Clinical Cancer Research CTRC-AACR San Antonio Breast Cancer Symposium Infection and Cancer: Biology, Therapeutics, and Prevention
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 Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
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 Takata, R.
Right arrow Articles by Nakamura, Y.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Takata, R.
Right arrow Articles by Nakamura, Y.
Clinical Cancer Research Vol. 11, 2625-2636, April 2005
© 2005 American Association for Cancer Research

Cancer Therapy: Preclinical

Predicting Response to Methotrexate, Vinblastine, Doxorubicin, and Cisplatin Neoadjuvant Chemotherapy for Bladder Cancers through Genome-Wide Gene Expression Profiling

Ryo Takata1,7, Toyomasa Katagiri1, Mitsugu Kanehira1,7, Tatsuhiko Tsunoda2, Taro Shuin3, Tsuneharu Miki4, Mikio Namiki5, Kenjiro Kohri6, Yasushi Matsushita7, Tomoaki Fujioka7 and Yusuke Nakamura1

Authors' Affiliations: 1 Laboratory of Molecular Medicine, Human Genome Center, Institute of Medical Science, The University of Tokyo, Tokyo, Japan; 2 SNP Research Center, RIKEN (Institute of Physical and Chemical Research), Yokohama, Japan; 3 Department of Urology, Kochi Medical School, Nankoku, Japan; 4 Department of Urology, Kyoto Prefectural University of Medicine, Kyoto, Japan; 5 Department of Urology, Kanazawa University Graduate School of Medicine, Kanazawa, Japan; 6 Department of Nephro-urology, Nagoya City University Graduate School of Medical Sciences, Nagoya, Japan; and 7 Department of Urology, Iwate Medical University, Morioka, Japan

Requests for reprints: Yusuke Nakamura, Laboratory of Molecular Medicine, Human Genome Center, Institute of Medical Science, The University of Tokyo, 4-6-1 Shirokanedai, Minato-ku, Tokyo 108-8639, Japan. Phone: 81-3-5449-5372; Fax: 81-3-5449-5433; E-mail: yusuke{at}ims.u-tokyo.ac.jp.


   Abstract
Top
 Abstract
Materials and Methods
 Results
 Discussion
 References
 
Purpose: Neoadjuvant chemotherapy for invasive bladder cancer, involving a regimen of methotrexate, vinblastine, doxorubicin, and cisplatin (M-VAC), can improve the resectability of larger neoplasms for some patients and offer a better prognosis. However, some suffer severe adverse drug reactions without any effect, and no method yet exists for predicting the response of an individual patient to chemotherapy. Our purpose in this study is to establish a method for predicting response to the M-VAC therapy.

Experimental Design: We analyzed gene expression profiles of biopsy materials from 27 invasive bladder cancers using a cDNA microarray consisting of 27,648 genes, after populations of cancer cells had been purified by laser microbeam microdissection.

Results: We identified dozens of genes that were expressed differently between nine "responder" and nine "nonresponder" tumors; from that list we selected the 14 "predictive" genes that showed the most significant differences and devised a numerical prediction scoring system that clearly separated the responder group from the nonresponder group. This system accurately predicted the drug responses of 8 of 9 test cases that were reserved from the original 27 cases. Because real-time reverse transcription–PCR data were highly concordant with the cDNA microarray data for those 14 genes, we developed a quantitative reverse transcription–PCR–based prediction system that could be feasible for routine clinical use.

Conclusions: Our results suggest that the sensitivity of an invasive bladder cancer to the M-VAC neoadjuvant chemotherapy can be predicted by expression patterns in this set of genes, a step toward achievement of "personalized therapy" for treatment of this disease.

Key Words: prediction system • cDNA microarray • quantitative RT-PCR • personalized therapy

Bladder cancer is the second most common genitourinary tumor in human populations, having an incidence of ~261,000 new cases each year worldwide; about a third of those are likely to be invasive or metastatic disease at the time of diagnosis (1). Although radical cystectomy is considered the gold standard for treatment of patients with localized but muscle-invasive bladder cancer, about 50% of such patients develop metastases within 2 years after cystectomy and subsequently die of the disease (2).

Neoadjuvant chemotherapy is usually prescribed for muscle-invasive bladder cancer to treat micrometastases and to improve resectability of larger neoplasms (3, 4). Regimens involving methotrexate, vinblastine, doxorubicin, and cisplatin (M-VAC) followed by radical cystectomy are more likely to eliminate residual cancer than radical cystectomy alone and improve survival among patients with locally advanced bladder cancer (5, 6). In some clinical trials, downstaging with drugs before surgery was shown to have significant survival benefits (6, 7); moreover, patients who respond to neoadjuvant chemotherapy may preserve bladder function and enjoy an improved quality of life.

However, because no method yet exists for predicting the response of an individual patient to chemotherapies such as M-VAC, some patients will suffer from adverse reactions to the drugs without achieving any benefit in terms of any positive effect, often losing opportunity for additional therapy when their physical condition deteriorates. Hence, development of an accurate method to predict the effectiveness of a specific therapy is of critical importance for patients with bladder cancer. Certain factors are known to be associated with chemosensitivity or prognosis, but information concerning only one or a few of those factors have failed thus far to reliably predict individual responses; clearly, a larger body of information is required.

Profiling of gene expression patterns on genome-wide cDNA microarrays enables investigators to perform comprehensive analyses of unusual molecular activities in cancer cells. Systematic analysis of expression levels among thousands of genes is also a useful approach for identifying molecules related to response to anticancer drugs or radiation. We have been attempting to establish systems based on gene expression profiling that would allow accurate prediction of responses to chemotherapeutic agents in diseases such as acute myeloid leukemia and chronic myeloid leukemia (8, 9).

In the study reported here, we established a system for predicting response to M-VAC neoadjuvant chemotherapy among patients with invasive bladder cancer, using genome-wide information obtained for 27 cases on a cDNA microarray consisting of 27,648 transcribed elements in combination with laser microbeam microdissection of the tumors to obtain pure populations of cancer cells for analysis. We identified 14 genes that showed significantly different levels of expression between responders and nonresponders among patients with bladder cancer treated with a neoadjuvant M-VAC regimen. We suggest that such information may lead ultimately to our goal of "personalized therapy".


   Materials and Methods
Top
 Abstract
 Materials and Methods
Results
 Discussion
 References
 
Patients, tissue samples, and neoadjuvant chemotherapy. Tissue samples from surgically resected bladder cancers and corresponding clinical information were obtained from five hospitals (Kochi Medical School, Kyoto Prefectural University of Medicine, Nagoya City University Graduate School of Medical Sciences, Kanazawa University Graduate School of Medical Sciences, and Iwate Medical University) after each patient had provided written informed consent. A total of 27 cancer samples (7 women and 20 men; median age, 66; range, 53-77 years; Table 1) that had been confirmed histologically as transitional cell carcinoma of the bladder were selected for this study. Clinical stage of each patient was judged according to the Union International Contre Cancer tumor-node-metastasis classification; we enrolled only patients who had no node metastasis at the clinical stage of T2aN0M0 to T3bN0M0 and were expected to undergo radical cystectomy without prior radiation therapy. Participants were required to have no serious abnormality in renal, hepatic, or hematologic function, with Eastern Cooperative Oncology Group performance status (PS) judged to be ≤2.


View this table:
[in this window]
[in a new window]
  		Table 1. Clinicopathologic features of patients with bladder cancer examined

 
Three to five pieces of cancer tissue had been taken from each patient at the time of biopsy before neoadjuvant chemotherapy. These samples were immediately embedded in TissueTek OCT compound (Sakura, Tokyo, Japan), frozen, and stored at –80°C. The frozen tissues were sliced into 8-µm sections using a cryostat (Sakura) and then stained with H&E for histologic examination. Bladder cancer cells were selectively enriched for our experiments using the EZ-cut system with a pulsed UV narrow beam focus laser (SL Microtest GmbH, Germany) according to the manufacturer's protocols.

All patients were examined by chest X-ray, computed tomography, and magnetic resonance imaging of their abdomen and pelvis, and confirmed to have neither lymph node nor distant metastases. Patients were given two 28-day cycles of M-VAC neoadjuvant chemotherapy as follows: methotrexate (30 mg per square meter of body surface area) on days 1, 15, and 22; vinblastine (3 mg per square meter) on days 2, 15, and 22; and doxorubicin (30 mg per square meter) and cisplatin (70 mg per square meter) on day 2.

Extraction of RNA and T7-based RNA amplification. Total RNAs were extracted from each population of microdissected cancer cells, as described previously (10). To guarantee the quality of RNAs, total RNA extracted from the residual tissue of each case were electrophoresed on a denaturing agarose gel, and quality was confirmed by the presence of rRNA bands. Extraction of total RNA and T7-based RNA amplification were done as described previously (11), except that we used RNeasy Micro Kits (Qiagen, Valencia, CA). After two rounds of RNA amplification, we obtained 30 to 100 µg of amplified RNA for each sample. As a control, normal human bladder polyadenylated RNA (BD Biosciences, Palo Alto, CA), was amplified in the same way. RNA amplified by this method accurately reflects the proportions in the original RNA source, as we had confirmed earlier by semiquantitative reverse transcription–PCR (RT-PCR) experiments (10), in which data from the microarrays were consistent with results from RT-PCR regardless of whether total RNAs or amplified RNAs were used as templates.

cDNA microarray. To obtain cDNAs for spotting on the glass slides, we did RT-PCR for each gene, as described previously (10). The PCR products were spotted on type VII glass slides (GE Healthcare, Amersham Biosciences, Buckinghamshire, United Kingdom) with a high-density Microarray Spotter Lucidea (GE Healthcare, Amersham Biosciences); 9,216 genes were spotted in duplicate on a single slide. We prepared three different sets of slides (a total of 27,648 gene spots), on each of which the same 52 housekeeping genes and two negative control genes were spotted as well. The cDNA probes were prepared from amplified RNA in the manner described previously (11). For hybridization experiments, 9.0 µg of amplified RNAs from each cancerous tissue and from the control were reverse transcribed in the presence of Cy5-dCTP and Cy3-dCTP (GE Healthcare, Amersham Biosciences), respectively. Hybridization, washing, and detection of signals were carried out as described previously (11).

Quantification of signals. We quantified the signal intensities of Cy3 and Cy5 from the 27,648 spots and analyzed the signals by substituting backgrounds, using ArrayVision software (Imaging Research, Inc., St. Catharines, Ontario, Canada). Subsequently, the fluorescence intensities of Cy5 (tumor) and Cy3 (control) for each target spot were adjusted so that the mean Cy5/Cy3 ratio of the 52 housekeeping genes became one. Because data derived from low signal intensities are less reliable, we determined a cutoff value on each slide as described previously (12) and excluded genes from further analysis when both Cy3 and Cy5 dyes yielded signal intensities lower than the cutoff (13). For other genes we modified our previous method that calculated Cy5/Cy3 as a relative expression ratio using the raw data of each sample, because if either Cy3 or Cy5 signal intensity was lower than the cutoff value the Cy5/Cy3 ratio might provide an extremely high or low reading and lead to selection of false-prediction genes. To reduce that bias, if either Cy3 or Cy5 signal intensity was less than the cutoff value we calculated the Cy5/Cy3 ratios using half of each cutoff value plus the Cy5 and Cy3 signal intensities of each sample.

Identification of discriminating genes for chemosensitivity. We applied a random permutation test to identify genes that were expressed at a significantly different level between the two groups, that is, tumors with good response and those with poor response to the chemotherapy. Mean (µ) and standard deviation ({sigma}) were calculated from the log-transformed relative expression ratios of each gene in responder (r) and nonresponder (n) cases. A discrimination score (DS) for each gene was defined as follows:

We carried out permutation tests to estimate the ability of individual genes to distinguish between responders and nonresponders; samples were randomly permutated between the two groups 10,000 times. Because the DS data set of each gene showed a normal distribution, we calculated a P value for the user-defined grouping (14). For the initial analysis, we applied the expression data for 18 cases, 9 responders and 9 nonresponders that were obtained at an earlier stage of the study.

Calculation of prediction score. We calculated prediction scores according to procedures described previously (14). Each gene (gi) votes for either responder or nonresponder depending on whether the expression level (xi) in the sample is closer to the mean expression level of responders or nonresponders in reference samples. The magnitude of the vote (Vi) reflects the deviation of the expression level in the sample from the average of the two classes:

We summed the votes to obtain total votes for the responders (Vr) and nonresponders (Vn), and calculated PS values as follows:

reflecting the margin of victory in the direction of either responder or nonresponder. PS values range from –100 to 100; a higher absolute value of PS reflects a stronger prediction.

Evaluation of classification and leave-one-out test. We calculated the classification score (CS) using prediction scores of responders (PSr) and nonresponders (PSn) in each gene set, as follows:

A larger value of CS indicates better separation of the two groups by the predictive-scoring system. For the leave-one-out test, one sample is withheld, the permutation P value and mean expression levels are calculated using remaining samples, and the class of the withheld sample is subsequently evaluated by calculating its prediction score. We repeated this procedure for each of the 18 samples.

Quantitative reverse transcription–PCR. We microdissected 5,000 to 10,000 tumor cells per sample to perform a quantitative RT-PCR analysis. Total RNA extracted from microdissected tissues was subjected to two rounds of T7-based RNA amplification followed by cDNA synthesis using 0.5 µg of random hexamer (Roche, Basel, Switzerland) and SuperScript II (Invitrogen, Carlsbad, CA) according to the supplier's protocol. Expression of 14 predictive genes and 3 endogenous control genes was measured by quantitative RT-PCR (TaqMan Gene Expression Assays products, on an ABI Prism 7700 Sequence Detection system) as described previously (15). The sequences of the primers and fluorogenic TaqMan MGB probes are shown in Table 2. To normalize the expression of each gene, we selected as internal controls chaperonin-containing TCP1, subunit 6A (CCT6A), isoleucine-tRNA synthetase (IARS), and heat shock 90-kDa protein 1, ß (HSPCB) from among the 52 housekeeping genes because they showed the smallest Cy5/Cy3 fluctuations in our microarray data. Because normalization to these three endogenous control genes (CCT6A, IARS, and HSPCB) led to similar conclusions (data not shown), we subsequently recorded only the data normalized according to levels of CCT6A expression. For generation of standard curves we used a mixture of mRNAs derived from the bladder cancer samples. Quantitative RT-PCR experiments were done in duplicate for all 14 "predictive" genes, and relative expression ratios of each sample were calculated. The normalized gene expression values were log-transformed (on a base 2 scale), in a manner similar to the transformation of microarray-based hybridization data.


View this table:
[in this window]
[in a new window]
  		Table 2. List of primer sets and TaqMan probes

 

   Results
Top
 Abstract
 Materials and Methods
 Results
Discussion
 References
 
Clinical response to neoadjuvant chemotherapy and selection of discriminating genes. We enrolled 27 patients with bladder cancer whose clinicopathologic features are summarized in Table 1. According to their responses to the treatment, we categorized the patients into two groups: "responders," patients who achieved downstaging (≤pT1 or ≤T1) after two courses of M-VAC neoadjuvant chemotherapy, and "nonresponders," who had not achieved downstaging (≥pT2 or ≥T2) after the two courses of drug treatment. Downstaging was judged according to the specimens obtained during cystectomy and/or through three-dimensional diagnostic imaging using computed tomography and magnetic resonance imaging. We compared microarray-expression profiles of tumors from nine responders and nine nonresponders as a step toward establishing a prediction scoring system for chemosensitivity to M-VAC neoadjuvant chemotherapy. First we identified genes that distinguished the two groups in accordance with two criteria: (1) signal intensities higher than the cutoff level in more than 60% of samples of at least one group; (2) Medr – Medn ≥1.0, where Med indicates the median derived from log-transformed relative expression ratios in responders or nonresponders. Then we carried out a random permutation test to select genes that might be associated with the drug response (see Materials and Methods). We identified 50 genes that showed permutation P values of <0.0001 (Table 3). As shown in Fig. 1, expression levels of 25 genes were increased and those of the remaining 25 were decreased in the nonresponder group, as compared with the responder group. Among them, RELA and TOP2A had already been implicated in chemosensitivity (16–22).


View this table:
[in this window]
[in a new window]
  		Table 3. List of 50 discriminating genes

 


View larger version (46K):
[in this window]
[in a new window]
  		Fig. 1. Expression patterns of the 50 genes that discriminated responders from nonresponders among 18 patients with bladder cancer. Horizontal rows represent individual genes; vertical columns represent individual samples. Each cell in the matrix represents the expression level of a single transcript in single sample, with red and green indicating transcript levels, respectively, above and below the median for that gene across all samples. Black represents unchanged expression or slight expression (intensities of both Cy3 and Cy5 under the cutoff value). Color saturation is proportional to the magnitude of the difference.

 
Selection of predictive genes. Using the expression profiles of the genes that seemed to distinguish the two groups, we calculated the prediction score of each sample by the weighted-vote method described previously (14). Then we rank-ordered these candidates on the basis of the significance of their permutation P values (Table 3) and calculated prediction scores by the leave-one-out test for cross-validation, using the top 7 to 30, 35, 40, 45, and 50 genes in the rank-ordered list. For the leave-one-out test, we withheld one sample and calculated the permutation P values and mean expression levels using the remaining samples to identify genes that were the most powerful for separating the responder and nonresponder groups. Some of the candidate genes, such as mitogen-activated protein kinase kinase 3 (MAP2K3) and ubiquilin 2 (UBQLN2), were lowered on the ranked ordered list (i.e., P > 0.0001; Table 3).

We calculated CS using the prediction scores of nine responders and nine nonresponders in each gene set (see Materials and Methods), and obtained the best separation of the two groups when we used the top 14 genes in our candidate list (Fig. 2A and B; Table 4). A hierarchical clustering analysis using this set of genes with Cluster and Treeview software (http://rana.lbl.gov/EisenSoftware.htm) yielded good separation of the two groups with regard to sensitivity to M-VAC neoadjuvant chemotherapy (Fig. 2C).



View larger version (17K):
[in this window]
[in a new window]
  		Fig. 2. A, optimization of the number of discriminating genes. The CS was calculated using the prediction scores of responders (PSr) and nonresponders (PSn) in each gene set, as follows: CS = {µ (PSr) – µ (PSn)} / {{sigma} (PSr) + {sigma} (PSn)}. A larger value of CS indicates better separation of the two groups by the predictive scoring system. B, prediction score for individual patients using the top 14 discriminating genes. High absolute values show high confidence. R, responders; NR, nonresponders. C, clustering analysis of 14 predictive genes. All samples fell appropriately into one of two "trees" according to their sensitivity to M-VAC neoadjuvant chemotherapy.

 

View this table:
[in this window]
[in a new window]
  		Table 4. List of the 14 predictive genes

 
Finally, to verify the prediction scoring system based on expression data for this set of 14 genes, we examined 9 "test" cases (5 responders and 4 nonresponders) that were reserved for verification from the 18 "learning" cases used for construction of the prediction system. We investigated gene expression profiles in each of the nine test cases and then calculated a prediction score for each sample. As shown in Fig. 3, eight of nine cases fell correctly into place according to their response to M-VAC. Among the 15 cases (including learning and test cases) with positive prediction scores, 14 cases showed downstaging after M-VAC therapy, whereas all the cases with negative prediction scores showed poor or no tumor response.



View larger version (24K):
[in this window]
[in a new window]
  		Fig. 3. Distribution of prediction scores for 27 patients. Blue squares and red triangles indicate scores in cross-validation cases of patients whose expression data were used for selecting discriminating genes (learning). Light blue squares and light crimson triangles represent scores for nine additional cases (test). High absolute values show high confidence.

 
Confirmation of expression measurements. To confirm the results of cDNA microarray analysis, we carried out real-time quantitative RT-PCR for the 14 predictive genes and the 3 quantitative-control genes, CCT6A, IARS, and HSPCB (see Materials and Methods), using 15 of the 18 original (learning) cases. We observed significant correlation between results from the cDNA microarray and those of the quantitative RT-PCR experiments. As shown in Table 5, Pearson and Spearman rank correlations were positive for all 14 genes and statistically significantly for 12 of them. In particular, RASL11B (RAS-like, family 11, member B) was significantly more highly expressed in responders than in nonresponders (Fig. 4A). On the other hand, expression of PHKA2 (phosphorylase kinase, {alpha} 2) was significantly lower in the responders than in nonresponders (Fig. 4C). The quantitative RT-PCR data were significantly concordant with microarray expression data for both genes (RASL11B: r = 0.90, P < 0.01; PHKA2: r = 0.92, P < 0.01; Fig. 4B and D).


View this table:
[in this window]
[in a new window]
  		Table 5. Correlation of microarray expression data with quantitative–PCR derived values

 


View larger version (29K):
[in this window]
[in a new window]
  		Fig. 4. Comparison of microarray-expression data for two genes with quantitative RT-PCR data. A, microarray and quantitative RT-PCR data for RASL11B. Black bars in the histogram represent expression of responders; white bars represent expression of nonresponders. B, correlation between microarray data for RASL11B and values derived from quantitative RT-PCR experiments. C, microarray and quantitative RT-PCR data for PHKA2. D, correlation between microarray data for PHKA2 and values derived from quantitative RT-PCR experiments.

 
Establishment of a quantitative reverse transcription–PCR-based prediction scoring system. To examine the possibility of adapting our prediction system for clinical use, we attempted to establish a scoring system based on quantitative RT-PCR results. We did quantitative real-time RT-PCR of the 14 predictive genes for 15 learning cases (8 responders and 7 nonresponders from the original 18 learning cases), and calculated the prediction score for each case. When we estimated these scores by the leave-one-out cross validation test, all cases were placed correctly according to their response to M-VAC (Fig. 5A). We showed further that the response of 8 of 9 additional (test) cases (5 responders and 4 nonresponders) were also predicted with accuracy. Because two genes showed low significant correlation between microarray and quantitative RT-PCR (TOP2A and TCTA; Table 5), we examined whether the prediction scores based on the remaining 12 genes can predict the clinical responses. As shown in Fig. 5B, this scoring system was also able to predict eight of nine cases accurately, although predictive power is slightly reduced.



View larger version (16K):
[in this window]
[in a new window]
  		Fig. 5. A, prediction scores for 27 cases using values derived from quantitative RT-PCR experiments with 14 predictive genes. Blue squares and red triangles indicate scores for selecting discriminating genes (learning). Light blue squares and light crimson triangles represent scores for nine additional (test) cases. B, prediction scores with 12 predictive genes.

 

   Discussion
Top
 Abstract
 Materials and Methods
 Results
 Discussion
References
 
cDNA microarrays are now widely used to analyze expression of thousands of genes simultaneously in cancer tissues. However, in our view, adequate attention has not been paid to the quality of the materials and experiments. For example, clinical samples (surgically resected tissue or biopsy materials) usually consist of various cellular components, and the proportions of cancer cells in a given tissue can vary enormously from one tumor to another (from 10% to nearly 90%). Hence, most microarray data published previously are likely to be influenced significantly by heterogeneity among tumor preparations. To obtain cancer cell–specific expression data, we applied a laser microbeam microdissection system to purify as much as possible the populations of cancer cells from biopsy specimens of 27 invasive bladder cancers, with a view to establishing a scoring system to predict response to M-VAC treatment among patients with that disease.

Despite recent advances, 50% of patients with bladder cancer who receive M-VAC chemotherapy show no or very poor response in terms of staging, and a large proportion of them suffer from adverse events such as myelosuppression and/or gastrointestinal and renal toxicity (23–25). Although certain factors are known to be associated with chemosensitivity or prognosis of bladder cancer (26–30), information from only one or a few of these factors has failed thus far to reliably predict individual responses, indicating a need for a more precise method for predicting response to anticancer drugs.

This study was designed to develop a prediction system for M-VAC neoadjuvant chemotherapy on the basis of gene expression profiles of purified populations of bladder cancer cells. We identified 50 genes whose expression was significantly different between the responders and nonresponders and ranked them by statistical significance of the permutation test (P < 0.0001; Table 3). Moreover, we calculated a false discovery rate by the Benjamini-Hochberg method (31). All of the 50 candidate genes still showed strong statistical power (P < 0.0001 means false discovery rate <0.006), indicating that these genes have significant power to discriminate the two groups. Then we selected 14 genes that showed the smallest P values and showed by the leave-one-out cross validation method that scores calculated on the basis of these 14 genes provided the best separation of responders from nonresponders. Furthermore, our scoring system was able to predict accurately the response of eight of nine additional cases to M-VAC neoadjuvant chemotherapy. It is notable that the one patient (BC01027) who was given a positive score by this prediction system but was classified as nonresponder was in fact confirmed significant reduction of the tumor size through three-dimensional diagnostic imaging using computed tomography and magnetic resonance imaging after two courses of M-VAC therapy. In addition, no macroscopic observation of tumors was observed in the cystectomy specimen of this patient, but we judged this case as a failure of the scoring system because this case was not satisfied with our downstaging criteria by pathologic examination (pretreatment, T2b; posttreatment, pT2a). However, regarding the tumor size, our scoring system could predict the response to the M-VAC therapy correctly.

The list of 50 genes that showed significant differences between the two groups might provide insight into the biological mechanism(s) underlying sensitivity to M-VAC chemotherapy. Among those 50 genes, we found that topoisomerase IIa (TOP2A) was down-regulated in the nonresponder group (Fig. 1). TOP2A is an essential nuclear enzyme with a role in the maintenance of DNA topology and reportedly a target for several anticancer drugs including doxorubicin (18–22). Decreased expression of TOP2A has been observed in a variety of cell lines that were resistant to a range of chemotherapeutic drugs (18, 19, 22). TOP2A-targeting anticancer drugs stabilize the TOP2A-DNA cleavable complex, prevent annealing of DNA strands, and cause subsequent double-stranded DNA breakage that leads to cell death. Low expression of TOP2A reduces formation of the TOP2A-DNA cleavable complex that is stabilized by anticancer drugs (21). Thus, the level of TOP2A expression in a tumor could influence the extent of drug-induced DNA damage and in turn modify cytotoxic effects. Hence, down-regulation of TOP2A might contribute to resistance to M-VAC neoadjuvant chemotherapy among patients with bladder cancer.

On the other hand, in our experiments genes encoding v-rel reticuloendotheliosis viral oncogene homologue A (RELA, p65), solute carrier family 16 (monocarboxylic acid transporters), member 3 (SLC16A3) and p53-associated parkinlike cytoplasmic protein (PARK) were up-regulated in nonresponders (Fig. 1). RELA binds with NF{kappa}ß (16, 17), which is an antiapoptotic factor expressed in certain neoplastic cells in response to chemotherapy; RELA might affect resistance to M-VAC drugs through interaction with NF{kappa}ß. SLC16A3, a proton-linked monocarboxylate transporter that catalyzes rapid transport across the plasma membrane of many monocarboxylates such as lactate and pyruvate (32, 33), is often overexpressed in cancer cells. SLC16A3 is closely associated with CD147 (32, 33), a glycoprotein that is enriched in tumor cells where it can induce several genes associated with drug resistance (34, 35). Hence, up-regulation of SLC16A3 might influence resistance to M-VAC neoadjuvant chemotherapy via its association with CD147 protein. PARK, a parkinlike ubiquitin ligase, is a cytoplasmic anchor for p53-associated protein complexes (36, 37). Overexpression of PARK promotes cytoplasmic sequestration of ectopic p53 and inhibits its tumor-suppressor function (36). Thus, up-regulated expression of PARK might contribute to resistance to M-VAC neoadjuvant chemotherapy through inhibition of p53.

Recently other groups have predicted prognosis or chemosensitivity of tumors based on quantitative RT-PCR results for expression of genes selected on microarrrays (38, 39). To confirm the reliability of microarray data and open the possibility of more convenient prediction strategies for routine clinical use, we also did quantitative RT-PCR experiments using the 14 predictive genes and 15 learning cases of bladder cancer selected after microarray analysis. We confirmed significant correlation between the data obtained for those 15 paired samples on the microarray and results of quantitative RT-PCR (Table 5; Fig. 4). Moreover, we verified that our quantitative RT-PCR–based prediction system could also correctly classify eight of nine our subsequent test cases with regard to their drug responses.

Prediction power using the 14 gene set was reduced when the same gene set was applied for measurement by quantitative RT-PCR because the absolute values of the prediction scores seems to be weak. The quantitative RT-PCR results of two genes, TOP2A and TCTA, did not correlate so well with the microarray data (Fig. 5B). We suspected that the cause of this discrepancy was mainly due to differences in the measurement range of the copy number of the transcript between microarray-based and quantitative RT-PCR–based system; especially, the small copy number, particularly that below the cutoff value, by microarray measurement was inaccurate. In the microarray analysis, if both Cy3 and Cy5 signal intensities were less than the cutoff value, we calculated the log-transformed Cy5/Cy3 ratios as zero, whereas relative ratios of raw data of each sample were calculated in quantitative RT-PCR experiments. However, when we applied the quantitative RT-PCR–based scoring system using the 12 genes that showed good concordance in the two measurement methods, prediction powers were still sufficient enough to separate the two groups (Fig. 5B). Although there is a space to improve our prediction system, the concordance of the two independent systems should encourage us to establish the personalized medicine.

To evaluate the possibility of making our prediction system convenient for routine clinical use, we compared the prediction score of each sample from whole-bladder cancer tissue with that obtained with microdissected bladder cancer cells. When the population of cancer cells in whole tissues was 84.5% (BC01015, responder) or 87.6% (BC01021, responder), we found a slight difference in prediction scores between whole tissue and microdissected cells (whole tissues/microdissected cells = 53.54:58.03, and 64.44:32.76, respectively). However, for nonresponder samples BC02003 and BC02014, containing 52% and 8% cancer cells in whole tissue, respectively, the prediction scores were not so concordant with scores obtained with microdissected cells (whole tissues/microdissected cells = –12.80:–68.69 and –9.14:–41.42, respectively; Table 6). These results suggested that using whole cancer tissue blocks for calculating prediction scores should be very limited and that in general we need to purify cancer cell populations to provide better prediction for individual patients.


View this table:
[in this window]
[in a new window]
  		Table 6. Comparison of prediction scores between whole-tumor tissues and microdissected cells

 
In conclusion, we are confident that our prediction system for sensitivity of bladder cancers to M-VAC therapy, whether based on microarray-derived expression profiles or quantitative RT-PCR results, should provide opportunities for achieving better prognosis and improved quality of life for patients, although a larger scale study is certainly warranted. Our data suggest that the goal of "personalized medicine," prescribing the correct treatment regimen for each patient, may be achievable by selecting specific sets of genes for their predictive values according to the approach shown here.


   Acknowledgments
 
We thank Yuka Ishizaki for cDNA microarray and real-time RT-PCR experiments; Noriko Sudo, Saori Osawa, Sachiyo Ikeda, Miwako Ando, Keiko Shigeta, and Mieko Takahashi for fabricating the cDNA microarray; Emi Okutsu-Ichihashi and Kazuyuki Hayashi for analysis of data; Tae Makino and Noriko Ikawa for preparation of tissue sections by cryostat; and Dr. Shingo Ashida and Dr. Seiya Imoto for helpful discussions.


   Footnotes
 
Grant support: Japan Society for the Promotion of Science Research for the Future Program grant 00L01402.

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.

Received 9/27/04; revised 12/26/04; accepted 1/ 3/05.


   References
Top
 Abstract
 Materials and Methods
 Results
 Discussion
 References
 

  1. Parkin DM, Pisani P, Ferlay J. Global cancer statistics. CA Cancer J Clin 1999;49:33–64, 1.[Abstract/Free Full Text]
  2. Sternberg CN. The treatment of advanced bladder cancer. Ann Oncol 1995;6:113–26.[Free Full Text]
  3. Fagg SL, Dawson-Edwards P, Hughes MA, Latief TN, Rolfe EB, Fielding JW. Cis-diamminedichloroplatinum (DDP) as initial treatment of invasive bladder cancer. Br J Urol 1984;56:296–300.[Medline]
  4. Raghavan D, Pearson B, Coorey G, et al. Intravenous cis-platinum for invasive bladder cancer. Safety and feasibility of a new approach. Med J Aust 1984;140:276–8.[Medline]
  5. Advanced Bladder Cancer Meta-analysis Collaboration. Neoadjuvant chemotherapy in invasive bladder cancer: a systematic review and meta-analysis. Lancet 2003;361:1927–34.[CrossRef][Medline]
  6. Grossman HB, Natale RB, Tangen CM, et al. Neoadjuvant chemotherapy plus cystectomy compared with cystectomy alone for locally advanced bladder cancer. N Engl J Med 2003;349:859–66.[Abstract/Free Full Text]
  7. Splinter TA, Scher HI, Denis L, et al. The prognostic value of the pathological response to combination chemotherapy before cystectomy in patients with invasive bladder cancer. European Organization for Research on Treatment of Cancer-Genitourinary Group. J Urol 1992;147:606–8.[Medline]
  8. Okutsu J, Tsunoda T, Kaneta Y, et al. Prediction of chemosensitivity for patients with acute myeloid leukemia, according to expression levels of 28 genes selected by genome-wide complementary DNA microarray analysis. Mol Cancer Ther 2002;1:1035–42.[Abstract/Free Full Text]
  9. Kaneta Y, Kagami Y, Katagiri T, et al. Prediction of sensitivity to STI571 among chronic myeloid leukemia patients by genome-wide cDNA microarray analysis. Jpn J Cancer Res 2002;93:849–56.[CrossRef][Medline]
  10. Kitahara O, Furukawa Y, Tanaka T, et al. Alterations of gene expression during colorectal carcinogenesis revealed by cDNA microarrays after laser-capture microdissection of tumor tissues and normal epithelia. Cancer Res 2001;61:3544–9.[Abstract/Free Full Text]
  11. Okabe H, Satoh S, Kato T, et al. Genome-wide analysis of gene expression in human hepatocellular carcinomas using cDNA microarray: identification of genes involved in viral carcinogenesis and tumor progression. Cancer Res 2001;61:2129–37.[Abstract/Free Full Text]
  12. Ono K, Tanaka T, Tsunoda T, et al. Identification by cDNA microarray of genes involved in ovarian carcinogenesis. Cancer Res 2000;60:5007–11.[Abstract/Free Full Text]
  13. Saito-Hisaminato A, Katagiri T, Kakiuchi S, Nakamura T, Tsunoda T, Nakamura Y. Genome-wide profiling of gene expression in 29 normal human tissues with a cDNA microarray. DNA Res 2002;9:35–45.[Abstract]
  14. Golub TR, Slonim DK, Tamayo P, et al. Molecular classification of cancer: class discovery and class prediction by gene expression monitoring. Science 1999;286:531–7.[Abstract/Free Full Text]
  15. Yamanaka Y, Tamari M, Nakahata T, Nakamura Y. Gene expression profiles of human small airway epithelial cells treated with low doses of 14- and 16-membered macrolides. Biochem Biophys Res Commun 2001;287:198–203.[CrossRef][Medline]
  16. Perkins ND. The Rel/NF-{kappa}B family: friend and foe. Trends Biochem Sci 2000;25:434–40.[CrossRef][Medline]
  17. Wang W, Cassidy J. Constitutive nuclear factor-{kappa}B mRNA, protein overexpression and enhanced DNA-binding activity in thymidylate synthase inhibitor-resistant tumour cells. Br J Cancer 2003;88:624–9.[CrossRef][Medline]
  18. Harker WG, Slade DL, Drake FH, Parr RL. Mitoxantrone resistance in HL-60 leukemia cells: reduced nuclear topoisomerase II catalytic activity and drug-induced DNA cleavage in association with reduced expression of the topoisomerase IIß isoform. Biochemistry 1991;30:9953–61.[CrossRef][Medline]
  19. Giaccone G, van Ark-Otte J, Scagliotti G, et al. Differential expression of DNA topoisomerases in non-small cell lung cancer and normal lung. Biochim Biophys Acta 1995;1264:337–46.[Medline]
  20. Zhou Z, Zwelling LA, Kawakami Y, et al. Adenovirus-mediated human topoisomerase II{alpha} gene transfer increases the sensitivity of etoposide-resistant human breast cancer cells. Cancer Res 1999;59:4618–24.[Abstract/Free Full Text]
  21. Topcu Z. DNA topoisomerases as targets for anticancer drugs. J Clin Pharm Ther 2001;26:405–16.[CrossRef][Medline]
  22. Allen KA, Williams AO, Isaacs RJ, Stowell KM. Down-regulation of human topoisomerase II{alpha} correlates with altered expression of transcriptional regulators NF-YA and Sp1. Anticancer Drugs 2004;15:357–62.[CrossRef][Medline]
  23. Sternberg CN, Yagoda A, Scher HI, et al. Preliminary results of M-VAC (methotrexate, vinblastine, doxorubicin and cisplatin) for transitional cell carcinoma of the urothelium. J Urol 1985;133:403–7.[Medline]
  24. Sternberg CN, Yagoda A, Scher HI, et al. Methotrexate, vinblastine, doxorubicin, and cisplatin for advanced transitional cell carcinoma of the urothelium. Efficacy and patterns of response and relapse. Cancer 1989;64:2448–58.[CrossRef][Medline]
  25. Herr HW, Bajorin DF, Scher HI. Neoadjuvant chemotherapy and bladder-sparing surgery for invasive bladder cancer: ten-year outcome. J Clin Oncol 1998;16:1298–301.[Abstract/Free Full Text]
  26. Esrig D, Elmajian D, Groshen S, et al. Accumulation of nuclear p53 and tumor progression in bladder cancer. N Engl J Med 1994;331:1259–64.[Abstract/Free Full Text]
  27. Cote RJ, Dunn MD, Chatterjee SJ, et al. Elevated and absent pRb expression is associated with bladder cancer progression and has cooperative effects with p53. Cancer Res 1998;58:1090–4.[Abstract/Free Full Text]
  28. Oosterhuis JW, Schapers RF, Janssen-Heijnen ML, Smeets AW, Pauwels RP. MIB-1 as a proliferative marker in transitional cell carcinoma of the bladder: clinical significance and comparison with other prognostic factors. Cancer 2000;88:2598–605.[CrossRef][Medline]
  29. Stein JP, Ginsberg DA, Grossfeld GD, et al. Effect of p21WAF1/CIP1 expression on tumor progression in bladder cancer. J Natl Cancer Inst 1998;90:1072–9.[Abstract/Free Full Text]
  30. Inoue K, Slaton JW, Karashima T, et al. The prognostic value of angiogenesis factor expression for predicting recurrence and metastasis of bladder cancer after neoadjuvant chemotherapy and radical cystectomy. Clin Cancer Res 2000;6:4866–73.[Abstract/Free Full Text]
  31. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc B 1995;57:289–300.
  32. Halestrap AP, Price NT. The proton-linked monocarboxylate transporter (MCT) family: structure, function and regulation. Biochem J 1999;343 Pt 2:281–99.
  33. Izumi H, Torigoe T, Ishiguchi H, et al. Cellular pH regulators: potentially promising molecular targets for cancer chemotherapy. Cancer Treat Rev 2003;29:541–9.[CrossRef][Medline]
  34. Yang JM, Xu Z, Wu H, Zhu H, Wu X, Hait WN. Overexpression of extracellular matrix metalloproteinase inducer in multidrug resistant cancer cells. Mol Cancer Res 2003;1:420–7.[Abstract/Free Full Text]
  35. Misra S, Ghatak S, Zoltan-Jones A, Toole BP. Regulation of multidrug resistance in cancer cells by hyaluronan. J Biol Chem 2003;278:25285–8.[Abstract/Free Full Text]
  36. Nikolaev AY, Li M, Puskas N, Qin J, Gu W. Parc: a cytoplasmic anchor for p53. Cell 2003;112:29–40.[CrossRef][Medline]
  37. O'Brate A, Giannakakou P. The importance of p53 location: nuclear or cytoplasmic zip code? Drug Resist Updat 2003;6:313–22.[CrossRef][Medline]
  38. Lossos IS, Czerwinski DK, Alizadeh AA, et al. Prediction of survival in diffuse large-B-cell lymphoma based on the expression of six genes. N Engl J Med 2004;350:1828–37.[Abstract/Free Full Text]
  39. Ma XJ, Wang Z, Ryan PD, et al. A two-gene expression ratio predicts clinical outcome in breast cancer patients treated with tamoxifen. Cancer Cell 2004;5:607–16.[CrossRef][Medline]



This article has been cited by other articles:


Home page
 Clin. Cancer Res.Home page
A. B. Als, L. Dyrskjot, H. von der Maase, K. Koed, F. Mansilla, H. E. Toldbod, J. L. Jensen, B. P. Ulhoi, L. Sengelov, K. M.E. Jensen, et al.
Emmprin and Survivin Predict Response and Survival following Cisplatin-Containing Chemotherapy in Patients with Advanced Bladder Cancer
Clin. Cancer Res., August 1, 2007; 13(15): 4407 - 4414.
[Abstract] [Full Text] [PDF]

Home page
 Cancer Res.Home page
M. Kanehira, T. Katagiri, A. Shimo, R. Takata, T. Shuin, T. Miki, T. Fujioka, and Y. Nakamura
Oncogenic Role of MPHOSPH1, a Cancer-Testis Antigen Specific to Human Bladder Cancer
Cancer Res., April 1, 2007; 67(7): 3276 - 3285.
[Abstract] [Full Text] [PDF]

Home page
 Cancer Res.Home page
Y. Matsui, S. Ueda, J. Watanabe, I. Kuwabara, O. Ogawa, and H. Nishiyama
Sensitizing Effect of Galectin-7 in Urothelial Cancer to Cisplatin through the Accumulation of Intracellular Reactive Oxygen Species
Cancer Res., February 1, 2007; 67(3): 1212 - 1220.
[Abstract] [Full Text] [PDF]

Home page
 Molecular Cancer TherapeuticsHome page
D. M. Havaleshko, H. Cho, M. Conaway, C. R. Owens, G. Hampton, J. K. Lee, and D. Theodorescu
Prediction of drug combination chemosensitivity in human bladder cancer
Mol. Cancer Ther., February 1, 2007; 6(2): 578 - 586.
[Abstract] [Full Text] [PDF]

Home page
 Ann OncolHome page
R Sawhney, D Bourgeois, and U. Chaudhary
Neo-adjuvant chemotherapy for muscle-invasive bladder cancer: a look ahead
Ann. Onc., September 1, 2006; 17(9): 1360 - 1369.
[Abstract] [Full Text] [PDF]

Home page
 JCOHome page
D. W. Lin, I. M. Coleman, S. Hawley, C. Y. Huang, R. Dumpit, D. Gifford, P. Kezele, H. Hung, B. S. Knudsen, A. R. Kristal, et al.
Influence of Surgical Manipulation on Prostate Gene Expression: Implications for Molecular Correlates of Treatment Effects and Disease Prognosis
J. Clin. Oncol., August 10, 2006; 24(23): 3763 - 3770.
[Abstract] [Full Text] [PDF]

Home page
 Toxicol SciHome page
J. G. Affleck, K. Neumann, L. Wong, and V. K. Walker
The Effects of Methotrexate on Drosophila Development, Female Fecundity, and Gene Expression
Toxicol. Sci., February 1, 2006; 89(2): 495 - 503.
[Abstract] [Full Text] [PDF]

Home page
 JCOHome page
D. Spentzos, D. A. Levine, S. Kolia, H. Otu, J. Boyd, T. A. Libermann, and S. A. Cannistra
Unique Gene Expression Profile Based on Pathologic Response in Epithelial Ovarian Cancer
J. Clin. Oncol., November 1, 2005; 23(31): 7911 - 7918.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
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 Takata, R.
Right arrow Articles by Nakamura, Y.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Takata, R.
Right arrow Articles by Nakamura, Y.

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