| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Letters to the Editor |
Children’s Cancer Research Institute St. Anna Kinderspital Kinderspitalgasse 6 A-1090 Vienna Austria
Letter
In the September issue of Cancer Research, Mora et al. presented a study intended to clarify the origin of the Schwann cell stromal component in neuroblastic tumors, i.e., NB,1 ganglioneuroblastoma, and GN. The authors used LCM for subsequent allelotype analyses; LSC for morphological, immunophenotypical, and DNA content analyses; and fluorescence in situ hybridization on touch imprints obtained after manual microdissection. Because those cells interpreted as Schwann cells showed identical genetic features as the neuroblastic cells, the authors concluded that the Schwann cells in neuroblastic tumors were of tumoral origin. Thus, this study is inconsistent with previous ones (1, 2, 3, 4) . We would like to demonstrate a series of points that we consider as weaknesses and mistakes of their study.
Our first criticism concerns the selection of the material from which the authors claimed to have analyzed 500 and 
1000 Schwann cells. It is well known that Schwann cells are rare or even absent in untreated stage 4, MYCN-amplified tumors, and tumors diagnosed in children <1 year of age. Nevertheless, the authors have chosen a series of 28 tumors for LCM that consisted of 21 Schwann cell stroma-poor tumors, among them, untreated stage 4 tumors, MYCN-amplified tumors, and tumors diagnosed in patients <1 year of age. Moreover, two of the seven tumors designated as Schwann cell stroma rich were diagnosed in infants (0.3 and 0.4 years), an age group in which these tumors are usually not found (5)
.
The methods selected to answer the authors’ questions represent another debatable point: (a) in LCM, although a powerful technique, the selection of the target cells is exclusively done on morphological grounds using cryosections without any immunological verification of the target cells; therefore, it seems to be impossible to isolate 500 and 1000 Schwann cells from the set of tumors the authors have used; (b) LSC was performed on touch preparations, and Schwann cells were identified by morphological and/or immunofluorescence criteria. However, assessment of morphology on touch imprints is insufficient to identify Schwann cells. Under these conditions, many cell types may display elongated or spindle morphology reminiscent of Schwann cells, including neuroblastic cells, and the procedure itself may lead to artifacts (as shown in Fig. 4); (c) furthermore, the immunological staining with GD2 and S-100 monoclonal antibodies, which are markers for NB and Schwann cells, respectively, is unclear. The authors’ descriptions and conclusions would imply a simultaneous or consecutive double-staining procedure on the same cells. This would, however, be unacceptable by using the described procedure. Therefore, we assume that the authors carried out the immunological staining on different touch preparations. In this case, their conclusion about the GD2 and S-100 coexpression would not be justified; and (d) seven cases were used for fluorescence in situ hybridization analyses. Although the authors investigated stroma-poor and stroma-rich areas separately on touch slides after manual dissection, they necessarily obtained a mixture of cells, tumor cells and Schwann cells, from those samples derived from the "stroma-rich" or "ganglioneuromatous" areas (see International Neuroblastoma Pathology Classification; Ref. 6
). Therefore, the manual dissection of the different areas without further characterization of the cells on the touch preparations cannot be regarded as proof that cells with three hybridization signals are really Schwann cells. By using this procedure, the authors do not circumvent the problem of "cellular heterogeneity," which they believe to be a problem of paraffin sections. In fact, paraffin sections are the preferable tools to solve this problem because of the good morphology and preservation of the histological context even after in situ hybridization.
We are also concerned about several contradictions between data presented for identical tumors. One tumor (case 2052-rel) is mentioned with a triploid DNA content in Table 1 but with a diploid in Table 3. The 1p36 status, which was used as a genetic "marker" to confirm the authors’ hypothesis, is indicated as intact in the whole tumor of three GNs (in Table 1), but in Table 2, 1p36 LOH is described in the Schwannian stroma of the same tumors. However, in GNs, the results should not diverge, because these tumors are Schwann cell stroma dominant. Another case, 1641-PC, was described as having 1p36 LOH in a previous publication (7) but has allelic imbalance in this paper. An additional case, 1436-rel, was indicated as showing "retained heterozygosity" (8) but shows LOH in this paper. An additional case was described as being derived from relapse in one study (case 1564; Ref. 8 ), yet in this publication, it is derived from diagnosis.
The next point concerns inconsistencies in the description of the results. The authors state that 27 of 28 tumors analyzed by LCM showed the same genetic composition in Schwannian stroma and neuroblastic cells. However, results for both cell populations are given for only 17 cases, 13 of which were Schwann cell stroma poor, a fact that leads to the problems mentioned above. Of the remaining 4 stroma-rich tumors, 2 have been diagnosed in infants, and for one, contradicting data concerning 1p36 status exist (case 1641, see also above). In the Abstract, the authors mention that in 20 cases studied by LSC, Schwann cells identified by morphology and S-100 immunostaining had DNA content and GD2 staining pattern identical to their neuroblastic counterparts. Besides the fact that only 19 cases are listed in Table 3, only 8 cases are indicated as S-100 positive. The other cases were designated as nonevaluable but in a previous publication as negative (Schweisguth prize-winning paper, International Society of Pediatric Oncology).
Additional shortcomings are evident in the interpretation of the data and the conclusions. The authors not only use aberrations but also intact alleles for the conclusion of a common tumoral origin. However, the lack of a genetic aberration cannot be judged as an indication of a common origin. What is also misleading is that the authors interpret GD2-positive cells as Schwann cells. Schwann cells are known to express S-100, but they do not express GD2 (9) . Finally, the authors’ conclusion, i.e., that the late relapses seen in stage 4 cases postchemotherapy could be attributable to viable non-neuroblastic stromal cells, appears worrisome. Therefore, they propose that in such cases, new therapies addressing specifically the stromal compartment of tumor cells might be important. This would imply the necessity of treating GNs, which mainly consist of Schwannian stromal cells, with cytotoxic agents. This is, however, not the case as clinical practice has shown for many decades.
Altogether, the data presented, including the inappropriate selection of materials and methods, the inconsistencies of data presentation, and the inadequate data interpretation do not allow the conclusion that neuroblastic and Schwann cells in neuroblastic tumors are derived from genetically identical neoplastic cells.
FOOTNOTES
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.
1 The abbreviations used are: NB, neuroblastoma; GN, ganglioneuroma; LCM, laser-capture microdissection; LOH, loss of heterozygosity; LSC, laser-scanning cytometry. 
Received 10/18/01. Accepted 3/18/02.
REFERENCES
Department of Pathology and Laboratory Medicine Childrens Hospital Los Angeles Los Angeles, California 90027
Letter
In the September 2001 issue of Cancer Research, Mora et al. presented a paper entitled "Neuroblastic and Schwannian Stromal Cells of Neuroblastoma Are Derived from a Tumoral Progenitor Cell" (1) . In this study, they concluded that both neuroblastic cells and SS1 cells in neuroblastic tumors are derived from genetically identical neoplastic cells based on the analysis of 17 stage IV tumors and 23 locoregional tumors. Although this conclusion supports the classical paradigm of neuralcriptopathy and a hypothesis that neuroblastic tumors are derived from neural crest cells with a potential of differentiation along multiple lineages, I find fundamental difficulties in the design of this study.
At this stage, I am very hesitant to accept the conclusions described by the authors in this article. Because their conclusions, if scientifically sound and reasonable, could potentially have a huge clinical implication for the management of highly aggressive NB cases, critical assessment of the methods used in the study is extremely important.
FOOTNOTES
The abbreviations used are: SS, Schwannian stromal; LCM, laser-capture microdissection; NB, neuroblastoma; GN, ganglioneuroma; GNB, ganglioneuroblastoma; LSC, laser-scanning cytometry.
Received 11/ 6/01. Accepted 3/18/02.
REFERENCES
Department of Hematology and Oncology Hospital Sant Joan de Deu de Barcelona Barcelona, Spain
Department of Pediatrics Memorial Sloan-Kettering Cancer Center New York, New York 10021
Department of Pathology Memorial Sloan-Kettering Cancer Center New York, New York 10021
Reply
We thank Dr. Shimada for his thoughtful comments and Dr. Ambros et al. for their careful critical reading of our report, and we appreciate the opportunity to respond. We are somewhat surprised by the comments that Schwannian stromal cells are difficult to identify by morphology alone, because the basis for current neuroblastic tumor classification is Schwannian stromal cell content as determined by morphology (1) . The presence of Schwannian stromal cells in neuroblastic tumors is also not uncommon. In fact, the published recommendations of the International Neuroblastoma Pathology Committee for classification states that for neuroblastoma, differentiating subtype: "... Schwann cells are commonly present.... Furthermore some tumors show substantial Schwannian stroma formation ... " (1) . Perhaps there is some misunderstanding in that our cases were not randomly selected neuroblastic tumors but were chosen with the prerequisite that there was adequate stromal tissue for our studies. We certainly agree that significant stromal differentiation in MYCN amplified tumors is rare. We identified 3 of 36 MYCN amplified cases (8.3%) in our collection that we felt contained adequate Schwannian cells for analysis. These were classified as neuroblastoma, differentiating subtype. This frequency is in line with the published morphological classification of MYCN amplified tumors from the Children’s Cancer Group with 7 of 117 classified as neuroblastoma with differentiation and, interestingly, 2 classified as stroma-rich ganglioneuroblastoma [total, 9 of 117 (7.6%); Ref. 2 ]. These are not significant differences in overall classification of MYCN amplified tumors.
We evaluated touch imprints of neuroblastic tumors by laser-scanning cytometry, scoring at least 3000 cells/case. In the analysis of GD2- and S100-positive subsets, we set a stringent criterion (described in the legend to Table 3 ; Ref. 3 ) that required at least 30% of the cells to be immunoreactive. This stringent criterion resulted in a number of cases that were not evaluable because of insufficient S100-positive cells present in the preparation. We believe that the low S100-positive cell count in imprints was attributable to the fact that Schwannian stromal cells tend to be more cohesive and less likely to be dislodged by touch preparation, not that the stromal cells were nonreactive for S100. This explains our difficulty in achieving technically useful touch preps from ganglioneuromas. We relied on immunophenotype for identification of Schwann cells in touch preps. These S100-positive cells were found to have DNA content that overlapped that of GD2-positive cells. Although we did not perform double labeling, the DNA indices of S100-positive cells and GD2-positive tumor cells were similar in all cases, indicating a common derivation. We should also point out that GD2 is expressed in several cell types other than neuroblasts, including Schwann cells, and can be readily detected by immunohistochemistry (4 , 5) .
|
Dr. Ambros et al. point out several seemingly contradictory results reported for individual tumors. Although these few cases do raise interesting questions concerning individual heterogeneity of neuroblastic tumors, they do not have an impact on the conclusions drawn from the entire data set. For example, the ploidy status of case 2052 was categorized as triploid, based on a major triploid peak detected by flow cytometry of disaggregated nuclei from whole tumor samples. This analysis also revealed a minor diploid peak that was considered to be contaminating normal cells, as is commonly the case. However, when the DNA index of the isolated populations was evaluated by laser scanning cytometry, a predominant diploid population was identified, raising the issue of the effect of sampling on ploidy analysis. Likewise, the determination of 1p status for 3 cases showed allelic ratio differences between analysis of bulk tumor and microdissected components. This is not an unusual occurrence in analysis of other malignant tumors and is attributable to tumor heterogeneity and the effect of sampling. We should not expect neuroblastic tumors to be exempt from this type of heterogeneity, although our study suggests that it is relatively rare.
It is unfortunate that Ambros et al. found our description of the results confusing; however, it is obvious based on their meticulous review that the results are presented in sufficient detail that they can be clearly evaluated, and readers are able to make their own interpretations. The samples that Ambros et al. believe to be reported differently in previous publications are simply technical misinterpretation on the part of the reader. The category of allelic imbalance (an altered allelic ratio not to the level of that characteristic of loss of heterozygosity) was not used in our previous publication. It was included in the present study because ploidy was known, and microdissection procedures allow fairly stringent calculations to be used. Therefore, case 1641 demonstrated allelic ratios within the range of imbalance, based on these criteria. Case 1436, although showing retained heterozygosity at D1S548 as described previously, was further evaluated at additional loci in the present study and demonstrated extensive loss at more proximal regions of 1p. Case 1564 was classified as a diagnostic sample, based on the lack of any therapy before sample collection, the critical criterion for that assignment in this study, although this was collected at a second surgical procedure. A final apparent discrepancy is a case that is missing from Table 3 . This was the last case in the table of our submitted manuscript and inadvertently omitted in the publication. We missed this in proof and present the entire table here.
Our suggestion in the discussion that chemotherapy could provide a selective pressure for certain differentiated cell types, including those with stromal differentiation which are resistant to therapy, is hypothetical. However, the fact that cytotoxic therapy often results in extensive stromal differentiation in neuroblastoma and the well- described phenomenon of transdifferentiation in cell lines derived from neuroblastic tumors suggests that this possibility should be considered. A more slowly proliferating differentiated cell type is likely to be less sensitive to antiproliferative agents, thereby providing a reservoir for recurrence. In the absence of the selective pressure, residual tumor cells may revert to a more proliferative, less differentiated state. This concept should not be ignored and should be considered in therapeutic approaches to this disease. The fact that cell lines with stromal differentiation are not easily isolated from neuroblastic tumors does not mean that such a cell type does not exist but likely reflects the selection biases of culture conditions.
We agree with Dr. Shimada that the question of differentiation potential in neuroblastic tumors is clinically important and one that requires careful study. It is for that reason that we applied three separate technical approaches to the question rather than accepting prior conclusions drawn from less rigorous studies. We hope that our findings encourage further investigation of this critical issue.
Received 1/22/01. Accepted 3/18/02.
REFERENCES
This article has been cited by other articles:
|
|
 |
|
  J. J. Molenaar, M. E. Ebus, J. Koster, P. v. Sluis, C. J.M. van Noesel, R. Versteeg, and H. N. Caron Cyclin D1 and CDK4 Activity Contribute to the Undifferentiated Phenotype in Neuroblastoma Cancer Res., April 15, 2008; 68(8): 2599 - 2609. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Cancer Research | Clinical Cancer Research |
| Cancer Epidemiology Biomarkers & Prevention | Molecular Cancer Therapeutics |
| Molecular Cancer Research | Cancer Prevention Research |
| Cancer Prevention Journals Portal | Cancer Reviews Online |
| Annual Meeting Education Book | Meeting Abstracts Online |
