Sil overexpression in lung cancer characterizes tumors with ... - Nature

15 downloads 38 Views 272KB Size Report
Apr 26, 2004 - University, Israel; 8Geneic Branch, National Cancer Institute, NIH, ... Center, University of Pittsburgh Medical Center, Pittsburgh, PA, USA.
Oncogene (2004) 23, 5371–5377

& 2004 Nature Publishing Group All rights reserved 0950-9232/04 $30.00 www.nature.com/onc

Sil overexpression in lung cancer characterizes tumors with increased mitotic activity Ayelet Erez1,7, Marina Perelman2,6, Stephen M Hewitt3, Gadi Cojacaru4,5,6, Iris Goldberg2, Iris Shahar4, Pnina Yaron6,7, Inna Muler1, Stefano Campaner8, Ninette Amariglio4, Gideon Rechavi1,4,7, Ilan R Kirsch8, Meir Krupsky6,7, Naftali Kaminski*,4,6,7,9 and Shai Izraeli*,1,7 1

Department of Pediatric Hemato-Oncology, The Chaim Sheba Cancer Research Center, Sheba Medical Center, Tel-Hashomer, Israel; 2Department of Pathology, The Chaim Sheba Cancer Research Center, Sheba Medical Center, Tel-Hashomer, Israel; 3 Advanced Technologies Center, NIH, Bethesda, MD, USA; 4Department of Functional Genomics, The Chaim Sheba Cancer Research Center, Sheba Medical Center, Tel-Hashomer; 5Compugen LTD, Tel-Aviv, Israel; 6Department of Pulmonary Medicine, The Chaim Sheba Cancer Research Center, Sheba Medical Center, Tel-Hashomer, Israel; 7Sackler Faculty of Medicine, Tel-Aviv University, Israel; 8Geneic Branch, National Cancer Institute, NIH, Bethesda, MD, USA; 9The Dorothy P and Richard P Simmons Center, University of Pittsburgh Medical Center, Pittsburgh, PA, USA

Sil (SCL interrupting locus) was cloned from the most common chromosomal rearrangement in T-cell acute lymphoblastic leukemia. It is an immediate early gene whose expression is associated with cell proliferation. Sil protein levels are tightly regulated during the cell cycle, reaching peak levels in mitosis and disappearing on transition to G1. A recent study found Sil to be one of 17 genes whose overexpression in primary adenocarcinomas predicts metastatic spread. We hypothesized that Sil might have a role in carcinogenesis. To address this question, we utilized several approaches. Using a multitumor tissue array, we found that Sil protein expression was increased mostly in lung cancer, but also at lower levels, in a subset of other tumors. Microarray gene expression analysis and immunohistochemistry of lung cancer samples verified these observations. Sil gene expression in lung cancer correlated with the expression of several kinetochore check-point genes and with the histopathologic mitotic index. These observations suggest that overexpression of the Sil gene characterizes tumors with increased mitotic activity. Oncogene (2004) 23, 5371–5377. doi:10.1038/sj.onc.1207685 Published online 26 April 2004 Keywords: Sil; mitosis; lung cancer; carcinogenesis

Introduction The Sil gene (SCL interrupting locus) was cloned from the most common chromosomal rearrangement in T-cell acute lymphoblastic leukemia. In this rearrangement, Sil promoter assumes control of a downstream gene, SCL, *Correspondence: S Izraeli; E-mail: [email protected]; N Kaminski, Dorothy P & Richard P Simmons Center for Interstitial Lung Disease, Pulmonary, Allergy and Critical Care, University of Pittsburgh Medical Center NW 628 MUH, 3459 5th Av., Pittsburgh, PA 15261, USA; E-mail: [email protected] Received 2 October 2003; revised 15 January 2004; accepted 2 March 2004; published online 26 April 2004

also known as Tal1. The resulting aberrant expression of SCL leads to the development of leukemia (Aplan et al., 1990, 1991). The human Sil gene encodes a large cytosolic protein with unknown function (Collazo-Garcia et al., 1995; Golling et al., 2002). Sil is induced in an immediate early fashion and is tightly regulated during the cell cycle. Its mRNA expression is higher in rapidly proliferating cells and tissues, and it decreases rapidly during terminal differentiation (Izraeli et al., 1997). The Sil protein accumulates and reaches peak levels in mitosis and then degrades upon entrance to the next cell cycle (Izraeli et al., 1997). Sil importance to cell growth and survival is supported by the phenotype of mouse and zebrafish embryos lacking a functional Sil protein (Izraeli et al., 1999; Golling et al., 2002). In both species, the loss of Sil causes embryonic lethality and is associated with marked apoptosis of the developing nervous system. Analysis of Sil knockout embryos and further genetic experiments (Izraeli et al., 2001) suggested that Sil is genetically required for the Sonic Hedgehog (Shh) pathway, but that is also required for other growth pathways since the Sil/ mice die earlier and suffer from more extensive growth defects than the Shh/ mice. The critical requirement of Sil for cell growth, proliferation and survival during embryonic development, and its regulation during the cell cycle prompted us to hypothesize that Sil might have a role in tumorogenesis.

Results Sil expression in multiple tumors Immunoblot of Sil in various cell lines derived from several solid tumors and leukemias revealed variable protein expression (Figure 1). To extend this observation, we stained the TARP multitumor tissue arrays composed from approximately 500 samples from a

Sil gene in lung cancer A Erez et al

5372

variety of tumors. Abundant expression of Sil was detected in all tumors, including melanoma, lung, breast, colon, prostate, lymphoma and ovary. As in the protein blot, Sil was not detected in glioblastoma multiforme (Figure 2a, panel c). Consistent with previous observations (Izraeli et al., 1997), Sil protein was detected in normal bone marrow and testes (data not shown). A closer look at the tissue array confirmed the variability of expression shown in the blot between different types of solid tumors, and also showed variable expression between different specimens of the same type of tumor. The fractions of tumors expressing Sil protein range from 11% in melanoma to 28% in lung cancer (Figure 2b). Thus, Sil protein is highly expressed in a variety of tumors. It is overexpressed in only a subset of each tumor type with the highest expression being in lung cancer.

compared with 90.1 AU in normal and carcinoid tissues (Figure 3b). Eight of 26 (30.7%) of the samples in Sheba and 69 of 191 (36.1%) NSCLC in the MIT database expressed Sil mRNA levels that were higher than twofold of the average of control lungs. These results closely fit with the fraction of lung cancer specimens on the tissue array

Sil mRNA and protein levels in lung cancer To evaluate Sil expression in lung cancer, we looked at Sil expression at the gene-microarray results of 26 nonsmall-cell lung cancer (NSCLC) tumors and 10 normal lung tissues. The Sil average expression signal was significantly higher in the tumors, 438.41723.5 arbitrary units fluorescence (AU), compared with normal lungs, 101.472.6, P ¼ 0.00043 (Figure 3a). These results were confirmed by real-time RT–PCR on six tumor and two normal independent samples (data not shown). To extend these observations to a larger dataset, we downloaded the raw microarray data (Cell Files) from the study of Bhattacharjee et al. (2001) (available at http:// www-genome.wi.mit.edu/cgi-bin/cancer/datasets.cgi). This dataset contained gene expression profiles for 217 lung tumor samples (170 lung adenocarcinomas, 21 squamous cell lung carcinomas, 20 pulmonary carcinoids and six small-cell lung carcinomas) and 17 normal lung samples. Sil mRNA expression level was significantly higher in noncarcinoid lung cancers (including adeno, squamous and small-cell lung cancer) compared to 17 normal lungs (P-value o0.00005 for Info and t-test). The average level of Sil in tumors was 157.97 AU

Figure 1 Western blot of Sil in different cell lines: (a) A polyclonal ‘243’ antibody directed against the carboxy terminal of Sil was reacted against protein lysates from: a. 293T transfected with Sil; b. 293T transfected with empty vector; c. Dami – acute megakaryocytic leukemia; d. U87 – glioblastoma; e. LS – colon cancer; f. HeLa – carcinoma of cervix; g. K562 – erythroid blast crisis of chronic myeloid leukemia. (b) The same membrane was stripped and blotted again with anti-Cbl antibody (Santa- Cruz) as a loading control Oncogene

Figure 2 Sil protein expression in tumors; (A) tissue array of various tumors stained with antibody for Sil reveal variable levels of Sil protein in different tumors. The magnification is  100 for c,e,g and  400 for a,b,d,f,h. (a, b) Cell lines fixed in parafomaldehyde and embedded in paraffin as controls for the antibody. (a) 293T transfected with Sil, transfected 293T cells (arrows) display intense cytosolic staining, nontransfected cells shows low background staining; (b) H1299 lung adenocarcinoma cells express high level of endogenous SIL; (c) glioblastoma multiforme; (d) melanoma; (e) normal colon; (f) colon carcinoma; (g) lung carcinoma; (h) metastatic prostate adenocarcinoma in a lymph node. Note that the staining is specific to the tumor cells and that the infiltrating cells are negative. (b) Sil is differentially expressed in different tumor types: The fraction of tumors demonstrating positive staining for Sil on a TARP tissue array

Sil gene in lung cancer A Erez et al

5373

that stained strongly for Sil (13 of 46, 28.2%, Figure 2b). To further examine the correlation between Sil mRNA and protein expression, we stained sections from five representative tumors from our center for which microarray data was available. The immunostaining was performed in a blind manner, the pathologist (MP) was not aware of the microarray results. The pathologist distinguished correctly between the two samples highly expressing Sil mRNA to the three with lower expression (Table 1 and Figure 4). Thus the analysis of Sil mRNA and protein expression by DNA microarray demonstrates that Sil is overexpressed in about a third of NSCLC. Characterization of lung tumors overexpressing Sil

Since Sil expression is associated with cell proliferation (Izraeli et al., 1997), we examined the correlation of its expression in lung cancer with the expression of Ki67, a general proliferation marker. Only a modest correlation was found between these two genes, (Pearson ¼ 0.50). Immunohistochemistry of the lung tumors for both genes sharpened the difference and showed that Ki-67 protein was expressed also in the absence of Sil expression (Figure 4 panel Ac vs Ad). These results suggest that Sil expression in lung cancer is not simply a marker of cell proliferation. Similarly, despite our previous published evidence that Sil is required for the cellular response to the cytokine Shh (Izraeli et al., 1999; Izraeli et al., 2001), there was no correlation with the expression of Shh responsive genes – Ptch and Gli1 (Pearson ¼ 0.073, 0.078, respectively).

There was no significant correlation between the level of Sil and the histological type of the tumor (P ¼ 0.0711 and 0.1283 for the Sheba and the MIT data, respectively). There was also no significant correlation between the degree of differentiation and the level of Sil (P ¼ 0.639, the Sheba dataset).

Figure 3 Microarray gene expression analysis reveals that SIL mRNA levels are increased in lung tumors. (a) Sheba database – Sil expression levels are increased in lung adenocarcinoma (AD, n ¼ 8, P-value ¼ 0.011), and lung squamous cell carcinoma (SQC n ¼ 19, P ¼ 0.039) compared to normal lung (n ¼ 10). (b) Verification of increased Sil expression levels using the previously published MIT lung cancer dataset. Sil expression levels were increased in lung adenocarcinoma (n ¼ 170, P ¼ 0.00002) and lung squamous cell carcinoma (n ¼ 21, P ¼ 0.00003), in small-cell carcinoma (n ¼ 6, PValue ¼ 0.0167) but not in lung carcinoid (n ¼ 20) compared to normal lung (n ¼ 17). Expression levels are output expression value defined as signal by MAS 5 and is expressed in arbitrary fluorescence units. * ¼ Po0.05, ** ¼ Po0.00005

Figure 4 Sil expression in lung cancer: (A) Immunohistochemistry of SIL in lung cancer; a. lung squamous cell carcinoma; b. lung adenocarcinoma with high expression Sil; c. lung adenocarcinoma with low expression of Sil; d. adjacent section of the same sample shown on c demonstrates high level of KI-67 expression. Magnification  100. (B) Variability in Sil mRNA expression level in different samples of lung carcinoma. The scale is in Affymetrix units

Table 1 Correlation between the mRNA and the protein level of Sil Case 1 2 3 4 5

mRNA level (Affymetrix units) 358.1 808.2 158.2 76.1 146.6

Tumor type

Differentiation

Adenocarcinoma Squamous cell Squamous cell Squamous cell Large cell

Well diff. Moderately diff. Poorly diff. Well diff. Undifferentiated

% of stained cells

Staining gradea

Staining intensity

71% 65% 21% 8% 24%

3+ 3+ 1+ 7 1+

Strong Strong Mild to moderate Mild Mild to moderate

A sample of lung tumors stained with Sil 243 Ab. and graded blindly by the pathologist. a7(0–5%); 1+ (6–25%); 2+ (26–50%); 3+ (51–75%) Oncogene

Sil gene in lung cancer A Erez et al

5374

Computational analysis of the lung cancer data (Sheba and MIT) revealed that the five most correlated genes common to the MIT data and Sheba were cell cycle genes, four of which function during mitosis in regulation of the kinetochore (spindle) mitotic checkpoint (Table 2). Although Sil expression was correlated with the general proliferation marker Ki-67, its statistical power was at least three orders of magnitude less than the significance of the correlation with mitoticspecific genes. To verify this analysis by a different approach, we used the ORFisher program (Compugen Inc.). ORFisher identifies genes (ORFs) with similar expression to the center of the kernel, which are not associated with any of the known pathways but have an expression pattern similar to that of the genes belonging to a known pathway (based on the assumption that coregulated genes are likely to share functional characteristics providing hints on gene function). The analysis revealed four kernels that could distinguish between lung cancer samples and normal lungs. One of the four kernels was populated with genes functioning in the kinetochore mitotic checkpoint based on their Gene Ontology (www.geneontology.org) annotation. Sil expression was shown to be mostly correlated with this kernel (r ¼ 0.85), a correlation almost similar to the kinetochore-associated proteins (Table 3). Thus, two different computational analyses of the gene expression profile of lung cancer found Sil to be coexpressed with mitotic checkpoint genes. Owing to the association between the expression of Sil and mitotic checkpoint genes and the regulation of Sil expression during mitosis, we counted the mitoses and calculated the histopathologic mitotic index in sections from all the 26 lung tumors analysed in Sheba. The tumor samples were divided into two groups based on

the level of Sil expression. The average number of mitoses in high-power field (hpf) in high Sil expressors was 17.55, compared to 7.3 in low expressors (Table 4) (P ¼ 0.006, two-tail t-test). Recently, Sil was reported as one of the 17 genes that were predictive of solid tumor metastases (Ramaswamy et al., 2003). Of the 26 Sheba patients, 10 developed metastases; 5/8 patients with high levels of Sil (62.5%), compared to 5/18 patients (27.7%), with low levels of Sil, P ¼ 0.093 (Table 4). This pattern supports the study by Golub et al. that Sil expression may be associated with metastatic potential.

Discussion Sil is a cell-cycle-regulated gene, associated with cell proliferation (Izraeli et al., 1997). Since tumors are highly proliferative tissues, we expected they might express high levels of Sil. To our surprise, we found Sil protein to be overexpressed in no more than 30% of epithelial cancers. There was a wide variability in Sil expression between different kinds of tumors and within Table 3

Sil correlates with the kinetochore gene kernel center with the same high correlation as this kernel known genes

Locus link

Gene name

LL:6491 LL:701 UG:Hs.334562 LL:4085 LL:7272 LL:890

Sil Bub1 Cdc2 Mad2 Ttk Cyclin A

Similarity to kernel center 0.85 0.87 0.89 0.88 0.88 0.64

Cyclin A, although a cell cycle gene, does not belong to this kernel, and indeed gets a lower score. The values were calculated using Compugen statistical analysis algorithm.

Table 2 Sil mRNA expression correlates with the kinetochore checkpoint genes Gene I.D.

Gene name

Gene function

Pearson corr. Sheba

Pearson corr. MIT

T-test P-value for NSCLC tumors with low vs high levels of Sil Sheba N ¼ 18 low vs 8 high

L47276

Topoisomerase (DNA) II alpha

M86699

Ttk protein kinase AF067656 Zwint, Zw10 interacting protein AF091433 Cyclin E2 A5000186 Mad2 X65550

Ki-67

Controls and alters the topologic states of DNA during transcription, essential for proper formation of centromere/kinetochore structure Dual specificity serine/threonine and tyrosine kinase, mitotic spindle checkpoint Phosphorylated during mitosis G1/S-specific cyclin, CDK kinase regulatory subunit that interacts with CDK2 and CDK3 Linked to mitosis through cyclin B Monitors the completeness of the spindlekinetochore attachment S phase marker, related to proliferation

MIT N ¼ 121 low vs 70 high

0.705518

0.633217

4.74E-06

4.46E-08

0.909067

0.5958

1.82E-06

2.7E-09

0.773572

0.593491

0.005041

1.13E-09

0.739969

0.586583

0.92586

0.000405

0.755215

0.466842

2.64E-05

5.74E-07

0.495223

0.479905

0.013674

0.00034087

The columns show the gene accession number, its known function and the Pearson correlation results computed for the Sheba and MIT separately. The P-value in the last column represents the statistical significance of the difference in the expression of the specific gene between tumors with high and low expression of Sil using two-tailed t-test on the two data bases. The cutoff between ‘high’ and ‘low’ Sil expressors was the average level in all samples including the normal ones, plus two standard errors. This analysis emphasizes that Sil correlation with the mitosis checkpoint genes is more significant than its correlation with Ki-67 or cyclin E Oncogene

Sil gene in lung cancer A Erez et al

5375 Table 4 Sil expression level correlates with mitotic index and with a tendency for metastases Sil average level in tumors/ category 438.41723.5 101.472.6

No. of tumors

Mitotic Index*

No. of patients with metastases**

8 18

17.5571.43 7.3370.3

5 (62.5) 5 (27.7)

The average level of Sil was calculated with the standard mean error from the Sheba mRNA on lung tumor data. The histopathologic mitotic index was examined on the same samples from which the mRNA was taken. The mitotic index was calculated by dividing the number of mitoses in 50 high-power fields by 5. Two assessors evaluated the tumors on two separate occasions. No significant difference was found between their evaluations. The information concerning metastases was taken from the patients’ files. *P ¼ 0.006; **P ¼ 0.093; 7 ¼ standard error of mean

different specimens of the same kind of tumor. Although its highest expression level was in lung tumors, there was a marked variability both at the mRNA and at the protein levels. Computational analysis revealed that genes with expression patterns similar to Sil were mainly mitosis-associated genes, and specifically mitotic-spindle-regulatory genes. The association of Sil expression and mitosis at the tumor phenotype is further supported by the finding of elevated histopathological mitotic index in tumors with high expression of Sil. Four of the genes, expressed in the most similar fashion to Sil, function in the kinetochore checkpoint during mitosis. The kinetochore checkpoint monitors proper attachment of the spindle to sister chromatids to ensure euploidy of the daughter cells (Musacchio and Hardwick, 2002). The genes correlating with Sil belong to the family of proteins that associates with the kinetochore transiently. They are recruited in a sequential order at discrete times, prior to, or early in mitosis (Zhou et al., 2002). Topoisomerase II alpha might be one of the earlier markers for the kinetochore assembly as it is recruited to centromers at the onset of heterochromatin condensation during the late S–G2 period, and is essential for proper formation of centromere/kinetochore structure (Rattner et al., 1996). The other genes are all part of the kinetochore checkpoint. Ttk (serine/treonine and tyrosine kinase) is a dual specificity kinase, the human orthologue of the yeast Mps1 kinase, displaying maximal activity during M phase. Ttk phosphorylates Mad1, which binds Mad2 to the kinetochore. Activated Mad2 interacts and inhibits the anaphase promoting complex (APC) from prompting the entry to anaphase until there is a proper attachment of all sister chromatids (Musacchio and Hardwick, 2002) (Song-Tao Liu et al., 2003). Mad/ mice develop lung tumors at high rates after long latency (Michel et al., 2001). Zwint mediates the interaction between Zw10 and the microtubule motor at the kinetochore, to ensure equal tension on both sides of the sister chromatids (Starr et al., 2000). Similar to Sil, Zwint orthologues are found only in vertebrates (Starr et al., 2000) and not in lower eucaryotes. All these proteins are distributed diffusely throughout the cell, and localize to the centromers in mitosis.

Although Sil has been shown to be necessary to the Sonic Hedgehog response pathway (Shh), during early mouse development, we could not find any support that Sil exerted its role in lung cancer through the Shh. The expression of Sil was not correlated with any of the known targets of the Shh pathway. This pathway has been recently linked with small-cell lung carcinoma (Watkins et al., 2003), a tumor type that was underrepresented in our dataset (six samples out of 245 tumors in both datasets). The cellular role of Sil is presently unknown. We have previously demonstrated that Sil protein levels peak during mitosis and are degraded on transition to G1 (Izraeli et al., 1997). A similar regulation of Sil transcription during the cell cycle was recently reported by Whitfield et al. (2002). Interestingly, gene expression analysis of their data cluster Sil with Zwint, similarly to the findings reported here. Recently, we have demonstrated that Sil is phosphorylated during mitosis, in particular following activation of the spindle checkpoint, and interacts with the mitotic regulator Pin1 (Campaner, Izraeli and Kirsch, manuscript submitted). These independent observations combined with the analysis of gene expression in lung cancer reported here suggest a role for Sil in mitosis. One intriguing speculation is that Sil being a cytosolic protein may regulate kinetochore checkpoint proteins by serving as their cytosolic anchor. Since, like Zwint, the Sil gene is exclusive to vertebrates, it is possible that Sil might have evolved for a unique regulatory role in the vertebrates’ mitotic checkpoint. Its over expression in a subset of tumors characterized by increased expression of spindle checkpoint genes and by elevated mitotic index suggests that this role may be important in tumorigenesis. What is the functional significance of Sil expression in a subset of epithelial tumors and of its association with mitotic checkpoint genes and the mitotic index in lung cancer? Does its overexpression represents heightened mitotic activity, activation of the mitotic checkpoint, or conversely, malfunctioning checkpoint? Further studies that may include experimental manipulation of Sil levels in experimental tumor models are required. Interestingly, the mitotic checkpoint was reported to be often defective in lung cancer (Takahashi et al., 1999) and the mitotic index, and not Ki-67 expression, was reported recently to provide significant prognostic information in breast cancer (Lynch et al., 2002). A recent study found Sil to be one of the predictive genes for metastases in adenocarcinomas of different tissues (Ramaswamy et al., 2003). Our data support a correlation between Sil level and metastases. These observations suggest that Sil might be a marker for metastatic potential in lung cancer.

Materials and methods Immunohistochemistry Multitumor tissue arrays, containing approximately 300 tissue samples of gliomas, melanomas, lymphomas, ovarian adenocarcinoma, carcinoma of the breast, carcinoma of the lung, colonic adenocarcinoma and prostatic adencocarinoma, Oncogene

Sil gene in lung cancer A Erez et al

5376 as well as normal control tissues were obtained from the TARP Laboratory, NCI. Complete description of the arrays (TARP1) can be obtained at www.cancer.gov/tarp. The lung cancer tissues used for immunohistochemical study were derived from the same specimens analysed by microarrays. The type and grade of the carcinomas were determined according to World Health Organization criteria (Travis et al., 1999). Immunostaining was performed on formalin-fixed, paraffinembedded sections, 4 mm thick, which were thaw-mounted onto Fisher brand super frost slides. After air drying at 371C for 16 h and incubation for 30 min at 601C, slides were deparaffinized and rehydrated in graded alcohols. Antigen was detected with a labeled Avidin–Biotin (LAB) method using a commercial kit (Zymed Laboratories, USA) according to the manufacturer’s instructions; peroxidase was detected with amino ethyl carbazole substrate (AEC) (Zymed Laboratories, USA) and sections were lightly counterstained with hemotoxylin. Affinity-purified polyclonal antibody raised in rabbit against the Sil carboxy terminal were used diluted 1 : 750 for immunohistochemistry of Sil. The specificity of this antibody for various applications including immunostaining of fixed cells has been shown by us before (Izraeli et al., 1997). We used as controls Sil highly expressing cells – H1299 (a cell line derived from adenocarcinoma of the lung) and the 293T cell line transfected with Sil driven by the CMV promoter. One section of tumor and one section of nontumoral lung tissue were stained in each case. The grading of the immunostaining was given according to the percentage of immunoreactive cells (70–5%; 1 þ , 6–25%; 2 þ , 26–50%; 3 þ , 51–75%; 4 þ , 76–100%). Intensity was assessed as mild, moderate and strong. The score was giving in a blind manner, that is, the pathologist was not aware of the microarray results. The Ki-67 antigen was detected using a standard peroxidaseconjugated streptavidin biotin system with antigen retrieval by placing the section into a pressure cooker containing 0.01 M citrate buffer, pH 6.0 and maintaining them under pressure (15 psi) for 2 min. The anti-Ki-67 antibody MIBI (immunotech) was applied at a dilution of 1 : 100. Both positive and negative controls were included with immunohistochemical staining to monitor batch to batch consistency and nonspecific staining. Ki-67 positivity was assessed as the number of nuclei expressing Ki-67 per 1000 malignant cells.

Data analysis The microarrays were scaled to an average intensity of 150 and analysed with Affymetrix Microarray Suit Version 5 and D-chip. For statistical analysis, we used Scoregene Package (available at http://www.cs.huji.ac.il/labs/compbio/ scoregenes/). The raw microarray data (Cell Files) from the paper by Bhattacharjee et al. (2001) was downloaded from the data website of the Whitehead Institute Center for Genome Research (http://www-genome.wi.mit.edu/cgi-bin/ cancer/datasets.cgi). This dataset contained gene expression profiles for 217 lung tumors (170 lung adenocarcinomas, 21 squamous cell lung carcinomas, 20 pulmonary carcinoids and six small-cell lung carcinomas) and 17 normal lung samples. Similarity of gene expression was assayed either by calculating the Pearson correlation coefficients between Sil and the rest of the genes in the same data set using GeneRank in Scoregenes or by ORFisher algorithm, which measures the distance between a gene and a specific group of genes (‘kernel’) that is expressed in a similar way (Compugen LTD, Tel-Aviv). Detailed mathematical description of the ORF fisher algorithm is available upon request. In general, the algorithms includes four important steps: (1) assigning GO (www.geneontology.org) and LIGAND (www.genome.ad.jp/ligand/) annotation to all genes in the data set; (2) calculation of the similarity of gene expression (Pearson correlation) of genes with similar annotation, generating mini-kernels. The center of each mini-kernel is then calculated; (3) search for mini-kernels centers that are ‘close’ enough. Once such are found, the algorithm joins them to a kernel (joining the biological annotation as well) and calculates the theoretical center of the kernel; (4) assigning the biological role of the kernel (based on functions of known genes in the kernel) to unknown genes with similar expression in the center of the kernel. Statistical analysis Statistical analysis of microarrays was performed as described above – TNoM, Info and t-test were calculated using Scoregene package. The relationship between Sil expression level and the mitotic index was assessed using two tails t-test. The Chi-square test was used for the relations between Sil level and the development of metastases. Cell line and cell culture experiments

Histopathologic mitotic index The mitotic count was evaluated over 50 hpf, using an Olympus 2439873 microscope, which has a  40 magnification field diameter of 0.55 mm. The mitotic index was calculated by dividing the number of mitoses in 50 hpf by 5. Two assessors evaluated the tumors on two separate occasions. No significant difference was found between their evaluations.

Microarray experiment Sample preparation: RNA from 26 NSCLC samples, 10 normal lung and three metastases was obtained from the Sheba Medical Center lung cancer tissue bank and used as a template to generate cDNA and labeled cRNA according to Affymetrix protocols (available at http://jcsmr.anu. edu.au/group_pages/ brf/services/Microarray/ExprAnalysisOverview.htm). Labeled probes were hybridized to Human Genome U95AV2 arrays and analysed as previously described by us (Kaminski and Friedman, 2002; Ortiz et al., 2003).The study was approved by the Sheba medical center institutional review board. Oncogene

293T and HeLa cell lines (American Tissue Culture Collection) were grown in DMEM supplemented with 10% heatinactivated fetal bovine serum, penicillin–streptomycin (100 U and 100 mcg/ml, respectively) and 2 mM glutamine (Life Technologies). H1299 cells derived from human lung adenocarcinoma were grown in RPMI supplemented as above. For transfection, 293T cells were plated onto 100 mm2 tissue culture dishes (Falcon) at a density of 1.5  106 cells/dish in 10 ml growth media overnight and transfected the next morning with 10 mcg of plasmid, either with Sil or with empty vector (pcDNA3, Invitrogen), using Ca-Phosphate. At 48 h after transfection, cells were washed twice with cold PBS and harvested for protein analysis (see below). Protein extraction and Western analysis After centrifugation for 5 min at 1250 r.p.m., cells were resuspended in lysis buffer containing 1% Triton-X-100, 20 mM Tris-Hcl, 120 mM NaCl and protease inhibitors (Roche). The extract was pelleted by centrifugation at 14 000 g for 5 min. The lysate protein concentration was

Sil gene in lung cancer A Erez et al

5377 determined by the Bradford method using Pierce protein assay kit with BSA as internal standard. In all, 10 mg of protein from transfected 293T and 100 mg of protein from HeLa and H1299 were loaded. The affinity-purified anti- Sil (243) (Izraeli et al., 1997) was used at 0.4 mcg/ml for 2 h, the secondary horseradish peroxidase-coupled anti-rabbit IgG (Jackson labs–Bar Harbor, Maine, USA) was used at dilution of 1 : 20 000 for 1 h at room temperature. For detection of the signal, Pierce (Rockford, USA), ECL kit was used. Equal protein loading was verified by reacting the same membrane with an anti-Cbl antibody (Santa Cruz).

Acknowledgements We thank Sergey Nemez for the statistical analysis. This study was supported by research grants from the Israel Science Foundation, the Israel Cancer Research foundation and from Tel-Aviv University Recanati foundation to S Izraeli. N Kaminski is the Dorothy P and Richard P Simmons chair of Interstitial Lung Diseases at the University of Pittsburgh Medical School. G Rechavi holds the Gregorio and Dora Shapiro Chair for Hematology Malignancies, Sackler School of Medicine, Tel-Aviv University. This work was preformed in partial fulfillment of the requirements for the PhD degree of Ayelet Erez, Sackler Faculty of Medicine, Tel-Aviv University.

References Aplan PD, Lombardi DP, Ginsberg AM, Cossman J, Bertness VL and Kirsch IR. (1990). Science, 250, 1426–1429. Aplan PD, Lombardi DP and Kirsch IR. (1991). Mol. Cell. Biol., 11, 5462–5469. Bhattacharjee A, Richards WG, Staunton J, Li C, Monti S, Vasa P, Ladd C, Beheshti J, Bueno R, Gillette M, Loda M, Weber G, Mark EJ, Lander ES, Wong W, Johnson BE, Golub TR, Sugarbaker DJ and Meyerson M. (2001). Proc. Natl. Acad. Sci. USA, 98, 13790–13795. Collazo-Garcia N, Scherer P and Aplan PD. (1995). Genomics, 30, 506–513. Golling G, Amsterdam A, Sun Z, Antonelli M, Maldonado E, Chen W, Burgess S, Haldi M, Artzt K, Farrington S, Lin SY, Nissen RM and Hopkins N. (2002). Nat. Genet., 31, 135–140. Izraeli S, Colaizzo-Anas T, Bertness VL, Mani K, Aplan PD and Kirsch IR. (1997). Cell Growth Differ., 8, 1171–1179. Izraeli S, Lowe LA, Bertness VL, Campaner S, Hahn H, Kirsch IR and Kuehn MR. (2001). Genesis, 31, 72–77. Izraeli S, Lowe LA, Bertness VL, Good DJ, Dorward DW, Kirsch IR and Kuehn MR. (1999). Nature, 399, 691–694. Kaminski N and Friedman N. (2002). Am. J. Respir. Cell Mol. Biol., 27, 125–132. Liu ST, Chan GK, Hittle JC, Fujii G, Lees E and Yen TJ. (2003). Mol. Biol. Cell., 14, 1638–1651. Lynch J, Pattekar R, Barnes DM, Hanby AM, Camplejohn RS, Ryder K and Gillett CE. (2002). J. Pathol., 196, 275–279.

Michel LS, Liberal V, Chatterjee A, Kirchwegger R, Pasche B, Gerald W, Dobles M, Sorger PK, Murty VV and Benezra R. (2001). Nature, 409, 355–359. Musacchio A and Hardwick KG. (2002). Nat. Rev. Mol. Cell. Biol., 3, 731–741. Ortiz LA, Gambelli F, McBride C, Gaupp D, Baddoo M, Kaminski N and Phinney DG. (2003). Proc. Natl. Acad. Sci. USA, 100, 8407–8411. Ramaswamy S, Ross KN, Lander ES and Golub TR. (2003). Nat. Genet., 33, 49–54. Rattner JB, Hendzel MJ, Furbee CS, Muller MT and BazettJones DP. (1996). J. Cell. Biol., 134, 1097–1107. Starr DA, Saffery R, Li Z, Simpson AE, Choo KH, Yen TJ and Goldberg ML. (2000). J Cell Sci, 113, 1939–1950. Takahashi T, Haruki N, Nomoto S, Masuda A, Saji S, Osada H and Takahashi T. (1999). Oncogene, 18, 4295–4300. Travis WD, Colby TV, Corrin B, Shimosato Y and Brambilla E. (1999). Histological Typing of Lung and Pleural Tumours, 3rd edn Springer: Berlin. Watkins DN, Berman DM, Burkholder SG, Wang B, Beachy PA and Baylin SB. (2003). Nature, 422, 313–317. Whitfield ML, Sherlock G, Saldanha AJ, Murray JI, Ball CA, Alexander KE, Matese JC, Perou CM, Hurt MM, Brown PO and Botstein D. (2002). Mol. Biol. Cell, 13, 1977–2000. Zhou J, Yao J and Joshi HC. (2002). J. Cell Sci., 115, 3547–3555.

Oncogene