Determining Risk of Biochemical Recurrence in Prostate Cancer by Immunohistochemical Detection of PTEN Expression and Akt Activation

Prostate cancer is diagnosed by a prostate biopsy that, in most cases, is prompted by elevated serum levels of prostate-specific antigen (PSA; refs. 1–3). The majority of patients in the United States undergo prostatectomy as first-line treatment for prostate cancer. PSA levels decrease to undetectable levels in most patients following prostatectomy. For prostate-confined disease, it is estimated that the potential for cure, following prostatectomy, ranges between 80% and 96% (4). However, evidence of increasing PSA values (biochemical recurrence) following the PSA nadir occurs in ∼15% to 19% of patients and relates to recurrent prostate cancer (either local or metastatic), which is treated with secondary treatment modalities including androgen ablation, radiation, and chemotherapy (4). The ability to predict which patients will undergo biochemical recurrence would allow for focused examination and treatment of high-risk patients. Unfortunately, current diagnostic procedures, including measurement of PSA levels and histologic grading of the tumor (using the Gleason scoring method), do not distinguish between patients who will undergo biochemical recurrence versus those who will not. As a result, more men will be treated for prostate cancer than would be expected to die from the disease. Therefore, a number of laboratories, including ours, are currently focusing on the identification of predictive markers of recurrent prostate cancer.

We previously showed by immunohistochemistry in paraffin-embedded prostate tumor tissues obtained from patients with prostate cancer that increased phosphorylation of the serine threonine kinase Akt, a downstream target of the phosphatidylinositol 3-kinase (PI3K) pathway, is an important predictor of the risk of biochemical recurrence (5). We also showed increased Akt phosphorylation in high Gleason grade prostate cancer (6). In addition, Akt activation correlated with proliferation in human prostate tumors as estimated by the expression of the cell proliferation antigen Ki67 (7). Based on these results, we hypothesize an important role for Akt in prostate cancer recurrence.

Previous studies in prostate cancer showed that loss of expression of the tumor suppressor phosphatase and tensin homologue deleted on chromosome 10 (PTEN), a common event in prostate cancer, resulted in increased phosphorylation of Akt (8). PTEN, a dual lipid/protein phosphatase, acts as a tumor suppressor by inhibiting the kinase activities of critical tumor-promoting kinases such as PI3K. Loss of PTEN expression in primary prostate cancer correlated with high Gleason scores (9) and increased angiogenesis (10). As loss of PTEN expression was associated with increased phospho-Akt (pAkt) activation, we investigated whether the combination of these two factors was of greater predictive potential compared with Akt activation alone. Our results indicate that, although PTEN by itself is not a good predictor of PSA recurrence, loss of PTEN expression, in combination with increased Gleason scores and Akt phosphorylation, is an even better predictor of biochemical recurrence than Akt by itself.

Patients and tissues used. Prostate tumors were obtained from surgical specimens of patients at the South Texas Veterans Health Care System, Audie Murphy Veterans Administration Hospital, and the University Hospital, San Antonio, Texas. Clinical data were obtained from the Tumor Bank for Prostate Cancer, Department of Pathology, University of Texas Health Science Center at San Antonio, under an Institutional Review Board–approved protocol. A total of 65 specimens were used. PSA data available from patient follow-up were evaluated to assess “good” and “poor” outcome following radical prostatectomy. Cases with a poor outcome were defined as follows: PSA was detectable and increased to >0.6 ng/mL within 60 months following radical prostatectomy, and a second value obtained confirms the elevated PSA. The patient was also included if he had no second value but was judged by his physician to have recurrent disease and was treated for recurrent disease without a second PSA. Cases with a good clinical outcome were defined as follows: (a) PSA remained <0.2 ng/mL for at least 60 months; (b) there was no other evidence of recurrent disease at time of selection. Tissue arrays were prepared using a 4-mm punch biopsy, placing ∼20 cores in each block.

Antibodies used. The immunohistochemistry studies described used a rabbit monoclonal anti-PTEN antibody, clone 138G6, and a rabbit polyclonal anti–phospho-Akt (Ser⁴⁷³) antibody (Cell Signaling Technology). A rabbit polyclonal anti-Akt (Santa Cruz Biotechnology), which recognizes both the phosphorylated and unphosphorylated forms of Akt, was also used.

Immunohistochemistry. Sections were heated to 60°C and rehydrated in xylene and graded alcohols. Antigen retrieval was done with 0.01 mol/L citrate buffer at pH 6.0 for 10 min in a 121°C pressure chamber. Slides were allowed to cool for another 30 min, followed by sequential rinsing in TBS-T [50 mmol/L Tris-HCl (pH 7.6), 150 mmol/L NaCl, Tween 20 (0.1%)]. Endogenous peroxidase activity was quenched by incubation in TBS-T containing 3% hydrogen peroxide. Each incubation step was carried out at room temperature and was followed by three sequential washes in TBS-T. Sections were incubated in primary antibody diluted in TBS-T containing 1% bovine serum albumin and 0.01% sodium azide (overnight), followed by incubations with biotinylated secondary antibody for 15 min, peroxidase-labeled streptavidin for 15 min (LSAB-2 System, DakoCytomation Corp.), and diaminobenzidine-substrate for peroxidase-based immunohistochemistry (DakoCytomation). Slides were counterstained with hematoxylin and mounted. Staining for pAkt has been described earlier (5, 6). For PTEN, positive and negative controls were blocks made from cell pellets of PTEN+/+ and PTEN−/− embryonic fibroblasts obtained from control and PTEN−/− knockout mice, kindly provided by Dr. Hong Wu (Molecular and Medical Pharmacology, University of California, Los Angeles, CA) and described elsewhere (11). For negative control, rabbit immunoglobulin fraction was used in place of the primary antibody.

Quantification of staining intensity. The degree of staining was evaluated blindly by a pathologist. Staining intensity was scored as follows: the score is 0 when none of the cells stained positively; 1 when there was weak staining (10-20%); 2 when there was moderate staining (20-50%); and 3 when there was strong staining (>50%). Scoring for pAkt was previously described (6).

Statistical analysis. Contingency tables with Fisher's exact test were used to compare the expression of pAkt and/or PTEN in Gleason or outcome groups (recurrent versus nonrecurrent). Logistic regression analysis was used to evaluate the individual and joint predictive value of the proposed markers for the outcome (12). Stepwise selection was used to identify significant predictors of PSA recurrence. The odds ratio (OR) was calculated to determine the potential risk of developing PSA recurrence for one-point increase in the value of the factor tested. Predictors with OR > 1 are considered risk factors whereas those with OR < 1 are protective factors. At OR = 1, the factor does not alter the predicted outcome. The interval estimation is given as a 95% confidence interval (95% CI), which defines the reference range for that variable (12). To determine discrimination of the markers, the sensitivity and specificity of the given data were identified (12). The sensitivity of a test is defined as the true positive rate (disease present when the test is positive) whereas the specificity of a test is defined as the true negative rate (disease absent when test is negative). The diagnostic value of the potential biomarkers as predictors of PSA recurrence was evaluated with receiver-operator characteristic (ROC) curves. The area under the ROC curve was determined from the plot of sensitivity versus 1 − specificity (true positive rate versus false positive rate) and is a measure of the predictability of a test. An area under the ROC curve of 0.8 to 0.9 is considered excellent discrimination whereas an ROC value of 0.5 indicates no discrimination (12). The Hosmer-Lemeshow goodness-of-fit test was conducted to see how well the data predict PSA failure (good fit of raw data with predicted data if P > 0.05). The diagnostic accuracy was assessed with a classification table having a positive outcome threshold of 0.58 based on specificity and sensitivity. The analysis was carried out by STATA (StataCorp.). Stepwise analysis was conducted using SAS version 9.1 for Windows (SAS Institute, Inc.).

PTEN expression and immunostaining. We and others previously showed that increased phosphorylation (activation) of the serine/threonine kinase Akt is of significant predictive value in determining risk of biochemical recurrence in prostate cancer (5, 13). Because loss of the tumor suppressor PTEN, a common event in prostate cancer (14, 15), is an important cause of increased Akt phosphorylation (8), we investigated how the expression of PTEN affects the predictive potential of pAkt. Sixty-five prostate tumor specimens obtained by radical prostatectomy from men with prostate cancer were used for this study. The patient demographics are described in Table 1. Specimens were immunostained with a rabbit polyclonal antibody to phospho-Akt (Ser⁴⁷³) as described earlier (5, 6). PTEN expression was detected with a rabbit monoclonal anti-PTEN antibody (Rm) and scored on a scale of 0 to 3. Figure 1 shows typical examples of strong (3+), moderate (2+), weak (1+), and absent (0) staining. PTEN staining was observed only in the epithelial cells, with the stroma almost completely negative for PTEN. In the epithelial cells, PTEN was mainly membranous or cytoplasmic, with the nucleus being free of PTEN expression, as shown by others for prostate epithelial cells (9).

PTEN expression by itself is not a good predictor of biochemical recurrence. Of the 65 patients, 34 were classified as “nonrecurrent” whereas the other 31 patients were classified as “recurrent” based on the criteria described in Materials and Methods. The mean PTEN intensity for the cases that were nonrecurrent was 0.8823 ± 0.327, whereas for those that were recurrent the mean was 0.6774 ± 0.475, but the difference in these two groups was not significant (Fisher's exact P = 0.069). The majority of patients staining negative for PTEN (intensity score = 0) would develop recurrent disease (71.43%); however, 41.2% of those staining positive for PTEN (intensity score = +1, +2, +3) would have biochemical recurrence also (Fig. 2A). Univariate logistic regression analysis for PTEN alone gives an OR of 0.882 (P = 0.592; Table 2), indicating that expression of PTEN alone was not a significant protective factor for recurrent prostate cancer. The area under the ROC curve for PTEN, which estimates its discriminatory capacity, was 0.54 (Fig. 2B). Thus, PTEN alone is of limited predictive value for determining the risk of biochemical recurrence.

Loss of PTEN expression together with increased activation of Akt is an excellent predictor of biochemical recurrence. The 65 specimens were also stained for total and phospho-Akt (Ser⁴⁷³). As we showed earlier (5), the staining for total Akt did not significantly differ between the two groups (not shown). In contrast, the mean pAkt staining intensity on a scale of 0 to 3 was significantly higher for the recurrent cases, at 2.4322 ± 0.6798, compared with the nonrecurrent cases, 1.3059 ± 0.9355 (Fisher's exact P = 0.0001, for low versus high intensity staining). Staining for pAkt was mostly predictive of outcomes: 85.7% of the specimens staining weakly for pAkt (intensity = 0, 1) would not develop recurrent prostate cancer (Fig. 2C). However, the presence of high pAkt (intensity = 2, 3) correlated less with outcome: only 63.6% of those staining strongly for pAkt would later experience recurrence (Fig. 2C). The univariate logistic regression analysis of pAkt (0-3 staining intensity) alone gives us an OR of 4.67 (P < 0.0001), indicating that increased pAkt by itself is a significant risk factor (Table 2). The area under the ROC curve for pAkt only was 0.82, indicating very good discrimination (Fig. 2D). The model had a good fit (Hosmer-Lemeshow χ² = 1.73, P = 0.78), with 73.85% correctly classified.

We now extended these studies to see whether the combination of PTEN and pAkt together had increased predictive potential for determining the risk of biochemical recurrence. Significantly, 90% of all the cases which showed high pAkt and negative PTEN had biochemical recurrence (Fig. 3A, bottom left), whereas 88.2% of the cases showing low pAkt and positive PTEN were nonrecurrent (Fig. 3A, top right), suggesting that the combination of these two factors is a stronger predictor of biochemical recurrence than either alone. Low levels of pAkt ensured nonrecurrence in spite of negative PTEN (Fig. 3A, top left), indicating that of the two, pAkt was the more decisive factor. However, the discriminatory ability decreased for cases where pAkt was high (increased risk) and PTEN was high as well (decreased risk; Fig. 3A, bottom right), indicating that the presence of PTEN, at least partly, protected from recurrence in the presence of high pAkt. This is due to a significant interaction effect of PTEN and pAkt on PSA failure.

Prediction of probability of PSA recurrence. Multivariate logistic regression analysis was conducted to determine the combinatorial effect of these two variables together with other factors such as Gleason score that influence outcome. Gleason scores alone were not predictive of biochemical failure, with an OR of 1.16 (P = 0.35) and an area under the ROC curve of 0.54 (not shown), but in combination with pAkt and PTEN, had increased predictive value. Other factors that had been considered but did not have enough power were PSA levels (pre- and post-surgery), tumor grade, surgical margins, age, and ethnicity of patients. Multiple logistic regression using stepwise selection resulted in pAkt, Gleason (linear and quadratic), and the interaction of pAkt and PTEN being significant predictors of PSA failure (Table 2). The area under ROC for the combination was 0.89, indicating excellent discrimination (Fig. 3B). A Hosmer-Lemeshow goodness-of-fit test indicated a good fit (Hosmer-Lemeshow χ² = 13.62, P = 0.16; sensitivity = 71%, specificity = 85.3%; 78.5% correctly classified). Table 2 shows that the OR (probability of recurrence) for one-point increase in pAkt increases from 4.67 (95% CI, 2.18-9.98) by univariate logistic regression to 14.4 (95% CI, 3.63-57.3) by multivariate logistic regression, indicating that the absence of PTEN and high Gleason scores significantly enhanced the risk of recurrence predicted by pAkt.

The above data were then used to develop a model for predicting the probability of PSA recurrence using known values of pAkt, PTEN, and Gleason scores. Table 3 shows the individual contributions of pAkt and PTEN and the influence these two factors have on each other. For negative PTEN, the OR for pAkt was 14.4 as indicated above (Table 3). A positive PTEN value, on the other hand, is protective and the OR for one-point increase in pAkt remains at 4.52, similar to the OR of 4.67 for univariate analysis for pAkt alone (Table 2). These results show that the presence or absence of PTEN greatly influences the risk predicted by an increase in pAkt. Similarly, univariate logistic regression for PTEN alone shows that PTEN is not a significant protective factor when considered by itself (OR, 0.88; Table 2), but using the stepwise model, PTEN becomes highly protective as an interaction effect depending on pAkt in a multivariate logistic regression analysis, with an OR of 0.31 for one-point increase in PTEN when pAkt is high, but has less influence when pAkt is low (OR, 0.68; Table 3).

The model was then expanded to take into consideration the effect of increased Gleason scores on the predicted probability of PSA failure. Based on a classification table with cutoff point of 0.58 predicted probability (not shown), we see that, for constant Gleason scores, an increase in pAkt increased the probability of biochemical recurrence, whereas an increase in PTEN resulted in a decreased probability of the same (Fig. 3C). At low pAkt, PTEN is less important; hence, the predicted probability of failure is low irrespective of PTEN. With high pAkt, however, PTEN is a significantly protective factor, partly protecting against the effect of increased pAkt. However, high Gleason scores resulted in reduced protection from decreased pAkt and increased PTEN (Fig. 3C), indicating that Gleason >7 is an overriding risk factor that greatly enhances the effects of pAkt and negates the protective effect of PTEN. As examples, we show here the predicted probability of biochemical recurrence for identical values of PTEN and pAkt at Gleasons 4, 7, and 9. Figure 3C shows that the predicted probability of biochemical recurrence is lower at Gleason 4 (Fig. 3C, left) or Gleason 7 (Fig. 3C, middle), compared with Gleason 9 (Fig. 3C, right) for identical values of PTEN and pAkt. This model therefore shows that increased pAkt and loss of PTEN, together with Gleason scores, are of excellent predictive potential for determining risk of biochemical recurrence.

We previously showed that increased activation of the serine/threonine kinase Akt in the prostatic epithelium is a very good predictor of biochemical recurrence following prostatectomy (5). In this study, our results indicate that, although PTEN alone did not predict PSA recurrence, loss of PTEN expression, combined with increased phosphorylation of Akt and Gleason scores, is an even better indicator of biochemical recurrence than Akt activation alone. Previous studies in prostate cancer showed that loss of PTEN expression is a common event in prostate cancer (14, 15), which correlated with high Gleason scores (9), and in combination with loss of p27^Kip1, was predictive of tumor recurrence (16). However, to the best of our knowledge, this is the first report studying PTEN in combination with pAkt as a predictive marker for biochemical recurrence in prostate cancer.

PTEN, also called mutated in multiple advanced cancers (MMAC1) and transforming growth factor β–regulated and epithelial cell–enriched phosphatase (TEP1), was identified simultaneously by three different laboratories as a tumor suppressor (17–19). PTEN is a dual lipid/protein phosphatase that antagonizes the action of PI3K and suppresses the phosphorylation (activation) of its downstream targets, including Akt (8, 20, 21). Whereas PI3K phosphorylates phosphatidylinositol 3,4,-trisphosphate (PIP2) to phosphatidylinositol 3,4,5-trisphosphate (PIP3), PTEN dephosphorylates PIP3 to PIP2 (22). We and others (9) observe PTEN exclusively in the cytoplasm of prostate epithelial cells, although in other tissues PTEN expression has also been reported in the nucleus (23).

Loss of heterozygosity at the PTEN gene was found to be a common occurrence in prostate cancer but few mutations have been detected in the remaining allele (24–27). However, reduced or absent PTEN expression at the mRNA or protein level is frequently observed (9, 24) and may be attributed, among other reasons, to transcriptional suppression by DNA hypermethylation (24). PTEN inactivation was associated with increased angiogenesis (10). In several animal models, loss of PTEN alone (28) or in combination with other factors such as FGF8b (29) resulted in varying levels of prostatic neoplasia, from prostatic intraepithelial neoplasia to metastatic prostate cancer. PTEN, alone or in combination with p27^Kip1, was predictive of Gleason scores (30) and increased risk of recurrence (16). Another study reported that PTEN expression was predictive of advance prostate cancer versus localized disease, but did not predict hormone refractory versus sensitive prostate cancer (31).

In vitro studies showed that loss of PTEN expression corresponds to increased Akt activation (8). Our data indicate that at low levels of Akt phosphorylation, the majority of cancers examined were nonrecurrent, irrespective of PTEN levels and Gleason scores. However, high pAkt did not automatically predict recurrent prostate cancer, with a third of the patients with high pAkt having no recurrence of the tumor. On the other hand, loss of PTEN by itself was a good indicator of tumor recurrence but the presence of PTEN did not ensure nonrecurrence. Thus, in combination, pAkt and PTEN were better predictors of both recurrence and nonrecurrence. PTEN and Gleason scores are important decisive factors for patients with high pAkt. PTEN is a protective factor: for negative PTEN, almost all patients with high pAkt levels had recurrent disease, whereas for those with positive PTEN, only ∼50% of patients with high pAkt had tumor recurrence. Our data also suggest that high Gleason scores greatly increased the risk of recurrence. Surprisingly, there was little difference in outcomes for Gleasons 4 and 7 for identical values of PTEN and pAkt, whereas for Gleason 9, the risk of recurrence was much higher. This observation is supported by previous reports showing that patients with high Gleason cancer had a more significant risk for recurrence (32–34).

Our data also show that loss of PTEN does not always cause high pAkt, and the presence of PTEN does not always suppress Akt phosphorylation. Whereas PTEN loss is frequently cited as a cause of increased Akt phosphorylation (8, 35, 36), it must be realized that PTEN not only acts as a suppressor of the PI3K/Akt pathway but, as a dual lipid/protein phosphatase, is instrumental in suppressing other kinases as well. For example, PTEN would also inhibit other kinases downstream of PI3K such as the mammalian target of rapamycin/p70S6 kinase pathway (37, 38) as well as kinases in other pathways such as focal adhesion kinase (39). These pathways have been previously found to play important roles in prostate cancer (40, 41). At the same time, Akt is phosphorylated not only by the loss of PTEN but also by other factors such as activation of receptor tyrosine kinases by overexpression or increased growth factor expression (42–45). Tumors in humans are caused by multiple factors, and loss of PTEN causes effects other than Akt activation, whereas Akt phosphorylation indicates activation of several pathways; hence, together they predict better the chances of biochemical recurrence.

Previous studies had indicated that other factors, including pathologic or clinical staging, PSA and surgical margins, affected biochemical recurrence free survival (32–34, 46–49). In support of these reports (48), we did not see any effect of age or race on risk of biochemical recurrence (not shown). Previous studies had shown that preoperative PSA levels did not affect the outcome significantly (33, 47), although PSA >20 ng/mL may have had a statistically significant effect (34). Pathologic or clinical staging had also been shown to affect risk of biochemical recurrence (34, 47) but lacked power in our study to be of any significance. In addition, the presence of positive surgical margins was considered to be a significant risk factor (34, 47). Inclusion of surgical margins in multivariate logistic regression analysis in our study, however, did not have a significant effect on the predictive ability of pAkt and PTEN (by logistic analysis using pAkt, PTEN, and surgical margins; OR_multivariate for pAkt = 6.389, P = 0.002, but OR not significant for the other variables). The area under the ROC curve including margins was 0.8695. Hence, our analysis indicates that Gleason scores seem to be the largest contributor to the ability of PTEN expression and Akt phosphorylation to predict biochemical recurrence in prostate cancer.

In conclusion, we have shown that loss of PTEN combined with increased Akt activation is an excellent predictor of biochemical recurrence in prostate cancer patients following prostatectomy. It may be noted that the tissues used had been formalin fixed and paraffin embedded with no specific standardization technique. Previous studies indicate that formalin fixation preserves phosphorylation sites (50), which is reasonable, given that formalin preserves the entire tissue in an intact form. With the evolution of new phospho-specific antibodies developed for immunohistochemistry use, it is now possible to standardize staining procedures for routine examination of these two factors in prostate cancer patients, which would enable better prediction of the course of the disease and focused examination and treatment of patients at higher risk.

We thank Dr. Ralph W. deVere White (Department of Urology, and Director, Cancer Center, University of California, Davis) for critical reading of the manuscript, advice and helpful suggestions; Dr. Hong Wu for PTEN+/+ and PTEN−/− cells used as positive and negative control for immunohistochemistry; and Dr. Graciela Gonzalez-Farias (CIMAT Gto., Mexico) for help with the statistical analysis.