Aberrant Expression of Tumor Suppressor Proteins in the Ewing Family of Tumors

The Ewing family of tumors (EFTs), composed of Ewing sarcoma and primitive neuroectodermal tumor, represents a unified neoplastic entity occurring in bone and soft tissues.1 While Ewing sarcoma represents the most primitive end of the differentiation spectrum, primitive neuroectodermal tumor shows histologic and/or immunohistochemical evidence of neural differentiation.2 Chimeric fusion of the EWS gene on chromosome 22q12 to 1 of the 2 ETS family transcription factors—FLI1 or ERG—is considered to be the initiating molecular event for the pathogenesis of EFTs.3–5 The EFTs are aggressive tumors; metastatic disease is detectable in 15% to 20% of patients at diagnosis.6 Although modern chemotherapy and radiation therapy regimens have markedly improved the prognosis of localized disease, a significant proportion of metastatic cases continue to have an adverse outcome (ie, death or progressive disease).7 

Abrogation of the late G1 cell cycle checkpoint occurs in a variety of malignancies, often as a secondary genetic alteration during multistage progression.8,9 The key components of this checkpoint include several regulatory proteins, such as the retinoblastoma protein (pRB), p53, D- and E-type cyclins, cyclin-dependent kinases (cdks), and cdk inhibitors (eg, p16^INK4a, p21^WAF1, and p27^KIP1).10 Hypophosphorylated (active) pRB prevents the cell from entering S phase, and cdk inhibitors maintain pRB in its active form by inhibiting the phosphorylating (ie, inactivating) cdk-cyclin complexes.10 Thus, inactivation of pRB itself, or any of the cdk inhibitors by mutations, deletions, or promoter hypermethylation, shifts the balance toward continuous cell proliferation. The expression of the cdk inhibitor p21^WAF1 is inducible by wild type, but not mutant, p53, and p53 mutations may contribute to the down-regulation of p21^WAF1 levels.11 The level of p53 is regulated, in part, by p14^ARF, which is the second protein encoded by the INK4a locus on chromosome 9p21.10 The relationship between the 4 proteins included in our study is depicted in Figure 1.

Since the functional endpoint of genetic alterations occurs at the level of protein expression, immunohistochemical abnormalities can be detected irrespective of the underlying mechanism of alteration.12 Recently, commercial antibodies against various cell cycle regulatory proteins have become available that can be used to study protein expression using immunohistochemical techniques. Immunohistochemistry (IHC) offers a reliable means of assessing the presence or absence of the protein of interest specifically in malignant cells in pathologic tissues, with adjacent normal tissue serving as an internal positive control for adequacy of antigenic preservation.13 The objective of this study was to evaluate abnormalities of p16^INK4a, p14^ARF, p21^WAF1, and pRB expression by IHC in 20 pediatric EFTs and to correlate the status of tumor suppressor protein expression with clinical parameters.

Formalin-fixed, paraffin-embedded material from 20 pediatric EFTs was retrieved from the surgical pathology archives of the Children's Medical Center in Dallas, Tex. To circumvent the issue of artifactual immunohistochemical abnormalities arising from therapy-induced changes and the effects of prolonged decalcification, we used only pretreatment material or, in the case of bony primaries, nondecalcified diagnostic biopsies. Primary tumor was used for IHC in all cases except for one in which the primary tumor was not available, and pretreatment biopsy material from a liver metastasis was used. Demographic and clinical data, including survival and disease status at last follow-up, were retrieved from the tumor registry at Children's Medical Center.

Mouse monoclonal anti-p21^WAF1 antibody (clone SX118) was obtained from PharMingen Laboratories (Franklin Lakes, NJ). Mouse monoclonal anti-RB antibody 3C8 was purchased from QED (San Diego, Calif). Mouse monoclonal anti-p16^INK4a (Ab7) and anti-p14^ARF (Ab2) antibodies were obtained from LabVision/NeoMarkers (Fremont, Calif). Nonspecific mouse immunoglobulin G (IgG) was used as a negative antibody control. The Vectastain Elite ABC kit from Vector (Burlingame, Calif) was employed for the detection reactions.

Serial 5-μm unstained sections were cut onto capillary gap slides and heat-treated at 60°C for 20 minutes. Standard ABC peroxidase assays were performed to demonstrate the presence of pRB, p16^INK4a, and p21^WAF1, as described previously.13–15 A detailed protocol of the p14^ARF assay is to be published elsewhere (J. Geradts, MD, unpublished data, 2001). Briefly, after antigen retrieval, the sections were reacted with the anti-p14^ARF antibody (4 μg/mL) at 4°C overnight. The detection reactions were performed according to the Vector ABC Elite kit protocol. Negative controls consisted of tumor sections that were reacted with nonspecific mouse IgG under identical conditions. The following external positive controls were used: normal colonic mucosa for all 4 proteins, a phyllodes tumor for p14^ARF, and a p16^INK4a-positive lung cancer xenograft for p16^INK4a. In addition, nonneoplastic stromal cells served as internal positive controls. Seventeen of 20 EFTs were also examined for MIC2 reactivity using anti-CD99 (Signet Labs, Dedham, Mass).

The cases were scored by 2 of the pathologists (J.G. and A.M.) using previously described criteria.12,13 Briefly, sections were examined for evidence of nuclear staining above any cytoplasmic background; cytoplasmic staining itself was disregarded. If there was nuclear staining in a diffuse or mosaic distribution throughout the tumor, the tumor was considered positive (normal) for the respective protein. A tumor was considered negative for pRB, p16^INK4a, or p14^ARF if no tumor cells showed nuclear expression of the respective antigen, while admixed nonneoplastic cells reacted positively.13,15,16 For p21^WAF1, a 5% cutoff was chosen to separate the positive and negative cases.15 

Cross tabulations were made between p16^INK4a, p14^ARF, p21^WAF1, and pRB expression status, and 2 categorical variables: metastatic disease at presentation and outcome. Outcome was stratified as favorable (for no evidence of disease at last follow up) and adverse (for death due to disease or progressive disease despite therapy). Comparisons were made using the 2-tailed Fisher exact test. The log-rank test was used to compare groups with respect to their survival experience. Death due to disease was treated as endpoint of interest, and survival to the end of the study, with or without disease, was considered a censoring point. Survival plots were constructed according to the Kaplan-Meier method. All statistical testing was for the 2-sided alternative, and P values less than or equal to .05 were considered indicative of a statistically significant effect. Statistical analyses were performed using the SAS software (SAS Institute Inc, Cary, NC).

The male-female ratio was 1:1, and the patients had a median age at diagnosis of 12.5 years (range, 2–16 years) (Table 1). There were 16 bony primaries, and 4 extraosseous primaries (2 from the retroperitoneum and 1 each from the small intestine and nasal sinus). Five patients had metastatic disease at presentation, whereas 15 had localized disease. Clinical follow-up was available in all cases (median, 21 months; range, 5–103 months). Eight patients were disease free at last follow-up (favorable outcome), whereas 12 patients had either died of disease or had progressive disease despite therapy (adverse outcome). In five patients (cases 1, 5, 7, 16, and 18), there was a t(11;22) karyotype and/or presence of an EWS-FLI1 fusion transcript. Immunoreactivity for CD99 (MIC2) was present in 17 of 17 cases examined; this stain was not performed in 3 cases.

Twelve cases (60%) demonstrated abnormal expression of at least one tumor suppressor protein (Table 2). There were 11 cases (55%) with loss of p21^WAF1 expression, 4 (20%) with loss of p16^INK4a expression, 2 (10%) with loss of pRB expression, and one case (5%) with loss of p14^ARF expression. Absence of 2 or more of these proteins was seen in 5 cases, and included 3 with loss of p21^WAF1 and p16^INK4a expression, one case with loss of p21^WAF1 and pRB expression; and one case with loss of p16^INK4a, pRB, and p14^ARF expression. Eight cases (40%) retained normal levels of all 4 proteins. No significant correlation was observed between the expression patterns of the proteins. Examples of the different patterns of pRB, p16^INK4a, and p21^WAF1 expression are depicted in Figure 2.

The status of tumor suppressor protein expression was compared with 2 categorical variables: metastatic disease at presentation and outcome. There was a statistically significant correlation between loss of p16^INK4a expression and metastatic disease at presentation (P = .026, Fisher exact test), but not outcome (P = .41, Fisher exact test). Although p16^INK4a loss appeared to be associated with reduced survival (Figure 3, a), this effect did not reach statistical significance (P = .20, log-rank test), probably because of the relatively small number of cases in this study. The p21^WAF1 (Figure 3, b), pRB, and p14^ARF status was not significantly correlated with either metastatic disease at presentation or outcome and did not affect survival.

Genetic alterations of cell cycle regulators can be considered to target one of 2 critical pathways. The RB pathway negatively regulates G1/S transition, whereas the p53 pathway causes G1/S arrest and/or signals apoptosis in response to genomic damage.10 Although abrogation of RB function is sufficient to cause uncontrolled cell cycle progression, there are several interacting proteins upstream that negatively impact proliferation, inactivation of any one of which is sufficient to tilt the balance in favor of unchecked cell division. Two of these key upstream proteins, p16^INK4a and p21^WAF1, belong to a class of cell cycle mediators known as cdk inhibitors.17 p16^INK4a and p14^ARF are alternative products of the INK4a gene18 and interact negatively with cdk4 and MDM2, respectively.19 Amplification of MDM2 results in inactivation of p53, whereas cdk4 complexes with cyclin D1 to inactivate pRB.10 Abrogation of p16^INK4a leads to increased phosphorylation of pRB. In a more indirect fashion, loss of p14^ARF expression also affects the late G1 checkpoint (Figure 1), in addition to its effect on other p53-mediated functions. Another important, less specific cdk inhibitor, p21^WAF1, is partly under the transactivational control of p53, and thus, down-regulation of p21^WAF1 can potentially be mediated by alterations that target either the WAF1 gene itself or p53.11 

Translocation involving the EWS gene on 22q12 and one of the 2 ETS-related transcription factors—FLI1 or ERG—is considered to be the primary molecular abnormality in the majority of EFT resulting in tumor initiation.3,4 Abrogation of the late G1 cell cycle checkpoint has also been reported in EFTs and probably represents secondary genetic alterations acquired during tumor progression. Kovar et al20 reported homozygous deletions and/or mutations of the INK4a gene in 8 of 27 (30%) primary tumors and 12 of 23 (52%) cell lines. Two of 27 (7%) primary tumors and 10 of 23 (43%) cell lines in their series had p53 mutations. Loss of pRB expression was detected in one tumor cell line with wild type INK4a gene, whereas cdk4 and cyclin D1 amplification was not observed in any case, suggesting that alterations at these loci are uncommon in EFT.20 Recently Wei et al21 reported INK4a deletions in 7 of 41 (18%) patients. In a companion article, they also described aberrant p53 expression in 6 of 55 (11%) primary EFTs15; mutational analysis of the p53 gene was not performed in this study. In contrast, Park et al22 and Patino-Garcia and Sierrasesumaga23 failed to detect INK4a mutations or allelic deletions in their combined series of 70 primary EFTs, while the latter investigators found p53 mutations in only one of 43 (2.5%) cases. Additionally, Park et al22 did not detect INK4a promoter hypermethylation or loss of p16 or pRB expression by immunohistochemistry in their series of 27 tumors. Combining the somewhat conflicting data from these studies, INK4a alterations (mutations and/or homozygous deletions) have been detected in 15 of 138 (11%) primary EFTs, whereas p53 alterations (mutations and/or aberrant expression) have been reported in 9 of 125 (7%) primary EFTs. There are scarce data on the status of the other G1 checkpoint proteins in EFTs, and to the best of our knowledge, p14^ARF status has never been examined in these tumors.

We believe our study to be the first to concomitantly examine the expression of more than 2 tumor suppressor proteins in the same series of pediatric EFTs. The most common immunohistochemical aberration was down-regulation of nuclear p21^WAF1 expression, seen in 11 of 20 (55%) cases. Recently, de Alava et al15 reported absence of p21^WAF1 expression in 30 of 50 (60%) archival tumors, which is in concordance with our findings. Since p21^WAF1 is transactivated by normal (but not mutant) p53,11 it may be argued that loss of p21^WAF1 expression merely represents a surrogate for p53 inactivation. Although this is certainly true in a subset of cases,24 the reported rate of p53 aberrations in EFTs does not exceed 10%.15,23 Hence, in the majority of instances, loss of p21^WAF1 expression in EFTs possibly occurs via p53-independent pathways, such as WAF1 gene mutations or epigenetic inactivation secondary to promoter hypermethylation.25 We could not investigate p21^WAF1 and p53 expression simultaneously in our cases because of the paucity of archival tissue in the paraffin blocks, but such a study is definitely warranted and would help define the mechanism of p21^WAF1 inactivation. We failed to find a significant correlation between p21^WAF1 status and either metastasis at presentation or outcome; similarly, p21^WAF1 expression did not affect survival in our patients (Figure 3, b). However, loss of p21^WAF1 expression has been reported in several other tumor types; in these tumor types, abnormal p21^WAF1 function had a significant negative impact on prognostic factors such as disease-specific survival, tumor recurrence, and metastasis.26–28 A larger study of the impact of p21^WAF1 status may help elucidate this question further in the context of EFTs.

Four of 20 (20%) cases showed loss of p16^INK4a expression by IHC, which is consistent with the rate reported by Wei et al21 and Kovar et al20 using genomic analysis. Although Park et al22 and Patino-Garcia and Sierrasesumaga23 did not detect p16^INK4a aberrations in their series, this discrepancy may reflect geographic differences and inherent biologic heterogeneity of EFTs, as well as methodologic differences such as the use of varied antibodies and criteria for classifying tumors as negative. We detected a statistically significant correlation between loss of p16^INK4a expression and the presence of metastasis at diagnosis (P = .026, Fisher exact test). The Kaplan-Meier survival curve also suggested that loss of p16^INK4a was associated with shortened survival (Figure 3, a); however, statistical significance was not reached because of the relatively small number of cases in each strata. Notably, Wei et al21 have reported that INK4a gene deletions are an independent predictor of poor disease-specific survival in patients with EFTs. The significant negative impact of loss of p16^INK4a function on survival has also been reported previously in cancers at other anatomic sites such as lung,29 breast,30 gastrointestinal tract,31 bladder,32 and most recently, in pediatric osteosarcomas.13 

Loss of pRB expression was not common in EFTs; loss was noted in only 10% of cases in the current study, consistent with the reported low frequency of pRB alterations in these tumors.20 Abrogation of p14^ARF expression was also infrequent in our series; this protein was absent in only one case. Curiously, 3 of 4 cases with loss of p16^INK4a expression retained p14^ARF expression. Although p16^INK4a and p14^ARF are derived by alternative splicing of 2 exons (E1α and E1β) at the 9p21 INK4a locus to a common splice acceptor (exon 2),33 recent studies have reported selective loss of function of one protein that spares the other.34 For example, epigenetic inactivation by promoter hypermethylation of E1α exon results in isolated loss of p16^INK4a expression with retention of functional p14^ARF protein16; conversely, hypermethylation-associated inactivation of p14^ARF has been shown to be independent of both p16^INK4a function and p53 mutation status.35 

Eight of 20 (40%) of our cases showed normal levels of the 4 cell cycle regulatory proteins, and these tumors did not behave in a less aggressive fashion. It is possible that in some of them, the immunoreactive pRB or cdk inhibitor proteins may be nonfunctional.36,37 Alternatively, it is conceivable that this checkpoint is abrogated in the remaining tumors by abrogation of other cdk inhibitors such as p27^KIP1 or by overexpression of cyclins or cdks. The G1 checkpoint is an intricate nodal point for the interplay of multiple cellular proteins, and abnormal function of any one protein may be sufficient to tilt the balance toward cellular proliferation. Finally, it must be noted that the concordance between immunohistochemical and molecular abnormalities is not always perfect, and therefore, any study based purely on immunohistochemical analysis has the potential of underestimating the true frequency of genetic aberrations.

In conclusion, we examined 20 archival pediatric EFTs for multiple tumor suppressor protein abnormalities using IHC. Abrogation of the G1 cell cycle checkpoint was common in EFTs, and losses of p21^WAF1 and p16^INK4a expression were the most frequent, followed by losses of pRB and p14^ARF expression. Loss of p16^INK4a expression correlated with the presence of metastatic disease and may identify a subset of tumors with an aggressive biologic behavior. A larger series examining cell cycle abnormalities in EFTs may be useful for confirming the conclusions of this small study.