Main

RMS is the most common soft tissue tumor of childhood and accounts for 5% of all malignant tumors in patients <15 y of age(1). Although historically reported survival rates are poor, survival has significantly improved subsequent to the implementation of combined modality therapy that uses surgery, radiation, and chemotherapy(2). Despite improved forms of therapy that have resulted in remarkable increases in survival duration and ultimate cure, striking differences in outcome persist between the two major histologic subtypes of childhood RMS, embryonal and alveolar(3). Although specific criteria for categorization into one or the other category have only recently been precisely defined(4–6), historical data(3) and recent analyses(4–7) both support the disparity in survival of patients with ARMS versus ERMS. The underlying reasons for and correlates of this difference in biologic behavior are presently unknown.

Efforts to assess unique characteristics of these two major groups of RMS have recently focused on parameters such as degree of differentiation(8), cytoskeletal and other muscle proteins(9), and chromosomal or molecular genetic abnormalities(10). Molecular characterization of RMS has shown excellent correlation with alveolar and embryonal histology. Specifically, ARMS is characterized by a t(2;13) (q35-37;q14) translocation in 70% of cases(11) or a variant t(1;13) (p36;q14) translocation in a smaller percentage of cases(11). The genes that are disrupted by these translocations are the PAX3 on chromosome 2, PAX7 on chromosome 1, and FKHR on chromosome 13(10). Northern blot analysis and cloning of the cDNA in ARMS tumors revealed fusion PAX3/FKHR or PAX7/FKHR products that can be detected with reverse transcriptase PCR in 54-87% and 8-15% of ARMS tumors, respectively(10). On the other hand, ERMS exhibit frequent allelic losses of chromosome 11p15.5 as well as duplication of imprinted alleles in this region(12). Interestingly, recent studies have shown that a small percentage of ARMS exhibit alterations of the 11p15.5 region similar to those encountered in ERMS(12). Molecular genetic studies have also been applied to other childhood malignancies, the most representative being childhood neuroblastoma. The latter has been noted to possess chromosome 1 abnormalities, double minute chromosomes, or homogenous staining regions on one or more chromosomes, often correlated with amplification and/or overexpression of the N-myc oncogene(13). Chromosome 1 abnormalities and N-myc amplification have been correlated with adverse prognosis in neuroblastoma(14).

We report here the results of a study of six childhood RMS tumor cell lines derived from both ERMS and ARMS tumors. We have correlated the histologic type of RMS with growth in vitro, nude mouse tumorigenicity, propensity for pulmonary metastasis formation, and levels of c- and N-myc oncogene expression as well as with evidence of myc gene amplification or rearrangement. We find strong correlation between histologic subtype and evidence of aggressive biologic behavior in the nude mouse model correlating with patterns of oncogene expression.

METHODS

Cell lines. Six RMS cell lines were used in the present study. The characteristics of each line, its origin, the patient's age, the site of the tumor, and the pathology of the tumor are summarized in Table 1. Three ERMS (SMS-CTR, RH-36, RD) lines and three ARMS (A-204, TC-206, TC-212) lines were included in the study. The line designated SMS-CTR was provided by Dr. C.P. Reynolds, University of Southern California School of Medicine. It was established by Dr. Reynolds 8 y before its use in the present study and was frozen down. The primary tumor consisted of a 15-cm pelvic mass arising in the posterior wall of the urinary bladder. The cell line was established at the time of massive pelvic recurrence 1 y after the diagnosis and after chemotherapy and pelvic radiation. Our studies were performed at the 10th passage of SMS-CTR cells. Line RH-36 was established by Dr. P. Houghton at St. Jude Children's Research Hospital 3 y before its use in the study. It was frozen down and used at passage 5. The RH-36 cell line was established at the time of a second look surgery for a left paratesticular ERMS in a 15-y-old male who had also developed metastasis to the lung, thoracic spine, and retroperitoneal lymph nodes and was treated in an RMS protocol. The TC-212 and TC-206 lines were established by the authors at the National Cancer Institute 2 y before their use and had been frozen down. Passages used were 5 and 7, respectively. The TC-212 line was established from a testicular metastasis of a left maxillary sinus RMS 6 mo after its presentation in a 16-y-old male who had received chemotherapy and local radiation. The TC-206 cell line was established at autopsy from a lung metastasis in a 5-mo-old male infant who had presented with a left forearm soft tissue mass 5 mo before his demise. The patient had received chemotherapy and radiation to the primary site, ipsilateral axillary lymph nodes, and lung nodules. All but TC-212 have been subsequently used in other studies(15,16). The TC-212 line was lost to contamination shortly after xenografting into nude mice and was not available for the tail vein injection experiments. Lines RD and A-204 were obtained from the American Type Culture Collection [(ATCC) Rockville, MD]. RD and A-204 are by far the longest established lines in the study(17,18). Clinical and immunocytochemical information for RD and A-204 has been provided previously(19). Briefly, the RD line originated from a pelvic ERMS in 7-y-old female who had received treatment. The A-204 cell line is the only one established before treatment from a right upper arm soft tissue mass. The tumor was metastatic to the regional axillary lymph nodes. Cytogenetic and/or molecular genetic characterization of TC-212, TC-206, and SMS-CTR has been reported elsewhere(11,16).

Table 1 RMS cell lines: origin, clinical information, histology
Full size table

Original tumor specimens before treatment on all RMS cell lines and recurrent or metastatic posttreatment tumors on all lines but A-204 were obtained and reviewed by M.T. and T.J.T. Additional sections were evaluated by immunocytochemistry for the presence of appropriate muscle markers in all lines except RD, in which only H&E slides were available for review. The paraffin blocks and/or H&E slides on SMS-CTR, RH-36, RD, and A-204 were kindly provided by Dr. A. Weinburg of Dallas Children's Hospital, Dr. D. Parham of St. Jude Children's Research Hospital, Dr. B. Landing of Children's Hospital of Los Angeles, and Dr. L.P. Dehner of the University of Minnesota (now at Washington University, St. Louis), respectively. Additional information regarding patient history for A-204 was kindly provided by Dr. S. Aaronson of the National Cancer Institute and Dr. J. Kersey of the University of Minnesota.

In addition to the RMS cell lines, three cell lines obtained from normal fibroblasts (BUD 8, CCD 33 CO, and WSI) were purchased from ATCC and used as control lines. PBL were also used as control cells. The SAN N-myc amplified neuroblastoma line(20) (obtained from Dr. C.P. Reynolds) and TC-32, a c-myc-expressing PNET cell line (established in our laboratory and published elsewhere)(21), were used as positive and negative controls for N-myc and c-myc oncogenes, respectively. These lines were also grown as described above for the RMS cell lines.

Cell lines were grown in RPMI 1640 medium supplemented with 10% FCS (GIBCO BRL, Gaithersburg, MD) and antibiotics (penicillin/streptomycin plus fungizone). Cultures were maintained at 37°C in a humidified atmosphere of 5% CO2.

Isolation of RNA and DNA from cultured cells. Confluent flasks from each cell line were used to isolate total cellular RNA and DNA. The medium was removed, cell monolayers washed twice in PBS, and the cell layers lysed after scraping with 4 M guanidine thiocyanate (Sigma Chemical Co., St. Louis, MO). Cells were then homogenized in a Dounce homogenizer, and high molecular weight DNA or total cellular RNA were recovered by centrifugation over a cushion of 5.7 M CsCI-0.1 M EDTA (pH 7.2) (GIBCO BRL) as previously described(22).

Northern blot analysis. Five microgram samples of total cellular RNA from all cell lines were denatured in 50% formamide, 0.2 M MOPS, and 2.2 M formaldehyde for 5 min at 60°C, electrophoresed through a 1% agarose-formaldehyde/MOPS buffered gel, and transferred to nitrocellulose membranes (Schleicher and Schuell, Inc., Keene, NH) by capillary action(22). These blots were hybridized with nick-translated 32P-labeled probes for the N-myc or c-myc oncogene, as described below, under high stringency conditions (42°C, 50% formamide and 5% dextran sulfate).

An L3-pHSR-1 cDNA labeled with 32P homology to the third exon of the c-myc oncogene (Lofstrand Labs Limited, Gaithersburg, MD) was used for detection of c-myc RNA, and an L3-PNB-1 cDNA 32P-labeled probe was used for detection of the N-myc RNA (Lofstrand). In addition, the relative amounts of RNA or DNA in each lane of each gel were assessed by hybridization with a β-actin probe obtained from Oncor (Gaithersburg, MD). The nitrocellulose membrane was washed in 0.2X SSC/0.1% SDS at room temperature for 1 h and at 50°C for 30 min, then incubated with X-AROMAT film (Kodak, Rochester, NY) in an x-ray cassette containing Lightening Plus intensifying screens at -80°C for varying periods of time. The N-myc RNA was detected on the same nitrocellulose membrane after washing with 0.1X SSC at 90°C for 10 min, followed by rehybridization with the L3-PNB-1 cDNA 32P-labeled N-myc probe. The same membrane was washed again and rehybridized with the β-actin probe. In separate experiments, an RNA blot was simultaneously incubated with both c-myc and N-myc probes.

Densitometric analysis of all autoradiograms was performed with an Ultrascan XL Laser Densitometer (LKB BROMMA, 2201). Variation in the amount of RNA in each lane was accounted for by comparison with the β-actin probe density for all lanes. A correction factor was applied to the measured density of c- or N-myc bands for all lanes, and these corrected values were compared.

To determine whether myc expression was independent of cell cycle, we also performed Northern blot analysis of the same cells after serum starvation for 4 d, as previously described(23).

Southern blot analysis. Ten micrograms of genomic DNA from RMS and control cell lines as well as commercially available DNA from human placenta (Oncor) and an additional case of ARMS in which fresh tumor tissue was available were digested to completion with EcoRI restriction endonuclease (GIBCO BRL) (5 U of enzyme per microgram of DNA), electrophoresed in 0.7% agarose gel, and transferred to nitrocellulose by membrane capillary blotting, as described(22).

DNA fragments blotted onto nitrocellulose with sequence homology to N- or c-myc were detected using the c-myc and N-myc specific cDNA probes described above. In various experiments, these were added both simultaneously and sequentially. Since one copy number of the N-myc gene is present in all normal cells, this was used as an internal control to evaluate the amount of DNA bound to the membrane.

Xenografts. To induce tumors in nude (nu/nu) mice, confluent 75-cm2 tissue culture flasks from each of the six RMS lines were trypsinized and the resultant cell pellets were suspended in PBS and injected subcutaneously into nu/nu mice. 1 × 106 cells in 0.2 mL of PBS were injected at each site. Six mice were injected with cells from each line. The mice were examined weekly for the appearance of subcutaneous nodules. The major diameter of any nodules identified was measured weekly. A mean value was obtained for all tumor nodules from a given cell line, and this value was plotted for each of the cell lines at each time point. The neoplastic nature of each nodule was confirmed histologically after death or after sacrifice of an individual mouse at each time point.

Pulmonary metastasis formation. Tail vein injections were performed with 2.5 × 105 cells from each RMS cell line injected into six mice per line as described elsewhere(24). All mice were killed after 6 wk and autopsied. Lung metastases were counted macroscopically and confirmed histologically. The TC-212 cell line was not available for tail vein injection experiments, as mentioned in the cell lines section of "Methods."

Statistical methods. The experimental units statistically compared are tumor lines. Within each tumor line, multiple mice were subsampled (that is, a nested design). To compare the tumor lines with regard to various end points reflecting tumor aggressiveness, we first summarized the multiple measurements within each tumor line by computing the average (e.g. the average tumor diameter at 6 wk for analyzing tumorigenicity) and then compared ARMS versus ERMS tumor lines using an exact version of the nonparametric Wilcoxon rank sum test. Qualitatively, ± tests yielded similar results but were deemed to be less reliable than the nonparametric test.

RESULTS

Growth in vitro. Each of the cell lines grew steadily in vitro over a 12-d period (Fig 1). However, there was no correlation with growth rate and histology of the RMS cell lines. The most rapidly growing tumor was, in fact, an ERMS (SMS-CTR), as was the least rapidly proliferating one (RH-36).

Figure 1
figure 1

RMS growth in tissue culture. Three ERMS cell lines (SMS-CTR, RH-36, and RD) were grown in culture identically to three ARMS lines (A204, TC-206, and TC-212). Initial seeding was approximately 1 × 106 cells per flask. Cell counts of triplicate flasks were taken at 3, 6, 9, and 12 d and the values plotted as indicated. Each value represents the mean of the replicate counts for each line at each time point. There is no correlation between tumor histology and growth in vitro; the most rapidly growing lines are both ARMS (A-204) and ERMS (SMS-CTR).

Full size image

Growth in nude mice. Tumorigenicity assayed in nude mice, in contrast, showed a noticeable correlation with histologic subtype. Of six lines studied, including three ERMS and three ARMS, three of the four most rapidly tumorigenic lines were ARMS (Fig. 2). In contrast, two of the three ERMS cell lines failed to produce any tumors until 4 wk, and the third, most aggressive ERMS cell line (RD) showed a growth rate comparable to the average of the three aforementioned ARMS. Despite this, the probability that these values were due to chance alone was low at all time points (3 wk, p1 = 0.1; 4 wk, p1 = 0.2; 5 wk, p1 = 0.1; 6 wk, p1 = 0.1; 7 wk, p1 = 0.1).

Figure 2
figure 2

RMS growth in nude mice. The same six cell lines were xenografted into the subcutis of nu/nu mice. Six animals were injected with each line. The values plotted are the mean major diameter of the six or the animals remaining at each time point. Note that all three ARMS were tumorigenic by 3 wk. From ERMS, only RD was tumorigenic at 3 wk and showed growth comparable to ARMS throughout the time course of the experiment. At 6 wk, the mean tumor diameter for ARMS was 42.3 mm; for ERMS, 17.0 mm (weighted average for each line).

Full size image

Pulmonary metastasis formation. Of five RMS cell lines studied, three were ERMS and two were ARMS. In no case did any ERMS cell line give rise to pulmonary metastasis, including RD, the most tumorigenic in the xenograft assay system (Table 2). In contrast, both ARMS cell lines produced pulmonary metastasis; A-204 gave rise to five metastasis and TC-206 to seven. Both produced bilateral metastases.

Table 2 Pulmonary metastasis formation by embryonal and alveolar RMS
Full size table

Oncogene expression and amplification. The c-myc oncogene was routinely expressed in all of the RMS cell lines studied (Fig. 3A). It was also expressed at low levels by normal fibroblasts and by the PNET cell line, known to overexpress c-myc based on previous studies(18). In contrast, the neuroblastoma cell line, known to overexpress N- but not c-myc, showed no evidence of c-myc expression. Further, the same blot rehybridized with N-myc probe failed to show evidence of N-myc expression in any of the RMS cell lines or normal fibroblasts; N-myc was detected only in the neuroblastoma control (Fig. 3B). The same membrane was washed again and rehybridized with β-actin probe for comparison (Fig. 3C).

Figure 3
figure 3

Myc oncogene expression in RMS. Total mRNA from each of the RMS was analyzed by Northern blotting and successive hybridization with c-myc (A), N-myc (B), and β-actin (C). The identity of the lanes is indicated by 1, SMS-CTR; 2, RH-36; 3, RD; 4, A-204; 5, TC-206; 6, TC-212. Control cell lines include normal fibroblast lines BUD 8, WSI, and CCD 33 CO in lanes 7-9, respectively; N-myc amplified neuroblastoma (SAN) in lane 10; a c-myc amplified PNET (TC-32) in lane 11; and normal PBL in lane 12. Note that despite lesser amounts of mRNA from the ARMS (panel C, lanes 4-6), the level of c-myc expression is markedly greater (panel A, lanes 4-6) than among ERMS (lanes 1-3); only RD (lane 4) approximates the levels observed in ARMS. N-myc was not expressed by any RMS or other cells except the positive neuroblastoma control (lane 10).

Full size image

These data were further analyzed quantitatively. The results are graphically displayed in Figure 4. Here, the level of myc expression is related to the mean level of expression by normal fibroblasts in culture that is arbitrarily quantified as 1. All three fibroblastic lines showed homogeneous c-myc expression. All RMS tumor cell lines showed greatly elevated levels of expression of the c-myc oncogene compared with the fibroblasts (range 4-83). More importantly, all three ERMS cell lines showed lesser levels of expression (range 4-8) than any of the three ARMS cell lines (range 16-83) (p1 = 0.05). Interestingly, the RD line showed intermediate-level expression of c-myc, paralleling the greater tumorigenicity in nude mice. The c-myc levels in ERMS remained lower than in ARMS even after serum starvation, suggesting that the differences were not cell-cycle related (data not shown). High N-myc expression was detected only in the control neuroblastoma line.

Figure 4
figure 4

Quantification of myc expression in RMS. Scanning densitometry of each lane was performed, and the values were expressed as multiples of the value obtained from normal fibroblasts (BUD 8, lane 7; WSI, lane 8; CCD 33 CO, lane 9) after correction for the amount of mRNA actually present on the blot, based on the density of the actin probe. The value from the fibroblasts was arbitrarily defined as 1. No ERMS (SMS-CTR, RH-36, or RD; range 4-8) approached the level of c-myc expression observed in ARMS (A-204, TC-206, and TC-212; range 16-83). RD (lane 4) was again intermediate but decidedly less than ARMS. Comparable results were obtained in multiple replicates and after serum starvation. The figure also shows the high N-myc expression in the control neuroblastoma cell line (SAN) as opposed to control PBL.

Full size image

Evidence of gene rearrangement or amplification was assessed by Southern blot analysis of the EcoRI digested genomic DNA. Although evidence of amplification of the 12.5-kbp DNA fragment containing the c-myc oncogene was detected, there was no correlation with histologic type. Two ERMS and two ARMS were noted to display amplification without rearrangement (Fig. 5).

Figure 5
figure 5

Myc oncogene amplification in RMS. Myc hybridization of genomic DNA extracted from each RMS line and control cells with c-myc and N-myc cDNA probes yielded the results observed here. No N-myc gene amplification was observed in any cell line except the known N-myc amplified neuroblastoma SAN (lane 8). In contrast, frequent amplification of c-myc was observed but did not correlate with either ERMS (lanes 1-3; SMS-CTR, RH-36, and RD, respectively) or ARMS (lanes 4-6; A-204, TC-206, and TC-212, respectively). Amplification of c-myc in two ERMS (SMS-CTR and RH-36) and two ARMS (A-204 and TC-206) was detected. No amplification was detected in an additional ARMS tumor (lane 7), PBL (lane 10), or placenta control (lane 11) but was detected in the PNET line (TC-32, lane 9). The density adjacent to lane 6 is an artifact; actual band density is observed to the left of this spot.

Full size image

We also found no evidence of amplification of the 2-kbp EcoRI digested DNA fragment containing the N-myc gene when hybridized with the N-myc probe (Fig. 5). Only the known positive control neuroblastoma (SAN) displayed such amplification.

DISCUSSION

The present study indicates that there is a correlation between alveolar versus embryonal histology of childhood RMS and in vitro levels of c-myc oncogene expression, as well as tumorigenicity and metastasis formation in nude mice. Specifically, ARMS exhibited high levels of c-myc in vitro and rapidly growing metastatic tumors in nude mice in contrast with ERMS that expressed lower c-myc levels and resulted in slowly growing nonmetastatic tumors in nude mice. This is the first substantial experimental evidence supporting a more aggressive biologic behavior for ARMS than for ERMS.

Biologic differences between ARMS and ERMS have long been suspected because of their distinct clinicopathologic features and have recently been verified by the detection of characteristic molecular genetic markers, such as the chromosomal translocations t(2;13) (q35;q14) and t(1:13) (p36;q14) in ARMS(10,11,25) and allelic losses at the 11p15.5 chromosomal region in ERMS(10,12). ARMS has been historically considered more aggressive than ERMS(3), but the predictive validity of histology in patient outcome has not always been easy to confirm in clinical studies(26). Absence of well-defined histologic criteria distinguishing ARMS from ERMS may have contributed to this, since more recent studies that make use of better-defined histologic criteria have detected excellent correlation between histology and behavior in RMS(4–7). The present study supports that ERMS is intrinsically less aggressive than ARMS and substantiates a recent clinical decision in the Intergroup Rhabdomyosarcoma Study III to intensity therapy only for patients with unfavorable histology (ARMS and not ERMS) in clinical groups I and II(27). Two of the three ERMS cell lines were remarkably less tumorigenic in nude mice and failed to produce pulmonary tumor nodules by intravenous injection, in contrast with ARMS cell lines. Further, they expressed lower levels of c-myc when compared with the ARMS cell lines. It is further noteworthy that in vitro growth characteristics of these cell lines bore no relationship to the histologic type, clinical behavior, or tumorigenicity of these tumors. Therefore, the formation of subcutaneous tumors or the generation of pulmonary metastasis in nude mice appears to be more reliable correlates of histologic subtype and prognosis.

In general, all RMS cell lines showed routine overexpression of c-myc oncogene, similar to many other human tumor systems described(28), including sarcomas in which c-myc has been reported to be almost universally overexpressed(29). However, c-myc overexpression, although not lineage-specific, has been associated with the proliferative potential of the cell(30) and has been used to refine categories of acute myeloblastic leukemia into subtypes with prognostic significance(31). On the basis of these studies, we speculated that possible differences at the levels of c-myc expression may correlate with the biologically distinct categories of ARMS and ERMS. Indeed, our data show that ARMS express higher levels of c-myc compared with ERMS in vitro (p1 = 0.05). Since the expression of c-myc has been cell-cycle related(32), we also analyzed levels of c-myc expression in serum-deprived RMS cells with thereby limited capability for proliferation. We found that c-myc expression was independent of cell cycle in RMS in vitro, since comparably high levels of c-myc expression were detected in the serum-deprived RMS cells. These data support that c-myc overexpression in RMS is due to true deregulation of gene expression and is not simply an index of rapid proliferation. However, c-myc gene expression does not always correlate with clinical profiles, as shown in colorectal carcinoma(33), and the in vivo significance of our findings is difficult to predict. Studies on a fair number of frozen tumor tissue specimens are required to address this question. Thus far, in vivo studies regarding the expression of c-myc in RMS have been limited to one case of a pelvic RMS and one case of a canine RMS(34), in contrast with studies on c-myc amplification/rearrangement that are easier to perform on archival tissues. The latter have shown that c-myc amplification and/or rearrangement is absent or only rarely encountered in RMS tumor tissues(35–37). The higher incidence of c-myc amplification in the present study may be the result of selective in vitro growth of RMS with amplified c-myc gene. Lung cancers with myc gene amplification have been previously correlated with increased xenotransplantibility(38). Overexpression of c-myc is usually the result of amplification of the gene(39). However, as the present study shows, overexpression of c-myc in RMS can be the result of mechanisms other than amplification of the gene, since the four RMS cell lines with amplified copies of c-myc did not have the highest c-myc expression, although they did have higher levels than the control fibroblastic cell lines. Further, the two other lines without c-myc amplification (RD and TC-212) expressed comparable or even higher levels of c-myc. Deregulation of c-myc expression in the absence of amplification or rearrangement of the gene has been described in a significant number of tumors(39,40) and may be due to posttranscriptional regulation resulting in increased stability of the c-myc mRNA, as reported in other cell types(41). Enhanced expression of c-myc in the absence of amplification has been reported in a canine RMS as well(34).

There was no evidence in the present study of N-myc oncogene expression or amplification in any of these six childhood RMS. These results are at variance with those reported by others(35,42). Specifically, N-myc amplification originally described in ERMS has since been considered a hallmark of ARMS(35,42) and has raised the possibility of important biologic differences between ARMS and ERMS(42). However, the finding of N-myc amplification in only a fourth of ARMS cases (10 of 40 total ARMS)(10) and the lack of association with clinical behavior(35,42) are not promising indications for a clinically prognostic candidate gene. Absence of N-myc amplification in the present small series of RMS cell lines can be explained by its rather low frequency in RMS. N-myc amplification was also absent from 25(36) and five RMS tumor specimens(37), respectively.

One may argue that treatment-induced differentiation may have influenced the behavior of some of these cells lines in nude mice and in vitro. In fact, progressive concentrations of actinomycin increased myogenic differentiation and decreased c-myc mRNA levels in RD in a previous study(43). However, c-myc expression is not always down-regulated in differentiated myoblasts(44). Further, none of the RMS tumors from which the RMS cell lines were established showed histologic signs of differentiation in recurrent or metastatic sites when reviewed for the purposes of the present study. Finally, since all but one (A-204) cell lines were established from posttreatment tumor material, it is unlikely that the observed differences in nude mice are the result of treatment. The fact remains that ARMS lines as a group were highly tumorigenic and metastatic, whereas none of the ERMS cell lines was metastatic and two of the three were much less tumorigenic than all three ARMS cell lines. The third ERMS line, RD, was a partial exception. This line formed nude mouse tumors similar to the three ARMS and displayed intermediate levels of c-myc expression but failed to form pulmonary metastasis. The ability of RD to form local and metastatic tumors in nude mice has been investigated in another study with similar results(45). In that study, metastasis was obtained only after pretreatment of the nude mice with cyclophosphamide. The fact that RD exhibited tumorigenicity comparable to that of the two ARMS cell lines argues that RD may possess additional genetic alterations that may contribute to its tumorigenicity. RD, for example, possesses an activated N-ras gene(46) that may compliment c-myc expression in the tumorigenic properties of this line, as previously reported in a human hepatoma cell line(47). In addition to genetic abnormalities common to all RMS, such as gene fusions in ARMS and allelic losses at 11p15.5 in ERMS, other alterations in oncogenes and/or growth-related genes have been detected in a fraction of RMS cases(10,19) and may account for some degree of heterogeneity within each one of the two major histologic subtypes of RMS. SMS-CTR has shown allelic losses of 11p(10,11). Both RD and SMS-CTR have also been reported to have p53 mutation and deletion(48). RH-36 has no p53 gene abnormality but expresses high MDM2 mRNA and protein that in turn results in binding and stabilization of the p53 protein(15). The p53 status of the ARMS lines used in the study is unknown. Recently, it has been proposed that common or variant gene fusions dictate aggressiveness in ARMS, with PAX3/FKHR conferring a more aggressive and less responsive phenotype than PAX7/FKHR(49). Previous studies have shown only the TC-212 line to have the classic t(2;13) (q37;q14) translocation. TC-206 exhibited a variant t(5;13) translocation with a 13q14 breakpoint but also a t(11;22) (q24;q12) translocation(11). The latter was subsequently verified by reverse transcriptase PCR and led to the concept of biphenotypic sarcomas with myogenic and neural differentiation(16). A small subset of histologic ARMS may not have either fusion product(10). There are no studies regarding the presence or absence of fusion products in A-204.

In summary, the studies reported here support that ARMS cell lines exhibit higher tumorigenicity and metastasis formation in nude mice and express higher levels of c-myc when compared with ERMS cell lines. However, our data on c-myc expression in RMS are not conclusive, and studies of a series of RMS tumors with proper c-myc expression analysis and appropriate follow-up data are warranted to address this issue.