We have previously isolated an oncogene EIF5A2 (eukaryotic initiation factor 5A2) from a frequently amplified region at 3q of a primary ovarian cancer cell line, and demonstrated its impact on prognosis in human ovarian cancer. Amplification of chromosome 3q has also been detected frequently in non–small cell lung cancer (NSCLC), however, abnormalities of EIF5A2 and its clinicopathologic significance in NSCLC haven't been studied. In our study, the methods of immunohistochemistry and fluorescence in situ hybridization were utilized to examine protein expression and amplification of EIF5A2 in 248 surgically resected NSCLCs (learning cohort) and another validation cohort of 120 stage I NSCLC patients. Overexpression and amplification of EIF5A2 was detected informatively in 48.7% and 13.7% of NSCLCs in learning cohort, 33.3% and 6.0% of NSCLCs in validation cohort. Overexpression of eIF5A-2 was found to correlate with gene amplification, increased cell proliferation and advanced T stage. In learning cohort, eIF5A-2 expression was evaluated as a strong prognostic factor on disease-specific survival, but in subgroup analyses, it only retained its stratified significance in stage I set (Hazards ratio = 2.799, p = 0.001). In validation cohort, the impact of eIF5A-2 expression on survival in stage I NSCLC patients was also observed (Hazard ratio = 2.097, p = 0.014). Our findings suggested that overexpression of eIF5A-2 correlates with local invasion of NSCLC, and might serve as an adverse prognostic marker of survival for stage I NSCLC patients.
Lung cancer is the first cause of cancer death in the world.1 Non–small cell lung cancer (NSCLC), which accounts for ∼85% of all lung cancers, often display considerable variability in recurrence and survival within the same tumor-node-metastasis (TNM) stages.2 Even for patients diagnosed as early as stage I, the 5-year survival rate ranged from 40 to 67%,3 which indicates the heterogeneity of prognoses within this population and the inadequacy of the TNM staging system in providing critical information that may influence the strategy of treatment. Since acquired chromosomal aberrations play an important role in tumor development and progression, the search for specific molecular and/or genetic alterations in NSCLC cells that can provide additional staging information would be of great value.4
Amplification of chromosome 3q has been detected frequently in several human malignancies,5 including lung cancer,6 suggesting an oncogene(s) localized in this region which might play important roles in carcinogenesis. We have isolated a novel candidate oncogene, EIF5A2 (eukaryotic initiation factor 5A2), from a frequently amplified region at 3q of a primary ovarian cancer cell line.7 Previous study showed that intracellular depletion of EIF5A2 could cause the inhibition of cell growth and induce cell apoptosis. Our results demonstrated that the cell growth in ovarian cancer cell line UACC-1598 could be inhibited substantially by the treatment of antisense DNA against EIF5A2 gene;8 and overexpression of eIF5A-2 was found to correlate with an ascending clinical stage and poor patient prognosis in ovarian cancer.9 Interestingly, up-regulated expression of eIF5A-2 was reported to be important in the acquisition of a recurrence phenotype in superficial bladder cancer.10 Despite these observations, the EIF5A2 expression and its clinicopathologic significance in human NSCLC remain unclear. As amplification of 3q, in which the EIF5A2 gene is located, was also frequently detected in NSCLC, our study is to evaluate its role in carcinogenesis and progression of NSCLC.
In our study, 248 NSCLC patients who received complete resection between February 1994 and January 1998 in Cancer Center, Sun Yat-sen University were enrolled (learning cohort). As a validation, 120 cases stage I NSCLC patient treated radical surgery in the same period from another institute: Guangdong Provincial People's Hospital (validation cohort) were also studied. The cases were selected consecutively based on the availability of resection tissue and follow-up data. Patients who have previous malignant disease or a second primary tumor, or those who received preoperative radiotherapy and/or chemotherapy were excluded. Tumor tissue samples from these patients were obtained from the surgical pathology archives of the Pathology Department of the two institutes. The study was approved by the medical ethics committee of the two institutes.
Data regarding stage according to the pathology Tumor-Node-Metastasis (pTNM) system (AJCC/UICC 2002)11, 12 and differentiation and histotype according to the World Health Organization (WHO) classification for NSCLC were retrospectively collected and shown in Table 1. Because adjuvant chemotherapy became a standard approach only after 2004, systemic chemotherapy using cisplatin-based combinations was administered to patients with stages III NSCLCs, no postoperative therapy was delivered to patients in stage I-II. Patients with stage III disease and pathological evidence of N2 disease received postoperative mediastinal radiotherapy.
Table 1. Correlation between eIF5A-2 expression and clinicopathologic variables
Construction of tissue microarrays (TMA)
The tissue microarrays (TMA) was constructed according to a method described previously.13 Briefly, the individual donor tissue block and the corresponding histological H&E stained slides were overlaid for tissue TMA sampling. The tissues (248 NSCLC tissues from learning cohort and 50 normal lung tissues taken from regions that were not affected of the same patients) were sampled using a tissue arraying instrument (Beecher Instruments, Silver Spring, MD); a 0.6-mm-diameter cylinder of tissue was removed. Subsequently, the tissue cylinder was re-embedded into a predetermined position in a recipient paraffin block. In our constructed lung tissue-TMA, 3 cores of sample were selected from each primary NSCLC and normal lung tissue. Multiple sections (5 μm thick) were cut from the TMA block and mounted on microscope slides. The same technique was performed for the 120 samples from validation cohort.
Immunohistochemistry (IHC) studies were performed using a standard streptavidin-biotin-peroxidase complex method.14 For antigen retrieval, TMA slides were microwave-treated and boiled in a 10 mM citrate buffer (pH 6.0) for 10 min. Nonspecific binding was blocked with 10% normal rabbit serum for 20 min. The TMA slides were incubated with either monoclonal mouse antihuman eIF5A-29, 10, 15 (kindly provide by Dr. Geng-Xi Hu, the Institute of Shanghai Science and Technology, Shanghai, China, 1:100 dilution) overnight at 4°C or anti-Ki-67 (Dako, Glostrup, Denmark, 1:100 dilution) for 30 min at 37°C in a moist chamber. The eIF5A-2 antibody used in our study has been recently used to examine the expression of eIF5A-2 by Western blot in a panel of fresh bladder cancer tissues, and a single band of ∼17 kDa was detected, which is the expected size for eIF5A-2.16 A negative control was obtained by replacing the primary antibody with normal murine IgG. Known immunostaining positive ovarian cancer slides were used as positive controls.
A semiquantitative scoring criterion for IHC of eIF5A-2 was used, in which both staining intensity and positive areas were recorded.9, 10, 15 A staining index (values 0–12), obtained as the intensity of eIF5A-2 positive staining (negative = 0, weak = 1, moderate = 2, or strong = 3 scores) and the proportion of immunostaining positive cells of interest (<25% = 1, 25–50% = 2, >50% to <75% = 3, ≥75% = 4 scores) were calculated.9, 10, 15 For Ki-67, the status of Ki-67 nuclear expression was assessed as the percentage of Ki-67-positive cells stained in each tumor.9, 10, 15 A minimum of 300 epithelial cells were counted for each normal and tumor case. Two independent pathologists (Dr. Xie D and Rao HL) blinded to the clinicopathologic information performed the scorings.
Fluorescence in situ hybridization (FISH)
Two-color FISH was performed using a Spectrum Orange-labeled BAC clone (RP11-115J24) at 3q26.2 containing the EIF5A2 gene and a Spectrum Green-labeled reference centromeric probe on chromosome 3 (Vysis, Downers Grove, IL). The FISH reaction was performed as described previously17 with slight modification. Briefly, deparaffinized TMA sections were treated with proteinase K (400 μg/ml) at 37°C for 25 min, followed by denaturing in 70% formamide, 2× SSC at 75°C for 6 min. Fifty nanograms of each probe were mixed in a 20 μl hybridization mixture (containing 55% formamide, 2× SSC, and 2 μg human Cot1 DNA), denatured at 75°C for 6 min and then hybridized to the denatured TMA sections at 37°C for 24 hours. The slides were counterstained with 1 μg/ml DAPI in an antifade solution and were examined with a Zeiss Axiophot microscope (Nussloch, Germany) equipped with a triple-band pass filter. A minimum of 300 tumor cells was evaluated per specimen. Amplification of EIF5A2 was defined as presence of either 6 (or more) EIF5A2 gene signals or at least three times as many gene signals than centromere signals of chromosome 3 in cells (Fig. 1c).9, 10, 15
Statistical analysis was performed with the SPSS software (SPSS Standard version 13.0, SPSS, Chicago, IL). Pearson's correlation coefficient (R) was used to assess the correlation between the scores judged by the two independent pathologists. Chi-square test was used to assess the statistical significance of the association of eIF5A-2 expression with the patient's clinicopathologic parameters and its correlation with EIF5A2 amplification. An independent sample t test was used to assess the expression of Ki-67 between groups with overexpression and normal expression of eIF5A-2. Disease-specific survival (DSS) was calculated from the date of diagnosis to the date of cancer-related death or last follow-up, assessed with the Kaplan-Meier method and compared by the log rank test. Multivariate Cox regression analysis was performed on all parameters that were found to be significant on univariate analysis. Two-tailed p value <0.05 were considered statistically significant.
eIF5A-2 expression in lung tissues
Positive expression of eIF5A-2 in epithelial tissue cells was primarily a cytoplasmic pattern (Figs. 1a and 1d). eIF5A-2 expression could be evaluated informatively in 224/248 (90.3%) of the learning cohort, 108/120 (90.0%) of the validation cohort, and 43/50 (86.0%) of normal lung tissues. The noninformative samples included unrepresentative samples, samples with too few tumor cells (<300 cells per case) and lost samples; such were not used in data compilation. To evaluate the reproducibility of eIF5A-2 IHC scoring, the interobserver agreements were assessed by Pearson's correlation analysis, and a significant correlation was observed (Pearson's r = 0.98, p < 0.001, Fig. 2c). The interobserver disagreements were reviewed a second time, followed by a conclusive judgment by both pathologists.
Because the expression of eIF5A-2 in each of 43 informative normal lung tissue was negative (0) or weak (1), the staining index in the normal lung tissues was determined to be less or equal to 3. Therefore, we designated the staining index of 0–3 as the normal expression of eIF5A-2, while staining index of 4–12 was depicted as overexpression of this protein (Figs. 1a and 1d).9, 10, 15 The distribution of eIF5A-2 scores of the two cohorts are displayed as Figures 2a and 2b. Using this designation, overexpression of eIF5A-2 was observed informatively in 109/224 (48.7%) of the NSCLCs in learning cohort, 36/108 (33.3%) of those in validation set with only stage I cases, and none of the normal lung tissues. Up-regulation of EIF-5A2 expression was also found to correlate with an ascending pT stage in both sets (p = 0.011, Table 1).
Relationship between clinicopathologic variables, eIF5A-2 expression and NSCLC patient survival
The mean observation period was 61.3 months (3.2–172.5 months) and 95.1 months (4.6–172.5 months) for learning and validation cohort; and 154 and 49 cancer-related deaths were observed, respectively.
Firstly, we assessed the impact of eIF5A-2 expression on DSS of NSCLC patients in learning cohort. Overexpression of eIF5A-2, elder age and advanced stage were evaluated to correlate with poor DSS for the whole cohort (Table 2, Fig. 3a). No significant association was found between DSS and other clinicopathologic variables, including gender, histology, tumor grade and receiving adjuvant radiotherapy or chemotherapy or not (data not shown). The parameters that were significant in univariate analysis were further examined in multivariate Cox regression analysis; eIF5A-2 expression and clinical stage were the only two significant independent prognostic factors on DSS. Further stratified analysis showed that eIF5A-2 expression could distinguish patient survival in stage I subgroup (HR = 2.799, p = 0.001), but not in stage II (HR = 1.549, p = 0.266) and III subgroups (HR = 1.430, p = 0.125) (Figs. 3b–3d).
Table 2. Cox proportional hazards regression analysis of DSS in learning cohort
For validation, we then consecutively selected 120 stage I NSCLC patients who received complete resection in the same period from another independent cohort, and got the similar results. eIF5A-2 expression (HR = 2.447, p = 0.002, Fig. 3e), age and pT stage were significant prognostic factors in univariate analysis; and eIF5A-2 expression was the most important predictor of survival (HR = 2.097, p = 0.014), even surpassing pT stage in multivariate analysis (Table 3).
Table 3. Multivariate Cox regression analysis of DSS in validation cohort
Correlation of eIF5A-2 expression and cell proliferation in NSCLCs
To address whether or not eIF5A-2 expression in NSCLC is associated with cell proliferation, the expression of Ki-67, a widely used cellular proliferation marker, was examined by IHC in our NSCLC -TMAs. Among the 368 NSCLCs, in 312 samples, eIF5A-2 and Ki-67 IHC was detected successfully and simultaneously. For the 123 cases with overexpression of eIF5A-2, an average of 28.5% of the NSCLC cells stained positive with Ki-67 antibody, a percentage of cancer cells that was significantly larger than that (16.4%) in the remaining 189 cancers with a normal expression of eIF5A-2 (p < 0.001, Fig. 2d).
Amplification of EIF5A2 in lung tissues
The FISH analysis was informative in 24/50 (48.0%) of the normal lung tissues and 53.8% (198/368) of the NSCLCs. Samples without FISH signal and samples with weak target signals or those with a strong signal background were the main reasons for most of the noninformative cases. FISH results demonstrated that the amplification of EIF5A2 was not detected in any of the normal lung tissues; but was detected informatively in 13.7% (18/131) and 6.0% (4/67) of the cases in each cohort. Totally, 175 informative cases were shown to be positive by both IHC and FISH simultaneously. Chi-square test indicated a significant correlation between overexpression of eIF5A-2 and amplification of EIF5A2 (p < 0.001, Supporting Information Table 1). In the majority (20/22, 90.9%) of carcinomas with EIF5A2 amplification, overexpression of eIF5A-2 protein was observed; whereas, eIF5A-2 overexpression can also be detected in 37.3% (57/153) of the NSCLCs without EIF5A2 amplification.
In our study, our results showed that the expression of eIF5A-2 in all of the normal lung tissue specimens was absent or at low levels. In many of our NSCLC specimens, in contrast, an overexpression of eIF5A-2 was frequently detected, and the frequency of eIF5A-2 overexpression increased with an ascending pT stage in NSCLC. Similar results were also observed in ovarian carcinoma.18 These findings suggest the possibility that up-regulated expression of eIF5A-2 may provide a selective advantage in local invasion in certain types of human solid tumors, including NSCLC.
It is noteworthy that overexpression of eIF5A-2 was an adverse predictor of survival for NSCLC patients in our learning cohort, independent of clinical stage. Thus, we wondered if eIF5A-2 expression could further distinguish patient survival within different stages. Interestingly, it was found to retain its stratified significance only in stage I set. Since it was retrospective conducted, some significant correlation based on subgroup analyses might be accidentally observed; we then consecutively selected 120 stage I cases from another independent cohort, in which this implication was validated. Our findings indicated a potentially important role of EIF5A2 as an underlying biological mechanism of tumor progression in early stage of NSCLC. As we know, despite an estimated 33–60% of patients diagnosed as early as stage I die of this disease every year,3 the use of adjuvant chemotherapy for these patients remains a matter of debate.2, 3 Our results suggested that the examination of eIF5A-2 expression by IHC may be used as an additional tool in distinguishing stage I NSCLC patients with unfavorable prognosis from those with better prognosis, which might provide some useful information for clinicians to optimize individual therapeutic strategies for stage I NSCLC patients. Obviously, further prospective studies are still needed before such implication can be put into clinical practice.
When come to the function of EIF5A, it has been well established to be essential for sustained cell proliferation in mammalian cells;19–21 while inhibiting of EIF5A activation would exert strong antiproliferative effects in various human cancer cell lines, and cause arrest of cell cycle progression.22, 23 In our study, we observed an overall significant positive association of up-regulated expression of eIF5A-2 and increased NSCLC cellular proliferation. Since fast cell growth and frequent triggering of cell division in tumor demands a general up-regulation of the protein synthesis machinery, our results suggest a potential important role of EIF5A2 in the control of cell proliferation, an activity that might be responsible, at least in part, for carcinogenesis and/or progression of NSCLC. In addition, our previous study has found that antisense DNA against the EIF5A2 gene could inhibit cell growth in ovarian cancer cell line UACC-1598 with amplification of EIF5A2.8 Other studies also indicated that inhibition of EIF5A activity could suppress the growth of HeLa cells,24 inhibit v-src-transformed NIH3T3 cells proliferation, and induce apoptosis.25 These findings provide evidence that the EIF5A2 may function as a proliferation-related oncogene in tumorigenic processes. Clearly, further work needs to be done to more precisely understand the potential oncogenic function of EIF5A2 in human cancer pathogenesis and progression.
With regard to the mechanism of up-regulated protein expression of eIF5A-2 in NSCLCs, it is known that overexpression of an oncogene is often caused by DNA amplification.26 To determine whether the overexpression of eIF5A-2 in NSCLCs was caused by gene amplification, the amplification status of EIF5A2 was examined by FISH. Our results indicated that DNA amplification might be only one of the mechanisms that cause eIF5A-2 overexpression, since overexpression of eIF5A-2 can also be observed in a portion of the NSCLCs without EIF5A2 amplification. Recently, Clement et al. reported that, unlike eIF5A-1, which mRNA and protein are constitutively expressed in human cells; in contrast, eIF5A-2 protein is hardly detectable in most human normal and cancer cell lines, even when EIF5A2 mRNA levels are clearly detectable.20 In a large series of human cancer cell lines they examined, eIF5A-2 protein was detectable only in two of the cell lines, colorectal-SW-480 and ovarian-UACC-1598.20 They demonstrated that the expression of eIF5A-2 seems to be regulated in a tissue- or cancer-specific manner, and the failure to detect eIF5A-2 protein even in EIF5A2 mRNA positive cells is, at least in part, due to inefficient translation.20 These data suggest that the regulation of protein expression of eIF5A-2 is complicated and it might be regulated not only by gene amplification, but also by other molecular mechanisms including transcriptional regulation and post-translational regulation (e.g., microRNA).27
In summary, the present study is the first one assessing protein expression and amplification patterns of EIF5A2 in NSCLC. Our results provide some evidences for the concept that overexpression of eIF5A-2 might play an important role in carcinogenesis of NSCLC, and might serve as a useful molecular marker for those with stage I disease.