow aspiration samples received in transport media were centrifuged, with all except 2 ml of the supernatant discarded, and lysed in 50 ml of ammonium chloride lysing solution (pH 7.4, 0.15 M ammonium chloride, 0.009 M sodium bicarbonate, and 0.001 M EDTA). The marrow samples were washed twice and enumerated. Cell viability was determined and cytospins were prepared as described above for tissue samples. The initial CD45− CD56very bright+ dual-color and CD45− CD56very bright+ NSEcytoplasmic+ three-color methods evolved into the three-color GD2 cell membrane method. A total of 1 × 106 isolated tissue or marrow aspirate cells were labeled with 15 μl of GD2FITC (G2005-15, clone 8.S.077, IgM kappa [US Biological, Swampscott, MA] or 870633, clone GMR7, IgM kappa, fluorescein isothiocyanate [FITC; Seikagaku America, Falmouth, MA]), 5 μl of 1:5 diluted CD56PE (R7127, clone MOC-1, phycoerythrin [PE], Dako, Carpinteria, CA), and 20 μl of CD45PerCP (340665, clone 2D1, peridinin chlorophyll protein [PerCP], Becton Dickinson, San Jose, CA). Depending on the lot, the optimum titer ranged from 2 to 15 μl for GD2 and from 5 μl of a 1:5 dilution to 3 μl of “neat” CD56 based on testing against the SK-N-DZ cell line. Some of the results successfully substituted optimally titered CD56 from Becton Dickinson or Beckman Coulter (Hialeah, FL). A similar cocktail of IgMFITC IgG1PE IgG1PerCP class/subclass controls was used to label another tube of patient cells. CD45FITC, CD45PE, and CD45PerCP compensator controls were used separately to label daily control peripheral blood cells from a normal donor to help fine-tune the flow cytometer compensation. Cells were labeled for 15 min at 4°C, washed in buffer, lysed a second time for 3–5 min at room temperature in 1 ml of lysing solution (349202 FACS lysing solution, Becton Dickinson), and fixed in 1% formaldehyde solution. Samples were analyzed on a FACScan, FACSort, or FACSCalibur flow cytometer (Becton Dickinson) using either Lysis II or CellQuest software. The analysis strategy was to maximize the count of suspected neuroblastoma cells, minimize data storage, minimize computer time for analysis of the acquired data, but still report the percent suspected neuroblastoma cells in terms of all tissue or marrow cells to help monitor the patient. Approximately 10,000 events were collected and stored on a forward scatter (FSC) versus side scatter (SSC) plot, an R1 gate was established around a lymphocytic/blast/neuroblastoma cell area, and 50,000 or more events were collected and stored only within the R1 gate. If positive events were seen, the cell acquisition count was increased and/or more cells were labeled to achieve the customary goal for rare event analysis of 100 positive events for statistical relevance. The percentage of cells of interest within the R1 gate multiplied by the percentage of the R1 gate (the 10,000 events described above) estimated the percentage of cells of interest in the entire sample. The International Society of Hematology and Graft Engineering (ISHAGE) method was used for stem cell assays. It was helpful at the end of the assay to “back-gate” any CD45− CD56very bright+ GD2+ events on a FSC/SCC light scatter plot to confirm a cluster of cells centrally/evenly located in the lymphocytic/neuroblastoma/blast area that did not nonspecifically cluster toward the perimeter of the light scatter gate (perhaps implying nonspecific events). A cluster of cells was considered positive whereas one to three cells arbitrarily was considered to be suspicious. A successful positive control for the assay could either be cryogenically stored/aliquoted neuroblastoma tissue (example not shown) or a routinely used SK-N-DZ neuroblastoma cell line from the American Type Culture Collection (Manassas, VA, CRL-2149) that successfully demonstrated CD45− CD56very bright+ GD2+ CD81+ expression (Fig. 1).
Tissue and marrow samples included in this study were either sent to us for evaluation for suspected leukemia/non-Hodgkin's lymphoma or to confirm neuroblastoma. From 1992 to the end of 2001, 36 patients were diagnosed at our institution with neuroblastoma. Counting bilateral marrow samples as one sample unless separately analyzed by flow cytometry, we received 300 samples from these 36 patients. In addition to DNA ploidy analysis, stem cell enumeration, and “cell holds,” 123 samples from 25 of these 36 patients (Table 1) were analyzed by either the dual-color assay (N = 48), the three-color NSE assay (N = 16), or the three-color GD2 assay (N = 59). The INSS stages of the 25 patients were Stage I (N = 2), Stage III (N = 3), Stage IV (N = 18), or Stage 4S (N = 2). Of the 25 patients, 15 are currently alive (15 received transplants).
Table 1. Concordance of the Obvious Morphology and Flow Cytometry Results
Neuroblastoma cells demonstrated an extremely bright expression of CD56 NCAM antigen (fourth decade on a four-decade scale) compared with natural killer (NK) lymphocytes (Fig. 2). This finding led to the evaluation of other neuroblastoma patients by a CD45−CD56very bright+ dual-color analysis approach. The dual-color approach evolved to a three-color CD45− CD56very bright+NSEcytoplasmic+ combination cell membrane/cytoplasmic assay and the three-color CD45− CD56very bright+GD2+ cell membrane assay for tissue (Fig. 3) and marrow samples (Fig. 4). As light scatter, CD56 intensity, and NSE expression of neuroblastoma cells can degrade over time, testing of fresh samples on the same day of collection was required.
In contrast to tissue biopsy and fluid samples, concordance analysis for the marrow samples was difficult. For example, the flow cytometry results from an aspirate sample could be positive, the morphology of the aspirate negative, with positive involvement of the clot or core biopsy. After excluding equivocal morphology reports including diagnoses such as cannot evaluate, unsatisfactory, crushed, aspicular, acellular, or suspicious, and also excluding cases with discrepancy between the morphology of the aspirate sample and clot or core biopsy, the combination of the two-color and three-color assay results from marrow aspirate/tissue biopsy/body fluid samples showed a 100% sensitivity and an 83% specificity with the morphology (Table 1).
The very bright CD56 antigen density was unique to neuroblastoma cells in comparison to other samples submitted for possible leukemia/non-Hodgkin's lymphoma. Other CD45− cells showed a dimmer/more usual density CD56 antigen expression (Fig. 2). These included normal cells such as brain parenchyma (N = 2), hematopoietic malignancies including cases of CD56+ CD45−CD38very bright+CD138+ multiple myeloma (N =9), and various solid tumors including undifferentiated small cell carcinoma (N =12), undifferentiated small cell/poorly differentiated neuroendocrine carcinoma (N = 2), neuroendocrine carcinoma (N = 3), poorly differentiated nonsmall cell carcinoma (N = 1), melanoma (N = 2), sarcoma (N = 1), epithelioid sarcoma (N = 1), and rhabdomyosarcoma (N = 2). The gating strategy described in Materials and Methods excluded some of these CD56+CD45− tumors that showed a higher SSC compared with neuroblastoma, such as some melanomas and epithelioid sarcomas.
Regarding other small, round, blue cell pediatric tumors, one case of Ewing's sarcoma and one case of PNET were negative for CD56, including one 2-day old sample with PNET. One cerebellar neuroblastoma/nodular medulloblastoma/PNET patient with neural differentiation was CD45−CD56very bright+GD2+ positive and one patient with differentiating ganglioneuroblastoma with residual islands of neuroblasts showed mostly CD45− CD56very bright+ GD2− positive cells with a few cells CD45−GD2+ CD56but only usual density+ positive. Another patient with differentiating ganglioneuroma showed a CD45−CD56usual density+ GD2− positive phenotype.
Of the 36 patients diagnosed at our institution from 1992 to 2001, 20 of 93 samples sent to the Children's Hospital of Los Angeles (CHLA) were reported positive by their IHC screening method of 100,000 cells using a cocktail of five antibodies (clones 390, 459, HSAN 1.2, 126-4, and BW575). Fourteen of these patients were also concurrently analyzed at our institution by the two-color or three-color flow cytometric methods from different samples. Five patients were negative by both methods, one positive by both methods, and three samples from two patients were positive by the flow cytometric method without histologic confirmation for the two to three cells of interest (these two patients either died or have residual disease). There were six patient samples suspicious by the flow cytometry method with only one cell of possible interest. These were obtained from five patients, three of whom have died of neuroblastoma, one awaiting transplant, and one alive with right abdominal and renal metastases.
Although testing various types of samples for leukemia or non-Hodgkin's lymphoma, especially from infants and children, clinical flow cytometry laboratories should be aware of CD56extremely bright+CD45− cells in the lymphocytic/blast light scatter. These cells may be due to the presence of neuroblastoma, as was serendipitously noted in our laboratory while evaluating samples of patient bone marrow (N = 2), pleural fluid (N = 2), liver (N = 1), fine-needle biopsy aspirates (N = 1), and retroperitoneal mass (N = 1). The preliminary flow cytometric results can help to expedite appropriate evaluation of the patient and selection of other laboratory testing. However, flow cytometry laboratories should be aware that other normal cells and malignancies can also be CD56+ but not extremely bright CD45− as was noted by Kussick et al. (8). The proceedings of the leukocyte typing meeting described a CD56 marker on nonhematopoietic cells such as normal brain cells, small cell carcinomas, neuronal-derived tumors, and CD45− hematopoietic malignancies (9). Although CD56 (such as the UJ13A clone) was described in a recent CAP check sample as a very helpful IHC reagent for evaluation of neuroblastoma (2) that correlates with morphology (10), it is not commonly implemented by histology laboratories. In contrast to flow cytometry, IHC laboratories cannot apply UJ13A as part of a simultaneous multicolor assay on an individual cell basis or easily quantitate antigen density. IHC does permit morphologic evaluation of the cells in question, but high nonspecific background staining and low specificity can make interpretation difficult.
Addition of GD2 to the flow cytometric assay cocktail increased the specificity of the assay. GD2 is expressed by neuroectodermally derived tumors such as melanoma, neuroblastoma, small cell carcinoma, glioma, with one reported association with T lymphocytic leukemia. It is also a normal component of the vertebral CNS system (11). Gangliosides represent a family of glycosphingolipids that have a ceramide (lipid) moiety covalently linked to a sialic acid containing oligosaccharide and is classified by structural differences in the carbohydrate portion (12). The lipid moiety is integrated into the cell membrane lipid bilayer and the oligosaccharide portion functions as a surface receptor (e.g., GM1 is the receptor for cholera toxin) that is involved in neuronal cell adhesion, neurite outgrowth, and cell differentiation. The presence of ganglioside on the cell membrane may be involved in the pathogenesis of neuroectodermal-derived tumors by protecting tumor cells from host immune surveillance. The GD2 described in this study is a disialoganglioside GD2 (II3 d (Neu Ac)2 Gg Ose3 Cer) that is undetectable in the normal peripheral blood (<2 pmol/ml) and is present only at a very low concentrations in normal brain tissue (less than 2% of total gangliosides). GD2 is not normally present in normal marrow samples, untreated leukemia, leukemia in remission, or solid tumors (13). Neuroblastoma cell lines generally have the highest level of ganglioside content (107 molecules per cell). However, not all neuroblastoma cell lines express GD2, such as SK-N-SH and others, which eliminates them as satisfactory positive controls for this flow cytometric assay (10, 13–16). GD2 is reportedly (12, 17) present in 36 of 36 neuroblastoma cases at different clinical stages (3–195 nmol/g of tissue) but is found at lower or undetectable concentrations in more differentiated forms such as ganglioneuromas and ganglioneurblastomas (<1–4 nmol/g of tissue). Because neuroblastoma can spontaneously mature, it was important to document that the GD2 flow cytometric assay was negative in ganglioneuroblastoma and ganglioneuroma. A composite ganglioneuroblastoma/neuroblastoma and a composite glioneuroma/neuroblastoma patient showed CD45− CD56very bright density+GD2+ positive staining by flow cytometry. A neuroblastoma patient with Stage IV disease (a left adrenal primary and a metastasis to the orbit) showed differentiation after 2 years to a more maturing ganglioneuroblastoma, with persistence of a CD45−CD56very bright+NSE+ positive phenotype in our study. Ewing's sarcoma is reportedly GD2 negative (16), which was also our experience. PNET, rhabdomyosarcoma, various small cell carcinomas, leukemia, and non-Hodgkin's lymphoma were also negative for GD2 in our study.
The flow cytometric assay recognized neuroblastoma in 15 of 15 tissue biopsies/body fluid samples confirmed by morphology to represent neuroblastoma. However, concordance analysis of marrow samples for specificity and sensitivity of the assay in neuroblastoma patients was problematic, as marrow morphology occasionally was equivocal or nondiagnostic. Even Homer Wright rosettes with central fibrillary material are not specific for neuroblastoma and can be seen in other malignancies such as medulloblastoma and PNET. There were instances in which flow cytometry aspirate results were clearly positive in the setting of negative, inconclusive, or suspicious aspirate morphologic results. One patient had bilateral marrow samples showing 17 and 121 positive flow cytometric events after an autologous transplant and later died. Another patient had bilateral marrow samples containing 280 and 68 flow cytometric positive events, confirmed morphologically with a later core biopsy, who received an autologous transplant and subsequently had disease recurrence. There were cases where clearly positive flow cytometry aspirate results encouraged deeper recuts of the marrow core biopsy to confirm the presence of neuroblastoma, or led to additional testing with confirmation of neuroblastoma. We suggest that the flow cytometry method can help resolve conflicting/inconclusive marrow aspirate samples, especially if a larger volume of marrow aspirate or a bone core biopsy sample can be analyzed. We have encountered cases where the flow cytometry aspirate samples were negative and the core morphology was positive. There have also been cases where small volume samples sent before harvest to CHLA were reported negative in five of five samples, with subsequent larger volumes of marrow sent at harvesting showing that four of five patients were positive. This discrepancy was possibly due to larger aspiration volumes or sample variability (18) and to the more sensitive and specific flow cytometry method compared with the IHC method. For example, the flow cytometry test results were positive in three samples from two patients or suspicious in six samples from five patients but negative by IHC at CHLA. These patients ultimately died with residual tumor or are alive with persistent tumor.
Many children with neuroblastoma present with disseminated disease, including bone marrow involvement. A sensitive and specific flow cytometric assay provides both diagnostic and prognostic information, monitoring data, and screening of purged marrow products for autologous transplant of neuroblastoma patients. IHC testing using GD2 has a reported sensitivity of 0.01% (one cell per 10,000 cells) and is superior to soft agar clonogenic assay and routine IHC stains (13, 18–20). The two-color and three-color flow cytometric methods have sensitivities of at least equal to the IHC method as shown in patients with 17 cells (0.02%), 34 cells (0.03%), 21 cells (0.01%), and 22 cells (0.01%) and is perhaps one log more sensitive in patients with 7 cells (0.002%), 9 and 6 cells from bilateral marrow samples (0.009%), and (0.005%), 14 cells (0.008%), and can be called suspicious at the level of one cell (0.001%). IHC methods (18) at Stage IV disease indicated a 0.1% sensitivity (100 neuroblastoma cells per 105 marrow nucleated cells) after three to four cycles of therapy. This is most likely the clinically useful threshold for patients. Children below this level of detection survive, whereas those above this level died despite autologous marrow transplantation. IHC has been used traditionally to evaluate peripheral blood samples before autologous bone marrow transplantation, but the 0.01% sensitivity of the IHC method may not be adequate. For example, 6 of 11 patients had disease recurrence although the stem cell products were noted to be free of tumor (21). To our knowledge, only two other reports using flow cytometric methods described the sensitivity of detecting neuroblastoma cells. The first report found that using a CD81 CD56 CD45 cocktail to detect neuroblastoma cells spiked into normal peripheral blood was superior to a CD9 CD56 CD45 cocktail and had a sensitivity of 0.005% or five cells per 100,000 (22). The second report found that a CD9 CD56 CD45 cocktail to detect neuroblastoma cells spiked in normal bone marrow mononuclear cell preparations had a sensitivity of 0.01–0.001% (23). The sensitivity of our results and the two other reports are comparable to reverse transcriptase polymerase chain reaction (RT-PCR) testing for tyrosine hydroxylase enzyme of catecholamine synthesis in neuroblastoma, which has a reported sensitivity of 0.001% or a detection level of one cell per 105–106 cells by using a patient mononuclear cell preparation (24) or a neuroblastoma IMR32 cell line artificially spiked into samples (25). Another RT-PCR method for detection of the neuroendocrine protein gene product 9.5 has a reported detection level of one cell per 107 cells using a SK-N-MC neuroblastoma cell line spiked into a peripheral blood mononuclear cell preparation (26). The mononuclear enrichment using gradient separation enriches for neuroblastoma cells (27) and decreases background activity, which would artificially increase the sensitivity of such assays.
Flow cytometry is an important adjunct to traditional morphology and IHC for bone marrow assessment of neuroblastoma patients and their prepurged and postpurged autologous transplant products (28). The flow cytometry test provides information on neuroblastoma disease status in an objective, quantitative, cost-effective, and expedient fashion, to allow for appropriate therapeutic decisions. The quantitative assessment of marrow involvement has been associated with clinical outcome and may define new prognostic subgroups within current clinical stages (13). Further improvements in the specificity and sensitivity of flow cytometric assays may involve adding additional parameters to the CD45 CD56 GD2 cocktail such as aminoactinomycin D (7-AAD) viability dye (personal communication, Mr. Mike Swartz, Children's Hospital of Orange County, CA) or other membrane/cytoplasmic antigens associated with neuroblastoma cells. Other antigens described on neuroblastoma cell lines (29–31) include CD81 (22) and CD9 (22, 23, 32). CD81 is an antiproliferative antibody that reacts with a 26-kDa type III transmembrane protein (TAPA-1 of the tetraspan family), but has a broad cell distribution on hematopoietic, neuroectrodermal, and mesenchymal cells (33). CD81 is strongly expressed on neuroblastoma cell lines by IHC. A CD81 flow cytometric assay is more sensitive and specific than a CD9 CD56 CD45 test for detecting spiked neuroblastoma cells in peripheral blood samples (22). CD9 is a 24-kDa IV transmembrane protein (MRD-1 of the tetraspan 4 super family) and is expressed on hematopoietic, neural, muscular, and epithelial cells. A CD9 CD56 CD45 flow cytometric method demonstrated suspected neuroblastoma cells in the peripheral blood of a patient who never achieved complete remission and sorting of CD9 CD56 CD45 cells from the peripheral blood and bone marrow from three other patients demonstrated neuroblastoma morphology and N-myc amplification by PCR (23). Confirmation of flow cytometrically defined neuroblastoma cells by another method such as PCR or FISH may further support flow cytometric impressions when morphologic features and IHC studies are nondiagnostic. In our laboratory, we successfully performed N-myc FISH analysis on sorted cells from the SK-N-DZ neuroblastoma cell line. Our current flow cytometric method now uses a four-color approach, incorporating CD81 to analyze all clinical samples from patients at our institution with a suspected diagnosis of neuroblastoma.
The authors extend their appreciation to Mr. John Roys (who performed most of the flow cytometric analysis), Mr. Richard West (for creating the Excel database and figures), Ms. Kimberly Leigh (who collected the archival pathology reports), and Ms. Sandy VanVels, Ms. Di Ann Staniulis, and Ms. Juli Funk (who transcribed the text).