OUP user menu

Detection of Isolated Tumor Cells in Neuroblastoma by Immunohistochemical Analysis in Bone Marrow Biopsy Specimens
Improved Detection With Use of β-Catenin

Chandra Krishnan MD, Clare J. Twist MD, Teresa Fu, Daniel A. Arber MD
DOI: http://dx.doi.org/10.1309/AJCPAJODRJYD3OB2 49-57 First published online: 1 January 2009


Evaluation of the bone marrow is a critical component of accurate staging and surveillance for recurrent disease in neuroblastoma. The value of routine immunohistochemical analysis of otherwise histologically negative bone marrow biopsy specimens has not been adequately evaluated. By using synaptophysin, chromogranin, and β-catenin, immunohistochemical analysis performed on otherwise histologically negative bone marrow specimens identified isolated tumor cells (ITCs) in 9.1%, 5.0%, and 10.0% of 220 biopsy specimens, respectively. Overall survival, as estimated by the Kaplan-Meier method, was not significantly different between patients with and without ITCs (P = .357). Of the immunohistochemical markers evaluated, β-catenin showed the greatest sensitivity for identifying ITCs in the bone marrow and showed reactivity in primary tumor samples. We found that the presence of ITCs identified by immunohistochemical analysis may predict the persistence of disease but does not show significant overall survival differences. We also identified β-catenin as a sensitive immunohistochemical marker of primary and metastatic neuroblastoma.

Key Words:
  • Neuroblastoma
  • Bone marrow
  • Metastasis
  • Minimal residual disease
  • Immunohistochemistry

Neuroblastoma is the most common extracranial malignancy in childhood. According to 2007 Surveillance, Epidemiology, and End Results data, nearly 7% of cancers in people younger than 15 years are attributable to neuroblastoma. Despite advances in supportive care, intensification of therapy, and improved risk-stratification strategies, the overall survival at 5 years for patients with high-risk disease is relatively unchanged for the past 2 decades.1 Evaluation of the bone marrow is a critical component of accurate staging and surveillance for recurrent disease. Along with patient age at diagnosis, accurate histologic classification, and evaluation of biologic features of the tumor such as MYCN gene amplification, accurate staging of neuroblastoma forms the basis of current risk-stratification models.25 Current recommendations include performing bilateral bone marrow aspirates and biopsies to minimize the chances of false-negative study results.6 Aspirate smears and trephine biopsy sections are used in conjunction to better evaluate for metastatic disease; each has its own advantages. Bone marrow aspirates allow for evaluation of a greater amount of marrow elements, and marrow biopsy specimens provide for greater ability to judge the relative amount of involvement by visualizing metastatic tumor in situ.

In contrast with immunocytochemical analysis of aspirate smears, which tends to involve a greater amount of manual labor and is beset with interoperator variability, immunohistochemical analysis of biopsies is a relatively rapid, consistent procedure that is automated in most pathology laboratories. Several markers are available for paraffin-embedded tissue samples that are sensitive for highlighting neuroblastoma. Among these are synaptophysin, chromogranin, and the neuroblastoma protein NB84.4,7,8 Although these antibodies have been the mainstay for immunohistochemical detection of neuroblastoma, the literature demonstrating their sensitivity against a large number of tumors is sparse. Nevertheless, immunohistochemical evaluation of marrow biopsy specimens may prove to reduce false-negative study results. To that end, the use of routine immunohistochemical analysis of otherwise histologically negative bone marrow biopsy specimens has not been adequately evaluated with respect to affecting long-term outcome.

In this study, we sought to identify the frequency with which isolated tumor cells (ITCs) are identified with immunohistochemical methods in histologically negative bone marrow biopsy specimens. We also sought to retrospectively correlate the presence of ITCs with long-term outcome for patients treated at a single institution using protocol-based treatment regimens. In addition, based on observations we made about the immunohistochemical usefulness of β-catenin as a marker of neural lineage (unpublished data, 2006), we postulated that β-catenin may be a useful marker of neuroblastoma. To provide perspective for this question, we compared the staining frequencies of metastatic neuroblastoma with the frequencies in the primary tumor and other “small round blue” cell tumors of childhood.

Materials and Methods

Case Selection

We identified 51 cases of neuroblastoma in the case files of Lucile Packard Children’s Hospital, Stanford University, Stanford, CA, between 1995 and 2007. The cases represented 29 boys and 22 girls. Of the 51 patients, 48 had at least 1 bone marrow biopsy specimen and clinical outcome data available for review. The median age of the patients was 29 months at the time of diagnosis (range, 0–138 months). A total of 325 biopsy specimens from 180 separate bone marrow evaluations were identified in this cohort. Bone marrow evaluations were defined as a single clinical encounter that resulted in unilateral or bilateral bone marrow biopsies. H&E-stained slides from all specimens were reviewed by one of us (C.K.), and clinical outcomes for all 48 patients were evaluated (median follow-up, 49 months; range, 6–148 months).

In addition to the bone marrow specimens, 30 primary pretreatment neuroblastoma cases and 13 specimens from 36 treated patients were also identified. These primary resection samples, along with 13 primitive neuroectodermal tumor family (Ewing sarcoma) tumors, 11 nephroblastomas, and 16 rhabdomyosarcomas were used to construct a tissue microarray (TMA) for immunohistochemical comparison.

Immunohistochemical Studies

Immunohistochemical studies using commercially available antibodies Table 1 to synaptophysin, chromogranin, and β-catenin were evaluated. The TMA and bone marrow biopsy specimens were stained with commercially acquired antibodies with different antigen retrieval techniques as defined in Table 1. Immunohistochemical analysis was performed on 4-μm, forma-lin-fixed, paraffin-embedded sections, after microwave-assisted antigen retrieval in 0.1 mol/L of citrate buffer (pH 6.0) and subsequent incubation with 3% hydrogen peroxide. For bone marrow biopsy specimens, ITCs were defined as single cells or small clusters of cells (usually <5 cells total) with strong cytoplasmic and membranous reactivity to the antibodies of interest and lacking any reactivity within the nucleus. Cases with ITCs were reviewed by two of us (C.K. and D.A.A.). Sources of false-positives that were not counted as ITCs included erythroid precursors that showed membrane-only staining or diffuse nuclear and cytoplasmic staining, osteoblasts that showed membranous staining and a light blush within the nucleus, and plasma cells that showed nonspecific cytoplasmic blush.

View this table:
Table 1

Immunohistochemistry Antibody Summary

β-Catenin141:1000.1 mol/L citrate, pH 6.0BD Biosciences, San Jose, CA
ChromograninDAK-A31:50NoneDakoCytomation, Carpinteria, CA
SynaptophysinPolyclonalPremixed0.1 mol/L citrate, pH 6.0Cell Marque, Austin, TX

TMA Assembly

A TMA of primary tumor samples was constructed by extracting 0.6-mm diameter cores of histologically confirmed areas of interest from the original paraffin blocks using a tissue core extractor (Beecher Instruments, Sun Prairie, WI)9 and reembedding these cores into a gridded paraffin block. The completed recipient block containing 210 tissue cores in total was arranged in two 12 × 11 sectors. Placental tissue samples were used as place markers to identify the beginning of each sector. Control tissue samples from cases of tubular adenomas and normal adrenal glands were placed in equal intervals within the array.


We examined 325 total bone marrow biopsy specimens from 180 individual evaluations Table 2. Of the 105 evaluations with morphologically negative results for metastatic tumor, 190 biopsy specimens (from bilateral and unilateral examinations) were available for review. An additional 30 biopsy specimens lacking metastatic tumor as determined by routine H&E evaluation were seen in conjunction with a contralateral biopsy specimen showing metastatic disease. H&E evaluation of the remaining 105 biopsy specimens revealed unequivocal involvement by metastatic neuroblastoma and encompassed 72 marrow evaluations. Of these unequivocally positive cases, 94 had enough material to evaluate immunohistochemically using all 3 antibodies. Three biopsy specimens from 3 bone marrow evaluations were deemed uninterpretable owing to lack of marrow tissue for examination.

View this table:
Table 2

Summary Data for Neuroblastoma Patients

  Average3.1 y
  RangeNewborn to 11.5 y
Total bone marrow biopsy325
  Histologically negative220
  Histologically positive105
Marrow evaluations180
  Negative results105 (190 biopsy specimens)
  Positive results72 (135 biopsy specimens)
  • * Data are given as number of patients or specimens unless otherwise indicated.

Immunohistochemical analysis routinely identified ITCs in a significant number of marrow examinations. β-Catenin highlighted ITCs in 10.0% of histologically negative marrow specimens, and synaptophysin and chromogranin did so in 9.1% and 5.0% of cases, respectively Image 1. When all marrow evaluations (including unilateral and bilateral marrow examinations) are considered together, β-catenin and synaptophysin identified ITCs in at least 1 of the histologically negative marrow specimens in nearly 20% of the cases. β-Catenin labeled ITCs in 3 cases that did not show ITCs with synaptophysin. Synaptophysin, on the other hand, identified ITCs in only 1 case that did not show ITCs with β-catenin. Of 94 marrow examinations with unequivocal metastatic disease, β-catenin, synaptophysin, and chromogranin showed immunoreactivity in 100%, 89%, and 71% of the cases, respectively Table 3. Sources of false-positive staining were apparent with synaptophysin and β-catenin Image 2. In particular, a subset of erythroid precursors showed membranous staining when using β-catenin. Given the proclivity of erythroid precursors to gather in discohesive clusters, they may mimic the tightly cohesive cells of neuroblastoma Image 3. With respect to synaptophysin, erythroid precursors showed a granular cytoplasmic and nuclear staining pattern that was in contrast with the diffuse cytoplasmic, nonnuclear pattern observed in neuroblastoma cells. This difference in the pattern of reactivity, in addition to smaller size compared with neuroblastoma cells, helps distinguish erythroid precursors from metastatic disease.

View this table:
Table 3

Summary of Immunohistochemical Results in 325 Bone Marrow Examinations

No. of Specimens*β-CateninSynaptophysinChromogranin
Histology negative marrow (ITCs)22022 (10.0)20 (9.1)11 (5.0)
Histology positive marrow9494 (100)84 (89)67 (71)
Negative marrow examinations (ITCs)10522 (21.0)19 (18.1)10 (9.5)
Pretreatment primary neuroblastoma3028 (93)27 (90)17 (57)
Posttreatment primary neuroblastoma1313 (100)11 (85)3 (23)
Wilms tumor119 (82)0 (0)0 (0)
Rhabdomyosarcoma1613 (81)1 (6)0 (0)
PNET/EWS1311 (85)0 (0)0 (0)
  • EWS, Ewing sarcoma; ITCs, isolated tumor cells; PNET, primitive neuroectodermal tumor.

  • * The number of biopsy specimens do not total 325 due to the lack of material available for evaluation by all 3 markers in 11 histologically positive bone marrow specimens.

Image 1

Isolated tumor cells in bone marrow biopsy specimens. A, β-Catenin; note occasional erythroid precursors with distinct cytoplasmic and nuclear reactivity (lower right) (×600). B, Synaptophysin; note the presence of cytoplasmic staining and absence of nuclear reactivity (×600). C, Chromogranin; note hemosiderin within macrophages mimicking true cytoplasmic reactivity (×400).

Image 2

False-positive reactivity using synaptophysin and β-catenin in bone marrow samples. A, In contrast with metastatic neuroblastoma (left), osteoclasts exhibit crisp membranous reactivity without cytoplasmic staining (right) (β-catenin, ×600). B, In contrast with metastatic neuroblastoma (left), erythroid precursors exhibit nuclear and cytoplasmic reactivity (right) (synaptophysin, ×400). C, Blood vessels show cytoplasmic reactivity within endothelial cells (β-catenin, ×400).

Image 3

Erythroid precursors are sources of false-positive staining using β-catenin and synaptophysin. A, Crisp membranous staining of erythroid precursors (β-catenin, ×400). B, Cytoplasmic and nuclear staining (synaptophysin, ×400). C, Typical membranous reactivity (glycophorin-A, ×400).

Of 32 patients who had resections of pretreatment, primary neuroblastoma, samples from 30 exhibited tumor cells of interest within the tissue array. Of these 30 pretreatment tumors, 28 (93%), 27 (90%), and 17 (57%) of the cases showed reactivity with β-catenin, synaptophysin, and chromogranin, respectively Image 4 and Image 5. In addition, 13 cases of posttreatment tumor samples were also available for immunohistochemical analysis. All 13 tumors showed immunoreactivity against β-catenin, in a pattern similar to that seen within the bone marrow and pretreatment samples. Of the 13 tumors 11 (85%) exhibited synaptophysin reactivity, and chromogranin was positive in only 3 (23%) of 13 post-treatment cases.

Image 4

β-Catenin reactivity in primary neuroblastoma. A, Poorly differentiated neuroblastoma. B, Undifferentiated neuroblastoma. C, Ganglioneuroblastoma, intermixed. Note the dense membranous reactivity in the ganglion-like cells and diffuse reactivity to the interspersed neuropil (A-C, ×400).

Image 5

Synaptophysin and chromogranin reactivity in primary neuroblastoma. Synaptophysin reactivity in undifferentiated (A) and poorly differentiated (B) neuroblastoma. Chromogranin reactivity in undifferentiated neuroblastoma (C) and ganglioneuroblastoma, intermixed (D) (A-D, ×400).

Other primitive tumors seen in a similar age group to that of neuroblastoma also showed cytoplasmic reactivity with β-catenin. Wilms tumor, rhabdomyosarcoma, and Ewing sarcoma/peripheral primitive neuroectodermal tumor showed cytoplasmic reactivity with β-catenin in 82% (n = 11), 81% (n=16), and 85% (n = 13) of cases, respectively Image 6. This finding is in contrast with findings for synaptophysin and chromogranin, which showed a near lack of staining in the tumors. Only 1 case of rhabdomyosarcoma showed focal cytoplasmic staining with synaptophysin; the remainder of the tumors did not stain with synaptophysin or chromogranin (Table 3).

Image 6

Other childhood small round blue cell tumors showing β-catenin reactivity. A, Peripheral neuroectodermal tumor. B, Rhabdomyosarcoma. C, Nephroblastoma (A-C, ×400).

Follow-up data, ranging from 6 to 148 months, were available for 48 of 51 patients. Survival at 120 months for patients without a history of ITCs was 68% and for patients with a history of ITCs was 48%. Of note, in only 1 of the 48 cases were ITCs observed in an initial staging bone marrow specimen that was histologically negative (2% total). This patient had clinical stage 4 disease at the time of diagnosis. Overall survival, as estimated by the Kaplan-Meier method, was not significantly different between patients with and without ITCs (P = .357), although patients without ITCs showed a trend toward better overall survival Figure 1. In patients with ITCs as detected by synaptophysin or β-catenin, bone marrow recurrence more often developed (31% [4/13]) than in patients without a history of ITCs (9% [3/35]) (P = .075).

Figure 1

Overall survival for patients with and without isolated tumor cells (ITCs). P = .357.


The identification of metastatic tumor in patients diagnosed with neuroblastoma is a critical component of accurate staging and disease surveillance. Aspirate smears and trephine biopsies have important roles in identifying metastatic neuroblastoma in the bone marrow. Both modalities face the challenge of nonuniform distribution of metastasis within the marrow space. The patchy nature of marrow involvement highlights a critical deficiency of marrow tissue biopsies, ie, they are susceptible to sampling bias that is highlighted by the well-known phenomenon of a histologically negative biopsy result in a patient with a concurrent marrow biopsy sample from the opposite side that is involved by metastatic tumor.

The current standard for defining bone marrow involvement by neuroblastoma remains the morphologic evidence of metastatic tumor by routine staining in aspirate or biopsy specimens.3 Although morphologic evaluation of the aspirate smears of H&E-stained biopsy sections is effective in identifying metastatic tumor, it is less sensitive to very small tumor burdens. Given that 60% to 70% of patients with high-risk neuroblastoma will ultimately experience relapse, there is considerable interest in identifying a reliable modality for detecting minimal residual disease in the bone marrow, which may be predictive of ultimate relapse. This type of tool may identify a subgroup of patients who might benefit from additional and/or novel therapies. Many studies have attempted to ascertain the value of adjunct special studies, such as molecular analysis and flow cytometry, in identifying minimal residual disease. Unfortunately, the bulk of these prior studies tended to highlight the technical feasibility of the method of interest, without correlation with long-term outcome.10,11

Some of the earliest studies confirmed the superior sensitivity of immunocytochemical methods to histologic evaluation alone and the prognostic relevance to the presence of marrow disease at the time of diagnosis. The reasons for interest in the usefulness of immunocytochemical methods include the fact that pooled aspirate material provides a greater volume of “tissue” to examine and that residual material can be used for immunocytochemical and flow cytometric studies. For example, the use of antibodies against disialoganglioside-D2 (anti-GD2) on aspirate smears has been shown to reliably identify metastatic tumor.1214 Use of this antibody is currently limited to immunocytochemical and flow cytometric studies, but is not appropriate for immunohistochemical analysis. In one of the largest evaluations of the use of immunocytochemical analysis as an adjunct to histologic evaluation of marrow metastases, Cheung et al15 demonstrated that the use of anti-GD2 in pooled aspirate material has greater sensitivity for identifying metastatic neuroblastoma, in comparison with histologic evaluation of concurrent bone marrow biopsy specimens. However, they stated that the risk of death in cases with marrow specimens with immunocytochemically positive aspirates was significantly lower than for histologically positive marrow specimens. Similarly, flow-cytometric analysis of aspirate material has been shown to reliably detect small numbers of neuroblastoma cells. Most often, an antibody panel consisting of CD56, disialoganglioside-D2, and CD81 is used in various combinations to optimize neuroblastoma cell detection. In spiked sample studies, however, the limits of detection by flow cytometry are still less than observed for immunocytochemical studies.16

One of the biggest advantages provided by tissue biopsies is the ability to perform immunohistochemical analysis on fixed tissue, which allows for histologic correlation and evaluation of disease burden. Immunohistochemical analysis of paraffin-embedded tissue samples is a time-honored, automated, and reproducible special study with great diagnostic and prognostic potential. Historically, the most commonly used antibodies for detecting or confirming the presence of neuroblastoma are synaptophysin, chromogranin, CD56, and neuron-specific enolase. These markers have demonstrated consistent reactivity to primary neuroblastoma, although they can exhibit poor specificity and reproducibility in decalcified tissue specimens.7,17

In one of the largest published studies, Wirnsberger et al8 compared the relative staining frequencies of synaptophysin and chromogranin in primary neuroblastoma. In their study, synaptophysin stained 29 of 31 cases, and chromogranin stained 21 of 34 cases. Our results show similar findings and support the fact that synaptophysin tends to have a higher sensitivity to neuroblastoma, compared with chromogranin. NB84 is an antibody with specificity to a yet-to-be characterized antigen specific to neuroblastoma tumor cells. Originally described in 1991, the antibody showed very high sensitivity in primary tumors.7 A major limitation of NB84 is its low sensitivity for detecting bone marrow metastases, thus decreasing its usefulness in identifying ITCs. Bomken et al18 demonstrated that only 62.5% of metastatic tumors were highlighted by NB84, in comparison with all pretherapy and posttherapy primary tumor samples.

In addition to these traditional neuroblastoma markers, we sought to identify the usefulness of β-catenin for detecting neuroblastoma. In a previous evaluation of β-catenin immunoreactivity among neural tumors, we observed strong, intense reactivity by β-catenin in a small number of neuroblastomas that were part of the study set (unpublished data). Historically, synaptophysin and chromogranin were the adjunct immunohistochemical stains used in our institution for evaluation of subtle neuroblastoma bone marrow metastases. The comparative staining frequencies of these established neuroblastoma markers within the bone marrow, however, are not well studied. With this in mind, we sought to compare the usefulness of β-catenin with these markers. Our findings suggest that β-catenin exhibits very high sensitivity as an immunohistochemical marker that is comparable to that of synaptophysin and superior to that of chromogranin. To better characterize the potential usefulness of β-catenin, we evaluated the staining frequency of other primitive small round blue cell tumors in the differential diagnosis of neuroblastoma. Specifically, rhabdomyosarcoma, nephroblastoma (Wilms tumor), and Ewing sarcoma/peripheral primitive neuroectodermal tumor are tumors of childhood, which have morphologic features that overlap those of poorly differentiated neuroblastoma, that do not generally react with synaptophysin. Unfortunately, β-catenin displays reactivity at a high frequency with all of these entities and is not a specific marker within this differential diagnosis (Table 3).

Our data show that routine H&E evaluation of bone marrow biopsy specimens for staging and surveillance in neuroblastoma fails to identify ITCs nearly one fifth of the time. Their presence, however, does not predict a significant difference in overall survival. ITCs do seem to be a marker of an increased risk of recurrence within the marrow. Although this latter finding is not statistically significant, it may identify a subset of patients in whom complete remission has not been fully attained and who may benefit from novel therapies that target minimal residual disease. Although our findings do not support the routine use of immunohistochemical studies in histologically negative staging or surveillance bone marrow biopsy specimens, larger studies are necessary to evaluate the validity of predicting recurrent disease through the identification of ITCs.


We express our deepest thanks to Amarjeet Grewal and Marylin Masek for expert assistance with histologic and immunohistochemical studies of the bone marrow biopsy specimens.


View Abstract