Nucleophosmin as a Candidate Prognostic Biomarker of Ewing's Sarcoma Revealed by Proteomics

Ewing's sarcoma is the second most common primary malignant bone tumor in children and adolescents. Despite significant progress regarding intensive chemotherapy protocols and local control measures, 30% to 40% of patients with localized Ewing's sarcoma and 80% of patients with metastatic Ewing's sarcoma at diagnosis die due to disease progression within 5 years (1). More intensified first-line chemotherapy regimens or combinations of chemotherapeutic agents improve clinical outcome compared with conventional chemotherapy (2, 3). However, such therapies may result in serious toxicity, including fatal gastrointestinal toxicity, grade 3 or 4 infections, and severe myelosuppression (4–6). The patients could thus benefit from less aggressive regimens by avoiding the higher risk of toxicity associated with overtreatment. Indeed, approximately two-thirds of patients with localized Ewing's sarcoma are cured with conventional therapy alone (7, 8). Therefore, the identification of prognostic factors may lead to the development of risk-adapted treatment strategies for Ewing's sarcoma.

Clinical factors currently evaluated have limited prognostic value; the presence of metastases at diagnosis, which is the most unfavorable prognostic factor for Ewing's sarcoma, concerns only ∼25% of Ewing's sarcoma patients (9). The prognostic value of other clinical and pathologic features that correlated with prognosis of Ewing's sarcoma, including the site and size of the lesion and the age of the patient (1, 10), has decreased following recent advances in treatment (11). For instance, in earlier studies, tumors >8 cm were associated with a worse prognosis (12), whereas tumor size is not assumed as definitive prognostic factor in studies using the more intensive EW92 protocol (13).

In recent years, high-throughput screening technologies such as array-based comparative genomic hybridization analysis and cDNA microarray technology have been used to identify up-regulated or down-regulated genes with prognostic value for Ewing's sarcoma (14–19). These comprehensive studies suggested the presence of a poor prognosis signature at diagnosis and identified several genes that may be involved in the process of invasion and metastasis in Ewing's sarcoma.

Emerging technologies that examine the overall features of the expressed proteins, that is, proteomics, have identified many candidate proteins associated with early diagnosis (20), differential diagnosis (21), prognosis (22, 23), and response to chemotherapy (24) in various diseases but have not been vigorously employed in the study of Ewing's sarcoma.

In this study, we performed a proteomic study using biopsy samples from Ewing's sarcoma patients. We found that nucleophosmin expression significantly correlated with progression of Ewing's sarcoma. Although aberrant expression of nucleophosmin has been implicated in various other malignancies (25–29), this proteomic study shows its aberrant expression and prognostic utility in Ewing's sarcoma.

Patients and clinical information. This study included a total of eight frozen incisional biopsy samples taken before treatment at the time of diagnosis from 8 Ewing's sarcoma patients treated between June 1996 and December 2006. These samples were snap-frozen in liquid nitrogen and stored at −80°C until use. The clinical information of the patients is summarized in Table 1. This project was approved by the ethical review board of the National Cancer Center after signed informed consent was obtained from all patients. All cases were reviewed and histopathologically diagnosed by a certified pathologist (N.T. and T.S.). Clinical staging was determined based on diagnostic imaging criteria according to the Musculoskeletal Tumor Society Surgical staging system (30). Primary tumor size was measured at the greatest tumor dimension on radiographic images, including computed tomography scans and magnetic resonance imaging.

From the 8 cases included in the study, 3 patients were alive and continuously disease-free (CDF) in the follow-up period of at least 3 years from diagnosis and 5 patients were dead of disease (DOD) within 2 years from initial diagnosis. All 5 patients in the latter group developed distant metastases within 7 months from initial diagnosis.

A previous report indicated that Ewing's sarcoma patients with early relapse, defined as relapse within 2 years after initial diagnosis, had shorter survival (31). We grouped the Ewing's sarcoma samples into two groups: the samples from patients that were alive and CDF over 3 years post-diagnosis were defined as the good prognosis group (Table 1, samples 1-3). The samples from patients that were DOD within 2 years were defined as the poor prognosis group (Table 1, samples 4-8).

For the nucleophosmin immunohistochemical expression study, we examined 34 tissues paraffin-embedded before treatment from 34 independent cases (Table 2, samples 9-42). These patients were treated between June 1981 and December 2005 at the National Cancer Center and Keio University Hospital. This project was approved by the ethical review boards of the National Cancer Center and Keio University Hospital after signed informed consent was obtained from all patients in this study. The clinical information concerning the cases used in the immunohistochemical study is summarized in Table 2.

Rearrangement analysis. Total RNA from tumors was extracted by the guanidinium thiocyanate method (ISOGEN, Nippon Gene). Samples were ground in a microcentrifuge tube. cDNA was generated using a first-strand cDNA synthesis kit (Pharmacia Biotech). Total RNA (1-5 μg) was transcribed. PCR was carried out in a 100 mL reaction mixture containing 1 to 7 μL cDNA template, 200 mmol/L deoxynucleotide triphosphates, 0.5 mmol/L of each oligonucleotide primer, and 2.5 units Taq polymerase in a 10 mmol/L Tris-HCl (pH 8.8) containing 50 mmol/L KCl and 1.5 mmol/L MgCl₂. The oligonucleotide primers used for the PCR were ESBP-1 (EWS specific), ESBP-2 (FLI-1 specific), and primers specific for ERG, E1AF, and ETV1 (32). PCR was done in 35 cycles under the following protocol: denaturation at 94°C for 1 min, annealing at 65°C for 1 min, and elongation at 72°C for 1 min. The amplified products were visualized on 1% agarose gels.

Protein expression profiling. Frozen samples were crushed to powder with a Multi-beads shocker (Yasui Kikai) with liquid nitrogen. The frozen powder was then treated with urea lysis buffer (6 mol/L urea, 2 mol/L thiourea, 3% CHAPS, 1% Triton X-100). After centrifugation at 15,000 rpm for 30 min, the supernatant was recovered and used in the subsequent protein expression studies.

Two-dimensional difference gel electrophoresis was done as described previously (33). In brief, the internal control sample was prepared by mixing a portion of all individual samples. Five micrograms of the internal control sample and of each individual sample were labeled with Cy3 and Cy5, respectively (CyDye DIGE Fluor saturation dye; GE Healthcare Biosciences) according to the manufacturer's instructions. The differently labeled protein samples were mixed and separated by two-dimensional difference gel electrophoresis. The first dimension separation was achieved using IPG DryStrip gels (24 cm length, pI range between 4 and 7; GE Healthcare Biosciences). The second dimension separation was achieved by SDS-PAGE on large-format gels (38 cm length, Bio-craft, Itabashi; ref. 33). The gels were scanned using laser scanners (Typhoon Trio; GE Healthcare Biosciences) at appropriate wavelengths (Fig. 1A). For all spots, the intensity of the Cy5 image was normalized by that of the Cy3 image in the identical gel so that gel-to-gel differences were compensated using the Progenesis PG240 software (Nonlinear Dynamics). System reproducibility was verified by comparing the protein profiles obtained from three independent separations of the same sample (Table 1, case 1). Scatter plot analysis revealed that the standardized intensity of >96.6% of the spots ranged within a 2-fold difference (R = 0.9103; Fig. 1B).

Protein identification by mass spectrometry. The proteins corresponding to the spots detected were identified using mass spectrometry according to our previous report (33). In brief, 100 μg Cy5- or Cy3-labeled proteins were separated by two-dimensional PAGE, recovered as gel plugs, and digested with modified trypsin (Promega). The trypsin digests were subjected to liquid chromatography (Paradigm MS4 dual solvent delivery system; Michrom BioResources) and mass spectrometry using a Finnigan LTQ linear ion trap mass spectrometer (Thermo Electron) equipped with a nano-electrospray ion source (AMR, Megro). The Mascot software (version 2.1; Matrix Science) was used to search for the mass of the peptide ion peaks against the SWISS-PROT database (Homo sapiens, 12867 sequence in Sprot_47.8 fasta file).

Functional classification of the identified proteins. Functional classification of the identified proteins was carried out according to their classification in Gene Ontology.⁶

Pathway analysis. Pathway analysis of the identified proteins was done using the MetaCore software analysis tool (GeneGo). MetaCore identifies networks based on a manually curated database containing known molecular interactions, functions, and disease interrelationships using proteome data. The pathways were identified by the probability that a random set of proteins with the same size as the input list would give rise to a particular mapping by chance. The identified networks were traced using the Metacore pre-filter tool. The Disease tab tool was used to automatically trace key proteins associated with disease networks stored in Metacore and to list the P value for each disease listed.

Immunohistochemical study. Nucleophosmin expression was examined immunohistochemically on paraffin-embedded tissues. In brief, 4-μm-thick tissue sections were autoclaved in 10 mmol/L citrate buffer (pH 6.0) at 121°C for 30 min and incubated with an antibody against nucleophosmin (sc-53175; Santa Cruz Biotechnology; 1:500 dilution) at room temperature. Immunostaining was done using the Envision Plus detection system (DAKO). Two observers (N.T. and K.K.) evaluated the staining in a blinded fashion for clinical data.

Statistical analysis. Hierarchical clustering was done using the Expressionist software (Genedata).

Statistical computations were done using the StatView version 5.0 statistical package (SAS Institute). Survival time was defined as the period from diagnosis to last follow-up (or death). Survival rate was estimated by the Kaplan-Meier method (34). The relationship between survival and other variables was investigated using the log-rank test for categorical variables and a score test based on the Cox proportional hazards model for continuous variables. A multivariate model was fitted using Cox regression with significant variables at the univariate level (P < 0.01; ref. 35). Following a large-scale cooperative study by the Japanese Musculoskeletal Oncology Group (36), in which age <16 years and tumor size <10 cm were shown to have a significantly worse clinical outcome by univariate survival analysis, we selected these cutoff values for this analysis.

We generated and compared the protein expression profiles between three good prognosis and five poor prognosis Ewing's sarcoma cases using two-dimensional difference gel electrophoresis. We detected 2,364 protein spots that appeared in all the images of the Cy3-labeled internal control sample. Among these 2,364 spots, 66 showed significantly (>2-fold ratio of means) different intensity between the two groups. The localization of the 66 spots on the two-dimensional image is shown in Fig. 1C. Using hierarchical clustering, the 66 spots were classified into two major groups, cluster A (7 spots) and cluster B (59 spots; Fig. 2). The intensity of the 7 spots belonging to cluster A was decreased, whereas that of the 59 spots of cluster B was increased in the poor prognosis group.

Mass spectrometric analysis resulted in the identification of 53 distinct gene products (6 proteins in cluster A and 47 proteins in cluster B) corresponding to the 66 protein spots (Fig. 2; Supplementary Table S1).

Although all six proteins in cluster A belonged to different functional categories as classified in Gene Ontology, most of the proteins in cluster B were divided into eight main categories: cytoskeletal/structural protein, transcription/translation, signal transduction, transport, antiapoptosis, response oxidative stress, acute-phase response, and cell proliferation (Fig. 2).

We further explored the biological significance of the altered protein expression pattern in cluster B based on a manually curated database containing known molecular interactions, functions, and disease interrelationships using MetaCore. We found that 31 (65.9%) of 47 proteins were functionally linked with each other and that the identified network of the 31 proteins was significantly related to bone tissue neoplasms (P < 0.000001; Fig. 2).

In the 31 proteins of cluster B, nucleophosmin was included. Although aberrant expression of nucleophosmin has been implicated in various other malignancies (25–29), its association with Ewing's sarcoma has not been reported previously. Therefore, we validated the correlation of nucleophosmin with prognosis using immunohistochemistry in an additional 34 Ewing's sarcoma cases. Two patterns of nucleophosmin-positive staining were observed, both nuclear: a dot-like pattern and a diffuse-like pattern (Fig. 3). Similar to previous reports (25, 37), cases with nuclear staining for nucleophosmin were considered as nucleophosmin positive (23 of 34 cases; Table 2), whereas cases without staining for nucleophosmin were considered as nucleophosmin negative.

In the follow-up period (median, 57.5 months; range, 8-179 months), 13 of 34 patients were alive and CDF and 21 patients were DOD (Fig. 4A).

The group of nucleophosmin-positive cases included a significantly higher number of DOD patients compared with the nucleophosmin-negative group (P < 0.01, log-rank test; Fig. 4B), showing that nucleophosmin expression correlates with prognosis.

Following univariate analysis, nucleophosmin positivity (P < 0.01) and clinical stage (presence of metastatic disease at diagnosis; P < 0.01) significantly correlated with shorter overall survival. No other factors examined, including the tumor size, age at diagnosis, sex, chemotherapy regimens, tumor resectibility, and primary site, were associated with overall survival (Table 3).

We investigated whether nucleophosmin expression significantly correlated with the original tumor site and the tumor respectability status. Of the 34 cases in this study, 21 originated at an extra-axial and 13 at an axial site. Nucleophosmin expression did not correlate with the original tumor site (P = 0.87, Fisher's test). Of the 21 extra-axial tumors, 14 were nucleophosmin positive, and nucleophosmin expression correlated with poor prognosis (P < 0.05, log-rank test). Of the 13 axial tumors, 9 were nucleophosmin negative; again, nucleophosmin expression significantly correlated with poor prognosis (P < 0.01, log-rank test). Therefore, nucleophosmin expression correlated with poor prognosis independent of the original tumor site.

There was no significant difference regarding the prognosis between the resectable (26 of 34) and nonresectable cases (P = 0.1219, log-rank test). Among the resectable tumors, 16 were nucleophosmin positive and had worse prognosis than the remaining 10 nucleophosmin-negative tumors (P < 0.01, log-rank test).

Multivariate analysis done on nucleophosmin staining and clinical stage, identified as significant prognostic factors by univariate analysis, revealed that both significantly correlated, as separate variables, with overall survival (P = 0.0063; relative risk, 7.768; 95% confidence interval, 1.783-33.841 and P = 0.0039; relative risk, 5.964; 95% confidence interval, 1.773-20.060, respectively; Table 3).

On a second univariate analysis, we found that nucleophosmin positivity was also a strong negative predictor of overall survival in the 29 of 34 patients that had localized disease at diagnosis (P < 0.01; log-rank test; Fig. 4C).

The identification of novel prognostic biomarkers for Ewing's sarcoma is required to improve the management of Ewing's sarcoma. Global genomic and transcriptomic expression studies conducted to identify prognostic biomarkers for Ewing's sarcoma resulted in the identification of MTA1, CDH11 (14), STEAP1, NKX2-2, and CCND1 (15), gains in chromosomes 1q, 8, and 12 and deletions of 1p as genetic lesions implicated in the progression of Ewing's sarcoma (16–18). Although these comprehensive studies may have the potential to further increase our understanding of the biology of Ewing's sarcoma and to lead to the development of practical tumor markers to support individualized therapy, practical prognostic biomarkers of Ewing's sarcoma are presently not used in a clinical setting.

Proteomic studies have unique advantages on other omics studies. The proteome is a functional translation of the genome, directly regulating cell phenotypes, and is thus a rich source of biomarkers. With this notion, we have established the gel-based proteomics system for cancer research and applied it to the Ewing's sarcoma proteomic study presented here. This is the first report using a proteomic approach to develop prognostic biomarkers for Ewing's sarcoma.

We identified 6 down-regulated and 47 up-regulated proteins in Ewing's sarcoma cases with poor prognosis. Functional classification revealed that these identified proteins belonged to a variety of functional pathways, including cytoskeletal/structural organization, transcription/translation, signal transduction, transport, antiapoptosis, response-oxidative stress, acute-phase response, and cell proliferation. The results of functional classification may suggest that the proteomic alterations observed may be a part of global series of functionally interconnected molecular lesions that include transcriptional and translational aberrations, which, taken together, include both the causes and the results of carcinogenesis and cancer progression.

This proteome study identified or confirmed the presence of several molecular aberrations concerning Ewing's sarcoma. The 47 proteins found to be up-regulated included neuron-specific enolase, which has been found to be associated with poor prognosis in Ewing's sarcoma (38, 39).

Nucleophosmin was included in the 47 proteins found to be up-regulated. Nucleophosmin overexpression has been related to carcinogenesis and tumor progression in prostate (25), gastric (26), colon (27), ovarian (28), and urinary bladder carcinomas (29). However, the association of nucleophosmin with Ewing's sarcoma has not been reported previously, including, importantly, in previous genomic and transcriptomic studies of Ewing's sarcoma (14–19). This may be due to discordance between mRNA and protein expression, the fact that a different patient population was studied, or, finally, the fact that transcriptome and proteome studies cannot uncover entire genome data. These results, therefore, also suggest that studies using proteomic tools are able to reveal unique molecular aspects of Ewing's sarcoma.

Network analysis showed that nucleophosmin is linked with four proteins (c-myc, nuclear factor-κB, Sp1 and p53), which have also been found to be implicated in poor prognosis in Ewing's sarcoma (Supplementary Fig. S1). c-myc has been identified as a potential EWS-ets target gene (40) and as promoting malignant progression of Ewing's sarcoma (41). Activation of nuclear factor-κB was found to contribute to resistance of Ewing's sarcoma cells to apoptosis (42). Coexpression of Sp1 with EWS-ets oncoprotein enhances activation of vascular endothelial growth factor, the expression of which was shown to be a negative predictor of survival in Ewing's sarcoma (43). Aberrations in p53 were found in ∼10% of Ewing's sarcoma cases and were associated with shorter survival (44, 45). Taken together, these observations suggest that nucleophosmin can be a single biomarker probably by linking these functionally different proteins. Ewing's sarcoma is characterized by a translocation between the EWS gene and a member of the ETS family of transcriptional factors (46). The EWS/ETS fusion protein has an altered transcriptional activity and modulates the expression of several downstream target genes (47, 48). The association between the EWS/ETS fusion protein and nucleophosmin should be further investigated.

Nucleophosmin may be used as a novel prognostic biomarker of patients with Ewing's sarcoma. We found that nucleophosmin expression correlated with clinical outcome in 34 Ewing's sarcoma patients. Univariate and multivariate analyses revealed that nucleophosmin expression along with clinical stage (presence of metastases at diagnosis) was an independent prognostic factor in Ewing's sarcoma patients. Furthermore, nucleophosmin expression was also a significant prognostic factor in patients with localized disease. Applying these findings in a clinical setting poses the next challenge. As the incisional biopsy is a procedure done routinely in establishing the diagnosis in Ewing's sarcoma, the immunohistochemical examination of nucleophosmin expression can be done without any additional invasive examinations.

Although previous reports have suggested a possible association between nucleophosmin and malignancies, the functional role of nucleophosmin in Ewing's sarcoma is still unclear. Nucleophosmin overexpression has been reported to be involved in human tumorigenesis (49, 50). In one study, it led to increased proliferation and inhibition of apoptosis in tumor cells; overexpression of nucleophosmin reduced the percentage of cells in the G₁ phase and increased the S-phase population in the p53-negative cells but induced cell cycle arrest in normal cells. Conducting further basic research on the function of nucleophosmin will pave the way for further understanding of the molecular background of Ewing's sarcoma and, hopefully, for novel diagnostic and therapeutic applications.

In conclusion, global protein expression profiling revealed the proteomic background of Ewing's sarcoma and identified novel associations of several proteins with progression of Ewing's sarcoma. Of the proteins with expression that may have prognostic value, we successfully validated the association of nucleophosmin expression with poor prognosis. The expression of the other proteins may still have prognostic value and further validation studies may prove it. Evaluation of nucleophosmin expression may allow the identification of poor prognosis Ewing's sarcoma patients who may benefit from highly effective treatment in the future.

No potential conflicts of interest were disclosed.

We thank Chizu Kina and Sachiko Miura for excellent technical support in the immunohistochemical study and Yukiko Fujie for excellent technical support in electrophoresis.