Gene expression analysis may help the management of cancer patients, allowing the selection of subjects responding to treatment. The aim of this study was the characterization of expression pattern of genes involved in gemcitabine activity in pancreas tumor specimens and its correlation with treatment outcome. The role of drug transport and metabolism on gemcitabine cytotoxicity was examined with specific inhibitors, whereas transcription analysis of human equilibrative nucleoside transporter-1 (hENT1), deoxycytidine kinase (dCK), 5′-nucleotidase (5′-NT), cytidine deaminase (CDA), and ribonucleotide reductase subunits M1 and M2 (RRM1 and RRM2) was done by quantitative reverse transcription-PCR in tumor tissue isolated by laser microdissection from surgical or biopsy samples of 102 patients. Association between clinical outcome and gene expression levels was estimated using Kaplan-Meier method and Cox's proportional hazards model. Transport and metabolism had a key role on gemcitabine sensitivity in vitro; moreover, hENT1, dCK, 5′-NT, CDA, RRM1, and RRM2 were detectable in most tumor specimens. hENT1 expression was significantly correlated with clinical outcome. Patients with high levels of hENT1 had a significantly longer overall survival [median, 25.7; 95% confidence interval (95% CI), 17.6-33.7 months in the higher expression tertile versus median, 8.5; 95% CI, 7.0-9.9 months in the lower expression tertile]. Similar results were obtained with disease-free survival and time to disease progression, and the multivariate analysis confirmed the prognostic significance of hENT1. This study suggests that the expression levels of hENT1 may allow the stratification of patients based on their likelihood of survival, thus offering a potential new tool for treatment optimization. (Cancer Res 2006; 66(7): 3928-35)
- Pancreatic cancer
- human equilibrative nucleoside transporter-1
Pancreatic cancer is one of the most lethal human tumors, and although the nucleoside pyrimidine analogue gemcitabine produces a clinical benefit response, the prognosis remains dismal, with 5-year survival rate of 1% to 4% ( 1). However, chemotherapy is administered to patients without knowledge of the genetic background of the disease, which may affect drug efficacy. Genetically determined variability of key enzymes has been shown to influence response and toxicity of cytotoxic agents, including 5-fluorouracil (5-FU), irinotecan, and 6-mercaptopurine ( 2). Potential candidates to predict which patients are likely to respond to treatment are genes encoding proteins involved in metabolism, transport across membranes, or target of anticancer drugs ( 3).
Gemcitabine is transported into the cell mostly by human equilibrative nucleoside transporter-1 (hENT1; ref. 4). Cells lacking hENT1 are highly resistant to gemcitabine ( 5), and pancreas cancer patients with hENT1-positive tumor tissue have significantly longer survival after gemcitabine chemotherapy than patients affected by tumors without detectable hENT1 ( 6). As a prodrug, gemcitabine must be phosphorylated to its active diphosphate and triphosphate metabolites that, respectively, inhibit ribonucleotide reductase and DNA synthesis. Deoxycytidine kinase (dCK) is the rate-limiting enzyme in the biotransformation of nucleoside analogues and the increase in dCK activity may improve the efficacy of gemcitabine ( 7). In contrast, high expression of the catabolic enzymes 5′-nucleotidase (5′-NT) and cytidine deaminase (CDA) is associated with cellular resistance to gemcitabine ( 8, 9). In addition to being incorporated into DNA, gemcitabine exerts its cytotoxicity by inhibiting ribonucleotide reductase. Non–small cell lung cancer (NSCLC) patients with low expression of the ribonucleotide reductase regulatory subunit M1 (RRM1) significantly benefited from gemcitabine/cisplatin neoadjuvant chemotherapy ( 10), although resistance to gemcitabine was associated with both RRM1 and RRM2 overexpression ( 11, 12). In addition to this, small interfering RNA targeting the RRM2 catalytic subunit enhanced the chemosensitivity to gemcitabine of pancreatic adenocarcinoma in vitro and in vivo ( 13).
Recent data indicate that modulation of cellular enzymes of gemcitabine metabolism and uptake may influence drug activity in vitro, suggesting the need for a rational choice of combination treatments guided by gene expression analysis ( 14).
The characterization of key genes that play a critical role in tumor sensitivity or resistance to anticancer drugs represents a new avenue for optimal drug use and enhanced therapeutic success that is worth exploiting. Moreover, experimental data may improve the success of pancreatic cancer treatment either by selecting responsive patients or by modulation of drug effect with rationally selected drug combinations. For these reasons, this study addressed the transcription analysis of hENT1, dCK, 5′-NT, CDA, RRM1, and RRM2 in pancreas cancer tissues to find possible associations between gene expression and gemcitabine efficacy.
Materials and Methods
In vitro Study on Pancreas Cancer Cells
Cell cultures. Cell lines (American Type Culture Collection, Rockville, MD) were grown in DMEM with 10% fetal bovine serum (FBS) and 2.5% horse serum (MIA PaCa-2), DMEM with 20% FBS (Capan-1), and RPMI 1640 with 10% FBS (PANC-1), glutamine, and penicillin-streptomycin at 37°C in 5% CO2 and 95% air and subcultured by harvesting with trypsin-EDTA.
Effect of Inhibition of Gemcitabine Metabolism and Transport on Cytotoxicity
Cells were plated in six-well plates at 105 per well, allowed to attach for 24 hours, and treated with gemcitabine (0.01 ng/mL-10 μg/mL) for 24 hours alone or in combination with 2′-deoxycytidine (to inhibit dCK substrate competition), tetrahydrouridine (CDA competitive inhibitor), and diethylpyrocarbonate (5′-NT noncompetitive inhibitor) at 10 μmol/L. The inhibition of drug transport was obtained by the specific and the aspecific hENT1 inhibitors nitrobenzylthioinosine and dipyridamole, at concentrations of 100 nmol/L and 10 μmol/L, respectively, as described ( 6). Cytotoxicity was determined from three separate experiments and expressed as the percentage of cells surviving relative to untreated cultures; the IC50 was calculated by nonlinear least-squares curve fitting (GraphPad Prism 4.0, San Diego, CA).
Clinical Studies on Pancreas Cancer
Patient characteristics and treatment. From December 2001 to October 2004, a total of 105 patients affected by pancreatic adenocarcinoma were enrolled. Median age was 65 years (range, 22-83); 53 were males and 52 females. Two (1.9%) patients had stage I disease, whereas 46.7% had stage II, 14.3% had stage III%, and 37.1% had stage IV.
Eighty-three patients with median age of 65 years (43 males and 40 females) received adjuvant or palliative treatments. Sixty-two patients had radical surgery (3 after neoadjuvant treatment) and 47 received adjuvant treatment. Forty-three patients had metastatic disease and 36 received palliative chemotherapy (3 after neoadjuvant treatment).
Adjuvant therapy consisted of 1,000 mg/m2/d gemcitabine on days 1, 8, and 15 every 28 days for two cycles followed by 300 mg/m2/wk gemcitabine plus concomitant radiation therapy to a total of 45 Gy. Palliative and neoadjuvant chemotherapy consisted of 1,000 mg/m2/d gemcitabine on days 1, 8, and 15 every 28 days. Patients were evaluated for response after completing at least two courses of treatment.
Tissue sampling and processing. Tissue sampling and genetic analysis were done according to a clinical protocol approved by the local ethics committee. Patients underwent pancreaticoduodenectomy, distal or total pancreatectomy, or biopsy and tissue samples were stored in liquid nitrogen. Frozen tissue sections (5 μm) were thawed, fixed in 75% ethanol, and dehydratated in 100% ethanol and xylene; neoplastic cells were then dissected using the laser microdissection Leica AS/LMD instrument (Leica, Wetzlar, Germany) as described ( 15). Laser-captured cells were pooled in lysis buffer and RNA was extracted with the QIAamp RNA Mini kit (Qiagen, San Diego, CA). RNA was dissolved in 10 mmol/L DTT and 200 units/mL RNase inhibitor in RNase-free water and measured by absorbance reading at 260/280 nm. In 17 cases, RNA extraction was also obtained from the whole tumor without microdissection.
Quantitative PCR analysis. After determining RNA quality by reverse transcription-PCR (RT-PCR) amplification of glyceraldehyde-3-phosphate dehydrogenase (GAPDH), a total of 102 microdissected samples were evaluable, whereas 3 subjects had no suitable RNA for gene expression analysis. From 50 to 500 ng RNA were reverse transcribed as described previously ( 16). The resulting cDNA was amplified with the 7900HT sequence detection system (Applied Biosystems, Foster City, CA). Forward and reverse primers and probes were designed with Primer Express 2.0 (Applied Biosystems) based on dCK (NM_000788), 5′-NT (NM_012229), and CDA (NM_001785) gene sequence obtained from the Genbank, whereas primers and probes for RRM1 (NM_001033), RRM2 (NM_001034), and hENT1 (NM_004955) were obtained from Applied Biosystems Assay-on-Demand products (Hs00168784, Hs0035724, and Hs00191940).
Validation experiments were carried out with cDNA obtained from QPCR Human Reference Total RNA (Stratagene, La Jolla, CA) to show that the efficiencies of amplification of the target and reference (GAPDH) genes were approximately equal. Therefore, data were expressed as GAPDH/target gene ratio. Specimens were amplified in triplicate with appropriate nontemplate controls, and the coefficient of variation was <2% for all replicates.
Immunohistochemistry and Western blot analysis. To assess the correlation between mRNA and protein expression, selected tumor samples were examined for RRM1 by immunohistochemistry and dCK by immunoblotting. Twelve frozen tissue sections were fixed with acetone for 15 minutes at 4°C and incubated overnight with monoclonal mouse anti-human RRM1 antibody (Chemicon, Hampshire, United Kingdom) at 1:30 dilution and stained with avidin-biotin-peroxidase complex (Vectastain ABC kit, Vector laboratories, Burlingame, CA). Sections were reviewed and scored blindly by comparing the staining of tumor cells versus adjacent normal tissue by one pathologist (D.C.). Negative controls were obtained by replacement of primary antibody with buffer. dCK immunoblotting was done in the same samples; proteins were separated on 12.5% SDS-polyacrylamide gels, transferred to nitrocellulose membranes, and probed with polyclonal mouse anti-human dCK antibody (Abnova, Taipei, Taiwan) at 1:500 followed by incubation with a goat anti-mouse antibody (Chemicon) conjugated to horseradish peroxidase (1:5,000). Immune complexes were detected by enhanced chemiluminescence (Amersham Pharmacia Biotech, Milan, Italy) and measured by densitometric scanning (Kontron, Eching, Germany) of the X-ray films.
Statistical methods. Data were expressed as mean ± SD. Demographic and clinical information was obtained from medical records. Response to treatment was evaluated using the Response Evaluation Criteria in Solid Tumors. Disease-free survival (DFS) was defined as the time from the date of diagnosis to the date of first relapse or last follow-up in radically resected patients. Time to progression (TTP) was calculated from the date of diagnosis to the date of first progression or last follow-up in metastatic patients. Overall survival (OS) was calculated from the date of diagnosis to the date of death or last follow-up. The Kaplan-Meier method was used to plot DFS, TTP, and OS, whereas the log-rank test was used to compare curves. Cox proportional hazards multivariate model was employed to examine the association of clinical and pathologic factors and the expression of genes potentially related to gemcitabine efficacy with OS. Multivariate analyses used a step-down procedure based on the likelihood ratio test. Data were analyzed using SPSS/PC+11.5 (SPSS, Inc., Chicago, IL). Statistical significance was set at P < 0.05.
The relationship between target genes and clinical outcome was evaluated by subgrouping patients into gene expression tertiles. The strategy of data stratification was adopted at the end of the study based on the following considerations: (a) patients were consecutively recruited, (b) gene expression levels were widely distributed across minimum and maximum values, and (c) subgrouping into three gene expression sets (i.e., low, medium, and high tertiles) was best suited to highlight the trend of the target variable ( 17). The expression data and clinical outcomes were reviewed and analyzed blindly by one statistician (C.O.). An additional analysis on the relationship between OS and genetic data was done by stratifying patients based on median gene expression value. Additional data analysis for specific subsets of data was done by Student's t test for paired data.
In vitro Studies
Modulation of dCK, 5′-NT, CDA, and hENT1 and gemcitabine cytotoxicity. 2′-Deoxycytidine increased the IC50s of gemcitabine from 12.29 ± 3.89 to 86.11 ± 6.34 ng/mL, from 53.43 ± 6.24 to 503.97 ± 23.40 ng/mL, and from 29.90 ± 0.66 to 272.53 ± 62.24 ng/mL in MIA PaCa-2, PANC-1, and Capan-1 cells, respectively. Similar results were observed with dipyridamole and nitrobenzylthioinosine; the IC50s increased to 102.36 ± 10.23 and 86.70 ± 6.74 ng/mL (MIA PaCa-2), to 122.30 ± 19.87 and 116.74 ± 8.32 ng/mL (PANC-1), and to 156.97 ± 14.31 and 104.25 ± 8.46 ng/mL (Capan-1), respectively, suggesting that hENT1 is actively involved in gemcitabine uptake and its function modulates the antiproliferative effect of the drug. Furthermore, 5′-NT inhibition resulted in a decrease of IC50s to 8.15 ± 2.21, 28.03 ± 2.07, and 13.71 ± 0.40 ng/mL in MIA PaCa-2, PANC-1, and Capan-1, respectively. Finally, CDA inhibition produced a decrease of IC50s to 7.54 ± 1.31, 10.52 ± 0.12, and 9.40 ± 0.70 ng/mL in MIA PaCa-2, PANC-1, and Capan-1 cells, respectively, suggesting that inactivating enzymes may play an additional role in gemcitabine cytotoxicity.
Treatment outcome and prognostic factor. Clinical data are available from 105 patients who were followed-up until April 30, 2005, with follow-up periods ranging from 0.4 to 32.1 months (median, 11.2 months) after surgery, and the median OS was 13.3 months [95% confidence interval (95% CI), 10.9-15.7].
Forty-seven radically resected patients received gemcitabine and radiation therapy as adjuvant treatment. At the time of the last follow-up, 23 (48.9%) patients had died, whereas 33 (70.2%) patients had disease recurrence and 14 (42.4%) of these received palliative gemcitabine. The median DFS and OS were 13.8 (95% CI, 11.7-16.0) and 21.0 months (95% CI, 13.6-27.6), respectively ( Table 1 ).
Fifteen radically resected and 7 metastatic patients did not receive chemotherapy treatment because of pathologic stage (1 patient, pT2N0M0), poor performance status (3 patients), advanced age (5 patients), refusal (1 patient), or early postoperative death (5 patients), whereas 7 metastatic patients did not receive chemotherapy treatment because of poor performance status (6 patients) or advanced age (1 patient) and all died.
Thirty-six patients received palliative treatment and 34 were evaluable for responses: 5 (14.7%) had partial response, 13 (38.2%) stable disease, and 16 (47.1%) progressive disease. Two patients were not evaluable for response, one because of early death and one because of refusal to continue chemotherapy. The median TTP was 8.4 months (95% CI, 5.4-11.4). Median OS was 12.4 months (95% CI, 9.3-15.5). Nine patients received a second line of chemotherapy for progressive disease consisting of 5-FU/leucovorin and oxaliplatin (4 patients), 5-FU/leucovorin and irinotecan (3 patients), and capecitabine (2 patients).
Six patients with locally advanced disease underwent biopsy for diagnosis and tissue collection; they received neoadjuvant therapy and three of six patients were then radically resected.
Treatment setting (adjuvant or palliative) and grading were significant prognostic factors of the OS on univariate analysis, whereas age, gender, and stage were not correlated with clinical outcome ( Table 1). Multivariate analysis indicated that the setting of therapy and disease grading were independent predictors of prognosis ( Table 2 ).
Gene expression in tumor tissue and correlation with protein levels. Figure 1A shows the variability of gene expression observed across the cohort of 102 patients subjected to transcription analysis. In microdissected samples, the quantitative assessment of the results showed that mRNA expression of dCK, 5′-NT, CDA, RRM1, RRM2, and hENT1 was detectable in all samples. Their variability (dCK, 1.031 ± 0.163; 5′-NT, 0.936 ± 0.149; CDA, 0.963 ± 0.109; RRM1, 0.994 ± 0.091; RRM2, 0.942 ± 0.079; hENT1, 1.226 ± 0.196) was good, suggesting a possible stratification of patients according to genotype to create homogeneous groups with different likelihoods of response to gemcitabine treatment. The analysis of 17 nonmicrodissected samples revealed a minor gene expression variability, with coefficient of variation values ranging from 3.74% (5′-NT) to 8.23% (hENT1) versus 4.52% (RRM2) to 14.27% (hENT1), potentially affecting the patient stratification. Furthermore, the Student's t test showed a significant difference (P < 0.001) in expression values for each gene compared with those observed in the respective microdissected specimens ( Fig. 1B).
At immunohistochemistry, a strong cytoplasmic RRM1 reactivity was observed in cells adjacent to pancreatic tumors, whereas neoplastic cells showed a variable staining, which was related to their gene expression. Indeed, tissues characterized by high mRNA levels presented a strong and diffuse staining ( Fig. 1C, top), whereas tumors with intermediate level of RRM1 mRNA showed a weak positivity ( Fig. 1C, middle). Finally, tumors with low RRM1 mRNA expression had only few scattered positive cells with a weak staining ( Fig. 1C, bottom). Similarly, dCK signal at immunoblotting, normalized to GAPDH, was significantly correlated to dCK mRNA (r2 = 0.65; P < 0.01; Fig. 1D).
Correlation between clinical outcome and gene expression levels. Patients were grouped by tertiles based on their gene expression levels and evaluated for clinical outcome after gemcitabine chemotherapy ( Tables 3 and 4 ). The survival curves according to tumor hENT1 expression are shown in Fig. 2A . The low hENT1 tertile had a significantly poorer prognosis than the high tertile (log-rank test, P < 0.001). Higher hENT1 levels associated with significantly longer median OS (25.69 months; 95% CI, 17.64-33.74) compared with 15.74 and 8.48 months in patients with hENT1 lower than 0.95 and 0.88, respectively. Similar results were obtained with patients dichotomized above and below median gene expression value (22.34 months; 95% CI, 16.34-28.34 versus 12.42 months; 95% CI, 8.18-16.66; Table 3).
There was also a significant difference among the Kaplan-Meier TTP curves ( Fig. 2B) of patients with hENT1 above 1.38, who had a median of 12.68 months compared with 10.09 and 5.85 months in patients with hENT1 below 1.38 and 1.06, respectively (P = 0.02). Likewise, a significantly longer (P < 0.01) median DFS was observed in patients with higher hENT1 (20.43 months; 95% CI, 13.27-27.60) compared with patients belonging to the lower hENT1 tertile (9.26 months; 95% CI, 3.86-14.67; Fig. 2C).
A significant difference in Kaplan-Meier survival curves was also found in patients with CDA above 0.99, who had a median OS of 12.91 months compared with 20.30 months in patients with CDA between 0.91 and 0.99 ( Table 3), but patients with low CDA had a median OS (13.75 months) similar to patients in the high CDA tertile. Moreover, there was no significant predictivity of CDA expression on DFS or TTP ( Table 4) as well as of dCK, 5′-NT, and RRM2 on OS, TTP, and DFS. Patients with RRM1 above 1.00 had a median TTP of 5.85 months compared with 13.30 and 9.92 months in patients with RRM1 below 1.00 and 0.95, respectively ( Table 4), whereas no significant predictivity of RRM1 was found on OS and DFS.
The multivariate analysis confirmed hENT1 expression as an independent factor associated with survival in patients stratified in both three ( Table 2) and two groups (hazard ratio, 4.21; P < 0.001).
In pancreatic cancer, the histopathologic grading has a significant prognostic value ( 18, 19), and this finding has been confirmed in this study; however, the outcome is poor even for patients with well-differentiated tumors ( 1).
The identification of molecular markers predictive of response seems to be critical for treatment optimization on individual basis. Indeed, the present study provides the first evidence of a significant correlation between gemcitabine chemotherapy outcome and hENT1 gene expression in pancreatic cancer.
The availability of new analytic techniques has improved the discovery of molecular markers useful for identifying biological processes involved in carcinogenesis ( 20) and for predicting response to chemotherapy ( 21, 22). The immunohistochemical analysis of pancreatic cancer showed that patients with high tumor expression of thymidylate synthase significantly benefited from 5-FU adjuvant therapy ( 23). Similar studies correlated ERCC1 and RRM1 mRNA levels, measured by quantitative-PCR, with prognosis of NSCLC patients treated with cisplatin and gemcitabine ( 10, 24).
Gemcitabine has a pivotal role in the treatment of locally advanced and metastatic pancreas cancer, either as a part of a curative strategy or with palliative intent, and has been compared favorably with 5-FU as the standard chemotherapy for advanced disease ( 25). Gemcitabine monotherapy produced clinical benefit in 20% to 30% of patients and the disappointing 1-year survival rate of 2% in 5-FU-treated patients was raised to 18% by gemcitabine ( 26). Other studies showed pain reduction, performance status improvement, and weight gain in 24% of patients receiving gemcitabine compared with 5% receiving 5-FU ( 1). Moreover, gemcitabine is expected to have a role in the adjuvant setting, and the results of the ESPAC-3 trial, which randomizes 990 pancreatic cancer patients to observation, 5-FU/leucovorin and gemcitabine, may provide a definitive answer ( 27). The clinical activity of gemcitabine against pancreatic cancer has been confirmed in the present work by the data on median OS, which were 12.4 and 21.0 months in the palliative and adjuvant treatment settings, respectively.
The transcription analysis of genes involved in gemcitabine transport and metabolism may unravel key information for its optimal use, which is worthy to be exploited with novel techniques for genetic analysis in pancreatic cancer specimens ( 3, 28). For these reasons, this study collected genetic data from a large group of pancreatic tumor specimens and examined them retrospectively to establish possible relationships between the pattern of tumor expression related to gemcitabine activity ( 10, 29) and survival.
In particular, this study focused on molecular determinants that have been studied previously in both preclinical investigations and clinical studies ( 6, 11), and the in vitro analysis of modulation of gemcitabine cytotoxicity by specific inhibitors confirmed that hENT1, dCK, 5′-NT, and CDA play a pivotal role in gemcitabine activity.
To facilitate the transition of molecular markers from the laboratory to the clinic, gene expression was studied after rigorous standardization of analytic methods and tissue banking. In particular, laser microdissection enabled the precise isolation of morphologically defined cancer cells from tissue specimens and, in combination with the highly sensitive quantitative RT-PCR technique, is a valuable tool for tumor cell–specific analysis of gene expression. All patients were recruited in one center and this was an additional factor to guarantee reliability of tissue sampling, preparation, and gene expression analysis.
Significant survival differences were observed in 83 gemcitabine-treated pancreatic cancer patients according to hENT1 gene expression. Moreover, univariate analysis showed that patients with high CDA levels had a significantly shorter OS compared with patients with intermediate gene expression but not in comparison with patients with low gene expression. A recent study has shown a correlation between high CDA expression in peripheral blood mononuclear cells and shorter survival in gemcitabine-treated patients with advanced pancreatic carcinoma ( 30). However, in the present study, no such relationship was found in multivariate analysis and the predictivity of CDA on clinical outcome after gemcitabine treatment deserves further investigation. Likewise, a significant association was detected for RRM1 and TTP values in the palliative setting, but no differences were observed in OS and DFS in patients in the adjuvant setting. On the contrary, patients with higher levels of hENT1 expression had significantly longer OS, DFS, and TTP than those with lower transcription levels. Moreover, in multivariate analysis, hENT1 gene expression has been proven to be an independent prognostic marker.
Similar results were described in a previous immunohistochemical study on tissues from patients with advanced pancreatic cancer. Patients with detectable hENT1 expression had significantly longer median survival from gemcitabine initiation than those lacking hENT1 in a proportion of adenocarcinoma cells (median survival, 13 versus 4 months; P = 0.01; ref. 6). Moreover, several in vitro studies showed that cell lines derived from human pancreatic adenocarcinomas incorporate gemcitabine mostly via the hENT1 transporter ( 31) and that the nucleoside transport inhibitors nitrobenzylthioinosine or dipyridamole reduced sensitivity to gemcitabine by 39- to 1,800-fold ( 5).
In the light of these findings, we conclude that hENT1 expression might be a possible new prognostic factor for chemosensitivity of pancreatic cancer to gemcitabine in both palliative and adjuvant settings. Therefore, if confirmed in prospective clinical trials, the analysis of hENT1 in microdissected tissue specimens may help clinicians to select those patients most likely to experience the survival benefit following gemcitabine therapy in pancreatic adenocarcinoma. Finally, the modulation of hENT1 expression, induced by thymidylate synthase inhibitors ( 16, 32), may represent a new way to explore effective modalities for the treatment of pancreatic cancer.
Grant support: Eli-Lilly (Indianapolis, IN) unrestricted research grants (M. Del Tacca, G. Bevilacqua, F. Mosca, and S. Ricci) and Ministero dell'Istruzione, Università e Ricerca (Rome, Italy) research grants (R. Danesi).
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 November 23, 2005.
- Revision received January 25, 2006.
- Accepted February 3, 2006.
- ©2006 American Association for Cancer Research.