Cancer Research Annual Meeting 2010 Sign up for Cancer Research eTOC's
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
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplementary Data
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Man, T.-K.
Right arrow Articles by Lau, C. C.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Man, T.-K.
Right arrow Articles by Lau, C. C.
[Cancer Research 65, 8142-8150, September 15, 2005]
© 2005 American Association for Cancer Research

Molecular Biology, Pathobiology and Genetics

Expression Profiles of Osteosarcoma That Can Predict Response to Chemotherapy

Tsz-Kwong Man1, Murali Chintagumpala1, Jaya Visvanathan1, Jianhe Shen1, Laszlo Perlaky1, John Hicks2, Mark Johnson3, Nelson Davino3, Jeffrey Murray4, Lee Helman5, William Meyer6, Timothy Triche7, Kwong-Kwok Wong1 and Ching C. Lau1

1 Departments of Pediatrics, Texas Children's Cancer Center; Departments of 2 Pathology and 3 Orthopedic Surgery, Texas Children's Hospital/Baylor College of Medicine, Houston, Texas; 4 Cook Children's Medical Center, Fort Worth, Texas; 5 Pediatric Oncology Branch, National Cancer Institute, Bethesda, Maryland; 6 University of Oklahoma Health Sciences Center, Oklahoma City, Oklahoma; and 7 Children's Hospital Los Angeles, Los Angeles, California

Requests for reprints: Ching C. Lau, Texas Children's Hospital, 6621 Fannin Street, MC 3-3320, Houston, TX 77030-2399. Phone: 832-824-4543; Fax: 832-825-4038; E-mail: cclau{at}txccc.org.


   Abstract
Top
 Abstract
Introduction
 Materials and Methods
 Results and Discussion
 References
 
Osteosarcoma is the most common malignant bone tumor in children. After initial diagnosis is made with a biopsy, treatment consists of preoperative chemotherapy followed by definitive surgery and postoperative chemotherapy. The degree of tumor necrosis in response to preoperative chemotherapy is a reliable prognostic factor and is used to guide the choice of postoperative chemotherapy. Patients with tumors, which reveal ≥90% necrosis (good responders), have a much better prognosis than those with <90% necrosis (poor responders). Despite previous attempts to improve the outcome of poor responders by modifying the postoperative chemotherapy, their prognosis remains poor. Therefore, there is a need to predict at the time of diagnosis patients' response to preoperative chemotherapy. This will provide the basis for developing potentially effective therapy that can be given at the outset for those who are likely to have a poor response. Here, we report the analysis of 34 pediatric osteosarcoma samples by expression profiling. Using parametric two-sample t test, we identified 45 genes that discriminate between good and poor responders (P < 0.005) in 20 definitive surgery samples. A support vector machine classifier was built using these predictor genes and was tested for its ability to classify initial biopsy samples. Five of six initial biopsy samples that had corresponding definitive surgery samples in the training set were classified correctly (83%; confidence interval, 36%, 100%). When this classifier was used to predict eight independent initial biopsy samples, there was 100% accuracy (confidence interval, 63%, 100%). Many of the predictor genes are implicated in bone development, drug resistance, and tumorigenesis.


   Introduction
Top
 Abstract
 Introduction
Materials and Methods
 Results and Discussion
 References
 
Osteosarcoma is the most common malignant bone tumor in children and accounts for ~60% of malignant bone tumors diagnosed in the first two decades of life (1). After the diagnosis is made by an initial biopsy, standard treatment involves the use of multiagent chemotherapy, definitive surgery of the primary tumor, and postoperative chemotherapy. At the time of definitive surgery, the resected tumor specimen is assessed for the degree of necrosis, which is a reliable and the only significant prognostic factor in patients with nonmetastatic disease and is used to guide the choice of postoperative chemotherapy. Patients whose tumors display ≥90% necrosis (good or favorable response) have an excellent prognosis and continue to receive chemotherapy similar to the preoperative regimen. Patients whose tumors display <90% necrosis (poor or unfavorable response) have a much higher risk of relapse and poor outcome even after complete resection of the primary tumor (2). To improve the outcome of the poor responders, attempts are usually made to use postoperative chemotherapy regimens that are different from the preoperative regimen by the addition or replacement of a chemotherapeutic agent. Such attempts in the past have been unsuccessful (1, 3) partly because the degree of necrosis is known only after 8 to 10 weeks of preoperative therapy. It is possible that resistant tumor cells have additional time to either metastasize to the lungs or evolve further during the period when ineffective preoperative chemotherapy is given. Therefore, there is a need to identify at the time of initial diagnosis the patients who are likely to have a poor response to standard preoperative therapy and therefore a poor outcome eventually. Therapies tailored to improve the outcome for those patients identified at the time of diagnosis to have a poor outcome can then be instituted at the outset when the chance for success is potentially higher. Although several other prognostic factors have been proposed for predicting the long-term outcome of osteosarcoma patients, most are still controversial or have not been tested in large prospective studies (4–11).

Recently, application of microarray technology to classify and diagnose various types of tumors has yielded promising results (12, 13). However, the use of this technology to predict response to chemotherapy in pediatric solid tumors is still in its infancy. In this study, we developed a multigene predictive model to classify good and poor responders of osteosarcoma in response to preoperative chemotherapy using gene expression profiling. We used a slightly different approach from previously published works (14). We first identified a molecular signature of chemoresistance by comparing the expression profiles of the definitive surgery samples of the good responders with those of the poor responders, which, in principle, have been enriched for resistant cells. We then tested the hypothesis that this predictor signature of chemoresistance could recognize the resistant cells present in the initial biopsy of some of the same cases used in the first analysis (definitive surgery samples), although these resistant cells might have constituted only a small fraction of the primary tumor. Finally, we tested the ability of this signature to predict chemoresistance in an independent set of initial biopsies and found that there was 100% accuracy in its prediction.


   Materials and Methods
Top
 Abstract
 Introduction
 Materials and Methods
Results and Discussion
 References
 
Patients and tumor samples. The clinical information of the osteosarcoma samples used in this study is summarized in Table 1. All samples were collected through institutional review board–approved protocols at four centers (Texas Children's Hospital/Baylor College of Medicine, Cook Children's Medical Center, Pediatric Branch of the National Cancer Institute, and University of Oklahoma Health Sciences Center) after informed consents were signed. All but three patients (10, 20, and 26) were enrolled in the same treatment protocol, the schema of which is shown in Fig. 1. All patients, including the three not enrolled in the protocol, received the same preoperative chemotherapy consisting of cisplatin, doxorubicin, and high-dose methotrexate, except patients 3, 4, 8, and 24, who received only cisplatin and doxorubicin. All pathologic diagnoses were centrally reviewed by a single pathologist (J.H.). A total of 34 samples (14 initial biopsies and 20 definitive surgery specimens) were included in this study, which were obtained from 28 individual patients, 18 males and 10 females. The age of the patients ranged from 9 to 22 years. Six patients contributed two samples each, both initial biopsies and definitive surgery specimens, whereas the remaining 22 patients contributed one sample each, either an initial biopsy or a definitive surgery specimen. The initial biopsy samples were obtained at the time of diagnosis before the initiation of preoperative chemotherapy. The definitive surgery samples were collected after the completion of preoperative chemotherapy. The good responders were defined as patients whose tumors had ≥90% necrosis in response to preoperative chemotherapy as determined by histologic examination at the time of definitive surgery and poor responders had <90% necrosis. In this study, the percentage necrosis in the poor responders ranged from 5% to 86%. Five of the 28 patients were diagnosed with metastatic disease at presentation. Immediately after collection, tumor specimens were snap frozen in liquid nitrogen and stored at –80°C until RNA extraction. All samples used for RNA extraction were immediately adjacent to the frozen sections used for diagnostic purpose and were representative of the corresponding tumors. All initial biopsy specimens were confirmed to contain >80% tumor cells.


View this table:
[in this window]
[in a new window]

  		Table 1. Clinical information of the osteosarcoma samples used in this study

 


View larger version (12K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1. Schema of the protocol for this study. Dexrazoxane is given as an adjuvant with doxorubicin for cardioprotection. In this protocol, all patients with localized disease and good response to preoperative chemotherapy will receive the same chemotherapeutic agents in postoperative chemotherapy. Poor-risk patients, including the poor responders and those that have unresectable primary tumors or metastatic disease, will receive the high-dose cytoxan and melphalan with peripheral stem cell rescue as postoperative therapy. CDDP, cisplatin; DOX, doxorubicin; MTX, methotrexate.

 
RNA extraction. Total RNA was extracted from tissues and cultured cells using TRIzol reagent (Invitrogen, Carlsbad, CA) following the manufacturer's protocol. Osteosarcoma samples were homogenized in TRIzol reagent using 1 mL per 50 to 100 mg tissue. Normal human osteoblast primary culture cells obtained from Clonetics (San Diego, CA) were used as reference for microarray experiments and were lysed in 1 mL TRIzol reagent per 5 x 106 to 10 x 106 cells. RNA concentrations were determined by spectrophotometry, and RNA integrity was assessed by either agarose gel electrophoresis or Bioanalyzer (Agilent Technologies, Palo Alto, CA).

T7 RNA amplification. RNA amplification of total RNA was done based on the modified Eberwine protocol (15). For first-strand cDNA synthesis, 1 µg total RNA from each sample was mixed with T7-oligo(dT) primer (5'-GCATTAGCGGCCGCGAAATTAATACGACTCACTATAGGGAGATTTTTTTTTTTTTTTTTTVN-3'), where V and N represent mixtures of G/C/A and G/C/A/T, respectively. The RNA/primer mixture was denatured followed by first-strand synthesis using Moloney murine leukemia virus (Invitrogen) at 37°C for 75 minutes. Second-strand synthesis was done using Escherichia coli DNA ligase, E. coli RNase H, and E. coli DNA polymerase I (Invitrogen). The mixture was incubated at 16°C for 2 hours and then at 70°C for 15 minutes. cDNA was purified using ChromaSpin TE-200 spin columns (BD Biosciences, Palo Alto, CA), dried, and resuspended in 8 µL DEPC water. In vitro transcription of double-stranded cDNA was done using the Ampliscribe kit (Epicentre Technologies Corp., Madison, WI) according to the manufacturer's instructions. Amplified RNA was purified using RNeasy Mini kit (Qiagen, Valencia, CA) and quantified by fluorescence spectroscopy using a RiboGreen RNA quantitation assay (Molecular Probes, Eugene, OR).

Microarray experiments. Labeling of T7-amplified RNA samples was carried out using an amino-allyl labeling kit (Ambion, Inc., Austin, TX). In brief, 0.5 µg tumor or reference (normal human osteoblasts) amplified RNA was mixed with spiked control RNA (see below; Perkin-Elmer, Boston, MA) and random decamers. The mixture was denatured and reverse transcribed with amino-allyl dUTP. Cy3 or Cy5 monoreactive dye (Amersham Pharmacia Biotech, Arlington Heights, IL) was then added for labeling. Labeled tumor and reference cDNAs were combined and purified by NucAway spin columns (BD Biosciences), dried, and redissolved in hybridization solution (240 µL H2O, 250 µL of 20x SSC, 10 µL of 10% SDS, 500 µL formamide per 1 mL solution) and combined with human Cot-1 DNA (10 mg/mL) and yeast tRNA (4 mg/mL). Probes were denatured and hybridized to cDNA microarrays that had been prehybridized in buffer containing 5x SSC, 0.1% SDS, and 1% bovine serum albumin at 42°C for 45 minutes. After 16 to 24 hours of hybridization in a humidified chamber, slides were washed in 0.2% SDS plus 1x SSC at 42°C for 4 minutes, 0.2% SDS plus 0.1x SSC at room temperature for 4 minutes, and 0.06x SSC at room temperature for 4 minutes. Slides were then scanned using a ScanArray 4.0 scanner (Packard Bioscience, Meriden, CT) using both Cy3 and Cy5 channels and quantified using the ScanArray Express software.

The cDNA microarrays used in this study were fabricated with the Easy-to-Spot Human UniGene 1 PCR products (Incyte Genomics, Palo Alto, CA) that consist of a total of 9,216 unique elements. The PCR products were printed in duplicate along with three Arabidopsis cDNA for detecting spike-in controls using an OmniGrid Accent Arrayer (GeneMachines, San Carlos, CA) equipped with 24 SMP3 pins (TeleChem International, Sunnyvale, CA). All PCR products were resequenced before printing to validate their annotation.

Quantitative real-time PCR. Total RNA samples were pretreated with DNase for real-time PCR analysis. Synthesis of cDNA was carried out using random hexamers and SuperScript reverse transcriptase II (Invitrogen). Quantitative real-time PCR was then carried out using the ABI Prism 7000 Sequence Detection System (Applied Biosystems, Foster City, CA) following the manufacturer's protocol and using gene-specific primers. All log ratio values were corrected for rRNA and referenced to normal human osteoblast cells using {Delta}{Delta}CT method (Applied Biosystems). The Pearson and Spearman correlation coefficients were calculated using SPSS statistical software (SPSS, Inc., Chicago, IL).

Data analysis. Raw quantification output of all array experiments were subjected to data analysis using BRB Array Tools 3.1.0 developed by R. Simon and A. Pang Lam (http://linus.nci.nih.gov/BRB-ArrayTools.html). Fluorescence intensities of the arrays were preprocessed and filtered by removing spots with low signal intensity and low sample variance (P > 0.01) as well as those that were missing in >50% of the experiments. (For details of sample variance calculations, refer to http://linus.nci.nih.gov/BRB-ArrayTools.html.) Intensities were then normalized by intensity-dependent local weighted regression (LOWESS) method. This method has been shown to perform better than other methods for normalizing cDNA microarrays (16). After normalization, intensity ratios were log transformed before any further analysis.

Slightly >3,000 informative spots (3,018) remained after preprocessing and filtering and were used for supervised classification. The classification algorithms tested include compound covariate predictors (CCP; ref. 17), k-nearest neighbor, nearest centroid, support vector machine (SVM), and diagonal linear discriminant analysis (LDA; refs. 18, 19). These algorithms explore different aspects of the data to perform classification. For example, CCP and LDA use linear combinations of different weighted expression ratios for classification, whereas k-nearest neighbor and nearest centroid are nonlinear and nonparametric methods (17, 18). SVM finds the optimal hyperplane that is able to separate the data by projecting them into a high-dimensional space (19). The feature extraction (predictor gene selection) was done using two-sample t test and cutoff of P ≤ 0.005. We divided our 34 samples into two sets: a training set, which contained 20 definitive surgery samples from different patients, and a validation set, which contained 14 initial biopsy samples. Six patients contributed paired initial biopsy and definitive surgery samples (see Table 1). The leave-one-out cross-validation (LOOCV) was done to test the robustness of our classifiers using the training set. To prevent overly optimistic estimation of prediction error, honest assessment of the classification error was carried out by repeating the feature selection for each step of the cross-validation (17, 18, 20). After the training and LOOCV analysis, we picked the best classifier, the SVM classifier, for validation. A two-step validation procedure was carried out to test the ability of the SVM classifier to predict chemoresistance at the time of diagnosis. This was first done using six initial biopsy samples that have corresponding definitive surgery samples in the original training set (paired samples). The final validation was done by using the SVM classifier to predict the response of eight independent initial biopsy samples that neither have been seen by the SVM classifier nor have their corresponding definitive surgery samples been used in the training set (independent samples). Gene ontology analysis of the predictor genes was done using FatiGO (http://fatigo.bioinfo.cnio.es; ref. 21). Gene symbols of the 45 predictor genes were used to create the gene ontology plot at level 4 of biological process. Kaplan-Meier analysis of the prognostic significance of each gene was computed using the initial biopsy samples of the nonmetastatic cases (n = 12). Overall survival was compared between the high expression group (n = 6) and the low expression group (n = 6) using the median expression of the gene as cutoff. The analysis was done using SPSS and the significance was calculated using log-rank test.


   Results and Discussion
Top
 Abstract
 Introduction
 Materials and Methods
 Results and Discussion
References
 
Classifying good and poor responders based on definitive surgery specimens. With the goal to identify molecular signatures that can predict chemoresistance of osteosarcoma, we first attempted to establish the expression profiles of resistant versus sensitive osteosarcoma cells. We hypothesized that the definitive surgery samples from the poor responders should be enriched for resistant tumor cells. Using expression profiles from definitive surgery samples would therefore enhance the sensitivity and power to detect the difference between chemosensitive and resistant cell populations compared with using initial biopsy samples in which resistant cells may only be present as a small fraction. Therefore, we predicted that using definitive surgery samples would increase the chance of identifying a molecular signature of chemoresistance and that this signature could be used to identify the good and poor responders in the initial biopsy samples. We further hypothesized that the nonspecific perturbations in the expression profiles caused by preoperative chemotherapy would affect both good and poor responder samples and therefore would not contribute to the determinants of the chemoresistance signature. To test these hypotheses, we examined if we could classify good and poor responders using definitive surgery specimens only. We divided the definitive surgery samples from 20 patients into two groups: good (n = 7) and poor (n = 13) responders. We first identified a set of 45 predictor genes that could discriminate the two classes (good and poor responders) in the definitive surgery samples using two-sample t test with a significant cutoff (P = 0.005). Figure 2 shows the relative expressions of these 45 predictor genes in good and poor responders. Most of these genes (91%) were overexpressed in poor responder specimens.



View larger version (21K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2. The 45 predictor genes in the chemoresistance signature were selected based on two-sample t test to distinguish between good and poor responders in 20 definitive surgery samples (P = 0.005). Forty-one genes were overexpressed in poor responders, whereas only 4 genes were overexpressed in good responders. Two genes do not have symbols yet in the public databases, and their accession numbers are used instead. Bottom, the color scale represents log2 expression ratios of the genes.

 
Various supervised classification algorithms (CCP, k-nearest neighbor, nearest centroid, SVM, and LDA) were then applied to the training set to test if they could classify good and poor responders using P of 0.005. LOOCV method was used to test the robustness of each classifier in the training set (Table 2). Instead of using resubstitution estimation, which underestimates the classification error, we used a honest estimation method (external LOOCV) for the classification error (see Materials and Methods). The correct classification rates of LOOCV using these algorithms were 65% to 70% (Table 2). Among the six algorithms, SVM had one of the best performances (70% correct classification). Three good responders and three poor responders were misclassified using SVM. Two of the good responders (300 and 394) that were misclassified as poor responders by the SVM classifier developed recurrent disease 11 and 9 months, respectively, after completion of therapy. This suggests that there may be some residual resistant cells in the definitive surgery specimens of these two cases that were recognized by the algorithm based on the predictor gene set. One of the poor responders (680) that was misclassified as a good responders by the multigene predictor remains free of disease after 30 months of follow-up. The other misclassified poor responder (761) had 86% necrosis, which is very close to our cutoff for good response.


View this table:
[in this window]
[in a new window]

  		Table 2. LOOCV of 20 definitive surgery osteosarcoma samples

 
SVM is a powerful pattern recognition technique with the ability to build robust predictive models when the number of training samples is small but the dimensionality is high. SVM has many appealing properties for classification of microarray data, including measures to prevent overfitting and local minima that are associated with other classification algorithms. Recently, it has been used successfully to classify microarray data (22–24). Therefore, although the LOOCV showed only moderate success in classifying the training set samples, we decided to use the SVM classifier to classify initial biopsy samples. Furthermore, LOOCV is of low bias but high variance and tends to select for more conservative models (18). Applying the SVM classifier to an independent test set would be a better test of its accuracy.

Use of multigene classifier to predict response to preoperative chemotherapy in initial biopsy. To test the SVM classifier, we divided our 14 initial biopsy samples into two groups. The first group consisted of six samples, which had corresponding definitive surgery samples included in the training set (paired samples). Using these six cases, we attempted to verify that our classifier built from definitive surgery samples could predict the chemoresistance of the corresponding initial biopsy samples based on the hypothesis that the molecular signature of chemoresistance as recognized in definitive surgery samples was already present in the initial biopsy at the time of diagnosis. The second group consisted of eight initial biopsy samples that did not have matched definitive surgery samples included in the training set, thus representing a totally independent set of samples that had not been used in building the classifier.

The SVM classifier misclassified one sample (of six) in the first group of paired samples, with a correct classification rate of 83% (confidence interval, 36%, 100%; Table 3; see Supplementary Table S1 for details of prediction by all algorithms). The only misclassified sample was from a patient (410) who was classified as a good responder based on histologic response but was predicted to be a poor responder by the multigene classifier. Interestingly, this patient initially presented with localized disease but eventually developed recurrent disease in the lungs 25 months after completion of therapy, suggesting that there were resistant cells present in the initial biopsy that were recognized by the multigene classifier; presumably, these resistant cells metastasized to the lungs before definitive surgery and subsequently gave rise to the recurrent tumor. Ironically, the multigene predictor classified this patient's definitive surgery sample (452) as good responder (Table 2), implying that either the definitive surgery sample used in our analysis was not representative of the primary tumor in that it did not contain the resistant cells or the resistant cells had already metastasized before definitive surgery and therefore were no longer detectable in the primary tumor.


View this table:
[in this window]
[in a new window]

  		Table 3. Classification of initial biopsy samples (paired and independent) using SVM classifier (see Table 2 legend)

 
In the second group of independent initial biopsy samples, the classifier correctly predicted eight of eight of the samples (100% correct; confidence interval, 63%, 100%). These eight samples included five poor responders and three good responders. These results further indicate that the gene expression signature of the resistant cells in the definitive surgery samples was already present in the initial biopsy samples at the time of diagnosis. Our result is consistent with the notion proposed by Ramaswamy et al. (25) that the metastatic signature of metastatic tumors is already present in the primary tumor. The high accuracy of our multigene classifier to identify good and poor responders from two separate groups of initial biopsy samples suggests that response to chemotherapy can potentially be predicted at the time of diagnosis. This can significantly affect the design of future therapeutic studies of osteosarcoma, in which intensified therapy could be given at the time of diagnosis to those patients who are predicted to be poor responders to standard therapy to improve their outcome.

Predictor genes. Gene ontology analysis of the predictor genes suggest that most of them can be grouped into cell growth and/or maintenance (38.24% of total predictor genes that have gene ontology annotation at level 4 of biological processes); nucleobase, nucleoside, nucleotide, and nucleic acid metabolism (32.25%), macromolecule metabolism (29.41%); response to stress (14.71%); regulation of metabolism (14.71%); and signal transduction (14.71%; Supplementary Fig. S1). In particular, ubiquitin-mediated proteolysis and cell cycle regulation are two major pathways that some of the predictor genes are grouped under. More detailed analysis revealed that many of the predictor genes have interesting properties that are related to bone development, cancer biology, and drug resistance (see Table 4 for the detailed information of the 45 predictor genes). For instance, TWIST1, which encodes a helix-loop-helix transcriptional factor, has been implicated in Saethre-Chotzen syndrome, radial aplasia, Robinow-Sorauf syndrome, and craniosynostosis (26–29). Mice with heterozygous Twist1 mutation showed defects in craniofacial and limb development (30), which resemble those found in Saethre-Chotzen syndrome. Although the role of TWIST1 in bone development is still under study, it has been reported that TWIST1 affects CBFA1/RUNX2 expression (31) and DNA-binding ability of RUNX2 (32), an important regulator of osteoblast differentiation and proliferation (33). Thus, dysregulation of TWIST1 expression may play a role in the pathogenesis of osteosarcoma. In addition, several studies indicated that TWIST1 functions as a potential oncogene, cooperating with MYC and MYCN in suppressing p53-dependent apoptosis pathways (34, 35). Recent findings also showed that TWIST1 was involved in Taxol resistance and metastasis, further implicating the important role of this gene in chemoresistance and tumor invasion in osteosarcoma (36, 37). It is also interesting to note that, in addition to having antiapoptotic properties, overexpression of TWIST1 is associated with characteristics of osteoprogenitor cells that have decreased proliferation and less mature phenotype (38). All of these are predicted properties of cancer stem cells (39), suggesting that the chemoresistant cells that show overexpression of TWIST1 may arise from cancer stem cells.


View this table:
[in this window]
[in a new window]

  		Table 4. Detailed information of 45 predictor genes in the chemoresistance prediction model

 
TMPO belongs to a group of ubiquitously expressed nuclear proteins that play an important role in nuclear envelope organization and cell cycle control. It is up-regulated in medulloblastoma when compared with normal cerebellum (40). One form of the TMPO transcripts, TMPO-ß, is highly expressed in human neuroblastoma cell lines compared with normal nerve tissue (41). UBE2D2 encodes a member of the E2 ubiquitin-conjugating enzyme family also known as UBCH5B or UBC4. Inhibition of UBE2D2 inhibits human papillomavirus E6–stimulated p53 degradation (42). MCM2 is part of the minichromosome maintenance (MCM) complex that is involved in initiation of eukaryotic genome replication. It regulates the helicase activity of the MCM complex. High expression of MCM2 is associated with poor survival in prostate cancer (43).

Six of the predictor genes (AMPD2, CDC20, NDUFS5, SDHB, CDC2L2, and DDAH1) are located on chromosome 1p, which is a recurrent region of chromosomal gain in osteosarcoma as reported by our group and others (44–46). The occurrence frequency of chromosome 1p in the predictor list (14%) is almost thrice higher than the expected frequency of 1p in the 3,018 filtered genes (5%). The relationship between DNA copy number changes and gene expression levels as well as their correlation with clinical outcome is currently under investigation.

Quantitative reverse transcription-PCR validation. To validate the microarray results, transcript levels of seven predictor genes were measured by quantitative reverse transcription-PCR using unamplified RNA from seven samples that had sufficient quantities of RNA. These seven samples were made up of both initial biopsy and definitive surgery samples from good and poor responders (Table 5). The seven predictor genes were selected randomly and exhibited various degrees of statistical significance. All seven genes tested showed positive correlation between microarray and quantitative reverse transcription-PCR results. Two of the genes, PDCD5 and TWIST1, also showed significant difference using Pearson correlation (P < 0.05). Although validation results of in-house cDNA microarrays usually have less correlation with quantitative reverse transcription-PCR when compared with other commercial microarray platforms (47), our validation results are comparable with or better than other published studies (48, 49).


View this table:
[in this window]
[in a new window]

  		Table 5. Validation of array expression data by quantitative real time-PCR

 
Predictor genes and overall survival. Because histologic response to preoperative chemotherapy is only a surrogate marker of survival, one way to cross-validate the chemoresistance signature is to test if the predictor genes in the resistance signature are also associated with the long-term outcome of osteosarcoma patients. Ultimately, we would like to directly identify prognostic markers that can predict long-term outcome of osteosarcoma patients. We did Kaplan-Meier analysis on all 45 predictor genes individually using only the initial biopsy samples from the nonmetastatic cases in our prediction set. Despite the small sample size (n = 12), three of the genes (HSPA4, PIGF, and LMO6) were found to be statistically significant in the prediction of survival (log-rank test, P < 0.05) and one gene (TREM2) showed a strong correlation (P = 0.0784; data not shown). Interestingly, the expression pattern of the first three genes confirms the prediction that high expression of these genes in poor responders will correlate with poorer prognosis. TREM2 is the only gene that shows correlation between low expression level and poor prognosis, which also matches our predictor model (i.e., lower expression of TREM2 in poor responders). A recent article reported that down-regulation of TREM2 may affect differentiation of osteoclasts, leading to impaired bone resorption (50). Additionally, HSPA4 has been found to be overexpressed in human hepatocellular carcinoma and showed antiapoptotic activities (51). We recognize the limitations of this preliminary analysis because of small sample size, relatively short duration of follow-up in some cases, and the problem with multiple testing. Therefore, these results need to be validated in a much larger sample set with long-term follow-up data.

Ochi et al. have recently reported a similar study of predicting chemotherapy response in a few osteosarcoma patients (14). Although our predictor gene list is different from theirs, most of the predictor genes in both studies had higher expression in tumors of poor responders than those of good responders. The different age distribution of the patients (more adults in their study), the relatively smaller sample size (n = 19 in their study compared with 28 in our study), use of different chemotherapeutic agents, their use of two rounds of RNA amplification, variations in the arrays, and application of different training and classification strategies and algorithms make it difficult to compare directly the predictor gene lists in these two studies. However, we would like to point out that the high prediction accuracy of independent samples in our study (eight of eight) suggests that the strategy of using definitive surgery samples for training is at least comparable with, if not better than, using initial biopsy samples in the study by Ochi et al. This further justifies the use of definitive surgery samples in our training set.

In summary, we have developed a multigene classifier that could predict at the time of diagnosis the response of osteosarcoma to preoperative chemotherapy. This chemoresistance signature can potentially be used to stratify patients to different preoperative chemotherapy regimens in future clinical trials, which could ultimately affect long-term survival. We also identified a set of predictor genes that could be used to predict clinical outcome. The encouraging results of this study warrant validation of the predictor gene set with a larger number of patients and eventually in a prospective trial to evaluate its utility in risk stratification.


   Acknowledgments
 
Grant support: NIH grant CA88126, Gillson Longenbaugh Foundation, and Cancer Fighters of Houston, Inc. (C.C. Lau).

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.

We thank Richard Simon, Lisa Wang, Alison Bertuch, Paul Meltzer, and Javed Khan for helpful discussions.


   Footnotes
 
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

Received 3/23/05. Revised 6/ 5/05. Accepted 7/ 6/05.


   References
Top
 Abstract
 Introduction
 Materials and Methods
 Results and Discussion
 References
 

  1. Link MP, Gebhardt MC, Meyers PA. In: Pizzo P, Poplack D, editors. Principles and practice of pediatric oncology. 4th ed. Philadelphia: Lippincott-Williams & Wilkins; 2002. p. 1051–89.
  2. Provisor AJ, Ettinger LJ, Nachman JB, et al. Treatment of nonmetastatic osteosarcoma of the extremity with preoperative and postoperative chemotherapy: a report from the Children's Cancer Group. J Clin Oncol 1997;15:76–84.[Abstract/Free Full Text]
  3. Meyers PA, Heller G, Healey J, et al. Chemotherapy for nonmetastatic osteogenic sarcoma: the Memorial Sloan-Kettering experience. J Clin Oncol 1992;10:5–15.
  4. Baldini N, Scotlandi K, Barbanti-Brodano G, et al. Expression of P-glycoprotein in high-grade osteosarcomas in relation to clinical outcome. N Engl J Med 1995;333:1380–5.[Abstract/Free Full Text]
  5. Bacci G, Longhi A, Ferrari S, et al. Prognostic significance of serum alkaline phosphatase in osteosarcoma of the extremity treated with neoadjuvant chemotherapy: recent experience at Rizzoli Institute. Oncol Rep 2002;9:171–5.[Medline]
  6. Gorlick R, Huvos AG, Heller G, et al. Expression of HER2/erbB-2 correlates with survival in osteosarcoma. J Clin Oncol 1999;17:2781–8.[Abstract/Free Full Text]
  7. Bielack SS, Kempf-Bielack B, Delling G, et al. Prognostic factors in high-grade osteosarcoma of the extremities or trunk: an analysis of 1,702 patients treated on neoadjuvant cooperative osteosarcoma study group protocols. J Clin Oncol 2002;20:776–90.[Abstract/Free Full Text]
  8. Davis AM, Bell RS, Goodwin PJ. Prognostic factors in osteosarcoma: a critical review. J Clin Oncol 1994;12:423–31.[Abstract]
  9. Feugeas O, Guriec N, Babin-Boilletot A, et al. Loss of heterozygosity of the RB gene is a poor prognostic factor in patients with osteosarcoma. J Clin Oncol 1996;14:467–72.[Abstract/Free Full Text]
  10. Franzius C, Bielack S, Flege S, et al. Prognostic significance of (18)F-FDG and (99m)Tc-methylene diphosphonate uptake in primary osteosarcoma. J Nucl Med 2002;43:1012–7.[Abstract/Free Full Text]
  11. Ulaner GA, Huang HY, Otero J, et al. Absence of a telomere maintenance mechanism as a favorable prognostic factor in patients with osteosarcoma. Cancer Res 2003;63:1759–63.[Abstract/Free Full Text]
  12. Liu ET. Classification of cancers by expression profiling. Curr Opin Genet Dev 2003;13:97–103.[CrossRef][Medline]
  13. Ramaswamy S, Golub TR. DNA microarrays in clinical oncology. J Clin Oncol 2002;20:1932–41.[Abstract/Free Full Text]
  14. Ochi K, Daigo Y, Katagiri T, et al. Prediction of response to neoadjuvant chemotherapy for osteosarcoma by gene-expression profiles. Int J Oncol 2004;24:647–55.[Medline]
  15. Van Gelder RN, von Zastrow ME, Yool A, et al. Amplified RNA synthesized from limited quantities of heterogeneous cDNA. Proc Natl Acad Sci U S A 1990;87:1663–7.[Abstract/Free Full Text]
  16. Park T, Yi SG, Kang SH, et al. Evaluation of normalization methods for microarray data. BMC Bioinformatics 2003;4:33.[CrossRef][Medline]
  17. Radmacher MD, McShane LM, Simon R. A paradigm for class prediction using gene expression profiles. J Comput Biol 2002;9:505–11.[CrossRef][Medline]
  18. Dudoit S, Friedman N. Classification in microarray experiments. In: Speed T, editor. Statistical analysis of gene expression microarray data. Boca Raton: Chapman & Hall/CRC; 2003. p. 93–158.
  19. Vapnik V, Chapelle O. Bounds on error expectation for support vector machines. Neural Comput 2000;12:2013–36.[CrossRef][Medline]
  20. Simon R, Radmacher MD, Dobbin K, McShane LM. Pitfalls in the use of DNA microarray data for diagnostic and prognostic classification. J Natl Cancer Inst 2003;95:14–8.[Free Full Text]
  21. Al Shahrour F, Diaz-Uriarte R, Dopazo J. FatiGO: a web tool for finding significant associations of Gene Ontology terms with groups of genes. Bioinformatics 2004;20:578–80.[Abstract/Free Full Text]
  22. Zhao Y, Pinilla C, Valmori D, Martin R, Simon R. Application of support vector machines for T-cell epitopes prediction. Bioinformatics 2003;19:1978–84.[Abstract/Free Full Text]
  23. Peng S, Xu Q, Ling XB, et al. Molecular classification of cancer types from microarray data using the combination of genetic algorithms and support vector machines. FEBS Lett 2003;555:358–62.[CrossRef][Medline]
  24. Furey TS, Cristianini N, Duffy N, et al. Support vector machine classification and validation of cancer tissue samples using microarray expression data. Bioinformatics 2000;16:906–14.[Abstract/Free Full Text]
  25. Ramaswamy S, Ross KN, Lander ES, Golub TR. A molecular signature of metastasis in primary solid tumors. Nat Genet 2003;33:49–54.[CrossRef][Medline]
  26. Howard TD, Paznekas WA, Green ED, et al. Mutations in TWIST, a basic helix-loop-helix transcription factor, in Saethre-Chotzen syndrome. Nat Genet 1997;15:36–41.[CrossRef][Medline]
  27. Yousfi M, Lasmoles F, El G, V, Marie PJ. Twist haploinsufficiency in Saethre-Chotzen syndrome induces calvarial osteoblast apoptosis due to increased TNF{alpha} expression and caspase-2 activation. Hum Mol Genet 2002;11:359–69.[Abstract/Free Full Text]
  28. Gripp KW, Stolle CA, Celle L, et al. TWIST gene mutation in a patient with radial aplasia and craniosynostosis: further evidence for heterogeneity of Baller-Gerold syndrome. Am J Med Genet 1999;82:170–6.[CrossRef][Medline]
  29. Kunz J, Hudler M, Fritz B. Identification of a frameshift mutation in the gene TWIST in a family affected with Robinow-Sorauf syndrome. J Med Genet 1999;36:650–2.[Free Full Text]
  30. Bourgeois P, Bolcato-Bellemin AL, Danse JM, et al. The variable expressivity and incomplete penetrance of the twist-null heterozygous mouse phenotype resemble those of human Saethre-Chotzen syndrome. Hum Mol Genet 1998;7:945–57.[Abstract/Free Full Text]
  31. Yousfi M, Lasmoles F, Marie PJ. TWIST inactivation reduces CBFA1/RUNX2 expression and DNA binding to the osteocalcin promoter in osteoblasts. Biochem Biophys Res Commun 2002;297:641–4.[CrossRef][Medline]
  32. Bialek P, Kern B, Yang X, et al. A twist code determines the onset of osteoblast differentiation. Dev Cell 2004;6:423–35.[CrossRef][Medline]
  33. Lian JB, Stein GS. Runx2/Cbfa1: a multifunctional regulator of bone formation. Curr Pharm Des 2003;9:2677–85.[CrossRef][Medline]
  34. Maestro R, Dei Tos AP, Hamamori Y, et al. Twist is a potential oncogene that inhibits apoptosis. Genes Dev 1999;13:2207–17.[Abstract/Free Full Text]
  35. Valsesia-Wittmann S, Magdeleine M, Dupasquier S, et al. Oncogenic cooperation between H-Twist and N-Myc overrides failsafe programs in cancer cells. Cancer Cell 2004;6:625–30.[CrossRef][Medline]
  36. Wang X, Ling MT, Guan XY, et al. Identification of a novel function of TWIST, a bHLH protein, in the development of acquired taxol resistance in human cancer cells. Oncogene 2004;23:474–82.[CrossRef][Medline]
  37. Yang J, Mani SA, Donaher JL, et al. Twist, a master regulator of morphogenesis, plays an essential role in tumor metastasis. Cell 2004;117:927–39.[CrossRef][Medline]
  38. Lee M-S, Lowe GN, Strong DD, et al. TWIST, a basic helix-loop-helix transcription factor, can regulate the human osteogenic lineage. J Cell Biochem 1999;75:566–77.[CrossRef][Medline]
  39. Dean M, Fojo T, Bates S. Tumor stem cells and drug resistance. Nat Rev Cancer 2005;5:275–84.[CrossRef][Medline]
  40. Yokota N, Mainprize TG, Taylor MD, et al. Identification of differentially expressed and developmentally regulated genes in medulloblastoma using suppression subtraction hybridization. Oncogene 2004;23:3444–53.[CrossRef][Medline]
  41. Weber PJ, Eckhard CP, Gonser S, et al. On the role of thymopoietins in cell proliferation. Immunochemical evidence for new members of the human thymopoietin family. Biol Chem 1999;380:653–60.[CrossRef][Medline]
  42. Rolfe M, Beer-Romero P, Glass S, et al. Reconstitution of p53-ubiquitinylation reactions from purified components: the role of human ubiquitin-conjugating enzyme UBC4 and E6-associated protein (E6AP). Proc Natl Acad Sci U S A 1995;92:3264–8.[Abstract/Free Full Text]
  43. Meng MV, Grossfeld GD, Williams GH, et al. Minichromosome maintenance protein 2 expression in prostate: characterization and association with outcome after therapy for cancer. Clin Cancer Res 2001;7:2712–8.[Abstract/Free Full Text]
  44. Lau CC, Harris CP, Lu XY, et al. Frequent amplification and rearrangement of chromosomal bands 6p12-p21 and 17p11.2 in osteosarcoma. Genes Chromosomes Cancer 2004;39:11–21.[CrossRef][Medline]
  45. Simons A, Janssen IM, Suijkerbuijk RF, et al. Isolation of osteosarcoma-associated amplified DNA sequences using representational difference analysis. Genes Chromosomes Cancer 1997;20:196–200.[CrossRef][Medline]
  46. Zielenska M, Bayani J, Pandita A, et al. Comparative genomic hybridization analysis identifies gains of 1p35 approximately p36 and chromosome 19 in osteosarcoma. Cancer Genet Cytogenet 2001;130:14–21.[CrossRef][Medline]
  47. Jarvinen AK, Hautaniemi S, Edgren H, et al. Are data from different gene expression microarray platforms comparable? Genomics 2004;83:1164–8.[CrossRef][Medline]
  48. Chang JC, Wooten EC, Tsimelzon A, et al. Gene expression profiling for the prediction of therapeutic response to docetaxel in patients with breast cancer. Lancet 2003;362:362–9.[CrossRef][Medline]
  49. Kihara C, Tsunoda T, Tanaka T, et al. Prediction of sensitivity of esophageal tumors to adjuvant chemotherapy by cDNA microarray analysis of gene-expression profiles. Cancer Res 2001;61:6474–9.[Abstract/Free Full Text]
  50. Paloneva J, Mandelin J, Kiialainen A, et al. DAP12/TREM2 deficiency results in impaired osteoclast differentiation and osteoporotic features. J Exp Med 2003;198:669–75.[Abstract/Free Full Text]
  51. Gotoh K, Nonoguchi K, Higashitsuji H, et al. Apg-2 has a chaperone-like activity similar to Hsp110 and is overexpressed in hepatocellular carcinomas. FEBS Lett 2004;560:19–24.[CrossRef][Medline]



This article has been cited by other articles:


Home page
 Clin. Cancer Res.Home page
S. Bruheim, Y. Xi, J. Ju, and O. Fodstad
Gene Expression Profiles Classify Human Osteosarcoma Xenografts According to Sensitivity to Doxorubicin, Cisplatin, and Ifosfamide
Clin. Cancer Res., December 1, 2009; 15(23): 7161 - 7169.
[Abstract] [Full Text] [PDF]

Home page
 JBJSHome page
T. A. Guise, R. O'Keefe, R. L. Randall, and R. M. Terek
Molecular Biology and Therapeutics in Musculoskeletal Oncology
J. Bone Joint Surg. Am., March 1, 2009; 91(3): 724 - 732.
[Full Text] [PDF]

Home page
 CarcinogenesisHome page
W. K. Kwok, M.-T. Ling, H. F. Yuen, Y.-C. Wong, and X. Wang
Role of p14ARF in TWIST-mediated senescence in prostate epithelial cells
Carcinogenesis, December 1, 2007; 28(12): 2467 - 2475.
[Abstract] [Full Text] [PDF]

Home page
 Clin. Cancer Res.Home page
A. Srivastava, B. Fuchs, K. Zhang, M. Ruan, C. Halder, E. Mahlum, K. Weber, M. E. Bolander, and G. Sarkar
High WT1 Expression Is Associated with Very Poor Survival of Patients with Osteogenic Sarcoma Metastasis.
Clin. Cancer Res., July 15, 2006; 12(14): 4237 - 4243.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplementary Data
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Man, T.-K.
Right arrow Articles by Lau, C. C.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Man, T.-K.
Right arrow Articles by Lau, C. C.

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