Role of eIF4E in human breast cancer

The eukaryotic translation initiation factor, eIF4E, is a 25-kD cytosolic cap-binding protein that recognizes and binds to the 7-methylguanosine cap in the 5'-untranslated regions (5'-UTR) of mRNAs during the initiation of protein translation (reviewed in [6, 7]). eIF4E may be considered the rate-limiting component in translation initiation because it is found in much lower amounts than other translation factors and is activated via mitogenic stimuli (serum, phorbol esters, tumor necrosis factor a, and lipopolysaccharide [6]). Several complex 5'-UTR mRNAs involved in cell division, cell growth, and angiogenesis, are known to be selectively translated via eIF4E, including ornithine decarboxylase (ODC) [8], vascular endothelial growth factor (VEGF) [9], c-Myc [10], cyclin D1 [11], and Tousled-like kinase 1B (TLK1B) which mediates radioresistance [12]. Furthermore, fibroblast cells transfected with eIF4E develop a malignant phenotype, whereas treatments aimed at inhibiting the level or activity of eIF4E result in inhibition of tumorigenic properties [13]. eIF4E is overexpressed in malignant breast cancer tumor lines MDA-MB-435, MDA-MB-231, and MCF-7, but not in non-tumor cells (MCF-10A) or epithelial cells from the milk of a nursing mother [14]. eIF4E protein expression is also elevated in a variety of human cancers including breast cancer but not in stroma or in benign tissue [13]. Furthermore, eIF4E expression is elevated during hypoxia [15], and at the invasive front in head and neck cancer and in invasive disease [16].

Based on these observations, clinical studies have been conducted to determine the relationship between eIF4E overexpression (quantitated by western blot analysis) and clinical outcome. The results indicated that patients with high eIF4E had a statistically significant higher rate of cancer recurrence (n = 38, p = 0.03 log-rank test) and cancer-related death (n = 38, p = 0.04 log-rank test) compared to those with low eIF4E overexpression in a 40-month follow-up [17]. These results were further supported in a prospective trial, in which patients with high eIF4E overexpression had a shorter disease-free survival (p = 0.004, log-rank test) and higher cancer-related deaths (p = 0.002) compared to those with low eIF4E overexpression. Furthermore, eIF4E protein expression correlated with increased VEGF levels and microvessel density [18]. Significantly, eIF4E expression was independent of ER, PR, HER-2/neu, or node status as determined by Cox proportional hazard model [18, 19].

Fresh-frozen vs formalin-fixed paraffin embedded tissue

As mentioned above, high eIF4E overexpression has been associated with a worse clinical outcome [17]. However, one of the limiting factors in that study was that it required western blot analysis of fresh-frozen tissue. Fresh-frozen tissue is typically scarce, especially in smaller tumors. Furthermore, in order to conduct a multi-institutional study to analyze enough samples for meaningful results, archived specimens will be essential. In addition, the use of paraffin-embedded archived samples would be useful for long-term follow-up. This will enable researchers and clinicians to establish eIF4E as a standard prognostic or diagnostic factor. Additionally, if eIF4E is determined to be a diagnostic factor, it may be used to personalize therapeutic care of the patient.

Tissue Microarrays

Yang and colleagues recently reported that eIF4E levels were moderately correlated with VEGF and cyclin D1 in a breast cancer TMA [20]. This TMA was obtained from TARP http://www.cancer.gov/tarp/. However, although complete histologic data was available for breast, only limited and incomplete clinical information was available. The goal of our present study was to validate our own in-house TMA's by comparing eIF4E expression with known downstream effector molecules, cyclin D1, c-Myc, VEGF, TLK1B, and ODC. We possess complete clinical information on each specimen, which will allow future TMAs to be constructed for further analysis.

Tissue procurement for western blot analysis

Breast cancer specimens of at least 100 mg were obtained from the tumor core at the time of surgery from each patient per IRB approved protocol. The specimens were verified by the study pathologist to be invasive mammary carcinomas. The specimens were then immediately frozen in liquid nitrogen and stored at minus 70°C for subsequent assay preparations.

Construction of TMAs

The archived H&E slides used for diagnosis were reviewed by the pathologist on the team for confirmation of diagnosis and selection of appropriate paraffin-embedded tissue blocks for the construction of TMAs. Slides with appropriate tissue of interest were selected and mapped to define representative areas for construction of the TMA blocks using a 1.5 mm punch size. In all, 3 TMA blocks were constructed.

TMA block 1 consisted of the following specimens: 5 node positive breast ductal carcinoma, 3 node negative breast ductal carcinoma, 1 ductal carcinoma in-situ, and 1 benign breast tissue. The carcinomas (in-situ carcinoma and the invasive ductal carcinoma cases) were punched in triplicates and the patient-matched controls of corresponding benign breast tissues were punched in duplicates. For slide orientation and as additional tissue control, normal pancreas tissue (punched in duplicate) was also included in each TMA.

TMA block 2 consisted of the following specimens: 6 node positive breast ductal carcinoma, 6 node negative breast ductal carcinoma, 2 ductal carcinoma in-situ with matched, 2 benign breast tissues as benign controls from the 2 the patients with ductal carcinoma in-situ, and 1 benign breast tissue from a breast reduction surgery. The invasive carcinomas were punched in triplicates. The in-situ carcinoma cases and the matched benign controls were punched in duplicates.

TMA block 3 consisted of the following specimens: 38 invasive ductal carcinoma patients (40 cases punched but 2 had no tumor on the TMA), 3 patients with ductal carcinoma in-situ, and 3 normal breast tissues from breast reduction surgeries.

Immunohistochemistry

For the immunohistochemical analysis, 5 μm thick sections were cut, warmed to 60°C, de-paraffinized in xylene, and then rehydrated with graded ethanol. This step was followed by antigen exposure for 20 minutes in heated antigen retrieval solution and then the endogenous peroxide activity was inactivated by treating with 0.3% H₂O₂ in methanol. The sections were blocked for 20 min in protein block (normal goat serum in PBS, BioGenex), and incubated with primary antibodies against ODC (Sigma #O1136, diluted 1:500); eIF4E (monoclonal, BD Transduction Laboratories, 1:600 dilution), c-Myc (Abcam, ab31426, 1:500 dilution), TLK1B (from De Benedetti [21], 1:700 dilution), VEGF (Ab-3, JH121, NeoMarker-Labvision, 1:60 dilution), and cyclin D1 (Cell Signaling #2926, 1:100 dilution) for 1 h using an automated stainer (BioGenex I6000 Automated Staining System, San Ramon, CA). Samples were rinsed 5 times in washing buffer, and incubated in secondary antibody (MultiLink-BioGenex Super Sensitive Link-Label IHC Detection System) for 30 min. Samples were rinsed 3 times in wash buffer, and then incubated in horseradish peroxidase label (BioGenex) for 15 min. Samples were rinsed 3 times in wash buffer and then incubated in diaminobenzidine (Dako Cytomation Liquid DAB Substrate Chromogen System) for 5 min. Samples were rinsed 3 times in wash buffer and counterstained in hematoxylin (Dako Cytomation Automation Hematoxylin) for 2 min.

Western Blot

Specimens were analyzed for eIF4E and TLK1B as previously described [22, 23]. Briefly protein lysates from each specimen (5–10 μg protein) were separated using 12% denaturing gel Tris-HCL polyacrylamide gel electrophoresis [24]. The proteins were then electroblotted on a nylon membrane (Immun-Blot PVDF, Bio-Rad Laboratory, Hercules, CA) [25]. The membranes were blocked in 3% nonfat milk overnight. Primary incubation of the membranes was carried out using a 1:1000 dilution of monoclonal mouse anti-eIF4E antibody (610270; BD Biosciences, San Jose, CA) or rabbit anti-TLK1B antibody (1:1000 dilution, De Benedetti laboratory). Secondary incubation of the membrane was then carried out using a 1:5000 dilution of goat antimouse or anti-rabbit IgG tagged with horseradish peroxidase. The blot was developed using Opti-4CN substrate kit (Bio-Rad Laboratories, Hercules, CA). The blots were scanned using the Biophotonics system (Biophotonics Corp., Ann Arbor, MI). The band intensity was evaluated using the Intelligent Quantifier software (Bio Image, Ann Arbor, MI). The overexpression of eIF4E and TLK1B was quantified as x-fold over the samples of benign tissue from noncancer specimens run concurrently on the gel.

Analysis of TMAs

The first TMA (TMA1) was constructed to optimize antibody dilutions. The second TMA (TMA2) was designed with triplicate specimens to analyze intra-individual variability. In this regard, three separate plugs from each patient were taken from each original block and re-imbedded into TMA2. Replicate breast tumor specimens were analyzed for plug-to-plug reproducibility by staining the TMAs immunohistochemically and quantitating them using the ARIOL imaging system (described below). The third TMA (TMA3) was designed to compare eIF4E to its downstream effector proteins using a larger set of breast cancer specimens.

ARIOL Imaging

The ARIOL imaging system (Genetix, San Jose, CA) was used to quantify antibody staining of the TMAs. The specimens were scanned at a low resolution (1.25×) and high resolution (20×) using Olympus BX 61 microscope with an automated platform (Prior). The slides were loaded in the automated slide loader (Applied Imaging SL 50). The images with high resolution were used for training and quantification purpose. The system was trained to select the stained and unstained cells/nuclei by the color of staining and shape of nuclei such that brown staining was considered positive and blue staining was considered negative. The number of cells/nuclei stained was calculated and represented as percentage of total cells/nuclei stained positively. By measuring both immunostaining intensity and percentage, data obtained are reproducible, objective measurements of immunoreactivity. Because standardizing IHC, from the fixation of tissues to the analysis of IHC results is critical, all immunohistochemistry data were normalized to cytokeratin. To control for the variability in tumor cellularity from one patient to another, and to also control for variations in the number of tumor cells at different TMA spots (intra-tumoral variations), the number of epithelial (tumor) cells present at each TMA spot as highlighted by expression of cytokeratin 7, was used for normalization of each protein expression studied [26]. For each protein, a score was generated based on the area with and the intensity of the brown staining reaction. The scores were then exported to an Excel spreadsheet for analysis. A normalized value was calculated for each protein by dividing the recorded score for the protein by the recorded score for cytokeratin 7 at the corresponding spot.

ER, PR, HER-2/neu analysis

Immunohistochemical staining for estrogen receptor (ER), progesterone receptor (PR), and HER-2/neu was performed using automated processing and staining technology (BenchMark XT IHC/ISH, Ventana). Processes included deparaffinization, pretreatment, antibody incubation, counterstaining, and coverslipping. Levels of membranous/cytoplasmic immunostaining for Her-2/neu, were scored using an automated cellular image analysis system (ACIS) (Clarient, San Juan Capistrano). Values less than 1.9 are interpreted as negative and values ≥ 2.0 are interpreted as positive for HER-2/neu over-expression. Nuclear ER and PR expression was assessed using the ACIS; both the quantitative intensity of expression and percentage of cells showing positive expression were noted.

Statistical analysis

Intra-individual coefficient of variations (CV) was calculated as ratio of standard deviation over mean × 100. The mean CV% and SD of CV for each marker was also added. The correlation among the expression levels of eIF4E, c-Myc, cyclin D1, ODC, TLK1B, VEGF, ER, PR, and HER-2/neu were calculated by the Spearman rank correlation method. These correlation coefficients were test against 0. All two-sided p-values < 0.05 were considered as statistically significant. The strength of correlation among the markers were classified as strong, moderate and weak for the correlation coefficient > 0.8, 0.4–0.8, and < 0.4 respectively. The statistical software used for the current study was SAS 9.1.3. SAS Institute Inc., Cary, NC.

Construction and analysis of TMAs

The first TMA was constructed in order to optimize the immunohistochemical staining techniques and to train the ARIOL imaging system. The criteria for successful staining included appropriate staining to the subcellular compartment, lack of staining in the absence of primary antibody, increase in staining when higher concentrations of primary antibodies were used, low staining in non-epithelial derived tissue (such as stroma or fat), and low staining in the negative controls (benign tissue). An example of the construction of TMA3 is shown in Figure 1. The ARIOL system first images the entire slide to show each plug. Higher resolution images can be made by zooming in on each plug. As shown in Figure 2, the ARIOL system can be trained to distinguish between cytoplasmic and nuclear staining. For example, ODC typically stains in the cytoplasm, leaving the counter-stained nuclei predominantly blue (Figure 2). The computational software can then scan and analyze each plug for positive staining.

Intra-individual coefficients of variances

TMA-IHC analysis: Correlation of eIF4E with downstream effector proteins

In TMA3, eIF4E expression levels correlated strongly with the downstream effector proteins, c-Myc, cyclin D1, ODC, TLK1B, and VEGF (Figures 3, 4). In Figure 3, we show a set of human breast carcinoma specimens from TMA3 that were either low or high for eIF4E (as measured by IHC). Their positions on TMA3 are marked in Figure 1. Then, the IHCs for the same specimens are also shown for the downstream effector proteins. Generally, specimens that possessed high eIF4E protein expression also exhibited high expression of c-Myc, cyclin D1, TLK1B, VEGF, and ODC. Likewise, specimens that expressed low amounts of eIF4E protein also expressed low amounts of c-Myc, cyclin D1, TLK1B, VEGF, and ODC. The rho values for eIF4E compared to c-Myc, cyclin D1, TLK1B, VEGF, and ODC were 0.880, 0.863, 0.729, 0.699, and 0.799 respectively, and all these comparisons were statistically significant at p ≤ 0.0001 (Figure 4A–E).

Western blot analysis: Correlation of eIF4E with TLK1B

We have previously shown by western blot analysis that the expression of eIF4E correlated with that of TLK1B [23]. As further validation of our TMA results, we also compared eIF4E with TLK1B using the corresponding fresh-frozen specimens from the same tumors as those used for TMA3 (Figure 4F). Due to limited amounts of fresh-frozen specimens, the other proteins were not analyzed. Protein expressions of eIF4E to TLK1B were positively correlated (rho value 0.485, p value 0.0054).

Non-correlation to independent markers