The aim of this study was to evaluate the expression of epidermal growth factor receptor (EGFR) and insulinlike growth factor 1 receptor (IGF-1R) proteins and IGF-1R gene copy numbers in pancreatic ductal adenocarcinoma in relation to patients' characteristics and prognosis.
Immunohistochemical staining was performed on formalin-fixed paraffin-embedded tissue derived from tumor specimens recovered during surgery. Slides were evaluated for membranous EGFR and IGF-1R staining using both the HercepTest and the semiquantitative H-score systems. Chromogenic in situ hybridization was performed to quantify IGF-1R gene copy number. The primary outcome was the association between EGFR expression, IGF-1R expression—in both neoplastic epithelial and stromal cells—or IGF-1R gene copy number and overall survival. Secondary outcomes included associations between EFGR and IGF-1R expression and pathologic variables.
A total of 105 patients were included. EGFR expression was present in 30.4% of cases and was associated with lymph node metastasis (P = .038). IGF-1R was overexpressed in 53% of tumors and correlated with higher tumor grade (P = .033). High membranous expression of EGFR (P < .001) and/or IGF-1R (P = .004), the cytoplasmic detection of EGFR (P = .027), and high expression levels of IGF-1R in the tumoral stroma (P < .001) were all associated with shorter overall survival, being significantly better in patients who simultaneously do not express membranous EGFR or stromal IGF-1R.
Pancreatic ductal adenocarcinoma is recognized as the fourth leading cause of cancer-related mortality, being currently responsible for almost 40,000 deaths per year in the United States.1 Beyond surgical resection, multiple therapeutic antitumor interventions have been tried over the past decade, but most of them have failed to accomplish any substantial benefit. This may be partially explained by the notorious complexity derived from the multiple genetic aberrations and mixed cellular population present in a pancreatic tumor,2, 3 which can also explain the heterogeneous clinical course observed in daily practice. It is expected that in the near future complete genomic profiling of each patient's tumor will become accessible, helping to tailor the best possible treatment for each patient. Meanwhile, practical and reliable prognostic and predictive markers are needed to identify high-risk patients, to better customize their treatments.
Among the many molecular pathways typically affected in epithelial cancers, deregulation and overexpression of growth factors are among the most significant. A growth factor receptor, or any of its downstream molecules, can be affected by many specific mutations ultimately resulting in an abnormally persistent activation of a specific signaling pathway. Alternatively, overexpression could either be because of gene amplification at the genomic DNA level or result from an increase in gene transcription or in protein translation, both of which result in increased protein expression. However, although these 2 phenomena—deregulation and overexpression—carry similar consequences, from a practical and diagnostic point of view it is more convenient to detect gene amplification or protein overexpression, which can be easily done in readily available paraffin-embedded tissues, rather than to test for individual mutations. A remarkable example of the potential practical application of the detection of growth factor overexpression is the use HER-2/neu assays in breast cancer.4
Epidermal growth factor receptor (EGFR) has been extensively studied in pancreatic ductal adenocarcinoma,5 and erlotinib, a small molecule inhibitor, was recently approved in combination with gemcitabine for the treatment of metastatic disease.6 EGFR amplifications and mutations in pancreatic ductal adenocarcinoma are rare,7, 8 but EGFR overexpression has been reported in 9% to 65% cases, with a variable correlation with survival.9, 10
IGF-1 is required by cells to progress through the cell cycle and has been shown to have both autocrine and paracrine effects at the organ level.11, 12 Moreover, these effects seem to be facilitated by an increase in the number of insulinlike growth factor 1 receptors (IGF-1Rs) in pancreatic cancer cells. Overexpression of IGF-1R, which has been reported in 38% to 64% of patients with pancreatic ductal adenocarcinoma,13 has been associated with higher tumor grade, antiapoptotic effect, proliferation rate, and angiogenesis.14-19 In addition, previous investigations have suggested a role of IGF-1 in the intricate relationship between pancreatic cancer cells and stromal cells.12, 20 No information, however, is available regarding the significance of IGF-1R gene copy number changes in pancreatic ductal adenocarcinoma.
Interestingly, EGFR and IGF-1R share common properties, and in some tumors including grade IV glioblastoma, hepatocellular carcinoma, and pancreatic ductal adenocarcinoma, IGF-1R has been identified as a potential source of resistance to EGFR therapies,21-24 suggesting that dual blockage may provide greater therapeutic effects.25-27 Consequently, multiple phase 1-2 clinical trials testing IGF-1R inhibitors in a diverse number of epithelial cancers, including pancreatic ductal adenocarcinoma, are currently ongoing.28-36
The aim of this study was to explore the prognosis and clinicopathologic significance of both EGFR and IGF-1R protein expression, evaluated with novel antibodies, as well as IGF-1R gene copy number changes in patients with pancreatic ductal adenocarcinoma.
MATERIALS AND METHODS
Patient Selection and Data Collection
We selected 105 patients with confirmed pancreatic ductal adenocarcinoma who had surgical resections performed at Thomas Jefferson University Hospital between 2002 and 2009, and for whom sufficient material was available for immunohistochemical and chromogenic in situ hybridization (CISH) analysis. All patients had consented to analyses of their tumors via a protocol approved by the Thomas Jefferson University Institutional Review Board. We subsequently reviewed the medical charts and contacted primary oncologists to obtain relevant clinical information. Vital status was obtained from medical records and verified by querying the Social Security death index.
The primary outcome was the association between 1) EGFR expression, 2) IGF-1R expression in both neoplastic epithelial and stromal cells, or 3) IGF-1R gene copy number, and overall survival defined from the day of surgery to the day of death or last follow-up. Secondary outcomes included associations between EFGR and IGF-1R expression and several pathologic variables (tumor size, stage, lymph node metastases).
The following variables were recorded from medical records: age, sex, tumor size and grade, number of lymph nodes resected with the surgical specimen, number of lymph nodes with metastases, resection margin involvement, and treatment received (type of surgery, radiation therapy, and chemotherapy). To ensure accuracy, dual data extraction was conducted. Data were subsequently verified between the 2 reviewers, and discrepancies were settled through consensus discussion. To minimize subjective judgment and selection bias, investigators were blinded to outcomes.
Immunohistochemical and Chromogenic In Situ Hybridization Analysis
Immunohistochemical staining was performed on formalin-fixed paraffin-embedded tissue derived from tumor specimens recovered during surgery. Staining was performed using the UltraView Universal DAB detection kit on a BenchMark Series automatic stainer (Ventana Medical Systems, Tucson, Ariz). Sections were deparaffinized at 72°C and subsequently conditioned in a buffer solution, pH 8.0 for 172 minutes (Ultra Conditioner #1, Ventana Medical Systems). For the EGFR stain, 1 drop of the CONFIRM anti-EGFR antibody (5B7, Ventana Medical Systems) at a concentration of 0.4 μg/mL was added to each slide. For the IFGF-1R stain, 1 drop of the CONFIRM anti–IGF-1R antibody (G11, Ventana Medical Systems) at a concentration of 1.7 μg/mL was used, and the slides were incubated for 32 minutes at 37°C. After incubation of the slides with 1 drop of antibody diluent (Ventana Medical Systems) for 12 minutes, the slides were counterstained with hematoxylin II (8-minute incubation).
Scoring of the immunohistochemical stains
Slides were evaluated for membranous EGFR and IGF-1R staining using both the HercepTest and the semiquantitative H-score systems. In the HercepTest, a score of 0 was given to cases with no staining or incomplete membranous staining in <30% of tumor cells, and 1+ to cases with faint complete or incomplete membranous staining in >30% of tumor cells. A score of 2+ was assigned when weak to moderate complete membranous staining was seen in >30% of tumor cells, and 3+ when strong, complete membranous staining was present in >30% of tumor cells.37 In the H-score system, the percentage of tumor cells in each intensity category (0, no staining; 1+, faint/barely appreciable staining; 2+, moderate/readily appreciable staining; 3+, strong staining) was determined. The H-score is calculated as the sum of the percentage of stained tumor cells by their staining intensity and ranged from 0 to 300.38 Cytoplasmic staining was evaluated for both EGFR and IGF-1R as either present or absent; stromal staining was scored for IGF-1R as either present or absent. Slides were interpreted by 2 independent scorers, and discrepancies were resolved by consensus discussion. For statistical analysis, EGFR and IGF-1R scores were dichotomized as low (0-2+, 0-200 H-score) versus high (3+, ≥201 H-score).
CISH was performed using an experimental Ventana IFG-1R probe. The IGF-1R gene (black signal) and centromere 15 (red signal) were counted in 50 nonoverlapping nuclei. A ratio >2.2 was considered amplified.
The primary outcome variable was overall survival. Univariate analysis was completed for each clinicopathological variable using Kaplan-Meier and univariate Cox models. The purpose of the analysis was to determine which approach to measuring EGFR and IGF-1R cell membrane expression would be most informative in a patient population. Thus, a base multivariate model was fit, using those factors significant at the univariate level as well as those known to be prognostic. In addition, first order interaction terms of these base factors were considered, and retained if P ≤ .05. Two multivariate models were then considered to compare the HercepTest to the H-score test, using the Akaike information criterion and Bayesian information criterion. Each model consisted of the base model with either the inclusion of HercepTests (model 1) or the inclusion of H-score tests (model 2; details of nonselected model not presented, but available from the authors). The model with the smaller Akaike and Bayesian information criteria is considered preferable, and is reported here as the final multivariate model. Results for the final multivariate model are reported as hazard ratio with confidence interval (CI) and P value. Linear combinations of the final multivariate model for specific covariate combinations are also reported. In all analyses, P ≤ .05 is considered statistically significant, and all P values are 2-sided. Analysis was completed in R version 2.11.1. For secondary outcomes, the association between pathologic variables and each biomarker (EGFR, IGF-1R, and IGF-1R–stroma) was analyzed using Fisher exact test and Wilcoxon 2-sample test.
A total of 105 patients were included in this study, with a median age of 66 years (range, 35-91 years) and a median follow-up of 39.78 months. The baseline clinicopathologic characteristics of the patients are summarized in Table 1. A total of 70 patients (67%) died during the period of follow-up, and the majority of the patients received gemcitabine-based chemotherapy regimens as well as local radiotherapy.
Table 1. Clinicopathologic Characteristics of the 105 Patients
No. of Patients (%)
Tumor size, T
Lymph nodes, N
Type of surgery
EGFR and IGF-1R Expression Results in Pancreatic Cancer
The EGFR and IGR-1R expression results are summarized in Table 2. High EGFR membrane expression by HercepTest was found in 30.4% of the cases (32 of 105; Fig. 1). Such high expression was statistically associated with the presence of lymph node metastases (odss ratio, 3.13; 95% CI, 1.13-8.69; P = .038). However, no significant association was found with tumor size, histological grade, or TNM stage. EGFR was present in the cytoplasm of neoplastic cells in 47 cases (45%), and the EFGR H-score was found to be high in 19% of the cases. Ninety percent of the patients with high H-score also had high HercepTest score (18 of 20). Neither cytoplasmic staining of EGFR nor high EGFR H-score was correlated with grade, tumor size, lymph node status, or TNM stage. No nuclear or stromal staining was seen with EGFR.
For HercepTest, low = 0-2, high = 3, medium is not applicable.
High IGF-1R membrane expression by HercepTest was found in 53% of the cases (56 of 105) (Fig. 1), and it was statistically associated with higher tumor grade (P = .003), but no association was found with tumor size, number of lymph nodes involved, or clinical stage. Cytoplasmic protein was detected in 52% of the cases (55 of 105) and was not associated with any pathological variable. Overall, 7% of the cases showed high IGF-1R by H-score. All of them were also classified as high using HercepTest score. Moreover, the IGF-1R gene was found to be amplified by CISH score ratio in 10 cases (Fig. 2), and in 7 of these 10 cases (70%) membranous IGF-1R protein expression was found to be high by immunohistochemistry. No nuclear IGF-1R staining was seen. IGF-1R was found to be expressed in the tumoral stromal cells in 51% of the cases (54 of 105; Fig. 3). No significant association was found between IGF-1R stromal expression and tumor size, grade, lymph node metastases, or TNM stage.
There was a strong correlation between the EGFR and IGF-1R expression as measured by immunohistochemistry in tumoral cells (P = .006).
As an exploratory analysis, expression of EGFR and IGF-1R in lymph node metastases (n = 8) and distant metastases to the liver (n = 4) were compared with the staining in the primary tumor. Expression of both markers was the same in 10 of 12 cases. In 2 cases (1 lymph node metastasis and 1 distant metastasis), staining was higher (3+ in the metastasis vs 2+ in the primary tumor).
Overall Survival Analysis
The median survival of the entire cohort was 18.02 months (95% CI, 14.8-31.4). High EGFR, as well as high IGF-1R, membrane expression by HercepTest were both associated with significantly shorter survival using Kaplan-Meier estimates (P < .001 and P = .002, respectively; Fig. 4). However, having a high membrane expression of both receptors did not add any additional risk. Moreover, the presence of IGF-1R in the tumoral stroma was also found to be highly correlated with poor outcome (P < .001; Fig. 5). Of note, patients who had low levels of EGFR membrane expression and did not show presence of IGF-1R in the tumoral stroma were found to have the longest survival (Fig. 6).
Cox proportional univariate analyses showed significant association between many of the variables investigated and overall survival (Table 3). High EGFR by HercepTest (P < .001), high IGF-1R by HercepTest (P = .004), and EGFR high H-score (P = .004) were all associated with poor survival. Detection of EGFR in the cytoplasm of the neoplastic cells was also associated with a worse prognosis (P = .027). As expected, presence of nodal metastases and larger tumor size were associated with shorter survival.
Table 3. Univariate Analysis Results for Overall Survival
There was no statistically significant association between increased IGF-1R gene copy number, IGF-1R cytoplasm expression, IGF-1R-H-score, age, or tumor grade and overall survival.
A multivariate Cox proportional regression model using the HercepTest rather than H-score (because this model showed lower Akaike and Bayesian information criteria results) was selected and is reported (Table 4). High levels of membrane EGFR expression (P < .001), cytoplasmic EGFR staining (P = .007), the presence of stromal IGF-1R (P < .001), and lymph node metastases (P = .003) remained as significant prognosis factors.
Table 4. Multivariate Cox Regression Analysis for Overall Survival
Pancreatic ductal adenocarcinoma is among the most lethal cancers, and to date no effective treatment beyond surgical intervention is available. Many patients are encouraged to participate in clinical trials; thus, it is important to be able to easily and accurately detect patients with worse prognosis from the beginning. This also emphasizes the critical need for identification of therapeutic targets, unique to a tumor, in a given patient. We have studied the expression of 2 growth factors, EGFR and IGF-1R, and found that they provide useful prognostic information.
Our study showed that high membranous expression of EGFR and/or IGF-1R, cytoplasmic detection of EGFR, and high expression levels of IGF-1R in the tumoral stroma are all associated with shorter overall survival, being significantly better in those patients who simultaneously do not express high levels of membranous EGFR or stromal IGF-1R (Figure 6). In addition, we found evidence that IGF-1R and EGFR expression persists in distal metastases. However, we did not have tissue material from a sufficiently large number of metastatic samples to make reliable conclusions. Previously, Ueda and colleagues,24 demonstrated that patients with increased membranous expression of IGF-1R had worse overall survival when compared with patients either with increased cytoplasmic IGF-1R or lacking IGF-1R overexpression. They also showed that patients with increased cytoplasmic EGFR were associated with worse overall survival when compared with patients with increased membranous EGFR or patients lacking EGFR overexpression. Cases with membrane-dominant IGF-1R and cytoplasm-dominant EGFR were associated with worse prognosis; inversely, cytoplasm-dominant IGF-1R with membrane-dominant EGFR expression was associated with favorable prognosis. We were unable to demonstrate a similar trend of reciprocal staining patterns with survival. The differences may be attributable to different patient populations (we used only surgical resection specimens from our Philadelphia-based population, whereas most of the reciprocal staining patterns noted in the Ueda study were noted in autopsy specimens from a Japanese population) or to the use of different antibody clones.
We found that tumoral cells tend to highly coexpress EGFR and IGF-1R in their membrane, explaining why the detection of both receptors does not add additional risk. In that sense overexpression of EGFR or IGF-1R could be seen as a surrogate of a complex phenotypic pattern associated with poor outcomes. Solid evidence derived from previous studies supports the idea that some tumoral clones are mainly driven by EGFR activation.39 However, the actual prognostic significance of EGFR overexpression is controversial, and the data should be interpreted with caution. Some authors have previously suggested that overexpression of this receptor implies a more aggressive phenotype5, 40 and poor outcomes.41 However, this concept has been challenged by additional studies.9 In our study, patients with high EGFR expression showed worse prognosis, even after adjusting for potential confounders. Regarding the value of EGFR detection as a marker of treatment response, many previous studies have shown lack of predictive significance.42-44 However, recent discoveries have given a better insight into the biology of the EGFR pathway in pancreatic cancer.45 Because many of the downstream effectors (i.e., Akt, MAPK) could be activated by other mechanisms, it seems evident that methods evaluating global EGFR pathway activation (such as gene expression profiling) may be the best predictors of response to EGFR inhibitors.
Conversely, IGF-1R may have not only prognostic but also predictive value. Preclinical data suggest that IGF-1R represents an important source of resistance to therapy,46 suggesting that dual pathway targeting might result in better clinical outcomes. This hypothesis is currently being tested in clinical trials.47 If the hypothesis holds true, patients who might benefit from dual pathway targeting could be selected by detection of IGF-1R expression levels in their tumor.
Concerns have been raised about the validity of IGF-1R antibodies. Nevertheless, recent data derived from lung cancer patients have confirmed the high correlation between the Ventana antibody and the mRNA expression.48 Moreover, the difference between the H-score and HercepTest seems to be related to the different cutoff values used in each test.
The role of IGF-1R in pancreatic tumors has not been extensively assessed in the clinical setting, and this study represents 1 of the first attempts to explore its clinical significance. A vast amount of preclinical evidence supports the concept that IGF-1R is involved in oncogenesis and tumor progression; this concept is supported by our findings. However, contrary to what has been previously found in NSCLC,49 IGF-1R gene copies did not appear to have any predictive or prognostic significance in our study. This may be because of the small sample size and the finding that the proportion of patients with IGF-1R gene amplification was also small (10%).
Previous in vitro studies have demonstrated increased IGF-1 and IGF-1R mRNA in pancreatic tumor stromal cells.12 To our knowledge, ours is the first study to find an association between IGF-1R stromal expression and overall survival in pancreatic ductal carcinoma. As little is known about the intrinsic mechanisms leading to the complex interaction between a tumor cell and its surrounding environment, this strong association found between tumoral stromal IGF-1R detection and overall survival deserves further exploration, confirmation, and clarification.
Finally, we believe this study demonstrates that distinct cell-type expression within a pancreatic tumor as well as subcellular localization of a biomarker is critical for its optimal utility. Thus, our work underscores both the intricacies in practical application of biomarkers, and at the same time the importance of these biomarkers for the individualized treatment of pancreatic cancer patients.