Clinical Cancer Research Grants Advances in Breast Cancer
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
Clinical Cancer Research 13, 7044, December 1, 2007. doi: 10.1158/1078-0432.CCR-07-1224
© 2007 American Association for Cancer Research
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 Google Scholar
Google Scholar
Right arrow Articles by Diallo, J.-S.
Right arrow Articles by Saad, F.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Diallo, J.-S.
Right arrow Articles by Saad, F.

Imaging, Diagnosis, Prognosis

NOXA and PUMA Expression Add to Clinical Markers in Predicting Biochemical Recurrence of Prostate Cancer Patients in a Survival Tree Model

Jean-Simon Diallo1, Abdulhadi Aldejmah1,4, Abdelali Filali Mouhim1, Benjamin Péant1, Mona Alam Fahmy1, Ismaël Hervé Koumakpayi1, Kanishka Sircar2, Louis R. Bégin3, Anne-Marie Mes-Masson1,4 and Fred Saad1,5

Authors' Affiliations: 1 Centre de Recherche du Centre Hospitalier de l'Université de Montréal and Institut du Cancer de Montréal, 2 Department of Pathology, McGill University Health Centre, 3 Service d'Anatomopathologie, Hôpital du Sacré-Coeur de Montréal, 4 Département de Médecine, and 5 Département d'Urologie, Université de Montréal, Montréal, Québec, Canada

Requests for reprints: Fred Saad, Centre de Recherche du Centre Hospitalier de l'Université de Montréal/Hôpital Notre-Dame/Institut du Cancer de Montréal, 1560 rue Sherbrooke est, Montréal, Québec, Canada, H2L 4M1. Phone: 514-890-8000, ext. 27466; Fax: 514-412-7620; E-mail: fred.saad.chum{at}ssss.gouv.qc.ca.


   Abstract
Top
 Abstract
Materials and Methods
 Results
 Discussion
 Conclusion
 References
 
Purpose: To assess the expression of proapoptotic NOXA and PUMA in prostate tissues and delineate their association with prostate cancer (PCa) recurrence.

Experimental Design: Normal, prostatic intraepithelial neoplasia (PIN), hormone-sensitive (HS) PCa, and hormone-refractory (HR) PCa tissues were used to build tissue microarrays encompassing a total of 135 patients. Two observers assessed the intensity of NOXA and PUMA immunohistochemical staining using a composite color scale. One hundred and eighty recursive partitioning and regression tree (RPART) models were generated to predict biochemical recurrence (BCR) within HS cancer patients using NOXA, PUMA, and clinical parameters. Models were then ranked according to the integrated Brier score (IBS).

Results: Increasing NOXA expression was associated with PCa progression, reaching the highest levels in HR PCa. Increased NOXA expression was observed in 68% of HS cancer patients and was predictive of BCR (LR = 8.64; P = 0.003). In contrast, PUMA expression was highest in HS cancer, and although 70% of HS cancer patients exhibited increased PUMA expression, PUMA alone could not predict the onset of BCR. Interestingly, the top-ranking RPART model generated [IBS = 0.107; 95% confidence interval (95% CI), 0.065-0.128] included surgical margin status and NOXA and PUMA expression, although recurrent prognostic classification schemes obtained in the top 10 models favored a survival tree model containing margin status, NOXA expression, and preoperative prostate-specific antigen (PSA) (IBS = 0.114; 95% CI, 0.069-0.142).

Conclusion: We conclude that NOXA and PUMA expression may be linked to PCa progression and propose further validation of a survival tree model including surgical margin status, NOXA expression, and preoperative PSA for predicting BCR.

Prostate cancer (PCa) remains a leading cause of cancer-related death in North American men (1). Although localized forms of the disease can often be successfully treated by surgery or radiotherapy, a significant proportion of patients having undergone such interventions are at risk of disease relapse. For this reason, considerable efforts have been made to discover new molecular markers that can accurately predict the onset of disease relapse and lead to better targeted and more effective treatment.

Androgen deprivation therapy is often used to treat recurrent PCa and can increase patient survival; however, this form of therapy eventually gives rise to androgen-independent PCa (or AIPCa; refs. 2, 3). Because the treatment of AIPCa remains palliative to date (4–6), much effort has been devoted to describing the molecular mechanisms associated with the transition of androgen-dependent PCa to an androgen-independent state. Many studies have established a role for androgen receptor (AR) signaling in this phenomenon (7, 8). However, increasing evidence suggests that other signaling pathways may also be important for progression to an androgen-independent state (9–14). At the convergence of many of these pathways, it has been suggested that PCa cells can become resistant to treatment-induced apoptosis through the up-regulation of antiapoptotic proteins such as BCL-2, BCL-XL, and MCL-1 (15, 16). Several studies have detected up-regulated BCL-2, BCL-X, and MCL-1 expression in high-grade PCa tumors and in AIPCa (17–21).

In theory, enhanced resistance to apoptosis can also be achieved by the down-regulation of proapoptotic proteins (22). To date, few studies have looked at the expression of proapoptotic proteins in PCa. Thus far, most studies addressing this question have focused on BAX, a proapoptotic protein that elicits its effect at the level of the mitochondrial outer membrane where it promotes mitochondrial depolarization, a key event in the intrinsic apoptotic pathway. Although it is clear from several immunohistochemistry (IHC) studies that BAX is expressed in the large majority of tumors, the association between BAX expression and PCa progression remains uncertain (18, 23–26).

NOXA and PUMA are two BH3-only proapoptotic proteins that act upstream of BAX/BAK to promote mitochondrial depolarization. NOXA is essentially thought to sensitize cells to the action of activator BH3-only proapoptotic proteins by disrupting their interaction with antiapoptotic proteins. Recent evidence suggests that NOXA specifically disrupts the interaction of MCL-1 with activator BH3-only proteins BID, BIM, and PUMA (27). In turn, activator BH3-only proteins such as PUMA and BID interact with the H{alpha}1 helix of BAX to induce conformational changes leading to the permeation of the mitochondrial outer membrane (28).

To date, few IHC studies have looked at PUMA or NOXA expression in cancer. In melanoma, weak PUMA expression was linked to poor patient survival (29), particularly in patients also showing elevated levels of phosphorylated AKT (30). In colorectal cancer, no relationship with clinical outcome was found, although 29% of tumors overexpressed PUMA (as opposed to 4% showing decreased expression; ref. 31). Similarly, NOXA expression was increased in 16% of colorectal tumors but was not associated with disease outcome (32). To date, neither NOXA nor PUMA has been studied in relation to PCa progression and clinical outcome.

To assist in the process of prognostic marker discovery, increasingly powerful statistical methods are being developed and applied. Of these methods, survival trees are particularly attractive when looking at multiple markers within one or more signaling pathways. Survival tree algorithms are based on recursive partitioning of the covariate space (33, 34), and their graphical output facilitates the visualization of prognostic groups reflecting multimarker interactions. In this study, we looked at the expression of NOXA and PUMA using tissue microarrays containing normal prostate tissue, primary PCa, and its adjacent non-neoplastic tissue, as well as specimens of androgen-independent PCa, representing a total of 135 patients. We then used survival trees to evaluate the ability of NOXA and PUMA, alone or in combination with clinical markers, to predict the onset of biochemical recurrence (BCR) in patients presenting primary PCa.


   Materials and Methods
Top
 Abstract
 Materials and Methods
Results
 Discussion
 Conclusion
 References
 
Patient cohort. A total of 51 normal prostate specimens were obtained from cancer-free patients. An additional 64 paraffin-embedded human primary PCa specimens from patients who had undergone radical prostatectomy between 1993 and 2000 were also used. Furthermore, transurethral resections of the prostate (TURP) specimens from 30 AIPCa patients were obtained. Regions of non-neoplastic and cancerous epithelial tissue were identified by two pathologists and subsequently spotted on tissue microarrays. In the sub-cohort of 64 primary PCa tumors, which was used for retrospective prognostic studies, no patient received preoperative hormone therapy, and all cases had a clinical follow-up of at least 5 years or until death (average follow-up of 72 months). No age difference was observed between the group of patients who relapsed and the group that did not. Postoperative PSA was available for all patients. The time to BCR was defined as the time elapsed between the date of surgery and the date where PSA first increased from undetectable levels to above 0.3 ng/mL and rising, consistent with previous studies (35–37). Non-relapsed patients had a PSA remaining below 0.3 ng/mL after radical prostatectomy. For PCa specimens, the final staging, grading, and histopathologic diagnosis was based on the pathology report in agreement with the review from an independent pathologist. Specimens were obtained from consenting patients, and the institutional ethics review committee approved the study.

Tissue array construction and verification. Tissue arrays containing a total of 613 1-mm-wide cores of prostate tissues were built and used for IHC studies. For the sub-cohort containing normal tissue cores obtained from 51 autopsied patients, two cores per patient were spotted on a tissue microarray. For the prognostic sub-cohort of primary tumors, two non-neoplastic and four cancerous cores per patient were spotted on tissue arrays. For the hormone-refractory (HR) TURP sub-cohort, four cores per patient were included on a tissue microarray. Following tissue microarray construction, 4-µm-thick cross-sections were put on glass slides and stained with H&E as well as for cytokeratin 34βE12 and reviewed by two pathologists. All cores were subsequently re-categorized as containing no epithelial cells, non-neoplastic epithelium, focal atrophy, PIN, or adenocarcinoma. Cores containing no epithelial cells or focal atrophy were not considered in the analysis. Following reclassification, the final specimen cohort consisted of 601 cores representing 43 patients with normal prostate tissues, 62 patients presenting primary PCa tissues, and 30 patients with HR PCa for a total of 135 patients. Patient characteristics are summarized in Table 1 .


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

  		Table 1. Patient cohort characteristics

 
Protein extraction. Confluent LNCaP, 22Rv1, PC3, and DU145 cells were scraped and washed twice with cold PBS, and pellets were frozen at –80°C. Subsequently, whole cell extractions were done by applying cold lysis buffer [10 mmol/L Tris-HCl (pH, 7.4), 150 mmol/L NaCl, 1 mmol/L EDTA, 1 mmol/L DTT, 1 mmol/L NaF, 0.5% NP40, 0.5 mmol/L phenylmethylsulfonyl fluoride, 0.2 mmol/L orthovanadate, 2 mg/mL of aprotinin, leupeptin, and pepstatin] on ice for 30 min. Whole cell extracts were collected after centrifugation in a Heraeus Biofuge (13,000 rpm for 10 min at 4°C) and were immediately stored at –80°C. Protein concentration was measured by Bradford assays (Bio-Rad Laboratories Inc.) according to the manufacturer's instructions.

Western blot analysis. For Western blot analysis, 50 µg of whole cell protein extract were resolved on a 12.5% polyacrylamide gel and then transferred onto polyvinylidene difluoride membranes (Millipore). Blots were blocked using 5% nonfat dry milk in PBS-Tween 0.05% buffer overnight at 4°C and probed using either a monoclonal antibody raised against recombinant glutathione S-transferase–tagged full-length NOXA (OP180, Calbiochem), polyclonal antibody raised against amino acids 2 to 16 of PUMA (PC686, Calbiochem) or actin B (ab6276-100, Abcam) for 1 h at room temperature in blocking buffer (1:500). Membranes were then incubated with secondary antibody conjugated to horseradish peroxidase (Amersham Life Sciences Inc.) in blocking buffer for 1 h at room temperature and developed with enhanced chemiluminescence (ECL) substrate (Amersham Life Sciences Inc.).

Immunohistochemistry. Samples were immunostained with either anti-NOXA antibody (OP180) at 50 ng/µL or anti-PUMA antibody (PC686) at 4 ng/µL diluted in PBS. Primary antibody detection was done using the LSAB 2 peroxidase system from DAKO Diagnostics Inc. Staining was done as described previously (35, 38–40). Briefly, tissue samples were deparaffinized, rehydrated, and treated with 0.3% H2O2 to eliminate endogenous peroxidase activity. An antigen retrieval step was done using 10 mmol/L citrate buffer (pH, 6.0) applied for 17.5 min at 95°C. All following steps were done at room temperature. The sections were blocked with a protein-blocking serum-free reagent (DAKO) and incubated with primary antibody for 60 min, followed by a 20-min treatment with the secondary biotinylated antibody (DAKO), washed 5 min in PBS, and then incubated for 20 min with streptavidin-peroxidase (DAKO). Following an additional 5-min PBS wash, reaction products were developed with diaminobenzidine (DAKO) containing 0.3% H2O2 as a substrate for peroxidase. Nuclei were counterstained with Harris hematoxylin (Sigma-Aldrich). No nonspecific secondary antibody staining was observed when PBS was used instead of the primary antibody.

Scoring procedure. For NOXA- and PUMA-stained tissues, digital pictures were taken of each core on an Olympus BX51 microscope using Q capture imaging software (Olympus). Two independent observers quantified epithelial staining intensity using a color scale (Fig. 2G) constructed from the various staining intensities observable in the digital pictures using the eyedropper tool in Adobe Photoshop 7.0. The observers assessed the percentage of epithelial cells representing each color of the scale (0-9), and an overall score was calculated from the sum of the products derived from the percentage (0-100%) multiplied by the scale value (0-9) for each core. Hence, all staining intensity values are on a continuous scale of 0 to 9. Notably, intraclass correlation coefficients (ICC; a measure of reliability between the two observers) were found to be excellent using this method (ICC > 0.75; ref. 41), in sharp contrast with initial estimates done using conventional microscopy methods (ICC < 0.5). Overall intensity values from each observer, obtained using the digital pictures, were then averaged and used for further statistical analyses.


Figure 2
View larger version (109K):
[in this window]
[in a new window]
[Download PPT slide]
  		Fig. 2. IHC staining of paraffin-embedded prostate tissues using anti-NOXA and anti-PUMA antibodies. A and B, normal prostate tissue probed for NOXA and PUMA, respectively. Note enhanced staining in the basal cell layer. C and D, HS prostate carcinomas probed for NOXA and PUMA, respectively. E and F, HR TURP specimens stained using anti-NOXA and anti-PUMA antibodies, respectively. G, color scale standard used for assessment of pictures obtained from tissue microarrays probed for NOXA and PUMA. Scale was constructed from several digital pictures evaluated in the study as described in Materials and Methods. Numerical values, associated intensity score.

 
Statistics. Mean staining intensities of cores from cancer-free patients, of non-neoplastic, PIN, and cancer cores from hormone-sensitive (HS) PCa patients as well as of cancer TURP cores from HR PCa patients were calculated. Kruskal-Wallis nonparametric tests were used to assess statistical significance of observed differences in mean staining intensity. All correlation coefficients were computed using Spearman's nonparametric test. Cutoff determination and survival tree construction was done using the recursive partitioning and regression tree (RPART) libraries (33), which extends the classification and regression trees (CART) routine (34). Model accuracy was assessed using the integrated Brier score (IBS) for censored data (IBS; ref. 42). We used 200 bootstrap (43) samples to compute the 95% confidence interval (95% CI) on the IBS. Survival tree growth was controlled using the minimum splitting (minsplit) criterion implemented in RPART. This parameter controls the minimum number of observations that must exist in a node for a split to be attempted. For combination models including NOXA and/or PUMA, as well as combinations of the four clinical markers, all possible RPART models were generated using three different values of minsplit (20, 25, and 30). This generated 180 combinations corresponding to 69 different unique tree models, which were ranked according to IBS. Kruskal-Wallis, Spearman, and Kaplan-Meier analyses were done using Statistical Package for the Social Sciences (SPSS) version 11 (SPSS, Inc.). Tree building and the calculation of IBS were carried out in the R (version 2.4.0; ref. 44) system for statistical computing,6 using rpart and ipred packages, respectively.


   Results
Top
 Abstract
 Materials and Methods
 Results
Discussion
 Conclusion
 References
 
NOXA and PUMA expression in PCa cell lines. We used Western blotting on whole cell extracts to assess the expression of NOXA and PUMA proteins in PCa cell lines. As shown in Fig. 1 , the antibodies targeting NOXA and PUMA detected the expected ~6-kDa and 23-kDa bands (respectively) and revealed variable but apparent NOXA and PUMA expression in all PCa cell lines. For PUMA, cell line expression levels were found to be highest in 22Rv1 followed by LNCaP and PC3 cells, with DU145 exhibiting the lowest PUMA expression. The NOXA expression levels were highest in DU145 cells, followed by PC3 and 22Rv1 cells, with the lowest expression levels in LNCaP cells.


Figure 1
View larger version (43K):
[in this window]
[in a new window]
[Download PPT slide]
  		Fig. 1. NOXA and PUMA expression in PCa cell lines. The Western blot was probed for NOXA, PUMA, and actin in whole cell extracts obtained from androgen-responsive (LNCaP, 22Rv1) and androgen-independent PCa cell lines. Because the anti-NOXA and anti-PUMA antibodies used here recognized their respective targets with little background, they were subsequently deemed adequate for IHC.

 
NOXA expression in prostate tissue subtypes. To determine whether NOXA expression could be linked to PCa progression, we stained prostate tissue microarrays using the antibody recognizing NOXA (same as used in Fig. 1). In general, we found that this antibody stained the cytoplasm of epithelial cells (Fig. 2A, C, and E ). In many normal cores from cancer-free patients and non-neoplastic cores found adjacent to cancer (hereby referred to as NA), we observed more intense staining in the basal cell layer of epithelial glands (Fig. 2A). To increase interobserver reliability and facilitate retrospective interpretation of the results obtained, we used a standard color scale (Fig. 2G) constructed from digital pictures of tissue cores as described in Materials and Methods. Overall, we found that cores taken from normal patients expressed significantly less NOXA than all other tissue subtypes obtained from PCa patients, including NA cores (Fig. 3A ). We also observed a slight but statistically insignificant decrease in PIN as opposed to NA cores (P = 0.09). Although HS cancer tissues exhibited higher mean NOXA expression than both NA and PIN cores (P < 0.001), HR TURP specimens exhibited the highest mean NOXA staining (mean = 5.09; P < 0.001). Notably, in the subgroup of patients for which we had both NA and HS cores (n = 51), 68% exhibited increased NOXA expression in HS cores.


Figure 3
View larger version (21K):
[in this window]
[in a new window]
[Download PPT slide]
  		Fig. 3. Average NOXA (A) and PUMA (B) expression in prostate tissue subtypes. Average was calculated over all the available cores in each subtype category. Normal, normal prostate tissue from autopsied patients (NOXA n = 94; PUMA n = 96 cores); NA, normal tissue found adjacent to cancer in radical prostatectomy specimens (n = 91 cores); PIN, PIN tissue obtained from radical prostatectomy (n = 43); HS cancer, HS cancer tissues obtained by radical prostatectomy (NOXA n = 225; PUMA n = 227); HR cancer, HR cancer tissue obtained from TURP specimens (n = 159). Two independent pathologists verified all core classifications. Error bars, SE. Associated P values were calculated using the Kruskal-Wallis nonparametric test. P values under 0.05 were considered significant.

 
PUMA expression in prostate tissue subtypes. Similarly to what was observed with the anti-NOXA antibody, we found that the antibody targeted against PUMA (same as used in Fig. 1) generally stained the cytoplasm of epithelial cells (Fig. 2B, D, and F). Basal cell staining was also apparent in several normal prostate cores as well as in NA cores (Fig. 2B). PUMA staining was subsequently evaluated using the same method employed for NOXA, and mean core intensity was calculated for each core subtype. As shown in Fig. 3B, we observed a significant increase in PUMA expression in NA cores as compared with normal prostate cores (P < 0.001). Although PUMA expression was similar in NA and PIN cores, HS cancer cores exhibited significantly higher PUMA expression than both NA/PIN (P < 0.001). It should be noted, however, that mean PUMA expression in HR cores was not found to be significantly different from that observed in NA cores (P = 0.654). Similarly to what was observed for NOXA, in the subgroup of patients for which we had both NA and HS cancer cores (n = 51), 70% exhibited increased PUMA expression in HS cancer.

NOXA and PUMA expression can predict the onset of BCR. Proapoptotic proteins such as PUMA and NOXA play a role in the initiation of cell death to various cellular stresses. Hence, we wondered whether NOXA and/or PUMA expression could be predictive of PCa re-emergence following radical prostatectomy. To address this question, we used BCR as a surrogate end point and determined whether NOXA and/or PUMA expression could be predictive of BCR. We employed the rpart function in R to assess whether NOXA or PUMA status alone could stratify patients in the function of BCR onset within the sub-cohort of 62 patients presenting HS cancer. Using optimal cutoffs obtained by rpart, corresponding to the primary splitter of the root node, we found that high NOXA expression (≥5.5 on a scale of 0-9) was associated with an earlier and more frequent onset of BCR (log rank or LR = 8.6; P = 0.003; Fig. 4A ). On the other hand, PUMA expression alone was not significantly predictive of the onset of BCR (LR = 2.5; P = 0.114; Fig. 4B). Interestingly, including both NOXA and PUMA in the RPART model revealed that low PUMA expression was associated to more rapid progression toward BCR and specifically when NOXA expression was also low (LR = 15.6; P < 5 x 10–4; Fig. 4C, Table 2 ). In this model, patients exhibiting high NOXA (≥5.5) expression were most likely to quickly undergo BCR, with ~77% (10/13) of these patients having undergone relapse before 3 years. Within the group of patients expressing low levels of NOXA, low expression of PUMA (<6.6) was associated to earlier and more frequent onset of BCR, with close to 46% (16/35) of patients having relapsed within 3 years. In contrast, patients exhibiting both low NOXA and high PUMA infrequently underwent BCR, with only 14% (2/14) having undergone relapse at 3 years. Because of potential overfitting due to the application of cutoffs obtained by RPART from the same test data set, Kaplan-Meier plots and associated LR P values should be considered as purely descriptive measures because the survival outcomes were used to define the prognostic groups.


Figure 4
View larger version (20K):
[in this window]
[in a new window]
[Download PPT slide]
  		Fig. 4. Kaplan-Meier plots for NOXA and PUMA categorized using optimal cutpoints obtained by RPART. A, NOXA expression in PCa is associated to BCR. Low (thin line), average patient NOXA staining intensity was <5.5; High (thick line), NOXA staining intensity ≥5.5. B, PUMA expression alone does not significantly predict the onset of BCR. Low (thin line), average patient PUMA staining intensity was <6.6; High (thick line), PUMA staining intensity ≥6.6. C, RPART model obtained for combined NOXA and PUMA. IBS = 0.155 (95% CI, 0.123-0.2073). Circled numbers, groups depicted in the associated Kaplan-Meier plot shown in the right. Immediately below circled numbers, fractions (in bold), number of patients that relapsed/number of patients in the group. LR, log rank; P, P value. Note that P value and log rank statistics should be considered as purely descriptive measures (see Materials and Methods).

 

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

  		Table 2. Brier scores and associated 95% CI for selected RPART models

 
NOXA and PUMA expressions predict the onset of BCR in combination with clinical markers. We next wondered whether NOXA and PUMA expression could help to predict BCR in combination with other clinicopathologic parameters such as preoperative PSA, Gleason score, pathologic stage, and resection margin status. We thus generated several RPART models using as input variables all possible combinations of the four clinical markers with NOXA and/or PUMA and ranked them according to the IBS, where lower IBS means greater accuracy. The top-ranking model stratified patients first on the basis of margin status (negative = good prognosis) and then on the basis of NOXA within negative margins (NOXA < 5.2 = good prognosis) and on the basis of PUMA within the positive margins (PUMA < 6.1 = good prognosis). Furthermore, PUMA also stratified negative margin patients expressing low levels of NOXA, with high PUMA expression (PUMA ≥ 6.1) being surprisingly associated with good prognosis (0/13 patients relapsed; Fig. 5A ). This model had an IBS of 0.107 (95% CI, 0.065-0.128; Table 2). Interestingly, we found that 9 of the top 10 RPART models exhibited an initial stratification according to resection margin status followed by that of NOXA expression in negative margins (Fig. 5B). In contrast, only 3 of the top 10 models included PUMA. In 7 of the top 10 models, positive margin patients as well as negative margin patients with low NOXA expression could be further stratified by preoperative PSA and/or Gleason score where patients exhibiting low preoperative PSA or low Gleason had a better prognosis. Notably, in all of these seven models, the best prognostic group (negative margin, low NOXA expression, and low preoperative PSA or Gleason) did not relapse. In one of these models with an IBS of 0.114 (95% CI, 0.069-0.142, Table 2), 0/9 patients exhibiting negative margins, low NOXA expression, and low preoperative PSA (<6.5) relapsed in the best prognostic group (Fig. 5C). In contrast, the top model that included only margin status and preoperative PSA had an IBS of 0.135 (95% CI, 0.093-0.157; Table 2), where 3/20 patients relapsed in the best prognostic group. Of the RPART models composed exclusively of clinical markers, the top-ranking model was one that included margin status and Gleason score and had an IBS of 0.132 (95% CI, 0.090-0.160; Table 2).


Figure 5
View larger version (22K):
[in this window]
[in a new window]
[Download PPT slide]
  		Fig. 5. Kaplan-Meier plots for RPART model including NOXA, PUMA, and clinical markers. A, top-ranking model obtained in the study (MGNXPU). IBS = 0.107 (95% CI, 0.065-0.128). Left, associated survival tree; numbers in circles represent MGNXPU groups depicted in the Kaplan-Meier plots (right). B, general survival tree structure determined from the top nine models ranked by IBS. Recurrent structures (bold line) were those present in all of the top nine ranking models. X and Y, node-splitting variables (discontinuous line) within the top nine models. In the top nine ranking models, splitting parameter X was either PUMA, Gleason, and preoperative PSA. One model exhibited no variable for X (no split). Splitting variable Y was either PUMA, Gleason, preoperative PSA, or NOXA. C, favored RPART model obtained from the top nine RPART models from this study (rank 6). IBS = 0.113 (95% CI, 0.069-0.142). Numbers within circles in the survival tree (left) represent groups depicted in the Kaplan-Meier plots (right). In (A) and (C), fractions below the colored circles, number of patients relapsed/number of patients in the group.

 

   Discussion
Top
 Abstract
 Materials and Methods
 Results
 Discussion
Conclusion
 References
 
To our knowledge, this is the first study describing the expression of NOXA or PUMA in a cohort of patients representing various histopathologic subtypes of PCa. In our overall cohort of 135 patients, we found that mean NOXA expression increased gradually going from normal prostate cores to NA and PIN cores, followed by HS cancer cores, and finally, to HR cancer cores, the latter expressing the highest levels of NOXA (Fig. 2A). These data suggest that increasing NOXA expression may be associated to PCa progression. In contrast with what has been previously observed in colorectal cancer (32), we found that increased NOXA expression is a frequent occurrence in PCa (16% in colorectal cancer versus 68% of PCa patients). In addition, we found that NOXA expression is associated with clinical outcome, which was not observed in the colorectal cancer study. In contrast with NOXA, the association between PCa progression and PUMA seems to be more complex. Although we found that 70% of HS patients showed elevated PUMA (as opposed to 29% in melanoma; ref. 31), our data suggest that PUMA expression does not further increase in HR PCa.

These findings are somewhat reflected in what was observed in PCa cell lines using Western blots probing for NOXA and PUMA expression. In Fig. 1, we can see that taken together, the two androgen-insensitive cell lines PC3 and DU145 express relatively higher levels of NOXA as compared with androgen-responsive LNCaP and 22Rv1 cells. In contrast, LNCaP and 22Rv1 cells seem to exhibit higher levels of PUMA. These findings are somewhat surprising because NOXA and PUMA have been found to have more than one transcriptional regulator in common, including p53 and E2F1 (45). However, we found that within cores, there was a generally strong correlation between PUMA and NOXA expression (Spearman's coefficient = 0.586; P < 10–6; data not shown). Altogether, these data may be indicative of the involvement of molecular pathways that lead to de-coupled PUMA/NOXA in HRPCa. Further investigation will be required to address this possibility.

To date, investigators have typically used linear Cox proportional hazard models to stratify patients' risk with respect to the expression of molecular markers. However, Cox models neither handle complex interactions among prognostic factors efficiently nor take into account nonlinear effects (46, 47). To overcome these limitations, tree-based methods offer an attractive alternative to Cox models (48). In this study, we used survival trees to assess the ability of NOXA and PUMA to predict BCR alone and in combination with clinical markers. Although this method is increasingly used for immunohistochemical analyses in cancer, its specific application to PCa cohorts has been thus far limited and generally focused on existing clinical parameters (49–52). However, one PCa study has recently applied the survival tree method to assess the prognostic ability of {alpha}-methylacyl CoA racemase detected by IHC as was done here for PUMA and NOXA (53).

Assessing the predictive performance measure and model selection criteria for prognostic models remains a matter of debate. For survival tree-based methods, the IBS is currently thought to be the most appropriate index (42, 54). Using IBS, we determined that the most accurate model for predicting BCR in our cohort was one that included surgical margin status, NOXA, and PUMA expression (IBS = 0.107). Although this model was particularly good at predicting which patients would not undergo BCR (0/13 in the best prognostic group, Fig. 5A and B), it presented a complex behavior of PUMA wherein its effect on BCR onset in negative margins was opposite to that found in positive margins. Because PUMA expression alone could not significantly predict BCR as shown in Kaplan-Meier analyses (Fig. 4B), it is unclear whether PUMA expression truly holds valuable clinical information. On the other hand, NOXA was a significant predictor of BCR when assessed alone (Fig. 4A) and was incorporated into the top nine RPART models where low NOXA expression was consistently associated with good prognosis. For this reason as well as for practical considerations pertaining to some degree of subjectivity in the Gleason grading, we favor the model presented in Fig. 5D and E, including NOXA expression, surgical margin status, and preoperative PSA, for future evaluation.

One limitation in the present study is the relatively small size of our prognostic cohort (n = 62), which may affect the reliability of the RPART algorithm. Nonetheless, we could observe recurrent and stable tree structures that were consistently present in the top nine models. Another relevant concern is that reapplying a RPART model established from one data set onto the same data set can lead to overfitting. To account for this, we used bootstrap resampling to calculate 95% CIs for the IBS as has been done elsewhere (55). Nonetheless, the results obtained here remain of an exploratory nature and will require subsequent external validation in another cohort. In addition, it will be of significant interest to determine whether NOXA/PUMA expression can also be useful prognostic markers at the biopsy level. In either case, the color scale method for measuring staining intensity devised here will likely be useful for the evaluation and classification of future samples.

Overall, our findings are somewhat at odds with the roles of NOXA and PUMA as proapoptotic factors. Although some have found that increasing apoptotic index is associated with disease recurrence (56), others have detected decreasing apoptosis in PCa progression (57). One possible reason for these discrepancies is that increased apoptosis may be counterbalanced by increased cell proliferation, leading to more rapid cellular turnover. Notably, others have found that increasing expression of the cell proliferation marker Ki67 correlates with Gleason grade and with decreased PCa patient survival (58, 59). Although we are currently investigating whether this is the case for our patient cohort,7 it is important to note that we did not find a significant correlation between Gleason score and either NOXA or PUMA in the present study (data not shown). As such, another possibility is that increased NOXA/PUMA expression is an indirect result of deregulated activation of factors mediating NOXA/PUMA transcription or protein stabilization. Potential candidates for this are varied and include p53, E2F, as well as forkhead family transcription factors (45, 60, 61).


   Conclusion
Top
 Abstract
 Materials and Methods
 Results
 Discussion
 Conclusion
References
 
We conclude that NOXA and PUMA expression may be linked to PCa progression. We also suggest that the assessment of NOXA expression may be particularly useful for PCa prognosis because it may extend the ability of existing clinical markers to predict BCR. We believe that a survival tree model including NOXA, surgical margin status, and preoperative PSA status deserves external validation in a larger cohort.


   Acknowledgments
 
We thank Dr. Armen Aprikian for allowing us to use some of the tissue specimens included in the tissue microarrays. We recognize the assistance of Jason Madore in the construction and cutting of the tissue microarrays. We are also grateful to Philippe O. Gannon, Dr. Laurent Lessard, and Dr. Cécile Le Page for critical review of the manuscript.


   Footnotes
 
Grant support: National Cancer Institute of Canada/Canadian Urologic Oncology Group/AstraZeneca research fellowship in prostate cancer and an Aventis Grant in Prostate Cancer Research. Dr. Fred Saad holds the Université de Montréal Chair in Prostate Cancer Research. J.-S. Diallo is the recipient of scholarships from the Fonds Marc Bourgie, the Fonds Canderel (Institut du Cancer de Montréal), the Faculté des EÅtudes Supérieures de l'Université de Montréal and the Fondation du Centre de Recherche du Centre Hospitalier de l'Université de Montréal. I.H. Koumakpayi holds an award from the Fonds de Recherche en Santé du Québec.

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.

Note: A. Aldejmah and A. Mouhim contributed equally to this work.

6 www.cran.r-project.org Back

7 Gannon PO et al., in preparation. Back

Received 5/18/07; revised 7/30/07; accepted 9/ 7/07.


   References
Top
 Abstract
 Materials and Methods
 Results
 Discussion
 Conclusion
 References
 

  1. Jemal A, Siegel R, Ward E, et al. Cancer statistics, 2007. CA Cancer J Clin 2007;57:43–66.[Abstract/Free Full Text]
  2. Landis SH, Murray T, Bolden S, Wingo PA. Cancer statistics, 1999. CA Cancer J Clin 1999;49:8–31, 1.[Abstract/Free Full Text]
  3. Denis L, Murphy GP. Overview of phase III trials on combined androgen treatment in patients with metastatic prostate cancer. Cancer 1993;72:3888–95.[CrossRef][Medline]
  4. Culine S, Droz JP. Chemotherapy in advanced androgen-independent prostate cancer 1990–1999: a decade of progress? Ann Oncol 2000;11:1523–30.[Abstract/Free Full Text]
  5. Crawford ED, Eisenberger MA, McLeod DG, et al. A controlled trial of leuprolide with and without flutamide in prostatic carcinoma. N Engl J Med 1989;321:419–24.[Abstract]
  6. Morris MJ, Scher HI. Novel strategies and therapeutics for the treatment of prostate carcinoma. Cancer 2000;89:1329–48.[CrossRef][Medline]
  7. Coughlin SS, Hall IJ. A review of genetic polymorphisms and prostate cancer risk. Ann Epidemiol 2002;12:182–96.[CrossRef][Medline]
  8. Shang Y, Myers M, Brown M. Formation of the androgen receptor transcription complex. Mol Cell 2002;9:601–10.[CrossRef][Medline]
  9. Ratan HL, Gescher A, Steward WP, Mellon JK. ErbB receptors: possible therapeutic targets in prostate cancer? BJU Int 2003;92:890–5.[CrossRef][Medline]
  10. Gavrielides MV, Frijhoff AF, Conti CJ, Kazanietz MG. Protein kinase C and prostate carcinogenesis: targeting the cell cycle and apoptotic mechanisms. Curr Drug Targets 2004;5:431–43.[CrossRef][Medline]
  11. Yang YM, Bost F, Charbono W, et al. C-Jun NH(2)-terminal kinase mediates proliferation and tumor growth of human prostate carcinoma. Clin Cancer Res 2003;9:391–401.[Abstract/Free Full Text]
  12. Edwards J, Krishna NS, Mukherjee R, Bartlett JM. The role of c-Jun and c-Fos expression in androgen-independent prostate cancer. J Pathol 2004;204:153–8.[CrossRef][Medline]
  13. Chesire DR, Isaacs WB. β-Catenin signaling in prostate cancer: an early perspective. Endocr Relat Cancer 2003;10:537–60.[Abstract]
  14. Libertini SJ, Tepper CG, Guadalupe M, et al. E2F1 expression in LNCaP prostate cancer cells deregulates androgen dependent growth, suppresses differentiation, and enhances apoptosis. Prostate 2006;66:70–81.[CrossRef][Medline]
  15. Catz SD, Johnson JL. BCL-2 in prostate cancer: a minireview. Apoptosis 2003;8:29–37.[CrossRef][Medline]
  16. Tantivejkul K, Loberg RD, Mawocha SC, et al. PAR1-mediated NF{kappa}B activation promotes survival of prostate cancer cells through a Bcl-xL–dependent mechanism. J Cell Biochem 2005;96:641–52.[CrossRef][Medline]
  17. Zellweger T, Ninck C, Bloch M, et al. Expression patterns of potential therapeutic targets in prostate cancer. Int J Cancer 2005;113:619–28.[CrossRef][Medline]
  18. Krajewska M, Krajewski S, Epstein JI, et al. Immunohistochemical analysis of bcl-2, bax, bcl-X, mcl-1 expression in prostate cancers. Am J Pathol 1996;148:1567–76.[Abstract]
  19. Augustin H, Hammerer PG, Graefen M, et al. Characterisation of biomolecular profiles in primary high-grade prostate cancer treated by radical prostatectomy. J Cancer Res Clin Oncol 2003;129:662–8.[CrossRef][Medline]
  20. Pollack A, Cowen D, Troncoso P, et al. Molecular markers of outcome after radiotherapy in patients with prostate carcinoma: Ki-67, bcl-2, bax, and bcl-x. Cancer 2003;97:1630–8.[CrossRef][Medline]
  21. McDonnell TJ, Troncoso P, Brisbay SM, et al. Expression of the protooncogene bcl-2 in the prostate and its association with emergence of androgen-independent prostate cancer. Cancer Res 1992;52:6940–4.[Abstract/Free Full Text]
  22. Viktorsson K, Lewensohn R, Zhivotovsky B. Apoptotic pathways and therapy resistance in human malignancies. Adv Cancer Res 2005;94:143–96.[CrossRef][Medline]
  23. Amirghofran Z, Monabati A, Gholijani N. Apoptosis in prostate cancer: bax correlation with stage. Int J Urol 2005;12:340–5.[CrossRef][Medline]
  24. Rubio J, Ramos D, Lopez-Guerrero JA, et al. Immunohistochemical expression of Ki-67 antigen, cox-2 and Bax/Bcl-2 in prostate cancer; prognostic value in biopsies and radical prostatectomy specimens. Eur Urol 2005;48:745–51.[CrossRef][Medline]
  25. Royuela M, De Miguel MP, Bethencourt FR, et al. IL-2, its receptors, and bcl-2 and bax genes in normal, hyperplastic and carcinomatous human prostates: immunohistochemical comparative analysis. Growth Factors 2000;18:135–46.[Medline]
  26. Chia SJ, Tang WY, Elnatan J, et al. Prostate tumours from an Asian population: examination of bax, bcl-2, p53 and ras and identification of bax as a prognostic marker. Br J Cancer 2000;83:761–8.[CrossRef][Medline]
  27. Kim H, Rafiuddin-Shah M, Tu HC, et al. Hierarchical regulation of mitochondrion-dependent apoptosis by BCL-2 subfamilies. Nat Cell Biol 2006;8:1348–58.[CrossRef][Medline]
  28. Cartron PF, Gallenne T, Bougras G, et al. The first {alpha} helix of Bax plays a necessary role in its ligand-induced activation by the BH3-only proteins Bid and PUMA. Mol Cell 2004;16:807–18.[CrossRef][Medline]
  29. Karst AM, Dai DL, Martinka M, Li G. PUMA expression is significantly reduced in human cutaneous melanomas. Oncogene 2005;24:1111–6.[CrossRef][Medline]
  30. Karst AM, Dai DL, Cheng JQ, Li G. Role of p53 up-regulated modulator of apoptosis and phosphorylated Akt in melanoma cell growth, apoptosis, and patient survival. Cancer Res 2006;66:9221–6.[Abstract/Free Full Text]
  31. Jansson A, Arbman G, Sun XF. mRNA and protein expression of PUMA in sporadic colorectal cancer. Oncol Rep 2004;12:1245–9.[Medline]
  32. Jansson AK, Emterling AM, Arbman G, Sun XF. Noxa in colorectal cancer: a study on DNA, mRNA and protein expression. Oncogene 2003;22:4675–8.[CrossRef][Medline]
  33. Therneau TAE. An introduction to recursive partitioning using the rpart routine. Technical Report. Rochester: Section of Biostatistics, Mayo Clinic; 1997.
  34. Breiman LFJ, Olshen RA, Stone CJ. Classification and regression trees. California: CRC Press; 1984.
  35. Fradet V, Lessard L, Begin LR, et al. Nuclear factor-{kappa}B nuclear localization is predictive of biochemical recurrence in patients with positive margin prostate cancer. Clin Cancer Res 2004;10:8460–4.[Abstract/Free Full Text]
  36. Koumakpayi IH, Diallo JS, Le Page C, et al. Low nuclear ErbB3 predicts biochemical recurrence in patients with prostate cancer. BJU Int 2007;100:303–9.[CrossRef][Medline]
  37. Le Page C, Koumakpayi IH, Alam-Fahmy M, Mes-Masson AM, Saad F. Expression and localisation of Akt-1, Akt-2 and Akt-3 correlate with clinical outcome of prostate cancer patients. Br J Cancer 2006;94:1906–12.[CrossRef][Medline]
  38. Lessard L, Mes-Masson AM, Lamarre L, et al. NF-{kappa}B nuclear localization and its prognostic significance in prostate cancer. BJU Int 2003;91:417–20.[CrossRef][Medline]
  39. Koumakpayi IH, Diallo JS, Le Page C, et al. Expression and nuclear localization of ErbB3 in prostate cancer. Clin Cancer Res 2006;12:2730–7.[Abstract/Free Full Text]
  40. Gannon PO, Alam Fahmy M, Begin LR, et al. Presence of prostate cancer metastasis correlates with lower lymph node reactivity. Prostate 2006;66:1710–20.[CrossRef][Medline]
  41. Edwards J, Traynor P, Munro AF, et al. The role of HER1–4 and EGFRvIII in hormone-refractory prostate cancer. Clin Cancer Res 2006;12:123–30.[Abstract/Free Full Text]
  42. Graf E, Schmoor C, Sauerbrei W, Schumacher M. Assessment and comparison of prognostic classification schemes for survival data. Stat Med 1999;18:2529–45.[CrossRef][Medline]
  43. Effron BTRJ. An introduction to the bootstrap. New York: Chapman and Hall; 1993.
  44. Team RDC. R: a language and environment for statistical computing. Vienna, Austria: R Foundation fro Statistical Computing; 2006.
  45. Hershko T, Ginsberg D. Up-regulation of Bcl-2 homology 3 (BH3)-only proteins by E2F1 mediates apoptosis. J Biol Chem 2004;279:8627–34.[Abstract/Free Full Text]
  46. Su X, Tsai CL. Tree-augmented Cox proportional hazards models. Biostatistics 2005;6:486–99.[Abstract/Free Full Text]
  47. De Rose A, Pallara A. Survival trees: an alternative non-parametric multivariate technique for life history analysis. Eur J Popul 1997;13:223–41.[CrossRef][Medline]
  48. LeBlanc M, Crowley J. Relative risk trees for censored survival data. Biometrics 1992;48:411–25.[CrossRef][Medline]
  49. Glass TR, Tangen CM, Crawford ED, Thompson I. Metastatic carcinoma of the prostate: identifying prognostic groups using recursive partitioning. J Urol 2003;169:164–9.[CrossRef][Medline]
  50. Steuber T, Graefen M, Haese A, et al. Validation of a nomogram for prediction of side specific extracapsular extension at radical prostatectomy. J Urol 2006;175:939–44; discussion 44.[CrossRef][Medline]
  51. Garzotto M, Park Y, Mongoue-Tchokote S, et al. Recursive partitioning for risk stratification in men undergoing repeat prostate biopsies. Cancer 2005;104:1911–7.[Medline]
  52. Garzotto M, Beer TM, Hudson RG, et al. Improved detection of prostate cancer using classification and regression tree analysis. J Clin Oncol 2005;23:4322–9.[Abstract/Free Full Text]
  53. Rubin MA, Bismar TA, Andren O, et al. Decreased {alpha}-methylacyl CoA racemase expression in localized prostate cancer is associated with an increased rate of biochemical recurrence and cancer-specific death. Cancer Epidemiol Biomarkers Prev 2005;14:1424–32.[Abstract/Free Full Text]
  54. Altman DG, Royston P. What do we mean by validating a prognostic model? Stat Med 2000;19:453–73.[CrossRef][Medline]
  55. Radespiel-Troger M, Hohenberger W, Reingruber B. Improved prediction of recurrence after curative resection of colon carcinoma using tree-based risk stratification. Cancer 2004;100:958–67.[CrossRef][Medline]
  56. Stapleton AM, Zbell P, Kattan MW, et al. Assessment of the biologic markers p53, Ki-67, and apoptotic index as predictive indicators of prostate carcinoma recurrence after surgery. Cancer 1998;82:168–75.[CrossRef][Medline]
  57. Zeng L, Rowland RG, Lele SM, Kyprianou N. Apoptosis incidence and protein expression of p53, TGF-β receptor II, p27Kip1, and Smad4 in benign, premalignant, and malignant human prostate. Hum Pathol 2004;35:290–7.[CrossRef][Medline]
  58. Keshgegian AA, Johnston E, Cnaan A. Bcl-2 oncoprotein positivity and high MIB-1 (Ki-67) proliferative rate are independent predictive markers for recurrence in prostate carcinoma. Am J Clin Pathol 1998;110:443–9.[Medline]
  59. Bubendorf L, Tapia C, Gasser TC, et al. Ki67 labeling index in core needle biopsies independently predicts tumor-specific survival in prostate cancer. Hum Pathol 1998;29:949–54.[CrossRef][Medline]
  60. You H, Pellegrini M, Tsuchihara K, et al. FOXO3a-dependent regulation of Puma in response to cytokine/growth factor withdrawal. J Exp Med 2006;203:1657–63.[Abstract/Free Full Text]
  61. Obexer P, Geiger K, Ambros PF, Meister B, Ausserlechner MJ. FKHRL1-mediated expression of Noxa and Bim induces apoptosis via the mitochondria in neuroblastoma cells. Cell Death Differ 2007;14:534–47.[CrossRef][Medline]




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 Google Scholar
Google Scholar
Right arrow Articles by Diallo, J.-S.
Right arrow Articles by Saad, F.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Diallo, J.-S.
Right arrow Articles by Saad, F.

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