MDM2 as a predictor of prostate carcinoma outcome

An analysis of Radiation Therapy Oncology Group protocol 8610

Authors


  • The contents of the current article are solely the responsibility of the authors and do not necessarily represent the official views of the National Cancer Institute or the U.S. Department of Defense.

Abstract

BACKGROUND

The MDM2 oncoprotein promotes p53 degradation via ubiquitin, establishing negative feedback control of p53 and consequently affecting cell cycle arrest and apoptosis. The authors evaluated the association between MDM2 expression and local failure, distant metastasis (DM), cause-specific mortality, and overall mortality in men treated in Radiation Therapy Oncology Group 8610 with radiotherapy, with or without androgen deprivation.

METHODS

Of the 456 eligible and analyzable patients (parent cohort), adequate archival diagnostic tissue specimens from 108 patients were available for MDM2 analysis (MDM2 cohort). Cox proportional hazards multivariate analysis (MVA) was used to determine the relation of MDM2 to the endpoints. MDM2 overexpression was manually classified as > 5% nuclear staining. An image analysis system was also used to quantify the proportion of tumor nuclei with MDM2 staining (ACIS index) and staining intensity.

RESULTS

Overexpression of MDM2 by manual counts was seen in 44% (n = 47) of the patients. In the manual count analysis, there was no significant relation between MDM2 overexpression and outcome. The ACIS index, using a cutoff point defined by the median value, ≤ 3% versus > 3%, was related to 5-year DM rates in univariate analyses (32.6% vs. 45.8%; P = 0.057) and MVA (P = 0.06). The intensity of MDM2 staining was not significant.

CONCLUSIONS

MDM2 expression quantified by image analysis was weakly associated with DM. The cohort examined was relatively small and with larger patient numbers, MDM2 overexpression may emerge as a more significant covariate. Cancer 2005. © 2005 American Cancer Society.

The MDM2 oncoprotein, a ubiquitin ligase, binds to several apoptotic proteins including E2F-1, pRb, and p53, but is principally a negative regulator of p53. It is induced by p53, binds to its amino terminal transactivation domain, and consequently inhibits transcription of genes responsible for cell cycle arrest and apoptosis.1–3 MDM2 oncogene overexpression has been seen in a variety of tumors,4 including prostate carcinomas,5–7 in which it has been observed in > 30% of men. It is associated with high-risk locoregional8 and hormone-refractory disease.9 In our experience, MDM2 suppression via antisense oligonucleotides sensitizes prostate tumor cells in vitro to radiotherapy (RT)10 and androgen deprivation (AD).11 Thus, MDM2 is a promising therapeutic target and the level of expression may be a useful marker of treatment outcome. To our knowledge, this is the first study to evaluate the predictive value of MDM2 overexpression in men with prostate carcinoma treated with RT.

Radiation Therapy Oncology Group (RTOG) protocol 8610 was a Phase III randomized clinical trial designed to compare the effect of RT plus short-term neoadjuvant and concurrent (STAD) with RT alone.12 The patients enrolled had locally advanced disease, with palpable tumors of surface area ≥ 25 cm2. Approximately one-third of the patients had Gleason score 8–10 disease and there was documented lymph node involvement in 8% of the patients. The purpose of the current analysis was to identify the relation of MDM2 expression to local failure (LF), distant metastasis (DM), cause-specific mortality (CSM), and overall mortality (OM).

MATERIALS AND METHODS

Patient Characteristics

RTOG 8610 has previously been described in detail.12 The pretreatment diagnostic samples were sectioned and reviewed by the study pathologist (DJG). Of the 108 patient samples available for MDM2 analysis, the distribution of patients by Gleason score was 27 with a Gleason score of 2–6 and 80 with a Gleason score of 7–10 (1 patient was missing a Gleason score). The distribution of patients by clinical T classification was 29 with T2 and 79 with T3 disease. Sixty-two and 46 patients were assigned to RT alone and RT/STAD, respectively.

Immunohistochemical Analysis

Sections best representing the tumor were cut 4-μm thick onto poly-L-lysine slides from paraffin-embedded, formalin-fixed tissue specimens. The tissue specimen was then deparaffinized in xylene and rehydrated in a series of ethanol washes (95%) to a final distilled water step. Slides were then pressure cooked in an antigen retrieval citrate buffer solution (pH 6.0) for 50 minutes. After rinsing with water, the slides were covered with 3% hydrogen peroxide for 5 minutes at room temperature, then rinsed in Tris buffer, and humidified. The primary monoclonal MDM2 antibody (clone IF2, Zymed Laboratories, Inc., South San Francisco, CA; 1:100 dilution) was then overlaid. The slides were rinsed in Tris buffer, then incubated with Biotin (LSAB II kit; Dako Corporation, Carpinteria, CA) for 10 minutes, rinsed again as before, then incubated with Streptavidin for 10 minutes. After rinsing again with Tris buffer, chromagen (diaminobenzidine [DAB]; Research Genetics, Huntsville, AL) was applied for 5 minutes. The slides were then counterstained with commercially prepared hematoxylin (Dako Corporation) for 5 minutes, dehydrated, and coverslipped. All staining was performed on a Dako Autostainer. Positive controls with prostate carcinoma tissue sections were used for comparison during tissue analysis.

Two investigators (L-YK and TA-S) reviewed the slides under a light microscope without knowledge of patient outcome. For the manual analysis, > 5% dark brown nuclear tumor cell staining was considered positive, indicating overexpression of MDM2. This was considered a reasonable cutoff point to use because previous analyses considered any positive staining,8 > 5% staining13 to ≥ 20% staining.7

The percentage of cells with nuclear staining (ACIS index) and the intensity of staining were also quantified using an image analysis system (ACIS, Clarient Inc., San Juan Capistrano, CA). A color threshold for brown (positive nuclei) and blue (negative nuclei) staining was set for every slide analyzed. When possible, ≥ 3 areas of interest in the tissue specimen visualized at × 40 magnification were designated for quantification. A final sample mean percent index (ACIS index) was derived by the computer software. Intensity of staining was scored on a gray scale of 0–255, in which 255 represented black.

The analysis of p53 by immunohistochemistry has been described previously in this patient population.14 The staining methods used were similar. p53 was deemed positive when > 20% of the tumor cells had nuclear staining, as quantified manually.

Definition of Endpoints

The four endpoints examined were LF, DM, CSM, and OM. The details of these endpoints have been described previously.12, 14, 15 Time to failure or death was measured from the date of randomization to the first reported date of failure.

Statistical Analysis

There were 456 assessable patients in the parent cohort of RTOG 8610.12, 16 The MDM2 study cohort comprised 108 patients analyzed both manually and by ACIS. As of June 30, 2000, the median follow-up of all surviving patients in the study cohort was 9.3 years and the median follow-up of all entered patients was 6.7 years. The distributions of patient characteristics and treatment assignments were compared by the Pearson chi-square test and the Yates correction factor. Estimates of OM were derived using the Kaplan–Meier method,17 whereas the cumulative incidence approach was used to estimate LF, DM, and CSM. Multivariate analysis (MVA) using Cox proportional hazard models was applied to each of the endpoints to identify the impact of MDM2.

There were 348 patients in the parent cohort in whom MDM2 was not quantified. Using the chi-square test, statistical comparisons were performed to assess whether the distributions of patients by prognostic factors were different between the groups.

The MDM2 ACIS index and ACIS intensity score were modeled as continuous and categorical (using a cutoff point at the median value) variables in Cox proportional hazards models.

The interaction between MDM2 and p53 also led us to include the p53 data described in a previous study on RTOG 8610.14 In that study, a cohort of 129 patients was analyzed for p53 positivity (overexpression) by immunohistochemistry. p53 overexpression was associated with an increase in the incidence of DM.

RESULTS

We determined MDM2 overexpression in 108 (23.7%) of the 456 eligible patients in RTOG 8610. Table 1 shows the distribution of patients for whom MDM2 was (MDM2 cohort) and was not (other assessable patients in RTOG 8610) determined, according to pretreatment characteristics and assigned treatment. There were no statistically significant differences in the distribution of patients by potential prognostic factors between these two groups. Table 2 displays the distribution of patients in the MDM2 cohort by MDM2 manual count results (5% cutoff point) and patient characteristics. The only significant finding was that MDM2 overexpression was significantly associated with higher Gleason scores. Forty (85%) patients with MDM2 overexpression had a Gleason score of 7–10, whereas 7 (15%) patients had a Gleason score of 2–6 (P = 0.029). MDM2 overexpression was not associated with age, clinical stage, assigned treatment, or p53 status.

Table 1. Distribution of all Patients by the Presence or Absence of MDM2 Data (n = 456)
CharacteristicsPresence (n = 108) (%)Absence (n = 348) (%)P valuea
  • GLSC: Gleason score; RT: radiotherapy; STAD: short-term androgen deprivation.

  • a

    Chi-square statistics.

GLSC   
 2–627 (25)102 (32)0.21
 7–1080 (75)220 (68) 
 Unknown1 (< 1)26 (7) 
T-classification   
 T229 (27)108 (31)0.41
 T379 (73)240 (69) 
Assigned treatment   
 RT alone62 (57)168 (48)0.10
 RT + STAD46 (43)180 (52) 
p53   
 Negative70 (86)36 (75)0.10
 Positive11 (14)12 (25) 
 Unknown27300 
Table 2. Distribution of Patients by MDM2 Manual Count Results
CharacteristicsNegative (n = 61) (%)aPositive (n = 47) (%)P valueb
  • GLSC: Gleason score; RT: radiotherapy; STAD: short-term androgen deprivation.

  • a

    One patient in the negative MDM2 group was missing the Gleason score.

  • b

    Chi-square statistics.

Age (yrs)   
 <7546 (75)35 (74)0.91
 ≥7515 (25)12 (26) 
GLSC   
 2–620 (33)7 (15)0.029
 7–1040 (67)40 (85) 
T-classification   
 T217 (28)12 (26)0.79
 T344 (72)35 (74) 
Assigned treatment   
 RT alone33 (54)29 (62)0.43
 RT + STAD28 (46)18 (38) 
p53   
 Negative34 (83)36 (90)0.35
 Positive7 (17)4 (10) 
 Unknown207 

The univariate analysis results for the MDM2 cohort are shown in Table 3. Although there was no significant relation between the MDM2 manual count results and outcome, MDM2 overexpression was associated with a 5-year DM rate in univariate analysis of 42.6% versus 28.6% when MDM2 was not overexpressed (P = 0.15) (Fig. 1). This observation may be clinically meaningful, given the relatively small sample of patients. The analyses, with respect to DM and the other endpoints tested, may not have been adequately powered to detect a difference in MDM2 expression. For the end point of DM, the power to detect the risk observed in the univariate analysis (relative risk [RR] = 1.49) was 31%. In the MVA (Table 4), controlling for Gleason score, p53 status, and assigned treatment, the association of the MDM2 manual count results with DM was slightly weaker (P = 0.17).

Table 3. Univariate Analysis Results for the MDM2 Cohort (n = 108)a
Local failure RR (95% CI)Distant metastasis RR (95% CI)Cause-specific mortality RR (95% CI)Overall mortality RR (95% CI)
  • RR: relative risk; 95% CI: 95% confidence interval.

  • a

    P values were derived from the chi-square test.

0.92 (0.49–1.76)1.49 (0.87–2.56)1.32 (0.70–2.49)1.12 (0.71–1.74)
P = 0.81P = 0.15P = 0.40P = 0.63
Figure 1.

Survival curve of distant metastasis by MDM2 manual count results, using cumulative incidence estimates. Solid line: negative MDM2 manual counts; dashed line: positive MDM2 manual counts.

Table 4. Multivariate Analysis of Distant Metastasis; MDM2 Manual Count results
VariableaGroupRR (95% CI)bP valuec
  • RR: relative risk; 95% CI: 95% confidence interval; GLSC: Gleason score; STAD: short-term androgen deprivation.

  • a

    All variables were dichotomous.

  • b

    A relative risk ratio of 1 indicates no difference between the two subgroups.

  • c

    P value was derived from the chi-square test using the Cox proportional hazards model.

MDM2Positive1.60 (0.82–3.10)0.17
GLSC7–102.66 (1.07–6.63)0.0353
STADYes0.89 (0.47–1.68)0.71
p53Positive2.67 (1.17–6.10)0.0199

The MDM2 manual count results were obtained using a > 5% cutoff point for overexpression. A range of cutoff points have been used in the past.7, 8, 13 The rationale for using this particular cutoff point was that it is clearly above background and has been used before.13 However, the results were not statistically significant and there is the possibility that it is not the optimal cutoff point. Hence, we proceeded to use an image analysis system to more precisely quantify the proportion of tumor cells with nuclear MDM2 staining (ACIS index). A median ACIS index of 3.0% (range, 0–26.0%) was obtained. Table 5 shows the relation of the ACIS index to the manual results. When we compared the 75% quartile cutoff point of 5% with the equivalent manual results, there were some discrepancies: 3 patients, scored negative in manual analysis due to extremes in staining intensity, were scored > 5% by ACIS. This is likely because of the ability of ACIS to more accurately score a wide range of staining intensities. Also, 15 of the 24 patients scored positive manually, were scored 5% exactly by ACIS.

Table 5. Distribution of Manual Vs. ACIS Index Results
ACIS indexManualTotalP valuea
Negative (n = 61)Positive (n = 47)
  • a

    Chi-square statistics.

≤ 1.028 (46)028< 0.0001
> 1.033 (54)47 (100)80 
≤ 3.054 (89)5 (11)59< 0.0001
> 3.07 (11)42 (89)49 
≤ 5.058 (95)24 (51)82< 0.0001
> 5.03 (5)23 (49)26 

The three cutoff points were then applied in univariate analysis to the four endpoints. A relation was seen between the median 3% ACIS index cutoff point and DM (Table 6). MDM2 overexpression in ≤ 3% of tumor cells was associated with a 5-year DM rate of 32.6% versus 45.8% when > 3% had overexpression (P = 0.057) (Fig. 2). A similar level of significance was seen in the MVA (RR = 1.85, P = 0.06) (Table 7). p53 positivity and a Gleason score of 7–10 were significantly associated with DM (RR = 2.68, P = 0.02; RR = 2.7, P = 0.03). When the MDM2 ACIS index was used as a continuous variable in MVA, no relation to DM or the other endpoints was observed in MVA.

Table 6. Univariate Analysis of the MDM2 ACIS Indexa
End pointMDM2 ACIS cutoff pointNo. of patientsFailuresRRb (95% CI)P valuec5-yr rate (%)5-yr (95% CI)
  • RR: relative risk; 95% CI: 95% confidence interval; LF: local failure; DM: distant metastasis: CSM: cause-specific mortality; OM: overall mortality.

  • a

    The MDM2 ACIS index indicator was coded as 0, cutoff point or lower; 1, higher than the cutoff point.

  • b

    A relative risk ratio of 1 indicates no difference between the two subgroups.

  • c

    P value was derived from the chi-square test using the Cox proportional hazards model.

LF≤ 3.059211.07 (0.57–2.03)0.8325.4(14.2–36.7)
LF> 3.04918  23.0(10.9–35.7)
DM≤ 3.059261.69 (0.98–2.91)0.05732.6(20.4–44.8)
DM> 3.04928  45.8(31.4–60.1)
CSM≤ 3.059191.29 (0.68–2.44)0.4318.6(8.6–28.7)
CSM> 3.04919  27.1(14.4–39.9)
OM≤ 3.059421.13 (0.72–1.77)0.5933.9(21.6–46.2)
OM> 3.04936  41.6(27.4–55.8)
Figure 2.

Survival curve of distant metastasis by MDM2 ACIS index at the 3% cutoff point, using cumulative incidence estimates. Solid line: ACIS index ≤ 3.0; dashed line: ACIS index > 3.0.

Table 7. Multivariate Analysis of Distant Metastasis; with the MDM2 ACIS Index Results
VariableaGroupRR (95% CI)bP valuec
  • RR: relative risk; 95% CI: 95% confidence interval; GLSC: Gleason score; STAD: short-term androgen deprivation;

  • a

    All variables were dichotomous.

  • b

    A relative risk ratio of 1 indicates no difference between the two subgroups.

  • c

    P value was derived from the chi-square test using the Cox proportional hazards model.

MDM2> 3.01.85 (0.97–3.56)0.06
GLSC7–102.70 (1.09–6.71)0.0328
STADYes0.89 (0.47–1.68)0.72
p53Positive2.68 (1.18–6.07)0.0181

Finally, MVAs for the MDM2 ACIS intensity score, modeled both as a continuous variable and by the median cutoff point (162 relative units), suggested a relation with DM when used as a continuous variable (continuous: P = 0.10; cutoff point: P = 0.97). The ACIS intensity score was not associated with any other end point.

DISCUSSION

MDM2 is a key regulator of apoptosis through its interactions with p53, E2F1, pRB, and other proteins.18, 19 We described recently that the apoptotic response of prostate carcinoma cells to AD and/or RT was significantly affected by the level of MDM2 expression10, 11 in prostate carcinoma cell lines. Antisense MDM2 is available as a potential therapeutic adjunct to AD and RT. We investigated the expression of MDM2 in men treated with RT, with and without STAD, to determine whether MDM2 overexpression is predictive of patient outcome, as a prelude to targeting men with antisense MDM2 in future trials.

To our knowledge, little is known regarding the abnormal expression of MDM2 in prostate carcinoma, as it relates to other prognostic variables and patient outcome. Osman et al.7 found MDM2 overexpression in 33% of 86 patients who received radical prostatectomy. MDM2 expression was not related to p53 expression, but was associated with advanced stage. No relation was observed between MDM2 expression and biochemical failure. Leite et al.8 found that MDM2 was overexpressed in > 40% of 118 men who underwent radical prostatectomy and such overexpression was associated strongly with increased tumor volume (P = 0.001) and weakly with a higher proliferation index (P = 0.046) and higher tumor stage (P = 0.054). MDM2 was not associated with p53. In our study, which is the first to investigate such relations in men treated with RT with or without STAD, nuclear MDM2 overexpression was observed in 44% of patients. MDM2 overexpression also was not related to p53 expression, but was related to a higher Gleason score and weakly to DM. The lack of a relation between p53 and MDM2 expression is possibly caused by a lack of feedback from mutant p53.

Our analysis of MDM2 expression was performed in 2 phases. First, we performed manual counts, assigning an incidence of > 5% tumor nuclear positivity to represent overexpression. Of the 3 previous reports of men with prostate carcinoma, the categorization of positive overexpression ranged from any nuclear staining to > 5% to ≥ 20%.7, 8, 13 A value of > 5% seemed reasonable, as this could be easily recognized as being above background. However, the cutoff point of 5% is somewhat arbitrary.

The determination of the relation of the percentage tumor cells demonstrating MDM2 staining was quantified more precisely using an image analysis system. The resulting ACIS index, while correlating with the manual count results, was more strongly related to the outcome measure of DM. The median ACIS index was 3% and this was chosen as the cutoff point. The ACIS index at the median cutoff point was related to DM in both univariate and multivariate analyses and MVA, although statistical significance at the P < 0.05 level was not obtained (P = 0.06). Likewise, the ACIS staining intensity was also related to DM, albeit more weakly. The results are promising in that such associations were seen even with relatively few patients and p53 included in the analysis.

The relation between MDM2 overexpression and DM in men treated with RT with or without AD may be clinically meaningful and should be further investigated in a larger cohort. The predictive value of MDM2 should also be investigated in a more contemporary group of men treated in the prostate-specific antigen era.

Ancillary