Expression of p16 in lymph node metastases of adjuvantly treated stage III colorectal cancer patients identifies poor prognostic subgroups

A retrospective analysis of biomarkers in matched primary tumor and lymph node metastases

Authors

  • Eva Karamitopoulou MD,

    1. Second Department of Pathology, University of Athens, Attikon University Hospital, Athens, Greece
    Search for more papers by this author
    • The first two authors contributed equally to this article.

  • Inti Zlobec PhD,

    1. Institute of Pathology, University Hospital Basel, Basel, Switzerland
    Search for more papers by this author
    • The first two authors contributed equally to this article.

  • Anna Koumarianou MD,

    1. Medical Oncology Unit, Second Department of Internal Medicine Propaedeutic, University of Athens, Attikon University Hospital, Athens, Greece
    Search for more papers by this author
  • Efstratios S. Patsouris MD,

    1. First Department of Pathology, University of Athens, Athens, Greece
    Search for more papers by this author
  • George Peros MD,

    1. Fourth Department of Surgery, University of Athens, Attikon University Hospital, Athens, Greece
    Search for more papers by this author
  • Alessandro Lugli MD

    Corresponding author
    1. Second Department of Pathology, University of Athens, Attikon University Hospital, Athens, Greece
    • Institute of Pathology, University Hospital of Basel, Schaenbeinstrasse 40, Basel, 4031, Switzerland
    Search for more papers by this author
    • Fax: (011) 41 061 265 3194


Abstract

BACKGROUND:

The objective of identifying protein biomarkers for patients with stage III and IV colorectal cancer is to improve risk stratification and, thus, to identify patients in the postoperative setting who may benefit from more targeted treatment. The objective of the current study was to determine the prognostic value of 19 protein markers assessed in primary tumors and matched lymph node (LN) metastases from patients with stage III and IV colorectal cancer.

METHODS:

Matched primary tumors and LN metastases from 82 patients with stage III and IV colorectal cancer were mounted onto a multiple-punch tissue microarray and were stained for 19 protein markers involved in tumor progression (β-catenin, E-cadherin, epidermal growth factor receptor, phosphorylated extracellular signal-regulated kinase [pERK], receptor for hyaluronic acid-mediated motility, phosphorylated protein kinase B, p21, p16, B-cell lymphoma 2, Ki67, apoptotic protease activating factor 1, mammalian sterile 20-like kinase 1, Raf kinase inhibitor protein, vascular endothelial growth factor, ephrin type-B receptor 2, matrix metalloproteinase 7, laminin5γ2, mucin 1 [MUC1], and caudal-related homeobox 2). The prognostic effects of biomarkers in both primary tumor and positive LNs were assessed.

RESULTS:

MUC1, pERK and p16 in LN (P = .002, P = .014, and P = .002, respectively) had independent prognostic value. In patients with stage III disease who received adjuvant treatment, negative p16 expression was associated with highly unfavorable outcomes overall (hazard ratio [HR], 0.26; 95% confidence interval [CI], 0.1-0.6; P = .005) when the analysis was stratified by pathologic tumor classification (HR, 0.25; 95% CI, 0.1-0.7; P = .005), age (HR, 0.23; 95% CI, 0.1-0.6; P = .004), and LN ratio (HR, 0.26; 95% CI, 0.1-0.7; P = .007); and, in multivariate analysis, it was associated with performance status and the receipt of folic acid treatment (HR, 0.29; 95% CI, 0.09-0.89; P = .03).

CONCLUSIONS:

The loss of p16 in LN metastases contributed to adverse outcomes in adjuvantly treated patients with stage III colorectal cancer independent of pathologic tumor classification, age, LN ratio, performance status, or folic acid treatment. The current results support the investigation of p16 as a prognostic and potential predictive biomarker for future randomized trials of patients with stage III colorectal cancer. Cancer 2010. © 2010 American Cancer Society.

Lymph node metastasis is considered the most important prognostic factor for patients with colorectal cancer.1 Studies have confirmed that a lymph node yield of a minimum of 12 lymph nodes is critical for the accurate staging of patients. Despite these guidelines, striking differences in the numbers of lymph nodes retrieved have been reported between institutions2; in particular, approximately 50% of Dukes B colorectal cancers appear to be under staged,3 and inadequate lymph node sampling has been discussed as a “high-risk” feature for patients with TNM stage II disease.4 The lymph node ratio (LNR), which is defined as the total number of positive/total number of lymph nodes collected, also has been highlighted as an important prognostic factor for patients with stage III (including stage IIIB and IIIC) colon cancer.5-8

In patients with metastatic lymph node disease, adjuvant therapy has become the standard of care.9 However, it is recognized that patients within this same disease stage may experience different clinical outcomes.10 The objective of identifying protein biomarkers for patients with stage III and IV colorectal cancer is to improve risk stratification and, thus, to identify patients in the postoperative setting who may benefit from more targeted treatment. To date, this goal largely has not been reached, although a series of markers has been associated with a greater probability of observing lymph node metastasis and has sparked interest as a possible indicator of poor outcome in patients with lymph node-negative disease.11-15

Prognostic biomarker studies in colorectal cancer clearly focus on the expression of proteins in primary tumor and correlating that expression with clinical endpoints. However, to date, very few reports have focused on the comparison of protein markers between primary tumor and corresponding lymph nodes,16 and a systematic evaluation of prognosis based on lymph node expression of protein markers has yet to be performed.

The first step in such an investigation is the retrospective study of potential biomarkers in tissues from patients with known outcomes.17 To this end, we evaluated 19 protein markers in both primary tumor and matched lymph node metastases from 82 patients with stage III and IV colorectal cancer. With patients stratified by disease stage and treatment, the objective of this study was to compare the prognostic performance of these markers according to tissue type and, thus, to identify prognostic and potential predictive biomarkers in patients who received adjuvant therapy. Protein markers were selected to represent the most important signaling pathways and cellular processes that contribute to colorectal cancer progression, prognosis, and response to therapy. For example, it has been demonstrated that the inactivation of Wnt signaling is linked to epithelial mesenchymal transition and is characterized by the overexpression of nuclear β-catenin and loss of membranous E-cadherin expression, both of which are studied here.18 RAS/mitogen-activated protein kinase (RAS/MAPK) signaling and protein kinase B/phosphatase and tensin homolog (AKT/PTEN) signaling are key regulators of colorectal progression.19 Most recently, mutations in key components of these pathways, including v-Ki-ras Kirsten rat sarcoma viral oncogene homolog (K-RAS), v-raf murine sarcoma viral oncogene homolog B1 (B-RAF), and phosphatidylinositol 3-kinase catalytic α polypeptide (PIK3CA) have been linked to nonresponsiveness in patients with metastatic disease who received antiepidermal growth factor receptor (anti-EGFR) therapy.20 Therefore, in the current study, we evaluated EGFR, phosphorylated extracellular signal-regulated protein kinase (pERK), receptor for hyaluronic acid-mediated motility (RHAMM), and phosphorylated protein kinase B (pAKT) for their prognostic effects in primary tumor and lymph nodes.21 Proteins involved in cell cycle regulation, apoptosis, or proliferation (p21, p16, B-cell lymphoma 2 [Bcl-2], Ki67, apoptotic protease activating factor 1 [APAF-1], and mammalian sterile 20-like kinase 1 [MST1]) can have a major effect on patient response to chemotherapeutic agents22; whereas proteins like Raf kinase inhibitor protein (RKIP), vascular endothelial growth factor (VEGF), ephrin type-B receptor 2 (EphB2), matrix metalloproteinase 7 (MMP7), laminin5γ2 (lam5γ2), mucin 1 (MUC1), and caudal-related homeobox 2 (CDX2) are linked to angiogenesis, lymph node positivity, or distant metastasis.23-27 To account for possible heterogeneity in tumor protein expression, multiple-punch tissue microarrays (TMAs) were used.

MATERIALS AND METHODS

Patients

Eighty-two nonconsecutive patients with primary colorectal cancer and TNM stage III or IV disease who were treated at the Fourth Department of Surgery, University of Athens Medical School between 2004 and 2006 were entered into the study. All histomorphologic data were reviewed from the corresponding hematoxylin and eosin (H&E)-stained slides, and clinical data were obtained from corresponding reports. Clinicopathologic information for all 82 patients included sex, age, greatest tumor dimension, histologic subtype, tumor location, pathologic tumor (pT) classification, pathologic lymph node (pN) status, pathologic metastasis (pM) status, vascular and lymphatic invasion, survival, and postoperative therapy. For 72 of the 82 patients, information also was available for chemotherapy regimen, performance status, and folic acid intake. Characteristics of the patients with matched primary tumors and lymph nodes are summarized in Table 1.

Table 1. Characteristics of Patients With Primary Colorectal Cancers and Matched Positive Lymph Nodes (n=82)
Clinicopathologic FeatureNo. of Patients (%)
  1. FOLFIRI indicates folinic acid (leucovorin), 5-fluorouracil, and irinotecan; FOLFIRIAVA, FOLFIRI and bevacizumab; FOLFOX, folinic acid (leucovorin), 5-fluorouracil, and oxaliplatin; FOLFOXAVA, bevacizumab and FOLFOX; FUFA, 5-fluorouracil and folic acid; IRFUFA, irinotecan and FUFA; XELIRIAVA, capecitabine, irinotecan, and bevacizumab; XELODA, capecitabine; XELOX, capecitabine and oxaliplatin; XELOXAVA, capecitabine and bevacizumab; CI, confidence interval.

Mean patient age at diagnosis in y [range], n=7966.2 [38-91]
Mean greatest tumor dimension in cm [range], n=814.6 [1.5-11]
Sex, n=82 
 Men39 (47.6)
 Women43 (52.4)
Diagnosis, n=82 
 Nonmucinous69 (84.2)
 Mucinous13 (15.9)
Tumor location, n=81 
 Left-sided52 (64.2)
 Right-sided11 (13.6)
 Rectum18 (22.2)
Pathologic tumor classification, n=82 
 pT1-pT24 (4.9)
 pT3-pT478 (95.1)
Pathologic lymph node status, n=82 
 pN134 (41.5)
 pN248 (58.5)
Pathologic metastasis status, n=80 
 pM067 (83.8)
 pM113 (16.3)
Tumor grade, n=61 
 1-235 (57.4)
 326 (42.6)
Venous invasion, n=82 
 Present26 (31.7)
 Absent56 (68.3)
Lymphatic invasion, n=82 
 Present69 (84.2)
 Absent13 (15.9)
Adjuvant therapy, n=82 
 None12 (14.6)
 Treated70 (85.4)
Chemotherapy46 (65.7)
Chemotherapy/radiotherapy24 (34.3)
Chemotherapy regimen, n=72 
 None11 (15.3)
 FOLFIRI and FOLFIRIAVA1 (1.4)
 FOLFOX9 (12.5)
 FOLFOXAVA2 (2.8)
 FUFA9 (12.5)
 IRFUFA4 (5.6)
 XELIRIAVA3 (4.2)
 XELODA4 (5.6)
 XELOX27 (37.5)
 XELOXAVA2 (2.8)
Performance status, n=72 
 033 (45.8)
 119 (26.4)
 218 (25)
 32 (2.8)
Folic acid 
 No47 (65.3)
 Yes25 (34.7)
5-Year survival rate, %: n=8138.9
 95% CI25-52
 Stage III45.5
  95% CI29-60
 Stage IV7.7
  95% CI0.1-29

Specimen Characteristics

Paraffin-embedded tissue blocks of primary tumors and corresponding positive lymph nodes were retrieved from the Second Department of Pathology, University of Athens Medical School (Attikon University Hospital), Greece. The use of this material was approved by the local ethics committee of the University of Athens. Two TMAs from these patients were constructed. For each patient, the H&E-stained slides of the primary tumor and lymph nodes from the corresponding whole tissue sections were evaluated, and representative areas of tissue were marked using a felt-tip pen for easy detection. To exclude bias because of possible tumor heterogeneity, each patient had multiple tissue and tumor punches taken from formalin-fixed, paraffin-embedded blocks using a tissue cylinder with a diameter of 0.6 mm, and the punches subsequently were transferred into 1 recipient paraffin block (3 × 2.5 cm) using a homemade, semiautomated tissue arrayer. Tissues were obtained from the tumor center, the invasive tumor front, the normal adjacent mucosa (if available), and the transitional zone (the region where tumor and adjacent normal tissue coincide), if available. Each patient on average had 5.1 tissue punches included on this array, including 4 tumor punches (2 from the tumor center and 2 from the invasive tumor front). The second TMA included single punches from matched metastatic lymph nodes in all 82 patients. Only 1 punch from affected lymph nodes was taken, and this was considered sufficient to study the expression of biomarkers in these affected tissues; the size and the distribution of tumor within the lymph nodes suggested that variations in protein expression were minimal.

Assay Methods

Five-micrometer TMA sections were dewaxed and rehydrated in distilled water. Endogenous peroxidase activity was blocked using 0.5% H2O2. After pressure cooker-mediated antigen retrieval in 0.001 M ethylene diamine tetraacetic acid (EDTA), pH 8.0, the sections were incubated with 10% normal goat serum (DAKO, Carpentaria, Calif) for 20 minutes. To determine mismatch-repair status, the tumor samples were incubated with primary antibody for mutL homolog 1 (MLH1) (clone MLH-1; BD Biosciences Pharmingen, San Diego, Calif; dilution, 1:100), mutS homolog 2 (MSH2) (clone MSH-2; BD Biosciences Pharmingen; dilution, 1:100), mutS homolog 6 (MSH6) (clone 44; Transduction Laboratories, Lexington, Ky), and postmeiotic segregation increased 2 (PMS2) (clone MRQ-28; Cell Marque Corp., Hot Springs, Ark; dilution, 1:100) and for APAF-1 (NCL-APAF-1; Novocastra, Newcastle, United Kingdom; dilution, 1:40), β-catenin (B-catenin-1; DakoCytomation, Carpinteria, Calif; dilution, 1:100), Bcl-2 (clone 124, DakoCytomation; dilution, 1:400), CDX2 (clone AMT28; AbCam, Cambridge, United Kingdom; dilution, 1:50), E-cadherin (NCH-38; DakoCytomation; dilution, 1:100), EphB2 (AF467; R&D Systems, Minneapolis, Minn; dilution, 1:200), EGFR (clone c3c6; Ventana Medical Systems, Oro Valley, Ariz; 3 mg/mL), Ki67 (clone MIB-1; DakoCytomation; dilution, 1:100), laminin5γ2 (clone 4G1; DakoCytomation; dilution, 1:25), MMP7 (clone 141-7B2; Oncogene, Cambridge, Mass; dilution, 1:100), MST1 (polyclonal; Cell Signaling Technology, Danvers, Mass; dilution, 1:200), MUC1 (clone 139H2; Cedarlane Laboratories, Toronto, Ontario, Canada; dilution, 1:100), p16 (clone, inhibitor of cyclin-dependent kinase 4a [INK4a]; MTM Laboratories, Westborough, Mass; dilution, 1:100), p21 (clone SX118; Novocastra; dilution, 1:20), pAKT (clone 244F9; Cell Signaling Technology; dilution, 1:100), pERK (clone 20G11; Cell Signaling Technology; dilution, 1:100), RHAMM (clone 2D6; Novocastra; dilution, 1:100), RKIP (polyclonal; Upstate Biotechnology, Lake Placid, NY; di8ution, 1:1000), and VEGF (clone A20; Santa Cruz Biotechnology, Santa Cruz, Calif; dilution, 1:100). Subsequently, the sections were incubated with horseradish peroxidase-conjugated secondary antibody (DakoCytomation) for 30 minutes at room temperature. For visualization of the antigen, the sections were immersed in 3-amino-9-ethylcarbazole plus substrate/chromogen (DakoCytomation) for 30 minutes and counterstained with Gill hematoxylin. The primary antibody was omitted for negative controls and was replaced by buffer. For a positive control, a TMA with various tissue samples was stained in parallel and included 30 carcinoma tissues (10 breast cancers, 10 prostate cancers, and 10 colon cancers) and 80 normal tissues (4 adrenal gland tissues; 8 colon tissues; and 4 tissues from normal endometrium, epidydymis, heart, kidney, lung, pancreas, parotid gland, placenta, prostate, skin, spleen, stomach, striated muscle, thymus, thyroid gland, lymph node, and tonsil).

Immunohistochemistry was evaluated by semiquantitatively assessing the percentage of positive cells per TMA punch. Because multiple tissue punches were obtained for all primary tumors, the average protein expression was obtained across all punches per patient and per tumor localization. For the matched positive lymph nodes, the same evaluation method was used. Staining intensity was not evaluated.28 Normal tissue punches as well as those from the transitional zone were not included in the analysis.

Statistical Analysis

Differences in the median protein expression for all markers between primary tumors and matched lymph nodes were analyzed using the Wilcoxon rank-sum test. To exclude the probability of observing a chance association because of multiple testing, a Bonferroni correction for multiple comparisons was performed. To reject the null hypothesis of no difference in survival time, only markers that resulted in survival with P values <.003 were considered significantly different. Negative and positive protein expression was assigned by classifying patients around the median expression value for primary tumors and lymph nodes. Differences in the length of survival were illustrated using the Kaplan-Meier method and were analyzed using the log-rank test. Cox proportional hazards regression analysis was carried out to determine prognostic differences between negative and positive protein expression after testing the proportional hazards assumption. Hazard ratios (HRs) and 95% confidence intervals (CIs) were obtained. Baseline hazards of 1.0 consistently were attributed to negative staining. Therefore, an HR >1.0 indicates a detrimental effect of positive staining on prognosis, whereas an HR <1.0 suggests a beneficial effect of positive staining on outcome. Multiple regression analysis of the markers that were identified as significant in univariate analysis was performed by including metastasis and adjuvant therapy. Next, these independent markers were entered into a multiple regression analysis using a forward selection process and a Type I error of α = .01. All P values were 2-sided. Analyses were performed using SAS software (version 9; SAS Institute, Cary, NC).

RESULTS

Association Between Clinicopathologic Features and Survival in Univariate Analysis

Survival information was available for 81 of 82 patients. The 5-year survival rate for the entire cohort was 38.9% (Fig. 1). Significantly poorer outcomes were observed for patients who had more advanced pN status (P = .006), M status (P < .001), vascular invasion (P < .001), lymphatic invasion (P = .008), adjuvant therapy (P < .001), chemotherapy regimen (P < .001), and higher performance status (P < .001); whereas marginal differences were noted for patients stratified according to histologic diagnosis (P = .056) and folic acid intake (P = .056). No differences in prognosis were observed according to sex, tumor location, pT classification, or tumor grade.

Figure 1.

Survival in months is illustrated for patients with lymph node-positive (stage III-IV) colorectal cancer. This Kaplan-Meier survival curve illustrates the probability of survival for patients (n = 81) and indicates the number of patients still at risk of death over time until 5-years of follow-up.

Protein Expression Differences in Primary Tumors Versus Matched Lymph Node Metastases

Of the 19 markers that we evaluated, significant changes in protein expression were observed between primary tumors and matched lymph nodes in 6 (P < .003; Bonferroni correction) (Table 2). Significant loss of expression in lymph nodes was observed for RHAMM (P = .001), Ki67 (P < .001), pAKT (P < .001), VEGF (P < .001), APAF-1 (P < .001), and MST1 (P < .001).

Table 2. Differences in Protein Expression Between Primary Tumors and Matched Positive Lymph Nodes (n=82)a
Protein MarkerPrimary Tumor, %Matched Lymph Nodes, %MeanMedianPb
MeanMedian
  • EGFR indicates epidermal growth factor receptor; c, cytoplasmic; RHAMM, receptor for hyaluronic acid-mediated motility; n, nuclear; pAKT, phosphorylated protein kinase B; Bcl-2, B-cell lymphoma 2; VEGF, vascular endothelial growth factor; APAF-1, apoptotic protease activating factor 1; MUC1, mucin 1; pERK, phosphorylated extracellular signal-regulated kinase; EphB2, ephrin type-B receptor 2; m, membranous; RKIP, Raf kinase inhibitor protein; MMP7, matrix metalloproteinase 7; CDX2, caudal-related homeobox 2; MST1, mammalian sterile 20-like kinase 1.

  • a

    Values are expressed as the mean and median percentage of immunoreactive tumor cells. Median protein expression values were used as cutoff scores for negative/positive staining of primary tumor and lymph node expression in subsequent analyses.

  • b

    Only P values <.003 were considered statistically significant after adjustment for multiple comparisons.

EGFR (c)18.75.813.40.005
RHAMM (c)16.816.312.310.001
B-catenin (n+c)32.928.138.730.168
P16 (n)16.513.513.810.188
P21 (n)17.113.811.710.004
Ki67 (n)46.64534.930<.001
pAKT (c)2.953.40<.001
Bcl-2 (c)17.81514.30.05
VEGF (c)27.818.110.110<.001
APAF-1 (c)32.43.14.810<.001
MUC1 (c)9.72.511.60.065
pERK (n)303.50.874
EphB2 (m)45.846.338.240.028
E-cadherin (m)48.152.940.440.025
RKIP (c)43.948.837.840.063
MMP7 (c+m)9.92.914.80.278
CDX2 (n)53.261.35060.832
Laminin5γ2 (c)18.213.328.125.069
MST1 (c)23.418.87.80<.001

Prognostic Differences in Primary Tumor and Matched Lymph Node Metastases

The expression of 3 proteins in primary tumors resulted in prognostic differences. A poorer outcome was observed for patients with tumors that had positive MUC1 expression (P = .048), positive pERK expression (P = .017), and loss of RKIP expression (P = .05) (Fig. 2). However, only pERK and RKIP were identified as independent prognostic factors after adjusting for the presence of distant metastasis and adjuvant therapy (pERK, P = .012; RKIP, P = .009) (Table 3).

Figure 2.

The expression of prognostic protein markers in primary tumors is illustrated. These Kaplan-Meier curves demonstrate significantly worse outcomes for patients with (A) mucin 1 (MUC1)-positive versus MUC1-negative tumors, (B) phosphorylated extracellular signal-regulated kinase (pERK)-positive versus pERK-negative tumors, and (C) loss of Raf kinase inhibitor protein (RKIP) expression versus RKIP-positive tumors.

Table 3. Hazard Ratios and Differences in Survival for Patients With Protein Expression in Either Primary Tumors or Lymph Node Metastasesa
ProteinbHR (95%CI)P
  • HR indicates hazard ratio; CI, confidence interval; EGFR, epidermal growth factor receptor; RHAMM, receptor for hyaluronic acid-mediated motility; pAKT, phosphorylated protein kinase B; Bcl-2, B-cell lymphoma 2; VEGF, vascular endothelial growth factor; APAF-1, apoptotic protease activating factor 1; MUC1, mucin 1; pERK, phosphorylated extracellular signal-regulated kinase; EphB2, ephrin type-B receptor 2; RKIP, Raf kinase inhibitor protein; MMP7, matrix metalloproteinase 7; CDX2, caudal-related homeobox 2; MST1, mammalian sterile 20-like kinase 1.

  • a

    HR>1.0 indicates that positive staining was linked to a poorer outcome, whereas HR<1.0 suggests an improved outcome with positive staining.

  • b

    In primary tumors, the expression of pERK and RKIP were independent prognostic factors after adjusting for metastasis and adjuvant therapy. In lymph nodes, of the 4 markers, only p16, MUC1, and pERK were prognostically independent of metastasis and adjuvant therapy.

Primary tumor  
 EGFR1.35 (0.7-2.5).351
 RHAMM1.33 (0.7-2.5).375
 β-catenin0.81 (0.4-1.5).525
 p160.61 (0.3-1.2).126
 p211.74 (0.9-3.4).113
 Ki670.55 (0.3-1.0).064
 pAKT0.69 (0.4-1.4).283
 Bcl-20.66 (0.4-1.2).194
 VEGF2.16 (0.7-7.1).204
 APAF-11.73 (0.9-3.3).09
 MUC12.04 (1.0-4.1).048a
 pERK2.53 (1.2-5.4).017a
 EphB21.76 (0.9-3.5).114
 E-cadherin0.72 (0.4-1.4).326
 RKIP0.5 (0.2-1.0).05a
 MMP70.75 (0.4-1.4).37
 CDX20.66 (0.3-1.4).262
 Laminin5γ21.04 (0.6-2.0).915
 MST10.83 (0.4-1.6).579
Positive lymph nodes  
 EGFR1.41 (0.7-2.8).34
 RHAMM1.13 (0.6-2.3).738
 β-catenin1.15 (0.6-2.3).7
 p160.42 (0.2-0.8).014a
 p210.73 (0.3-1.6).445
 Ki670.91 (0.51.8).78
 pAKT0.45 (0.06-3.3).426
 Bcl-20.86 (0.3-2.5).781
 VEGF0.98 (0.4-2.2).953
 APAF-10.77 (0.4-1.7).529
 MUC12.15 (1.1-4.3).03a
 pERK1.99 (1.0-4.1).049r
 EphB21.08 (0.6-2.1).818
 E-cadherin0.16 (0.1-0.9).013r
 RKIP0.97 (0.5-2.0).935
 MMP71.04 (0.4-2.7).937
 CDX20.52 (0.3-1.0).083
 Laminin5γ21.94 (0.9-4.2).086
 MST10.51 (0.2-1.5).215

The expression of 4 markers in lymph node metastases led to significant differences in prognosis. Poorer outcome was observed for patients who had lymph nodes with positive MUC1 expression (P = .03) and positive pERK expression (P = .049), whereas a more favorable prognosis was attributed to positive expression of p16 (P = .014) and E-cadherin (P = .013) (Fig. 3). In a multivariate analysis of each marker that included distant metastasis and adjuvant therapy, only MUC1 (P = .002), pERK (P = .014), and p16 (P = .002) were identified as independent prognostic factors.

Figure 3.

The expression of prognostic protein markers in lymph node metastases is illustrated. These Kaplan-Meier curves demonstrate significantly worse outcome in patients with (A) p16-negative versus p16-positive lymph nodes, (B) mucin 1 (MUC1)-positive versus MUC1-negative lymph nodes, (C) phosphorylated extracellular signal-regulated kinase (pERK)-positive versus pERK-negative lymph nodes, and (D) E-cadherin-negative versus E-cadherin-positive lymph nodes.

Comparison of the Prognostic Performance of Proteins in Primary Tumor and Lymph Node Metastases

To determine the most competitive protein marker expressed in primary tumor, multivariate survival analysis was performed by including into the model the following variables: pERK, RKIP, MUC1, metastasis, and adjuvant therapy (Table 4). By using a forward selection procedure, positive expression of pERK and adjuvant treatment maintained their significant effects on outcome. By using the same approach for protein expression in the lymph nodes and including p16, MUC1, pERK, metastasis, and adjuvant therapy in the multivariate analysis, we observed that all 5 variables independently predicted prognosis. When comparing the HRs, loss of p16 expression in positive lymph nodes was attributed to a 5.56-fold greater risk of death compared with positive p16 expression when the model was adjusted for all remaining prognostic factors, indicated that p16 was the most competitive prognostic marker in positive lymph nodes. Furthermore, we compared differences in survival according to pERK expression in tumors versus p16 expression in lymph nodes. The stratification of patients according to better or worse prognostic factors was more pronounced using p16. This is underlined by the larger relative risk of death for patients who had p16-negative tumors (HR, 0.32; 95%CI, 0.2-0.7; P = .003) compared with patients who had pERK-positive tumors (HR, 2.99; 95%CI, 1.3-6.9; P = .012) when the model was adjusted for distant metastasis and adjuvant therapy. Representative immunostains of p16 in lymph nodes from affected patients are shown in Figure 4.

Table 4. Multiple Cox Regression Analysis After Forward Selection (Type I Error, α=.01) of Independent Prognostic Protein Markers in Primary Tumors and Positive Lymph Nodesa
FeatureHR (95%CI)P
  • HR indicates hazard ratio; CI, confidence interval; pERK, phosphorylated extracellular signal-regulated kinase; MUC1, mucin 1.

  • a

    Variables that were entered into the model for primary tumors included pERK, Raf kinase inhibitor protein (RKIP), MUC1, metastasis, and adjuvant therapy. For positive lymph nodes, variables that were entered into the model were p16, MUC1, pERK, metastasis, and adjuvant therapy.

  • b

    The HR corresponds to a 5.6 times greater relative risk of death in patients with p16-negative lymph nodes compared with p16-positive lymph nodes.

Primary tumor  
 pERK3.73 (1.7-8.4).002
 Adjuvant therapy012 (0.1-0.3)<.001
Positive lymph nodes  
 p160.18 (0.1-0.5)b<.001
 MUC12.91 (1.3-6.7).013
 pERK5.27 (2.0-13.9)<.001
 Metastasis2.86 (1.3-6.5).012
 Adjuvant therapy0.07 (0.03-0.2)<.001
Figure 4.

These are representative photomicrographs of (Left) negative immunostaining and (Right) positive immunostaining of nuclear p16 expression in lymph nodes from patients with stage III and IV colorectal cancer.

p16 in Stage III Patients Treated With Adjuvant Therapy

Fifty-one patients had stage III disease, received systemic adjuvant therapy, and had corresponding information on p16 expression, including 13 of 51 patients (25.5%) who were negative (<10% staining in tumor cell nuclei) and 38 of 51 patients (74.5%) who were positive (≥10% staining in tumor cell nuclei) for p16 in lymph node metastases (Fig. 5). These included 24 patients who received combined capecitabine and oxaliplatin (XELOX); 8 patients who received combined folonic acid (leucovorin), 5-fluorouracil (5-FU), and oxaliplatin (FOLFOX); 2 patients who received FOLFOX plus bevacizumab; 8 patients who received 5-FU plus folic acid (FUFA); 4 patients who received FUFA plus irinotecan; 2 patients who received combined capecitabine, irinotecan, and bevacizumab; and 4 patients who received capecitabine. The overall survival for these patients is illustrated in Figure 5A. The highly unfavorable impact of negative p16 expression in lymph node metastases on outcome for this group was underlined by an HR of 0.26 (95%CI, 0.1-0.6; P = .005) (Fig. 5B).

Figure 5.

The prognostic effect of p16 expression is illustrated in lymph nodes from patients who had TNM stage III disease for which they received adjuvant therapy. In all stratifications, negative expression of p16 was linked to worse survival, highlighting the strong and independent prognostic effect of this immunohistochemical biomarker. These Kaplan-Meier survival curves illustrate the length of survival in months for (A) all patients with stage III disease who received adjuvant treatment and according to differences in p16 expression for (B) all patients; (C) patients with pathologic T3 (pT3) tumors; (D) patients with pT4 tumors; (E) patients with a performance status of 1, 2, or 3; (F) patients who did not receive folic acid treatment; (G) patients who received folic acid treatment; (H) patients who received combined capecitabine and oxaliplatin (XELOX); and (I) patients who received other chemotherapy regimens.

This result was maintained again when T classification was considered (HR, 0.25; 95%CI, 0.1-0.7; P = .005) (Fig. 5C,D). Because p16 expression reportedly is modified by patient age, we evaluated the independent prognostic effect of p16 expression adjusting for age at diagnosis. Both p16 (HR, 0.24; 95%CI, 0.1-0.6; P = .004) and patient age (HR, 0.94; 95%CI, 0.88-0.99; P = .045) were significant independent prognostic indicators for patients with stage III disease who received adjuvant therapy.

A multivariate analysis of p16 expression was performed again that included both performance status and folic acid intake in the model. Negative expression of p16 remained an independent prognostic factor (HR, 0.29; 95%CI, 0.09-0.89; P = .03), whereas performance status and folic acid intake no longer were linked significantly to outcome (P = .184 and P = .295, respectively). This effect is illustrated in Figure 5E-G. Because only 3 patients with a performance status of 0 were negative for p16 expression, no survival analysis could be performed. The adverse effect of p16 expression was maintained after adjusting for the LNR (HR, 0.26; 95%CI, 0.1-0.7; P = .007). Finally, p16 negativity led to a poorer outcome in patients who received either XELOX (P = .009) or chemotherapy other then XELOX (P = .046) (Fig. 5H,I).

DISCUSSION

The major findings of this study indicate that negative lymph node expression of p16 in patients with stage III disease who receive adjuvant therapy is a highly adverse prognostic factor in this subgroup. Moreover, the prognostic effect of p16 on outcome was independent of pT classification, patient age, LNR, performance status, treatment with folic acid, and chemotherapy regimen, underlining its value as a prognostic indicator.

In a first step, we investigated differences in the protein expression of 19 markers between primary tumors and matched lymph node metastases; and we identified RHAMM, Ki67, pAKT, VEGF, APAF-1, and MST1 as significantly decreased in lymph node metastases. It is noteworthy that all of these proteins play crucial roles in proliferation or apoptosis and are implicated directly both in the RAS/MAPK signaling pathway and in the pAKT signaling pathway.19, 29-31 Differences in protein expression between primary tumors and matched lymph nodes rarely have been documented. However, Deng and colleagues recently published their TMA work evaluating EGFR expression in both matched primary tumors and lymph nodes.16 Those authors reported no effect of EGFR expression in primary tumors on survival in patients with stage III or IV disease but noted an adverse prognostic effect with high EGFR expression in lymph node metastases. We could not confirm those findings in the current study despite the use of a similar sample size and similar methodology. This discrepancy is likely because of the different methods that were used for protein marker evaluation (semiquantitative scoring vs composite scoring system). A representative immunostain is shown in Figure 6.

Figure 6.

These are representative photomicrographs of (Left) negative immunostaining and (Right) positive immunostaining of membranous and cytoplasmic epidermal growth factor receptor expression in lymph nodes from patients with stage III and stage IV colorectal cancer.

Although significant differences between primary tumor and lymph node metastases were observed for 6 protein markers, none of these had effects on outcome when they were assessed in either sample type. Positivity for pERK and loss of RKIP in the primary tumor were linked to poorer survival independent of the presence of distant metastasis and adjuvant therapy. These findings are in agreement with several larger colorectal cancer prognostic studies.23, 32-34 Lymph node expression of pERK and MUC1 and loss of p16 expression also were identified as independent features of poor outcome. In particular, p16 appeared to play a more important role in predicting prognosis than any other marker, including pERK, in both primary tumors and lymph node metastases.

A member of the INK4 class of cell-cycle inhibitors, p16 functions as a tumor suppressor gene that is active in cell-cycle regulation and in the angiogenic switch.35 The main mechanism of p16 inactivation is promoter hypermethylation, which can be detected in 12.5% to 47% of colorectal cancers and appears to be age-dependent.36-41 In lymph node metastases, aberrant activation of p16 by methylation reportedly occurs in 14% of patients.42 Hypermethylation of p16 has important implications for prognosis, is linked to colorectal cancers with high microsatellite instability, and also is considered among the elements of a classic panel for detection of the CpG island methylator phenotype.37, 40, 43-49 The methylation of p16, as evaluated in DNA extracted from primary tumors, peritoneal lavage fluid, and serum from patients with colorectal cancer, has been related to more advanced tumor stage and to significantly poorer overall and disease-free survival rates.37, 45, 50, 51 It is noteworthy that protein expression of p16 is correlated inversely with the degree of methylation, supporting the argument that immunohistochemistry for the protein could be used as a surrogate marker for p16 epigenetic silencing.47 Furthermore, the use of p16 protein expression has been suggested as a screening tool for the selection of patients who undergo genetic testing for Lynch syndrome.47 Few studies have assessed the potential predictive value of p16 in colorectal cancer in vitro and in vivo.52, 53 However, Kamoshida and colleagues compared p16 and thymidylate synthase expression in preoperative biopsies and postsurgical resection specimens and observed significant overexpression of p16 in resection specimens from patients who had a response to 5-FU-based chemotherapy compared with nonresponders and controls.54 Our major novel findings suggest that negative p16 expression evaluated in lymph nodes, rather than in primary tumors, from patients with stage III disease who received adjuvant therapy is a highly adverse prognostic indicator. In fact, negative p16 expression was related to worse survival in both pT3 and pT4 disease and was independent of patient age, LNR, performance status, treatment with folic acid, and chemotherapy regimen.

In this study, we used the TMA technique with multiple tissue punches per patient to account for possible heterogeneity in terms of protein expression in the primary tumor. Each patient had an average of 4 tumor punches taken, including 2 samples from the tumor center and 2 from the tumor border, where changes in certain protein markers can be most prevalent. Therefore, tumor heterogeneity was minimized by such sampling and by evaluating the average protein expression across the total number of samples. Our results fulfilled the requirements set by Goethals and coworkers, who recommend taking at least 3 or 4 punches of primary tumor to account for possible biases in protein evaluation.55 For lymph nodes, the issue of heterogeneity is likely to be substantially less important because of the size and distribution of tumor within the lymph node itself. The evaluation of H&E-stained slides before TMA construction allowed us to distinguish positive lymph nodes from unaffected lymph nodes, to sample the most representative affected lymph nodes, and to identify the specific areas to be punched. Therefore, only a single punch of a lymph node was obtained and evaluated for protein marker expression. However, our study was limited to a relatively small number of patients with stage III and IV colorectal cancer who had matched primary tumors and lymph nodes (n = 82), and only a single tissue punch was obtained from the lymph nodes.

The current results provide evidence supporting the investigation of potential prognostic biomarkers in lymph node metastases from patients with stage III and IV colorectal cancer. Moreover, our findings indicate that, in patients with stage III disease and lymph node metastases who receive treatment with adjuvant therapy, negative expression of p16 is an independent prognostic factor in the postoperative setting. Although our current study did not allow us to determine the predictive value of lymph node p16 expression regarding response to adjuvant therapy, the results indicate that further investigation of this issue in randomized controlled trials is warranted.

Acknowledgements

We thank Drs. Panayiotides and Karakitsos for their support and critical revision of this article.

CONFLICT OF INTEREST DISCLOSURES

The authors made no disclosures.

Ancillary