Abstract
The serine/threonine protein kinase aurora B, a key regulator of mitosis, is emerging as a novel drug target for cancer treatment. Aurora B overexpression has been previously documented by immunohistochemistry in several types of human tumors. We assessed aurora B expression in a series of 160 non–small cell lung cancer (NSCLC) samples (60% stage I, 21% stage II, 11% stage III, and 8% stage IV). In addition, we determined the expression of survivin and p16, two molecules also involved in cell cycle control. Aurora B was expressed selectively in tumor cells compared with normal epithelium. Aurora B expression was significantly correlated with expression of survivin in the nucleus (P < 0.0001), but not with expression of p16 (P = 0.134). High aurora B expression levels were significantly associated with older age (P = 0.012), male sex (P = 0.013), squamous cell carcinoma histology (P = 0.001), poor tumor differentiation grade (P = 0.007), and lymph node invasion (P = 0.037), in the subset of radically resected patients in our series. In addition, aurora B expression predicted shorter survival for the patients with adenocarcinoma histology, at both univariate (P = 0.020) and multivariate (P = 0.012) analysis. Survivin expression levels were neither associated with patient clinicopathologic characteristics nor with survival. However, expression of survivin in the nucleus was preferentially detected in stage I and II than in stage III and IV (P = 0.007) in the overall series of NSCLC samples. Taken together, our results suggest that aurora B may represent a valid target in NSCLC. [Mol Cancer Ther 2006;5(11):2905–13]
- aurora B
- survivin
- p16
- inhibitors of apoptosis
- kinase inhibitors
Introduction
Lung cancer is the most common cause of cancer mortality worldwide (1). Approximately 80% of lung tumors are non–small cell lung carcinomas (NSCLC), for which surgery represents the main curative treatment modality. However, only 30% of NSCLC patients can undergo an operation. The 5-year survival of NSCLC is only ∼15% and did not improve in the past decades. Chemotherapy has been increasingly used in advanced NSCLC, and more recently also in the adjuvant setting. However, the effects of chemotherapy are modest, at the cost of significant toxicity. There is a major need for improvement in the systemic treatment, and the recent introduction of targeted therapies for NSCLC has had some success with the epidermal growth factor receptor tyrosine kinase inhibitors (2). Among potential targets for therapy are molecules involved in the regulation of mitosis. Examples of such targets are the members of the aurora kinase family, and inhibitors of aurora kinases are now being tested in the clinic (3).
The kinases aurora A/STK-15, aurora B/AIM-1, and aurora C/AIK-3 are members of a family of serine/threonine protein kinases (4), all described as mitotic regulators. Aurora A is normally localized in the pericentrosomal region, to where, during mitosis, it recruits several important components on the forming mitotic spindle (5). Thus far, aurora A is the only family member that has been reported to act as an oncogene (6). Aurora B is localized to the late-interphase chromosomes proximal to the centromere, from where, during mitosis, it regulates the phosphorylation and thus the mitotic activity of many substrates, including histone H3 (7, 8). Together with survivin, inner centromere protein, and borealin, aurora B constitutes a chromosome passenger complex that regulates several aspects of mitosis, such as the spindle checkpoint and cytokinesis (9). Aurora C has also been reported to localize to the chromosome passenger complex, but little is known about its function (10, 11).
Survivin is a member of the inhibitor of apoptosis protein family that possesses a single baculoviral inhibitory repeat domain. In addition to regulate apoptosis as the other inhibitor of apoptosis family members, survivin can also direct cell cycle progression during mitosis (12). The p16 protein is a cyclin-dependent kinase inhibitor that regulates the earlier G1-S step of the cell cycle by inactivating the cyclin-dependent kinases that phosphorylate retinoblastoma protein (13).
As defects in mitosis can lead to genomic instability, deregulation of the expression and activity of aurora kinase family members has been implicated in tumorigenesis. Recently, small-molecule inhibitors of aurora kinases have been developed that show promising antitumor effects (14). Unexpectedly, recent evidence from RNA interference experiments and overexpression of aurora kinase mutants (15) indicates that aurora B may be the critical target of aurora kinase inhibitors (16).
Determining the expression of the target may provide critical information to apply one novel targeted agent to cancer patients. Although aurora A expression has been studied in many different tumor types (17–23), aurora B protein expression has received attention only recently in studies on patient material, and has been found overexpressed in primary human colorectal (24), prostate (25), endometrial (26), and thyroid (27) cancers; in seminomas (28); and in astrocytomas (29), at various pathologic stages, with a tendency to group in higher grades of malignancy.
A recent report has shown a general up-regulation of aurora B mRNA levels compared with normal epithelium in the majority of samples of a small cohort of resected lung cancer patients (30). However, aurora B protein expression has not been previously characterized in human NSCLC by immunohistochemistry, and its role in NSCLC progression has not been well documented. The aim of our study was to investigate aurora B protein expression in a large series of NSCLC samples, and to test its correlation with clinical characteristics and patient outcome. To evaluate the effect on patient prognosis and clinical characteristics of two other key players of cell cycle regulation and well-known prognostic factors in NSCLC, we also did immunohistochemistry for survivin and p16 expression in parallel to aurora B. Our results show a marked up-regulation of aurora B in tumor cells, and a correlation with tumor differentiation, thus suggesting that aurora B might be a good drug target in NSCLC.
Materials and Methods
NSCLC Cell Lines, Drug Treatment, and Colony Assay
H460 and H322 NSCLC cells were grown in RPMI 1640 supplemented with 10% FCS, 50 units/mL penicillin, and 50 μg/mL streptomycin. A stock solution (1 mmol/L in DMSO) of the aurora kinase inhibitor ZM 447439 (Tocris Bioscience, Avonmouth, United Kingdom) was prepared and stored in aliquots at −20°C. The drug was freshly diluted in RPMI before being added to the cells. For the colony assays, ∼103 cells per well were plated on six-well trays. Forty-eight hours later, ZM 447439 was added at the indicated final concentrations. After 72 hours, cells were washed and allowed to grow in drug-free medium for a further 48-hour period. Finally, cells were fixed and stained with Giemsa to visualize the colonies. Cells growing on glass coverslips were treated in parallel and stained with the chromosome dye Hoechst 33258 to visualize the nuclei.
Patients and Specimens
The present study included 160 patients with NSCLC (60% stage I, 21% stage II, 11% stage III, and 8% stage IV), diagnosed at the Academic Hospital Vrije Universiteit of Amsterdam between January 1988 and December 1999. The patients, 129 males (81%) and 31 women (19%), ranged in age from 40 to 86 years (mean 64 years). Early-stage patients (from stage I to stage IIIA) underwent radical surgery of primary tumor and lymph nodes, between January 1988 and December 1995, and received no other treatment before or after surgery. Twenty-two patients were inoperable at diagnosis, and were treated between January 1993 and December 1999 with either neoadjuvant chemotherapy followed by surgery or radiotherapy (4 patients), or with palliative chemotherapy (18 patients).
The surgical specimens were classified according to the WHO criteria, into 49 (31%) adenocarcinoma (including bronchioloalveolar carcinomas), 79 (49%) squamous cell carcinoma, and 32 (20%) large-cell undifferentiated carcinomas, on the basis of histology and histochemistry (periodic acid-Schiff and Alcian blue stains). Furthermore, 106 tumors (66%) were poorly differentiated, 49 (31%) moderately differentiated, and only 5 (3%) were well differentiated. Staging was determined using the most recent classification (31).
Clinical data for the patients included in the study were previously reported (32, 33). The patient follow-up was updated; the median follow-up was 131 months and 130 patients were reported dead, 79 patients had relapsed, and 78 of them died. Overall survival was calculated from the date of surgery or beginning of chemotherapy to the last follow-up visit or death.
Immunohistochemistry
Formalin-fixed, paraffin-embedded tissue blocks were freshly cut in 4-μm sections and mounted on poly-l-lysine–coated slides. Slides were deparaffinized, rehydrated, and incubated in a solution of 0.3% H2O2 in absolute methanol for 30 minutes to block endogen peroxidase. Immunohistochemistry was done using the rabbit anti-survivin polyclonal antiserum NB 500-201 from Novus Biologicals (Littleton, CO) at dilution 1:3,000, a mouse anti aurora B monoclonal antibody from Transduction Laboratories (San Diego, CA) at dilution 1:3,000, and the mouse monoclonal 16P04 anti p16 antibody from NeoMarkers (Fremont, CA). Pretreatment was done in autoclave with 10 mmol/L citrate buffer solution (pH 6.0) in the case of the anti-survivin and anti–aurora B antibodies, and in microwave with Tris/EDTA buffer solution (pH 9.0) in the case of the anti p16 antibody. Antisurvivin and anti–aurora B antibodies were incubated overnight at 4°C, whereas anti-p16 was applied for 1 hour. Detection of the primary antibodies was done using the ChemMate DAKO EnVision Detection kit, peroxidase/3,3′-diaminobenzidine, rabbit/mouse (code K 5007, DAKOCytomation, Glostrup, Denmark) according to the protocol of the manufacturer. Finally, samples were counterstained with hematoxylin, dehydrated, and mounted. The use of the antibody against survivin for immunohistochemistry was already characterized in a previous publication from our group (33). Aurora B staining was validated in normal tonsil tissues and in acetone-fixed cell lines (MCF7, Jurkat, SKOV3). The specificity of the anti-survivin and anti-aurora B antibodies was further confirmed using an RNA interference and fluorescence microscopy approach in MCF-7 cells (not shown). Samples from melanoma tissues known to express high levels of p16 were used as positive control. The omission of the primary antibody in simultaneously incubated sections was used as a negative control. All the stained tumor sections were analyzed semiquantitatively. On each section, the percentage of immunoreactive tumor cells was assessed in agreement by two observers (J.J.O. and B.V.), evaluating at least five areas at ×400 magnification, blinded to knowledge of the clinical outcome of patients. Cases were considered positive when >5% of the cells were stained. Expression levels of aurora B were additionally classified as low and high, selecting as cutoff point aurora B median value of expression, as conventionally done in other studies (33).
Statistical Analysis
Analysis of the association between the expression of survivin, aurora B, and p16 with the major clinicopathologic patient characteristics and survival, was done considering the 138 operated patients only, and also considering the complete series. The Spearman's correlation coefficient (ρ) was used to assess the correlation between expression levels of aurora B and survivin. Correlation of aurora B or survivin expression and localization with clinical features was calculated using the χ2 test. The χ2 test was also used to calculate correlation of p16 expression with aurora B or survivin, and with clinical patients' features. Cumulative survival curves were estimated by the Kaplan-Meier method and compared by the log-rank test. Multivariate survival analysis was done according to the Cox regression model.
In all the cases, a value of P < 0.05 was considered statistically significant. Statistical analysis was done using the SPSS software program 9.0 (SPSS, Inc., Chicago, IL).
Results
Effect of the Aurora Kinase Inhibitor ZM447439 on NSCLC Cell Division In vitro and Immunohistochemical Analysis of Aurora B Expression in NSCLC Tissue Samples
The small-molecule aurora kinase inhibitor ZM 447439 has been shown to efficiently disrupt aurora B function, leading to defective cytokinesis in several types of tumor cells, including A549 NSCLC cells (34). Using a colony assay, we tested the effect of ZM 447439 on the NSCLC cell lines H460 and H322. As shown in Fig. 1A , the growth of colonies in both cell lines was dramatically reduced after 72 hours incubation with 1 μmol/L ZM 447439 compared with untreated control samples. No effect was noted when the drug was applied at 100 nmol/L. A more detailed analysis of the colonies using Hoechst dye to stain the nuclei (Fig. 1B) revealed striking differences between the control and the ZM 447439–treated colonies. In the samples treated with 1 μmol/L ZM 447439, the colonies contained a very low number of cells with large, multilobed nuclei, suggesting that the drug had induced inhibition of cell division and endoreduplication as previously reported (34). These data suggest that inhibition of aurora kinase activity with ZM 447439 profoundly affect division of NSCLC cells in vitro, and are consistent with recent results obtained with a different aurora kinase inhibitor, VX-680 (35).
Effect of the aurora kinase inhibitor ZM 447439 on NSCLC cell division in vitro. A, colonies of H460 and H322 NSCLC cells untreated (control) or incubated with 100 nmol/L or 1 μmol/L ZM 447439 (ZM). Approximately 103 cells per well were seeded and treated as detailed in Materials and Methods. The resulting colonies were stained with Giemsa. In both cell lines, 1 μmol/L ZM 447439 dramatically reduced colony growth. B, detail of single colonies stained with Hoechst to visualize the nuclei. In control samples, the colonies contained hundreds of cells, whereas in the samples treated with 1 μmol/L ZM 447439, the colonies contained only a few cells with large, multilobed nuclei. Bar, 100 μm in all cases.
We carried out immunohistochemical analysis of aurora B expression in a series of 160 tumor tissue samples from NSCLC patients. Aurora B staining was assessable in 158 cases. The number of tumor cells that scored positive for aurora B ranged from 0% to 80% (median value 20%). In our series, 123 samples (78%) were classified positive for aurora B. The staining was exquisitely limited to the tumor cells and was absent in the surrounding normal lung tissue (Fig. 2A ). Lymphocytes and fibrotic areas within the tumor were usually also negative. The localization of aurora B was exclusively nuclear (Fig. 2A–D), as already reported for other tumor types. Mitotic cells stained strongly.
Aurora B, survivin, and p16 overexpression in NSCLC. A, normal lung columnar epithelium completely negative for aurora B (arrowhead), adjacent to a tumor area positive for aurora B expression. B, enlargement of the area marked by the square in A. C to F, examples of NSCLC sample expressing high percentage of both aurora B–positive (C) and survivin-positive (E) cells. Aurora B and survivin stained the nucleus of tumor cells at similar levels. Enlargements of the areas selected by the squares are given for both aurora B (D) and survivin (F) immunohistochemistry. Mitotic figures were positive (D, arrowhead). G, sample of NSCLC positive for p16 expression in all the cells of the tumor area, both in the nucleus and in the cytoplasm. Columnar normal epithelium adjacent to the tumor is negative (arrowhead). Bar, 100 μm.
Aurora B in Relation to Survivin and p16 Expression in NSCLC
In vitro and in vivo studies have shown that aurora B colocalizes and functions together with survivin in mitotic cells. To the best of our knowledge, this is the first study to investigate the expression and localization of both aurora B and survivin in the same series of clinical samples. Survivin expression was assessable in 159 samples. Similar to aurora B, a variable percentage of the neoplastic population was survivin positive (range 0–100%, median value 30%), in the majority of cases (145 of 159 samples, 91% of the series), specifically in the tumor area and in cells undergoing mitosis. Survivin was predominantly localized in the nucleus (104 tumors, 65% of the samples); however, in some tumors, survivin was detected both in the nucleus and the cytoplasm of cells (27 tumors, 17% of the samples), or only in the cytoplasm (14 tumors, 9% of the series). Thirteen of 14 samples that were survivin negative were also aurora B negative. As shown in Figs. 2 and 3 , survivin expression in the nucleus paralleled aurora B expression (P < 0.0001). Interestingly, in 9 of 12 cases where survivin could be detected exclusively in the cytoplasm, aurora B staining was negative, and in 2 of 12 other cases where survivin stained only the cytoplasm, aurora B levels were below its median value of expression (Table 1 ).
Relationship between nuclear survivin and aurora B expression. Aurora B and nuclear survivin percentage scorings for each sample of the series of 160 NSCLC was plotted. The graph indicates linear relationship, with Spearman's correlation coefficient (ρ) = 0.641 and P < 0.0001.
Patterns of survivin localization in relation to aurora B staining in the overall series of 160 NSCLC samples
Expression of p16 (Fig. 1G) was assessable on 157 tumor samples. The majority of cases were negative, and 44 samples (28%) were classified as positive for p16 expression. The level of p16 expression was below 100% in only a few positive NSCLC samples. Localization of p16 was usually both nuclear and cytoplasmic within the same cell. Normal lung epithelium was usually negative, except in rare cases of severe inflammation. In our series of NSCLC samples, p16-positive staining was neither correlated with aurora B (P = 0.134) nor survivin (P = 0.513) expression.
Taken together, our data clearly show that aurora B and survivin, but not p16, are expressed at similar levels, and often colocalize in the nucleus of NSCLC cells.
Correlation of Aurora B, Survivin, and p16 Expression with Clinicopathologic Variables
The immunohistochemical analysis of aurora B, survivin, and p16 expression and its relation with patient clinicopathologic characteristics was initially done only on the 138 operated early-stage patients of this series. Because the results of this preliminary investigation suggested a correlation of aurora B overexpression with tumor progression, we extended the analysis to other 22 patients with advanced-stage disease in an attempt to better characterize aurora B expression as a potential biomarker of aggressive tumor behavior and prognostic factor in NSCLC. To overcome potential confounding effects due to the different patient treatment, here we present separately the data on correlation of protein expression and clinical features for the group of early-stage patients and the whole series.
In Table 2 , we show the correlation of all the biomarkers studied by immunohistochemistry with clinicopathologic patient characteristics. Aurora B expression significantly correlated with older patient age (P = 0.011) and with squamous cell carcinoma histology (P < 0.0001). Moreover, a tendency of aurora B positivity to increase beyond stage IIA was noted (P = 0.034). As shown in Table 3 , by dichotomizing the patients according to aurora B levels of expression (66 patients expressing low and 72 patients expressing high aurora B), we could observe the same significant correlation, in addition to sex (P = 0.013), lymph node invasion (P = 0.037), and tumor differentiation grade (P = 0.007). The analysis extended to all the 160 samples of the series confirmed the significant correlations of high aurora B expression with older age, squamous cell carcinoma histology, and undifferentiated morphology (data not shown).
Correlation of aurora B, survivin, and p16 expression with clinicopathologic characteristics of 138 operated NSCLC patients
High and low levels of expression of aurora B and clinicopathologic characteristics of 138 operated NSCLC patients
Survivin expression was not associated with any of the clinicopathologic variables studied (Table 2). For the analysis of survivin in relation to patient characteristics, the expression in the nuclei was also taken into account by using the same criteria that we previously reported (33). The patients were divided into two groups based on the presence or absence of nuclear survivin staining: the first group (nuclear survivin-positive) included 119 patients with tumors expressing survivin exclusively in the nucleus of cells or with a mixed population of cells with nuclear and cytoplasmic survivin, and the second group (nuclear survivin-negative) included 19 cases with no survivin expression or only cytoplasmic survivin expression. Like aurora B expression, nuclear survivin positive was correlated significantly with patient age (P = 0.038) and squamous cell carcinoma histology (P < 0.0001). These correlations were also observed in the analysis of the complete series of 160 NSCLC. In addition, the expression of nuclear survivin was significantly higher in early tumor stages (P = 0.007).
The analysis of the correlation of p16 expression with clinicopathologic characteristics is reported in Table 2. In contrast to aurora B and nuclear survivin staining, p16 positivity was observed more frequently in younger patients (P = 0.012) and in adenocarcinoma-type tumors (P = 0.011), also when the overall series of 160 NSCLCs was analyzed (data not shown).
Prognostic Value of Aurora B, Survivin, and p16 Expression in NSCLC
In Table 4 , the results of univariate analysis for survival are presented for the patients with resected tumors. Interestingly, when considering only the 38 operated NSCLC of adenocarcinoma histology, aurora B expression correlated with shorter overall survival, by both univariate (P = 0.031; Fig. 4 ) and multivariate analysis (P = 0.030; Table 5 ), and these data were confirmed for the tumors of adenocarcinoma histology of the overall series (P = 0.020 at univariate analysis; P = 0.012 at multivariate analysis).
Univariate analysis of survival in the series of 138 operated NSCLC patients
Kaplan-Meier curve of overall survival for adenocarcinoma patients with resected NSCLC, depending on the presence or absence of aurora B. Aurora B–positive cases (solid line) predicted shorter overall survival compared with aurora-negative cases (dotted line). Median survival time was 16 months for aurora-positive cases and 26 months for the aurora-negative (log-rank, P = 0.031).
Multivariate analysis for overall survival in 38 NSCLC operated patients with adenocarcinoma histology
p16 or survivin expression and localization were not correlated with survival in any subgroup analyzed. As aurora B correlated with survivin in our study, we also analyzed the influence of aurora B expression (positive and negative or high and low levels) in survivin-positive and survivin-negative cases. The same rankings for aurora B and survivin expression were also evaluated in p16-positive and p16-negative cases. However, none of the combinations of markers showed any correlation with patient survival in neither the resected nor the overall series (data not shown).
Discussion
This is the first study that investigates aurora B protein expression in NSCLC. Using immunohistochemical analysis of a large series of tumors, we show frequent expression of aurora B protein in NSCLC samples, whereas normal lung tissue is negative. This finding suggests a role for aurora B as tumor marker in NSCLC, and confirms at the protein level the recently reported up-regulation of aurora B mRNA in lung tumors (30).
In NSCLC cells, aurora B was localized in the nucleus, in line with previous reports in other tumor types (25, 27, 28). Interestingly, aurora B positivity was often correlated with expression of survivin, consistent with the results of in vitro studies showing that cells lacking survivin upon targeting by antisense oligonucleotide treatment have lower aurora B kinase activity (36). p16 has been reported to regulate the G1 cell cycle restriction point, but has never been investigated together with aurora B and survivin that control the later mitotic spindle checkpoint. The coexpression and nuclear colocalization of aurora B and survivin suggest that their role in NSCLC is closely related. Conversely, the lack of significant association of aurora B and survivin with p16 suggests that these pathways regulating the checkpoints at different steps of the cell cycle are likely independent.
We show that aurora B expression increases with decrease in tumor differentiation, consistent with what has been reported for the levels of aurora B mRNA quantified by real-time reverse transcription-PCR in colorectal cancer (24) and seminoma (28). These data, together with the significant correlation of aurora B levels with lymph node invasion, suggest a role for aurora B in tumor progression of NSCLC. The mechanisms by which aurora B has been proposed to promote tumor progression are loss of mitotic control and abnormal proliferation (7). In this regard, we found a positive correlation between aurora B levels and the tumor proliferation rate assessed by the immunostaining for the early proliferation marker Ki67 (data not shown), similar to the data published for cohorts of astrocytoma (29), endometrial carcinoma (26), and thyroid carcinoma (27).
In NSCLC and adenocarcinoma of the lung, survivin expression has been related to poor patient survival and tumor invasion (37, 38). In our study, survivin overexpression was not correlated with any patient characteristics of adverse disease. On the other hand, it has been previously reported that the expression of survivin in the nucleus correlates with favorable patient clinicopathologic variables (39, 40). In particular, we have recently shown that, in a series of 53 NSCLC samples, nuclear survivin predicts longer patient survival (33). In the present series, however, we did not observe an association between nuclear survivin localization and survival, although expression of survivin in the nucleus was detected more frequently in tumors of stages I and II compared with more advanced tumors.
Overexpression of aurora B has been related to poor prognosis in recent reports of endometrial carcinoma and mesothelioma (26, 41). In our study, among the biomarkers tested, only aurora B could predict prognosis, limited to the adenocarcinoma patients. This histologic subtype is on the increase and represents nowadays the most frequent subtype in North America and Northern Europe (42). Although it needs further validation, our finding might be important for the clinical management of these patients.
Targeted therapies alone and in combination with cytotoxic therapy have recently been successfully introduced in the treatment of hematologic malignancies and some solid tumors. Among the new targeted agents, aurora kinase inhibitors seem to have a very interesting preclinical activity (43), and are presently being tested in phase I studies in cancer patients (44). In line with previous observations in A549 NSCLC cells and other tumor cell lines (34), we found that the aurora kinase inhibitor ZM 447439, which seems to target primarily aurora B, profoundly affects cell division in the NSCLC cell lines H460 and H322. These in vitro results, together with the overexpression of aurora B shown in our study, provide support for testing of aurora kinase inhibitors in the treatment of NSCLC.
In conclusion, we have reported aurora B overexpression in a large series of NSCLC, and its correlation with adverse tumor features in the overall series, and with shorter survival in adenocarcinomas. These results suggest the involvement of aurora B in NSCLC malignancy and progression. Studies aimed at further understanding the role of aurora B in tumor progression in NSCLC are warranted, and clinical studies of aurora B inhibitors should be pursued in NSCLC and adenocarcinomas of the lung in particular.
Footnotes
Grant support: European Society for Medical Oncology Foundation (Lugano, Switzerland; B. Vischioni) and the Walter Bruckerhoff Stiftung (Zurich, Switzerland; J.A. Rodriguez).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Accepted September 11, 2006.
- Received May 22, 2006.
- Revision received August 8, 2006.
- American Association for Cancer Research