Loss of programmed cell death 4 expression marks adenoma-carcinoma transition, correlates inversely with phosphorylated protein kinase B, and is an independent prognostic factor in resected colorectal cancer

Authors

  • Giridhar Mudduluru MSc,

    1. Department of Experimental Surgery Mannheim/Molecular Oncology of Solid Tumors, Deutsches Krebsforschungszentrum and University Heidelberg, Mannheim, Germany
    Search for more papers by this author
    • The first three authors contributed equally to this article.

  • Fabian Medved,

    1. Department of Experimental Surgery Mannheim/Molecular Oncology of Solid Tumors, Deutsches Krebsforschungszentrum and University Heidelberg, Mannheim, Germany
    Search for more papers by this author
    • The first three authors contributed equally to this article.

  • Rainer Grobholz MD,

    1. Department of Pathology, University of Homburg, Saar, Germany
    Search for more papers by this author
    • The first three authors contributed equally to this article.

  • Camela Jost,

    1. Department of Experimental Surgery Mannheim/Molecular Oncology of Solid Tumors, Deutsches Krebsforschungszentrum and University Heidelberg, Mannheim, Germany
    Search for more papers by this author
  • Anette Gruber,

    1. Department of Experimental Surgery Mannheim/Molecular Oncology of Solid Tumors, Deutsches Krebsforschungszentrum and University Heidelberg, Mannheim, Germany
    Search for more papers by this author
  • Joerg H. Leupold PhD,

    1. Department of Experimental Surgery Mannheim/Molecular Oncology of Solid Tumors, Deutsches Krebsforschungszentrum and University Heidelberg, Mannheim, Germany
    Search for more papers by this author
  • Stefan Post MD,

    1. Department of Surgery, Mannheim Faculty, University of Heidelberg, Mannheim, Germany
    Search for more papers by this author
  • Aaron Jansen PhD,

    1. Gene Regulation Section, Basic Research, National Cancer Institute, Bethesda, Maryland
    Search for more papers by this author
  • Nancy H. Colburn PhD,

    1. Gene Regulation Section, Basic Research, National Cancer Institute, Bethesda, Maryland
    Search for more papers by this author
  • Heike Allgayer MD, PhD

    Corresponding author
    1. Department of Experimental Surgery Mannheim/Molecular Oncology of Solid Tumors, Deutsches Krebsforschungszentrum and University Heidelberg, Mannheim, Germany
    • Department of Experimental Surgery/Molecular Oncology of Solid Tumors (Collaboration Unit, German Cancer Research Center, Deutschen Krebsforchungszentrum-Heidelberg), Mannheim Medical Faculty, Ruprecht Karls University-Heidelberg, 68167 Mannheim, Germany
    Search for more papers by this author
    • Dr. Allgayer was supported by Alfried-Krupp-von-Bohlen-und-Halbach-Stiftung, Essen, Germany; Wilhelm-Sander-Stiftung, Munich, Germany; Auguste-Schaedel-Dantscher-Stiftung, Garmisch, Germany; and Hella-Bühler-Stiftung, Heidelberg, Germany; and Dr. Ingrid zu Solms Foundation, Frankfurt, Germany

    • Fax:(011) 49 621 383 3809/1938


  • This article is a US Government work and, as such, is in the public domain in the United States of America.

  • The current study contains the dissertation of Fabian Medved, which was performed in partial fulfillment of the “DrMed” degree at Mannheim Faculty, University of Heidelberg.

Abstract

BACKGROUND.

Programmed cell death 4 (Pdcd4) inhibits malignant transformation, and initial studies of Pdcd4 suggested the regulation of Pdcd4 localization by protein kinase B (Akt). However, supporting patient tissue data are missing, and the diagnostic/prognostic potential of Pdcd4 rarely has been studied. The objectives of the current were 1) to determine Pdcd4 as a diagnostic marker in the adenoma-carcinoma sequence, 2) to support phosphorylated Akt (pAkt)-mediated Pdcd4 regulation in vivo, and 3) to obtain the first prognostic evidence of Pdcd4 in colorectal cancer.

METHODS.

Tumor samples and normal tissues from 71 patients with colorectal cancer who were followed prospectively (median follow-up, 36 months) and 42 adenomas were analyzed for Pdcd4, Akt, and pAkt in immunohistochemical and Western blot analyses.

RESULTS.

A significant reduction in Pdcd4 was observed between normal mucosa and adenomas and between adenomas and tumor samples (P < .01 and P < .01, respectively). Normal mucosa demonstrated strong nuclear Pdcd4, which was reduced significantly in adenomas (P < .01) and almost was lost in tumors (P < .01). pAkt was correlated inversely with Pdcd4 and with the transition of Pdcd4 from nucleus to cytoplasm (P < .01). Kaplan-Meier analysis (using the Mantel-Cox log-rank test) indicated a significant correlation between the loss of total and nuclear Pdcd4 in tumors and overall survival (P < .05 and P < .02, respectively) and disease-specific survival (P < .01 and P < .01, respectively). In multivariate analysis, loss of total or nuclear Pdcd4 was an independent predictor of disease-specific or overall survival.

CONCLUSIONS.

To the authors' knowledge, this is the first study to demonstrate an independent prognostic impact of Pdcd4 and its expression pattern in colorectal cancer. Data from this study support the regulation of Pdcd4 localization by pAkt in vivo. Pdcd4 immunohistochemistry may be useful as a supportive diagnostic tool for the transition between normal, adenoma, and tumor tissues. Cancer 2007. Published 2007 by the American Cancer Society.

Programmed cell death 4 (Pdcd4) has been characterized as a new tumor suppressor gene. Its overexpression reportedly inhibited tissue polypeptide antigen-induced neoplastic transformation1, 2 and tumor promotion/progression in a transgenic mouse model.3 In the JB6 mouse epidermal system, it was identified as a 64-kilodalton (kD) protein, because it was expressed preferentially in tumor promoter-resistant cells.1 Murine Pdcd4 combinational DNA (identical to MA-3, TIS)4, 5 exhibits high homology to the human genes H731 and 197/15.6, 7 The human gene is located at chromosome10q248 and is expressed, for example, in human interleukin 12-induced natural killer cells and T cells,6 epithelial cells of the normal mammary gland,7, 8 normal human lung tissue,9 and senescent human fibroblasts.10 Pdcd4 has been linked to apoptosis in response to different inducers.4, 11 Molecules regulated by Pdcd4 include Waf/Cip1,12 Cdk4, ornithine-decarboxylase (at protein levels, not activity or transcriptional activation),2, 3 carbonic anhydrase-II,13 and C-Jun-amino-terminal kinase (JNK)/c-Jun/adaptor-related protein complex 1 (AP-1).14, 15 Pdcd4 interacts with translation-initiation factors eukaryotic initiation factor 4A (eIF4A)/eIF4G, inhibiting translation.16–18 It has been demonstrated that Pdcd4 is regulated by topoisomerase inhibitors,19 cyclooxygenase 2 inhibitors,11 and Myb.20

Recent studies also suggested that Pdcd4 is regulated by protein kinase B (Akt), one of the most important antiapoptotic regulators within the phosphatidylinositol 3-kinase pathway.21 In transfection studies, it has been observed that Akt phosphorylates overexpressed Pdcd4 at serine 67 (Ser67)/Ser457, and Ser457 specifically regulates the nuclear localization of Pdcd4.21 In addition, it has been demonstrated that Ser457 of Pdcd4 is important for the ability of Akt to interfere with the inhibition of AP-1 activity caused by Pdcd4.21 However, those studies were conducted with overexpressed Pdcd4 and, to date, have not been supported by data from, for example, tissues of cancer patients.

We and others recently demonstrated that the urokinase-receptor (u-PAR), a molecule that essentially promotes invasion, metastasis, and tumor progression,22, 23 is down-regulated in cultured colon cancer by Pdcd4. However, although it has been demonstrated convincingly that Pdcd4 suppresses transformation, tumorigenesis, and invasion,15, 24 with the exception of a single study in lung cancer,10 to date, Pdcd4 has not been investigated as a prognostic factor. The potential of Pdcd4 as a diagnostic marker to differentiate benign from malignant tissues also has been addressed initially in only a very few studies.12

Therefore, we performed this first prognostic study on Pdcd4 in a series of 71 patients with colorectal cancer who were followed prospectively to determine the levels of Pdcd4 protein expression in resected tumors and corresponding normal tissues independently with 2 different methods, putting emphasis on the cellular distribution pattern, as suggested by immunohistochemistry [IHC]. Given the recent in vitro studies on the regulation of Pdcd4 by Akt, we also performed Akt/phosphorylated Akt (pAKT) staining to support the regulation of Pdcd4 by data in a subset of resected tumors. Finally, we compared the levels of Pdcd4 expression in normal tissues, 42 colonic adenomas, and colorectal cancer. Consequently, we present the first clinical study to our knowledge suggesting Pdcd4, along with the loss of nuclear Pdcd4 staining, as a new and independent risk factor for recurrence and survival in patients with resected colorectal cancer and correlating Pdcd4 nuclear localization inversely with nuclear pAkt. Moreover, we suggest the intracellular Pdcd4 staining pattern as a diagnostic marker to define the colonic adenoma-carcinoma transition.

MATERIALS AND METHODS

Patients and Tumors

Seventy-one patients who were followed prospectively underwent surgery for colorectal cancer between March 1999 and October 2003. Patient/tumor characteristics are shown in Table 1. This population represents the usual distribution of colon cancer in Germany. Follow-up (physical examination, ultrasound, chest x-ray, tumor marker carcinoembryonic antigen) was performed 6 months, 12 months, 18 months, and 24 months after surgery and at 1-year intervals thereafter. Adjuvant therapy was administered to 17 patients. Recurrence was diagnosed by biopsy/explorative surgery. Nutritional status/performance scores were evaluated but were not associated with prognosis, Pdcd4 status, or pAkt status. Causes of death were evaluated by autopsy if feasible or thorough clinical and apparative examination by the physician in charge. Adenomas (endoscopy) of 42 cancer-free patients (29 men, and 13 women) were included: The majority of those patients (n = 41) had tubular adenoma, and 1 patient had a tubulovillous adenoma. Fourteen of the patients who had adenoma presented with clinical symptoms (diarrhea, weight loss, hemorrhage, obstipation). In addition, the patients with adenoma were screened for carcinoma, and all had negative results. Tissue screening was approved by the institutional ethics committee and was performed with patients' informed consent. Specimens were collected after verification by a pathologist and immediately were frozen in liquid nitrogen.

Table 1. Patient and Tumor Characteristics (n = 71)
CharacteristicAbsolute no.Relative, %
  1. UICC indicates International Union Against Cancer.

Sex
 Men4157.7
 Women3042.3
Localization
 Cecum/ascending colon2028.2
 Transverse colon22.8
 Descending colon/sigmoid colon55
 Rectum4462
Intention of surgical resection
 Curative (R0)5881.7
 Palliative (R1-R2)1318.3
UICC stage
 I1419.7
 IIA1723.9
 IIIA57
 IIIB1521.1
 IIIC710
 IV1318.3
Duke classification
 A1419.7
 B1723.9
 C2738.1
 D1318.3
Grade
 1–211.4
 25577.5
 2–31014.1
 322.8
 X34.2
Pathologic tumor classification (pT)
 pT134.2
 pT21926.8
 pT34664.8
 pT434.2
Pathologic lymph node classification (pN)
 pN03447.9
 pN12433.8
 pN21318.3
Metastasis classification (M)
 M05881.7
 M11318.3
Lymphangiosis carcinomatosa
 No3346.5
 Yes3853.5

IHC

IHC was performed as described previously.25 Briefly, expression levels of Pdcd4, Akt, and pAkt were detected by using rabbit Pdcd4 antibody (a gift from Nancy Colburn1), Akt (no. sc-1618; Santa Cruz Biotechnology, Inc., Santa Cruz, Calif), and pAkt (Ser473; no. 3787; Cell Signaling Technology, Beverly, Mass) at 4°C overnight followed by goat-antirabbit (biotinylated immunoglobulin G [IgG]) secondary antibody at room temperature for 30 minutes. Pdcd4 slides were processed with avidin-biotin Vector Elite Complex SP-2001 (Vector Laboratories, Burlingame, Calif) with 3,3′-diaminobenzidine (Sigma, St. Louis, Mo). Akt/pAkt slides were incubated with NovaRED according to the manufacturer's protocol (Vector Laboratories). Slides were counterstained with 10% hematoxylin. One section with isotype-nonspecific IgG and a section with C-terminal Pdcd4 peptide as a competitor served as negative controls, and tumors from the RKO cell line (no. CRL-2577; American TypeCulture Collection; Manassas, Va) with known Pdcd4, Akt, pAkt expression served as positive controls.

All slides were coded and evaluated without knowledge of the patient by 2 independent investigators (R.G. and F.M.). According to our previous experience,25 we classified the nuclear Pdcd4, Akt, pAkt staining results and the cytoplasmic Akt and pAkt staining results into 4 groups according to the percentage of positively stained nuclei: score 1, negative; score 2, ≤30% of nuclei stained positive [weak staining]; score 3, from 30% to 70% of nuclei stained positive [intermediate staining]; score 4, ≥70% of nuclei stained positive [strong staining]). Separate scores were applied to tumor cells and normal epithelial cells. Cytoplasmic Pdcd4 scores were given according to staining intensities (score 1, negative staining intensity; score 2, weak staining intensity; score 3, intermediate staining intensity; score 4, strong staining intensity) in reference to a strongly stained tissue as a control. For Pdcd4 and pAkt, a total score was calculated by the addition of the nuclear and cytoplasmic scores, and 4 groups were defined: negative (total score, 1 and 2), weak staining (total score, 3 and 4), intermediate staining (total score, 5 and 6) and strong staining (total score, 7 and 8).

Western Blot Analysis for Pdcd4 Levels

The same tissues were used for IHC and Western blot analyses. Samples and blots were prepared as described previously24; briefly, 50 μg protein were separated on a 10% acrylamide gel and transferred to a polyvinylidene fluoride membrane. The membrane was incubated with Pdcd4 antibody for 3 hours, then with by horseradish peroxidase-coupled goat-antirabbit secondary antibody for 2 hours, and chemiluminescence (ECL; Amersham). Band intensities were quantified by densitometry (AIDA; Germany) in reference to RKO cells.

Statistical Analysis

Analyses were performed by using SPSS software (version 14.0; SPSS Inc., Chicago, Ill) with a significance level of α = .05; a statistical trend was defined as α ≤ .1. Differences in quantitative variables between tissues from the same patients were tested with 2-sample paired Wilcoxon tests. The Mann-Whitney test was used to compare quantitative variables between groups. Spearman correlations among continuous variables were computed. Bonferroni-corrected chi-square tests were applied for grouped/dichotomized variables. Survival was estimated by using the Kaplan-Meier method and Mantel-Cox log-rank statistics. The primary endpoints were tumor-related death (disease-specific survival), overall survival (death), tumor recurrence (only for patients who achieved a curative resection [R0]). Multivariate analysis (Cox proportional hazard) included established clinical risk factors in colorectal cancer. Variables that were dichotomized included pathologic tumor stage (pT1/pT2 vs pT3/pT4), pathologic lymph node status (pN0 vs pN1-pN3), metastasis classification (M0 vs M1), histologic grade (grades 1, 1–2, and 2 vs grades 2–3 and 3), lymphangiosis carcinomatosa (present vs absent), and surgical curability (R0 vs noncurative resection).

RESULTS

Patient Results

Seventy-one patients (including 58 patients who underwent R0) (Table 1) with primary colorectal cancer and 42 patients with primary colonic adenoma were analyzed for Pdcd4. A subset of 16 patients with cancer was analyzed for total and pAkt. The median follow-up was 36 months (range, 1–72 months). Tumor recurrence was observed in 6 patients who underwent R0. In total, 13 patients died, and 12 of those deaths were caused by the tumor.

IHC for Pdcd4

IHC (Fig. 1a–c) was used to reveal whether Pdcd4 staining was nuclear or cytoplasmic. The median total Pdcd4 IHC score was 5.5; of 71 normal tissue samples, 68 samples had a total Pdcd4 score greater than the median. In contrast, only 3 tumor tissue samples had a total Pdcd4 score >5.5 (P < .01). Of the 71 normal colonic epithelial samples, all samples showed nuclear staining for Pdcd4, and the majority (68 samples) had high staining scores of 3 or 4 (Fig. 1a). In contrast, in the 71 tumor tissue samples, only weak nuclear Pdcd4 staining was observed in 13 samples, and the majority (58 samples) had a complete loss of nuclear staining (Fig. 1c). Moreover, the majority of tumor tissues had a weak cytoplasmic staining pattern for Pdcd4 (48 of 71 samples) in contrast to normal mucosa. In the adenoma samples, an intermediate situation was observed: 25 of 42 samples had weak to intermediate nuclear staining (score, 2/3), and 34 samples had intermediate cytoplasmic staining (Fig. 1b). Differences in exhibiting a nuclear or cytoplasmic staining pattern between tumors, adenomas, and normal tissues were significant: The ratio between nuclear and cytoplasmic staining decreased significantly from normal tissues to adenomas and tumor tissues (P < .01 and P < .01, respectively; chi-square test). An overview on total nuclear and cytoplasmic staining scores comparing normal tissues, adenomas and tumors is provided in Figure 2a–c.

Figure 1.

Programmed cell death 4 (Pdcd4) immunostaining of normal colorectal tissues, colorectal adenomas, and carcinomas (original magnification, ×40). (a) Most normal colorectal tissues had strong nuclear and cytoplasmic staining. (b) In adenoma tissues, weak nuclear Pdcd4 staining and intermediate cytoplasmic Pdcd4 staining is observed. (c) In carcinomas, no nuclear Pdcd4 is detected, but weak or intermediate cytoplasmic Pdcd4 staining is detected compared with the staining in normal and adenoma tissues. (d and e) Phosphorylated protein kinase B (pAkt) immunostaining was performed on a subset of 16 normal tissues (d) and their corresponding tumor tissues (e). In tumor tissues (e), strong nuclear and cytoplasmic staining is observed compared with the staining in normal tissue, in which pAkt staining is weak (d), and overall total pAkt staining is stronger in tumor tissue compared with normal tissue. Each staining was performed along with a negative control.

Figure 2.

Immunohistochemical (IHC) staining scores for programmed cell death 4 (Pdcd4) and phosphorylated protein kinase B (pAkt). (a) Total Pdcd4 staining decreased significantly in adenomas compared with normal tissues (P < .01). Further significant decreases were observed in colorectal tumors compared with adenomas (P < .01). Differences between tumors and normal tissues were significant (P < .01). (b) Nuclear Pdcd4 staining decreased in adenomas compared with normal tissues and was lost completely in tumors. Differences between all 3 groups were significant (P < .01). (c) Cytoplasmic Pdcd4 staining decreased in adenomas compared with normal tissues and disappeared in tumors compared with adenomas (P < .01). Nuclear staining in tumors and normal tissues differed significantly (P < .01). (d) Nuclear pAkt staining increased in tumors compared with normal tissues (P = .02).

The distribution of nuclear staining for Pdcd4 in normal mucosa was comparable to the pattern reported by Goke et al.,12 with strong nuclear staining observed especially in the middle and apical portions of crypts. In adenomas, we detected mostly weak nuclear staining and intermediate cytoplasmic staining. In cancer tissues, the distribution pattern changed, revealing an almost complete loss of stained nuclei in carcinoma cells and staining of the cytoplasm in carcinoma cells. Some fibroblasts, lymphocytes, smooth muscle cells, and endothelia had positive nuclear staining for Pdcd4 in the stroma of tumor, normal, and adenoma tissues, as reported by others,7, 12 although that staining was not considered for IHC scoring.

Western Blot Analysis for Pdcd4

In Western blot analysis, Pdcd4 was identified as a main band between 60 kD and 65 kD (Fig. 3). Most often, 2 additional bands were observed below, as reported by Cmarik et al.,1 and were correlated with the intensity of the main band in each sample. Because of limitations of material, Western blot could be performed in a subset of 20 patients with colorectal cancer. There was a significant correlation between high-intensity Pdcd4 staining observed in Western blots and high IHC staining scores observed as either nuclear or cytoplasmic staining (P < .01). Protein amounts (as measured by densitometry) of Pdcd4 detected in Western blots were significantly higher in normal tissues than in corresponding tumor tissues, and the differences between tumor and normal tissues were significant (P < .03).

Figure 3.

Western blot analysis of total programmed cell death 4 (Pdcd4) (representative data from 20 patients). High total Pdcd4 staining was detected in normal tissues (N) compared with the staining in tumor tissues (T). Differences between normal and tumor tissues in all 20 patients analyzed were significant (P < .03; Wilcoxon test). Total lysate cells from the RKO cell line (R) served as positive controls.

Correlation With Established Patient and Tumor Characteristics

To assess the association of Pdcd4 IHC with established clinical parameters, a Bonferroni-corrected chi-square analysis was performed with parameters classified as described above. The loss of Pdcd4 in tumors was correlated significantly with the presence of distant metastasis (P = .03), pT stage (P = .03), lymphangiosis (P = .02), an increased probability of noncurative surgical resection (P = .02), and a trend toward significance with pN status (P = .05), suggesting an association of loss of Pdcd4 with clinical tumor progression. No significant correlations were observed for sex, tumor localization, or other clinical tumor parameters.

IHC for pAkt and Inverse Correlation of Pdcd4 With pAkt

Normal and tumor tissue samples from 16 patients with colorectal cancer also were stained for total Akt and pAkt (Fig. 1d and e). Seven tumor samples demonstrated strong nuclear pAkt staining (score, 4), and 8 tumor samples showed weak or intermediate nuclear pAkt staining (score, 2 or 3) (Fig. 1e) compared with the corresponding normal tissue samples, which demonstrated mostly negative or weak nuclear pAkt staining (P = .02) (Figs. 1d, 2d). There also was a trend toward higher cytoplasmic pAkt staining scores in tumor samples compared with the scores in normal tissue samples (P = .07). In tumor samples, there was a significant gain of total pAkt (combined nuclear and cytoplasmic scores) compared with normal tissue samples (P = .03). Total Akt staining was stronger in the nuclei of tumor cells; however, no significant differences in staining scores between normal tissue samples and tumor samples were observed for total Akt (data not shown).

Next, this subset of patients was analyzed for correlations between Pdcd4 staining and pAkt staining. We observed a significant, inverse correlation between nuclear Pdcd4 and nuclear pAkt staining in tumor samples (P < .01), in that low nuclear Pdcd4 levels were associated with high nuclear pAkt levels. In addition, the transition from nuclear to cytoplasmic Pdcd4 staining (as expressed as the nuclear:cytoplasmic ratio) was associated significantly with high pAkt staining in the nucleus (P < .01) and cytoplasm (P < .01). In tumor samples, cytoplasmic Pdcd4 levels were correlated inversely with cytoplasmic pAkt (P < .01) and total pAkt (P < .01). There was no correlation between Pdcd4 and total Akt staining. Taken together, these data suggest that active pAkt is associated inversely with Pdcd4 in both cellular compartments and with the transition of Pdcd4 from the nucleus to the cytoplasm in colorectal tumor tissue.

Univariate Prognostic Impact of Pdcd4 in the Colorectal Cancer Series

In a univariate Kaplan-Meier analysis (Mantel-Cox log-rank test; median follow-up, 36 months), the decrease in total Pdcd4 measured by IHC (the total Pdcd4 score) was found to be associated significantly with poor recurrence-free survival (P < .05; R0 patients), disease-specific survival (P < .02) (Fig. 4a), and overall survival (P < .05) (Fig. 4b). In addition, total Pdcd4 protein, as measured by Western blot analysis in a subset of patients, also was associated with poor disease-specific survival (P = .03) and overall survival (P = .01) in all patients (Fig. 4c). Moreover, the transition of Pdcd4 from the nucleus to the cytoplasm in tumor tissue was correlated significantly with disease-specific survival (P < .01) (Fig. 4d) and overall survival (P < .01) for all 71 patients. These data suggest that the loss of total Pdcd4 and the transition from nuclear to cytoplasmic Pdcd4 confer a poor prognosis in patients with colorectal cancer.

Figure 4.

Kaplan-Meier survival analyses (Mantel-Cox log-rank test). (a) Disease-specific survival (70 patients) decreased with the loss of total programmed cell death 4 (Pdcd4) (immunohistochemistry [IHC]; P < .02). Patients were dichotomized at the median (3) of the total Pdcd4 IHC score (score ≤3 [41 patients]: deaths, 11; mean survival [MS], 56.88 months; standard error [SE], 3.88 months; score >3 [29 patients]: deaths, 1; MS, 63.28 months; SE, 1.68 months). (b) Overall survival (71 patients) decreased with the loss of total Pdcd4 (IHC; P < .05). Patients were dichotomized at the median (3) of the total Pdcd4 score (score ≤3 [41 patients]: deaths, 11; MS, 56.88 months; SE, 3.88 months; score >3 [30 patients]: deaths, 2; MS, 61.13 months; SE, 2.66 months). (c) Overall survival (n = 20 patients) according to total Pdcd4 protein levels in tumors compared with normal tissues (Western blot analysis). The Pdcd4 tumor:normal ratio was dichotomized at a median of 0.8 (P = .01). Top: Pdcd4 tumor:normal ratio >0.8 (n = 10 patients; deaths, 0; MS, 41.63 months; SE, 2.88 months). Bottom: Pdcd4 tumor:normal ratio <0.8 (n = 10 patients; deaths, 5; MS, 25.45 months; SE, 2.14 months). (d) Disease-specific survival (70 patients) decreased with the change from nuclear to cytoplasmic Pdcd4 staining (P < .01). Patients were dichotomized at the median (0.5) nuclear:cytoplasmic ratio for Pdcd4 immunostaining (ratio ≥0.5: n = 61 patients; deaths, 8; MS, 64.74 months; SE, 2.40 months; ratio <0.5: n = 9 patients; deaths, 4; MS, 27.39 months; SE, 4.90 months).

Multivariate Analysis of Pdcd4

Finally, we performed a multivariate analysis, which included established clinical risk factors in colorectal cancer, of immunohistochemical Pdcd4 staining. Parameters were included in the Cox proportional-hazards model if they were significant in the univariate analysis. Table 2 summarizes the results from multivariate analyses. Loss of total Pdcd4 was a new, independent predictor of disease-specific survival (P = .03; hazard ratio [HR], 1.313; 95% confidence interval [95% CI], 1.004–1.718). For overall survival, loss of total Pdcd4 showed a trend toward significance (P = .06; HR, 1.560; 95% CI, 1.047–2.664). Furthermore, loss of nuclear Pdcd4 staining in tumors was a new, independent prognostic factor for disease-specific survival (P < .01; HR, 1.474 [95% CI, 1.141–1.913]) and overall survival (P < .01; HR, 1.237 [95% CI, 1.054–1.451]) for all patients along with surgical curability (R0 vs noncurative resection), pTNM stage, and lymphangiosis carcinomatosa. Taken together, these data suggest that loss of total Pdcd4 and the transition from nuclear to cytoplasmic Pdcd4 are independent predictors of poor disease-specific and overall survival.

Table 2. Significant Multivariate Analyses of Pdcd4 Staining in Tumor Tissue*
ParameterPHR95% CI
  • HR indicates hazard ratio; 95% CI, 95% confidence interval; pT, pathologic tumor classification; pN, pathologic lymph node classification; NS, nonsignificant; M, metastasis classification; surgical curability, curative versus noncurative resection.

  • *

    Shown are either the total Pdcd4 values or the transition from nuclear to cytoplasmic Pdcd4 (Pdcd4 ratio: tumor core/cytoplasm). Trends are shown in parentheses.

Disease-specific survival: Total Pdcd4
 Total Pdcd4 (score 0 vs 1).0351.3131.004–1.718
 Lymphangiosis carcinomatosa.0321.2111.003–1.671
 pT(.054)2.6251.942–7.392
 pNNS
 MNS
 Surgical curabilityNS
Overall survival: Total Pdcd4
 Total Pdcd4 (score 0 vs 1)(.065)1.5601.047–2.664
 Lymphangiosis carcinomatosa.0441.2511.002–2.861
 pT.0751.7001.082–2.601
 pNNS
 MNS
 Surgical curabilityNS
Disease-specific survival: Pdcd4 ratio
 Pdcd4 ratio (score 0 vs 1).0031.4741.141–1.913
 Lymphangiosis carcinomatosa.0091.0141.003–1.392
 pT.0261.4521.052–2.003
 pN.0311.2741.023–1.572
 MNS
 Surgical curability.0231.5231.064–2.183
Overall survival: Pdcd4 ratio
 Pdcd4 ratio (score 0 vs 1).0091.2371.054–1.451
 Lymphangiosis carcinomatosa(.060)
 pTNS
 pNNS
 MNS
 Surgical curability.0111.3671.075–1.739

DISCUSSION

To our knowledge, this is the first prognostic study to implicate the loss of tumor suppressor Pdcd4 as a new and independent risk factor in colorectal cancer. In addition, the current results indicate that there is a significant loss of nuclear Pdcd4 from normal tissues to colonic adenomas and carcinomas, supporting the notion that IHC for Pdcd4 intracellular localization may be an additional, helpful diagnostic tool to discriminate benign from malignant colonic tissue. Furthermore, in this study, we demonstrated an inverse correlation between nuclear Pdcd4 and nuclear pAkt, which recently has been implicated in cell lines to regulate transfected Pdcd421 in an in vivo setting of resected patient tumors.

Pdcd4 has been identified as a suppressor of transformation,1, 2, 17 tumorigenesis, progression,3 invasion, matrix-metalloproteinase activation,15, 24 and tumor growth9 and as an inducer of apoptosis4, 11 and senescence.10 Specifically, we demonstrated previously that Pdcd4 suppresses the invasion- and progression-related molecule u-PAR at the transcriptional level and inhibits the metastatic process26 of cultured colon cancer at 2 different steps: invasion and intravasation.24 In addition, in a recent study, Dorrello et al.27 demonstrated that mitogen-induced activation of the S6-kinase1 pathway phosphorylates Pdcd4 at Ser67, leading to an interaction of Pdcd4 with the ubiquitin ligase βtransducin repeat-containing protein 1, which, in consequence, leads to ubiquitin-mediated Pdcd4 degradation. This degradation of Pdcd4 was necessary for efficient protein translation, cell growth, and proliferation, which are crucial prerequisites for tumor progression. These functions of Pdcd4, in terms of preventing carcinogenesis and progression,3 are supported considerably by the current work, which demonstrated an overall decrease in mean Pdcd4 protein levels and a transition from a nuclear staining pattern to a cytoplasmic staining pattern from normal tissue, to adenomas, and further to carcinomas, supporting a significant role of Pdcd4 in carcinogenesis. The significant clinical and prognostic value of Pdcd4, its overall loss, and its transition from the nucleus to the cytoplasm considerably support a fundamental biologic function of Pdcd4 in carcinogenesis and progression, contributing to the long-term clinical course of patients with colorectal cancer.

To our knowledge, the only prognostic study on Pdcd4 to date was performed for a series of lung cancer patients9 in whom the loss of Pdcd4 protein was associated with poor survival in a univariate analysis. That study also indicated a loss of Pdcd4 staining in lung cancer samples, corroborating our observation in colorectal cancer, with staining signals observed in nuclei and cytoplasm. However, the study evaluated IHC staining mainly in tissue-array samples, which may provide less of an overview on intralesional staining patterns than large tissue sections. Second, that study presented only 1 Kaplan-Meier analysis of overall survival and did not describe the impact on recurrence-free or disease-specific survival. Furthermore, those investigators did not perform a multivariate analysis, which would have addressed prognostic relevance independent of established parameters.

The loss of a nuclear staining pattern for Pdcd4 in tumor tissues, its negative correlation with pAkt staining in the nucleus, and potential mechanisms that may explain these phenomena merit discussion. There are reports corroborating our own observation of a loss of nuclear staining in tumors compared with normal tissues. For example, Goke et al.12 reported that 6 of 7 colon carcinomas examined showed a complete loss of nuclear Pdcd4 staining that was paralleled by additional cytoplasmic staining in 4 tumors. In the mouse JB6 preneoplastic cell line, Yang et al.17 observed both nuclear and cytoplasmic localization of Pdcd4 with a perinuclear emphasis of staining. With regard to potential mechanisms underlying nuclear or cytoplasmic Pdcd4 staining, it has been demonstrated in vitro that Pdcd4 has a predominantly nuclear localization under normal growth conditions but is exported to the cytoplasm upon serum withdrawal.28 Therefore, one hypothesis that partly explains our observation may be changes of metabolism or microenvironment conditions because of tumor transition. Palamarchuk et al.21 transfected 293 cells that stably expressed wild-type Pdcd4 and observed a primarily nuclear Pdcd4 staining pattern. However, an S457A mutant of Pdcd4 that was unable to be phosphorylated by Akt, in contrast to the wild type, was localized no longer to the nucleus but to the cytoplasm, indicating the importance of this amino acid and Akt phosphorylation site for the localization of Pdcd4. The antibody for IHC that we used in the current study recognizes an epitope at the C-terminus, including position 457 of Pdcd4, and leads to positive staining whether or not position 457 is mutated to A. Therefore, a second explanation for the loss of nuclear Pdcd4 staining may be a mutation of this amino acid of Pdcd4 in tumor tissues, although Jansen et al.3 did not report any evidence for Pdcd4 mutations in NCI60 cell lines. Another mutant of a second Akt phosphorylation site, S67A-Pdcd4, primarily localized to the nucleus and was less important for the change of Pdcd4 localization brought about by Akt phosphorylation. However, as discussed above, this phosphorylation site has been identified as essential for the induction of ubiquitin-mediated Pdcd4 degradation.27 This may be a general explanation for the finding that total Pdcd4 decreased in tumor cells in our study and was paralleled by an increase of pAkt. Specifically, although Akt-dependent phosphorylation at S67 can occur both in the cytoplasm and in the nucleus, proteasome-mediated degradation occurs only in the cytoplasm. Therefore, high pAkt levels in the cytoplasm generally would trigger low cytoplasmic Pdcd4 through degradation but also would trigger nuclear Pdcd4 localization through S457 phosphorylation, unless the Pdcd4 is degraded first in the cytoplasm through S67 phosphorylation. A further observation of Palamarchuk et al. was that Pdcd4 phosphorylation by active pAkt at both sites (S67/S467) inhibited the tumor suppressor ability of Pdcd4 in terms of repressing AP-1-responsive promoters, regardless of Pdcd4 nuclear or cytoplasmic localization.21 This is in line with our current observation of an increase in active pAkt as an antiapoptotic, prosurvival molecule, and this was paralleled significantly by a decrease of Pdcd4 in resected tumor tissues compared with normal tissues. Because Pdcd4 can inhibit the JNK/c-Jun pathway both in the nucleus and in the cytoplasm,21 this function would be inhibited by pAkt regardless of a nuclear or cytoplasmic localization, which would match our findings of significant, inverse correlations between Pdcd4 and pAkt both in the nucleus and in the cytoplasm. Finally, because translation-inhibitory activity of Pdcd4 needs to occur in the cytoplasm, another hypothesis may be that shuttling of Pdcd4 to the nucleus is done by cells when there is an excess, rendering nuclear Pdcd4 (which is associated with inactivation, as demonstrated by Palamarchuk et al.) a quick source to mobilize when needed. Cancer cells, because of an overall decrease in total Pdcd4, may not send it to the nucleus; therefore, the lack of nuclear Pdcd4 may represent another read-out of low total cellular Pdcd4 levels, which would easily be visible with IHC. In any event, we must emphasize that the study by Palamarchuk et al.21 was performed with overexpressed Pdcd4, and their results cannot address the question of whether this mechanism is obtained with endogenous Pdcd4. Therefore, to our knowledge, our current in vivo results in resected tissues provide the first correlative data on this issue.

It is interesting to note that, in the current study, pAkt in tumors was localized to the nucleus in a considerable number of samples. In general, Akt reportedly is localized predominantly to the cytoplasm; however, there have been reports of an additional nuclear presence29 and of membrane translocations that depend on different modes of growth factor activation.30 Ahmed et al.29 reported that retroviral v-Akt, which, in contrast to complementary c-Akt, was identified as highly oncogenic in nude mice when it was transfected into 5675 cells, was localized 30% in the nucleus. Therefore, in the context of our current data, it is reasonable to speculate that an oncogenic form of pAkt, which is localized to the nucleus and leads to the deactivation and/or cytoplasmic translocation of Pdcd4, may appear in colorectal carcinomas. Certainly, these hypotheses, along with investigations on further mechanisms associated with the shuttling of Pdcd4 and the interaction pAkt/Pdcd4, need to be addressed further in future studies.

Taken together, the current results offer the first comprehensive data demonstrating the prognostic impact of the Pdcd4 tumor suppressor on clinical prognosis, recurrence, and survival in patients with colorectal cancer. In addition, these data suggest Pdcd4 staining as a promising additional diagnostic tool to help discriminate between normal tissues, benign adenomas, and colorectal carcinoma tissues. Therapeutic tools supporting the physiologic expression and/or function of Pdcd4, or suppressing pAkt (especially concerning Pdcd4 inhibition), therefore, may be promising strategies for the prevention of colorectal carcinogenesis and for significantly improving recurrence and survival rates among patients with colorectal cancer.

Ancillary