A Macrophage Gene Expression Signature Defines a Field Effect in the Lung Tumor Microenvironment Robert S

Research Article

A Macrophage Gene Expression Signature Defines a Field Effect in the Lung Tumor Microenvironment Robert S. Stearman,1 Lori Dwyer-Nield,2 Michael C. Grady,1 Alvin M. Malkinson,2 and Mark W. Geraci1

Departments of 1Medicine/Pulmonary Sciences and Critical Care Medicine, and 2Pharmaceutical Sciences, University of Colorado Health Sciences Center, Denver, Colorado

Abstract with (cases) or without (controls) lung cancer, the greatest risk factors were current or exsmoker status (adjusted odds ratio, 13.5), One area of intensive investigation is to understand complex and an elevated odds ratio was present even 25 years after smoking cellular and signaling interactions in the tumor microenvi- cessation (4). Recently released data (1) covering 1975 to 2002, ronment. Using a novel, although straightforward, microarray showed a relatively constant 5-year survival rate of f15% after approach, we defined a gene expression signature from the lung cancer diagnosis, independent of race or gender, with a higher lung tumor microenvironment in the murine A/J-urethane 5-year survival rate (49%) if early stage disease is discovered before model of human lung adenocarcinoma. The tumor microen- detectable lymph node or metastasis involvement. In contrast, vironment is reflected by the composition of the cell types colorectal cancer had an overall 5-year survival rate of 64% over the present and alterations in mRNA levels, resulting in a ‘‘Field same study period. Although lung cancer is the most common Effect’’ around the tumor. The genes composing the Field cancer affecting both sexes, it had the third lowest 5-year survival Effect expression signature include proteases and their rate, exceeding only liver and pancreatic cancers, which account inhibitors, inflammation markers, and immune signaling for f50,000 cases per year combined. molecules. By several criteria, the Field Effect expression Survival studies indicate that early detection of lung cancer as signature can be attributed to the macrophage lineage, stage I disease is the most significant factor in providing the suggesting a qualitative change in the expression pattern of highest 5-year survival rate (1, 5, 6). Unfortunately, there are no tumor-associated macrophages (TAM) observed in lung effective low-cost options available for general annual screening of tumors. The protein expression levels for a number of Field the tobacco smoking population, estimated at 20% to 25% of the Effect genes were verified by Western blot analysis of lung adults in the United States (1). Two recent studies examining the homogenates, and for their expression in macrophages and utility of spiral computed tomography as an annual screening parenchymal cells outside of the tumors by immunohisto- method had conflicting results, although they used different chemistry. In addition, the Field Effect expression signature clinical outcome measures (7, 8). Both studies did agree that spiral was used to classify bronchoalveolar lavage (BAL) cells from computed tomography does increase early detection and treat- tumor-bearing or age-matched control mice. Using a variety of ment of lung cancer in the easily identifiable at-risk population statistical measures, the Field Effect expression signature (a smoking history and/or occupational exposures; age >40years). correctly classified the BAL cells >94% of the time. Finally, the The significance of improved survival rate versus decreased protein levels for several Field Effect genes were higher in cell- mortality must await the results from two randomized controlled free BAL fluid, indicating they may be secreted by the TAMs. trials currently under way by a Dutch-Belgian collaboration and the This work suggests that TAMs generate a unique gene National Cancer Institute. expression signature within the tumor microenvironment, One promise of gene expression microarray studies in lung and this signature could potentially be used for identifying cancer has been to develop novel approaches to diagnostics for lung cancer from BAL cells and/or fluid. [Cancer Res disease classification and patient stratification into high probability 2008;68(1):34–43] treatment outcome groups (9–12). This overall strategy has found considerable success in the treatment of breast cancer, moving Introduction from research tool to clinical diagnostics, to identify best treatment Lung cancer continues to be the number one cause of cancer outcome and recurrence using a 70-gene panel (MammaPrint; mortality in the United States, with >210,000 new cases and 160,000 ref. 13). Recent reports suggest a similar approach is feasible using deaths estimated for 2007 (1). Tobacco smoke is the most lung cancer tissue to identify high- and low-risk groups for recur- widespread risk factor accounting for lung cancer in 85% to 90% rence and survival (14–16), although these approaches do not of cases (2, 3). Environmental risks include occupational exposures address early detection methodologies. to asbestos, uranium, and other mining activities, second-hand As a first step toward identifying biomarkers of early lung cancer, tobacco smoke, and household radon. In a study of Iowan women we used the well-characterized A/J mouse-urethane model of human adenocarcinoma (12, 17). We reasoned that histologically appearing, normal lung tissue from urethane-treated mice would have an altered gene expression profile because of proximity to the Note: Supplementary data for this article are available at Cancer Research Online tumor microenvironment, when compared with normal lung tissue (http://cancerres.aacrjournals.org/). Requests for reprints: Robert S. Stearman, Box C272, Room BB 3B10, 4200 East from age-matched untreated animals. The altered gene expression Ninth Avenue, Denver, CO 80262. Phone: 303-315-2317; Fax: 303-315-5632; E-mail: could be derived from lung parenchymal cells and/or immune cells [email protected]. I2008 American Association for Cancer Research. infiltrating and responding to the tumor. In lung cancer, the tumor doi:10.1158/0008-5472.CAN-07-0988 microenvironment has been extensively studied, especially with

Cancer Res 2008; 68: (1). January 1, 2008 34 www.aacrjournals.org

Downloaded from cancerres.aacrjournals.org on September 25, 2021. © 2008 American Association for Cancer Research. Macrophage Expression in Lung Tumor Microenvironment regard to a wide variety of chemokine (C-C motif) signaling chitinase 3-like 1 (Chi3l1), cathepsin D (Ctsd), cathepsin K (Ctsk), cathepsin pathways after activation of nuclear factor-nB (18–20). Chemotac- Z (Ctsz), coagulation factor VII (F7), lipoprotein lipase (Lpl), leucine-rich tic factors produced by tumors recruit blood-derived monocytes, a-2-glycoprotein 1 (Lrg1), lymphocyte antigen 75 (Ly75), nephroblastoma and within the tumor microenvironment, they mature to tumor- overexpressed gene (Nov), prostaglandin-endoperoxide synthase 1 [Ptgs1; cyclooxygenase (COX)1], signal-regulatory protein a (previously designated associated macrophages (TAM; refs. 21–23). Ptpns1; Sirpa), secreted phosphoprotein 1 (osteopontin; Spp1). TAMs play a central role in many tumor-stromal interactions Murine model of human lung cancer. As previously reported (12), including angiogenesis, extracellular matrix remodeling, and male A/J mice (6–8 weeks old; The Jackson Laboratory) were given a single invasion/metastasis (21, 24, 25). The degree of macrophage 1-mg/g i.p. injection of urethane or saline vehicle (age-matched untreated infiltration into human non–small cell lung cancers is correlated controls). At two time points (early, 24–26 weeks; late, 42 weeks) after with the microvessel counts and inversely correlated with survival urethane or saline treatment, mice were sacrificed by lethal pentobarbital (21). In addition, alveolar-derived macrophages isolated by bron- injection, and their lungs excised for examination and molecular choalveolar lavage (BAL) from lung cancer patients have altered characterization. Tumors were dissected from neighboring tissue using a in vitro characteristics (26). In transgenic mouse models, elevating dissection microscope, and the adjacent tissue was placed in RNAlater the number of pulmonary macrophages overexpressing FGF10 (Ambion) for microarray analysis or in 20mmol/L HEPES (pH 7.4), 2 mmol/L EDTA, 2 mmol/L EGTA, 1 mmol/L DTT, 10% glycerol, 5 Ag/mL aprotinin, in type 2 and Clara cells induced spontaneous development of 10 Amol/L leupeptin, and 1 mmol/L 4-(2-aminoethly) benzenesulfony- lung tumors in the absence of exogenous carcinogens (27). Finally, fluoride for protein analysis. Lungs from age-matched untreated controls chromosomal sites containing genes that regulate macrophage underwent similar processing. behavior map to lung cancer susceptibility sites (28). TAMs receive RNA isolation and microarray analysis. RNA was isolated from lung signals from diverse cell types within the tumor microenvironment tissues or BAL derived cells (below) using Trizol Reagent (Invitrogen) and and release signals that affect a variety of different cell types, RNeasy kits (Qiagen), followed by quality control measures using UV spectra including epithelial and endothelial cells, fibroblasts, and lympho- characteristics (NanoDrop) and Bioanalyzer (Agilent) for size and integrity cytes (24, 29). TAMs alter epithelial cell phenotype by stimulating of the total RNA (12). Total RNA (2–5 Ag from lung tissues or 0.25–1 Ag from proliferation, inhibiting apoptosis, and/or changing their morphol- BAL cells) was converted to fragmented biotin-labeled cRNA by the ogy/differentiation status so that initiated epithelial cells adapt a Affymetrix kit reagents according to the manufacturer (Affymetrix). Labeled cRNAs were hybridized to MG-U74Av2 (lung tissue samples) or MOE430 more motile and less sessile configuration (30). Macrophages are 2.0(BAL samples) Affymetrix microarrays were washed, developed, and phenotypically heterogeneous, and extremes of this continuum are scanned using Fluidics and Scanner Workstations. The lung tissue called M1 (classic activation) and M2 (alternative activation; refs. microarray dataset consisted of four possible different samples: normal 23, 29). M1 macrophages are stimulated by IFNg and microbial lung tissues or tissues adjacent to tumor from two different time points, products (lipopolysaccharide), are bactericidal, and express induc- 24 to 26 (early) or 42 (late) weeks after urethane injection. The number of ible nitric oxide synthase (iNOS). The M2 state is induced by replicate mice used for lung tissue microarrays was 5 early normal, 8 early the cytokines interleukin (IL)-4 and IL-13, produces polyamines, adjacent, 7 late normal, and 7 late adjacent. Data quality characteristics of lymphocyte suppression, and arginase expression. In A/J mice, the MG-U74Av2 microarrays were as previously reported (12). The adjacent 4 alveolar macrophage phenotype changes from M0(expressing tissue datasets were deposited at Gene Expression Omnibus (31) under the small or negligible amounts of arginase or iNOS) in the absence series accession GSE2514 (12). Microarrays completed from BAL-derived cells were from 7 control and 11 urethane-treated A/J mice. Two control of tumors, to M2 (arginase expressing) in adenoma-bearing mice, BAL RNAs were combined (C6_C13 BAL) to give sufficient total RNA for to M1 (iNOS expressing) in mice with carcinomas (22). labeling, and 1 control sample (C36; indicated with an asterisk) was later In this study, we compared the gene expression profiles of identified as from a mouse with a spontaneous lung tumor. The MOE4302.0 normal-appearing lung tissue adjacent to tumors with age- microarrays had an average scaling factor of 5.0(SD = 1.0),average matched, untreated controls, and we identified a set of 46 genes background intensity of 77 (SD = 64), and average percent present called specifically altered by the tumor microenvironment. Supporting the 43% (SD = 5.5%). Affymetrix probe IDs from the various microarrays used in observed biology of TAMs, these genes are highly expressed in the this analysis were annotated and cross-referenced using NetAffx (32). The monocyte-macrophage cell lineage. The increased protein expres- new microarray datasets described in this work, including .cel and .exp files, sion for a number of these candidates was verified by immuno- have been deposited at the Gene Expression Omnibus under the series histochemistry of lung cancer tissue. Importantly, these genes had accession GSE7269. Processing and statistical analysis of the microarray datasets was completed using BRB-ArrayTools (v3.4.0a; 6/2006) freely strong predictive value when used to classify expression data from available from Dr. Rich Simon (33).5 The BRB-ArrayTools suite makes BAL-derived macrophages from urethane-treated A/J mice in an extensive use of parametric and permutation analyses, as well as estimation independent study. These results suggest that gene expression of significance by False Discovery Rate. Classifiers are crossvalidated by a information contained within samples from outside the tumor ‘‘leave one out’’ strategy by several different methods. More limited sets of itself (surrogate tissues) are informative for predicting tumor probeIDs intensity data were analyzed using Excel XP and GraphPad Prism status, and that TAMs play a central role in lung tumorigenesis. 4. Cluster and heatmap diagrams were generated in BRB-ArrayTools, which implements the algorithms according to Eisen et al. (34), using centered Materials and Methods genes, centered correlations, and average linkage settings. Measurement of RNA levels by quantitative real-time PCR. All Expression levels of specific genes verified by quantitative real-time reagent kits, primer/probe sets, and the GeneAmp 5700 Instrument were PCRor Western blot analyses. The following genes were verified by the from Applied Biosystems, Inc. and used according to the manufacturer’s methods described below and are listed here using their standardized gene instructions. Briefly, 100 ng of total RNA was converted to cDNA using the symbol notation3 in alphabetical order for reference: acid phosphatase 5 (Acp5; tartrate resistant), C-C motif ligand 6 (Ccl6), CD68 antigen (Cd68),

4 http://www.ncbi.nlm.nih.gov/geo/ 3 http://www.ncbi.nlm.nih.gov/gene 5 http://linus.nci.nih.gov/BRB-ArrayTools.html www.aacrjournals.org 35 Cancer Res 2008; 68: (1). January 1, 2008

High Capacity cDNA Reverse Transcription kit in 100 AL reaction and called the ‘‘Field Effect’’). The tumor microenvironment could diluted 2-fold with water. Five microliters of cDNA was used per primer/ produce a specific gene expression signature due to immune cell probe (1 concentration) set reaction in TaqMan Universal PCR MasterMix infiltration, endothelial cell alterations (e.g., angiogenesis), and/or A (20 L final volume) and run for 40cycles under standard conditions. the variety of cellular responsiveness to signaling pathways. In Mouse h-actin endogenous control primer/probe set was used on all addition, a Field Effect signature may aid in biomarker discovery samples and a plasmid DNA dilution standard curve validated amplification from a more distantly removed surrogate tissue source (38) such efficiencies (r = 0.99; >5 replicates). Ct values are averages of RNA derived from 5 age-matched or 12 urethane-treated A/J mouse BAL cells, each as BAL. sample run in duplicate. DCt value of each gene was calculated relative to Developing an expression signature of the tumor microen- h its -actin (lower DCt values indicate more abundant transcripts) and fold vironment. Our microarray dataset consisted of normal or increase observed in urethane treated versus controls as 2 ( DDCt). adjacent to tumor A/J mouse lung tissues from two different time Protein biomarker analysis and immunohistochemistry. Protein points, 24 to 26 or 42 weeks after urethane injection (early versus extract preparation, Western methodology, and antibodies used are late time points). Using a four-way classifier algorithm in BRB- described in the Supplementary Method. For immunohistochemistry, ArrayTools (testing for significance in four categories: normal, A/J mouse lungs were perfused by instilling saline through the pulmonary adjacent, early, and late), a total of 204 probeIDs were identified artery to clear the blood from the lungs. Lungs were inflated with 10% with parametric P values of V0.0001 (see supplementary Excel buffered formalin before formalin fixation. Fixed lungs were paraffin embedded and 4 Am sections were cut and affixed to glass microscope Spreadsheet for complete signal intensity and annotation data). On f slides. Tissue sections were rehydrated, endogenous peroxidases quenched the MG-U74Av2 microarray, with 12,000 probeIDs, <2 probeIDs in 1% H2O2 in methanol, and subjected to antigen retrieval in 100 mmol/L would be expected by random chance at this level of significance. sodium citrate buffer in a steamer. After blocking in PBS containing As shown in a cluster dendrogram (Fig. 1A), the 204 probeID 10% of the appropriate blocking serum, sections were incubated with classifiers generate a tree diagram with 2 main branches [early primary antibody (Supplementary Method) for 90min in a humidified (yellow) versus late (black) samples; Age] subdivided by proximity chamber at 37jC. Proteins were visualized using Vectastain kits according to lung tumors [normal (green) versus adjacent (blue); Field]. Two to the manufacturer’s instructions (Vector Laboratories) and diamino- criteria were jointly used to designate whether a probeID was benzidine. classifying samples based on their proximity to lung tumor (Field) BAL. The tracheas were cannulated with an 18-gauge angiocatheter, and or because the animals in this study were used at two time points 1 mL PBS+0.6 mmol/L EDTA was instilled and aspirated back into the syringe as described previously (35). For cell isolation, lungs were rinsed (Age). First, on a probeID basis, was the Student’s t test for signi- 5 times with 1 mL PBS, the rinses combined and centrifuged at 500 g for ficant differences in the log2 signal intensities with a P value of 10min. The rinses were then aspirated off, leaving f500 AL per sample, and <0.001, when comparing both potential pairs of samples (normal the cells transferred to a 1.5 mL microfuge tube, pelleted at 16,000 g for versus adjacent or early versus late). Second, which t test com- 1 min, and directly lysed in 400 AL Trizol Reagent (Invitrogen) for RNA parison was smaller (i.e., more statistically significant), the Field or purification. For protein analysis, BAL fluid (f600 AL/mouse) was Age comparison. Using both comparisons, the 204 probeIDs were centrifuged at 16,000 g to remove cells before being concentrated in a divided into 48 Field probeIDs (24%; see Table 1 for listing of Field f A Microcon (Millipore) concentrator with a 3 kDa cutoff to 50 L/sample. Effect genes), 139 Age probeIDs (68%), and 17 not clearly separated Protein concentration was determined and samples were diluted into by these criteria into either category (8%). The Field probeIDs had protein-loading buffer. Insufficient protein amounts of BAL samples from significant absolute signal intensities (>64 arbitrary units) and each vehicle-treated mice did not allow comparable SDS-PAGE analyses, so only the samples from 24- and 42-week urethane-treated animals were analyzed signal intensity distribution of the probeID was separable by for Ctsd, Ctsz, Chi3l1, Spp1, and Sirpa content. Four separate replicate mice sample type (Supplementary Fig. S2). If the Field Effect genes are were used to produce BAL fluid protein samples. The BAL protein yield reflective of the tumor microenvironment, the magnitude of the from the 24-week urethane-treated mice was substantially lower than that difference in signal intensities (adjacent-normal) should increase from the 42-week urethane-treated mice. Immunoblot analysis was as the tumor becomes more aggressive at the later time point. A performed as described above, except that blots were stripped with 1% DDSignal Intensity graph (Supplementary Fig. S3) shows that 25 of Tris-HCl (pH 6.7), 0.8% h-mercaptoethanol, and 2% SDS at 50jC for 30min the 43 up-regulated genes in the tumor microenvironment were before reuse. significantly increased, whereas 3 of the 5 down-regulated genes were significantly decreased at the late time point. The Field and Age t test comparisons gave unambiguous Results categorization, as the ratio of t tests (e.g., Field t test P value/Age We and others have used microarray technology to reveal t test P value) were clearly different (median Field probeIDs t test expression differences between lung cancer and its surrounding ratio, 30,000; median Age probeIDs t test ratio, 78,000). To visualize tissue, with the goals of tumor classification, therapeutic respon- the significance of these two sets of probeIDs, supervised clustering siveness, recurrence, and murine modeling of this disease state of the microarray datasets clearly identified the samples appropri- (10–12, 36, 37). As an extension of the analysis of the A/J-urethane ately [Fig. 1B (Field) and C (Age)]. The expression level heatmap murine model of human lung cancer (12), our approach here was to of the Field probeIDs shows the majority of genes (90%) are up- ascertain the influence of the tumor microenvironment on gene regulated (red) in the adjacent tissues, suggesting the presence of a expression in adjacent tissue by comparison to age-, sex-, and novel gene expression signature (Fig. 1D). environmental exposure–matched normal control A/J mice (Sup- Assessment of cell lineage characteristics within the tumor plementary Fig. S1). Given the extensive literature on the microenvironment. Two in silico analyses were done to ascertain interrelationships between the roles of inflammation processes, if the Field genes can be attributed to a likely candidate for cellular immune and stromal cells, and tumorigenesis (19, 21, 28), we origin. The murine Field probeIDs were converted to their appro- hypothesized that a novel gene expression signature would be priate human ortholog probeIDs for querying the SymAtlas data- found in what is histologically appearing normal lung tissue as a base of normal human tissue expression (39). Each probeID was result of the influence of the tumor microenvironment (herein scored for its expression level by tissue type and marked as

Cancer Res 2008; 68: (1). January 1, 2008 36 www.aacrjournals.org

Figure 1. Microarray analysis of the A/J-urethane lung cancer model identifies Field Effect classifiers. A, RNAs from murine A/J-urethane lung tissues from either adjacent (A or A42)or age-matched normal controls (N or N42)were analyzed using MG-U74Av2 microarrays (Affymetrix)and the data were classified usi ng a four-way comparison for age (24 or 42 wk after urethane treatment)or source of tissue (adjacent or normal control).The resulting 204 probeID classif ier was used for supervised clustering of the datasets and displayed in a dendrogram format. Color coding bars are used to indicate tissue source (green, normal; blue, adjacent) and age (yellow, early, 24 wk postinjection; black, late, 42 wk postinjection). The datasets are distributed into two main branches by age (early versus late), and then each branch subdivided by location of tissue (adjacent versus normal). B, the microarray datasets were analyzed by supervised clustering using the Field Effect probeIDs (n48; adjacent versus normal)identified from the four-way comparison classifier (n204).The data are displayed in a dendrogram format wit h color coding as in Fig. 1A. Two main branches clearly divide the microarray datasets by adjacent and normal lung tissues. C, the microarray datasets were analyzed by supervised clustering using the Age Effect probeIDs (n139; early, 24 wk postinjection; late, 42 wk postinjection)identified from the four-way comparison clas sifier (n204). The data are displayed in a dendrogram format with color coding as in Fig. 1A. Two main branches are found, clearly dividing the microarray datasets by early and late lung tissues. D, gene expression heatmap displays the degree of regulation for the Field Effect probeIDs with the probeIDs down-regulated in the adjacent tissue listed as the top five rows followed by the up-regulated probeIDs. Maximum color saturation was set at f3-fold change in gene expression. The data are displayed above the heatmap as a dendrogram with the color coding as in Fig. 1A.

‘‘overexpression’’ if the signal intensity of the tissue was signi- murine probeID orthologs of the human immune cell type specific ficantly higher than its median level (Supplementary Fig. S4). genes, characterized by Du et al. (40), were used for supervised A tissue was scored as overexpressing a gene when the signal clustering of the A/J normal and adjacent microarray datasets intensity was greater than the median of all tissues plus thrice the (Supplementary Fig. S5). In no case were the major immune cell SD of the measured signal intensities. The Field probeIDs were gene clusters able to distinguish the normal from adjacent samples. highly expressed in whole blood, specifically peripheral blood These results suggest a unique myeloid!monocyte!macrophage CD14+ monocytes and bone marrow CD33+ myeloid cells, the lineage for the cells found in the adjacent lung tissue as responsible precursors to tissue matured macrophages. As a further test, the for the gene expression signature. www.aacrjournals.org 37 Cancer Res 2008; 68: (1). January 1, 2008

Table 1. Field effect genes

ProbeID Description Gene symbol Fold change Parametric (adjacent/normal) P value

160564_at Lipocalin 2 Lcn2 4.26 1.57E05 97420_at Leucine-rich a-2-glycoprotein 1 Lrg1 3.65 2.00E07 98467_at Inter a-trypsin inhibitor, heavy chain 4 Itih4 3.19 7.10E06 101912_at RIKEN cDNA 1100001G20 gene 1100001G20Rik 2.98 <1E07 160406_at Cathepsin K Ctsk 2.61 1.30E06 92849_at Chemokine (C-C motif) ligand 6 Ccl6 2.53 9.20E06 100507_at Nephroblastoma-overexpressed gene Nov 2.36 2.00E07 97519_at Secreted phosphoprotein 1 Spp1 2.16 2.70E06 104308_at Integrin a X Itgax 2.11 1.36E05 96787_at Serine (or cysteine) peptidase inhibitor, Serpina102.0 7 8.99E05 clade A (a-1 antiproteinase, antitrypsin), member 10 95597_at Prostaglandin-endoperoxide synthase 1 Pigs1 2.07 6.40E06 92918_at Coagulation factor VII F7 2.03 3.20E06 98859_at Acid phosphatase 5, tartrate resistant Acp5 1.97 2.86E05 162198_f_at 1.94 <1E07 103258_at Lymphocyte antigen 75 Ly75 1.88 4.70E06 95784_at Paired-Ig–like receptor A1 Pira1 1.86 2.23E05 93354_at Apolipoprotein C-1 Apoc1 1.86 1.70E05 103016_s_at CD68 antigen Cd68 1.82 5.58E05 97413_at RIKEN cDNA 1600029D21 gene 1600029D21Rik 1.78 3.00E07 92762_at C-type lectin domain family 4, member a2 Clec4a2 1.75 9.60E06 92633_at Cathepsin Z Ctsz 1.74 7.91E05 161895_s_at cDNA sequence BC054822 BC054822 1.69 1.23E05 99071_at Macrophage-expressed gene 1 Mpeg1 1.67 7.03E05 100607_at Phospholipase D family, member 3 Pld3 1.66 6.31E05 99952_at Chitinase 3-like 1 Chi3I1 1.64 1.00E04 95611_at Lipoprotein lipase Lpl 1.58 7.92E05 93810_at Cathepsin D Ctsd 1.58 8.00E07 161458_at Coagulation factor VII F7 1.54 2.21E05 94747_at Colony-stimulating factor 2 receptor, Cst2rb1 1.47 2.11E05 h 1, low affinity (granulocyte-macrophage) 103070_at Protein tyrosine phosphatase, Sirpa 1.46 4.57E05 nonreceptor type substrate 1 98543_at Cathepsin S Ctss 1.44 1.50E05 101851_at CD200 antigen Cd200 1.43 1.10E05 161819_f_at Lysosomal-associated protein transmembrane 5 Laptm5 1.43 3.77E05 104300_at IQ motif containing GTPase-activating Iqgap1 1.40 8.00E05 protein 1 (lqgap1), mRNA 94056_at Stearoyl-CoA desaturase 1 Scd1 1.37 7.60E06 103210_at Colony-stimulating factor 2 receptor, Csf2rb2 1.35 1.07E05 h 2, low affinity (granulocyte-macrophage) 94713_at Myosin Vlla Myo7a 1.33 2.40E06 92232_at Suppressor of cytokine signaling 3 Socs3 1.31 9.02E05 101389_at Scavenger receptor class B, member 2 Scarb2 1.31 1.29E05 96657_at Spermidine/spermine N1-acetyl transferase 1 Sat1 1.27 3.49E05 103584_at RIKEN cDNA 5830471E12 gene 5830471E12Rik 1.27 2.62E05 100581_at Cystatin B Cstb 1.26 8.57E05 100611_at Lysozyme Lyzs 1.16 1.47E05 101585_at Progesterone receptor membrane component 1 Pgrmc1 0.84 7.20E06 102871_at Eph receptor B6 Ephb6 0.75 4.58E05 96812_at Smoothened homologue (Drosophila) Smo 0.75 1.50E06 92448_s_at Glial cell line–derived neurotrophic Gfra2 0.61 4.89E05 factor family receptor a 2 92449_at Glial cell line–derived neurotrophic Gfra2 0.52 2.01E05 factor family receptor a 2

NOTE: The gene descriptions, symbols, fold change (adjacent/normal), and estimated P value from the parametric statistical test in BRB-ArrayTools classifier algorithm are given for each probeID identified as a Field gene. The 48 probeIDs represent 46 different genes, with the majority of the genes (91%) being up-regulated (higher in the adjacent tissue than normal control). Previously published work (41) independently supports increased protein level of Ptgs1 (COX1), and our studies reported here validate increased protein expression of Ctsd, Ctsz, Chi3l1, Spp1, and Sirpa in adjacent lung tissue.

Cancer Res 2008; 68: (1). January 1, 2008 38 www.aacrjournals.org

(Supplementary Fig. S6), the tumor-stromal interface can be visualized. Generally, there is dark staining within the individual macrophages surrounding the tumors. The tumors show little staining for Cstz and Sirpa, some light staining for Cstd and Chi3l, but increased staining for Spp1. Interestingly, Spp1 is the only gene in the Field Effect classifier that was also found in our orthologous gene expression comparison of human lung adenocarcinoma and the A/J-urethane murine model (12). This suggests that Spp1 has an increasing expression gradient going from age-matched normal to adjacent to tumor tissue. Thus, 47 of the 48 Field Effect probeIDs are unique classifiers of the tumor microenvironment and are not the result of tumor cell contamination in the adjacent tissue or phagocytosis of tumor cells by the macrophages. Field Effect classifiers correctly predict BAL samples from tumor-bearing mice. If the Field genes are highly expressed in the vicinity of lung cancer due to infiltrating macrophages, we reasoned that the same genes would be useful in classifying microarray datasets derived from the cells recovered by BAL. As previously found (22), urethane-treated mice have a significant increase in total cells recovered by BAL compared with controls (3.7-fold; P < 0.0001). However, the cell type distribution was indistinguishable between urethane-treated and control mice, with macrophages the major cell type recovered (f95%) followed by lymphocytes (f5%) and neutrophils (f1%). RNA was prepared from cells recovered by BAL from 6 normal untreated controls and Figure 2. Western blot analysis for protein levels in whole homogenates. 11 urethane-treated A/J mice. These datasets were used to generate Whole homogenates from either normal or adjacent lung tissues were measured for the protein levels of five different Field genes (Ctsd, Ctsz, Chi3l1, Spp1, a BAL-specific classifier probeID lists (168 probeIDs; P < 0.0001), and Sirpa). In each case, the protein levels were higher in the adjacent tissues and a supervised cluster dendrogram using this list is shown in compared with the controls. The expression of Ctsz, Chi3l1, Spp1, and Sirpa were significantly increased (P < 0.05)as determined by densitometry Fig. 4A. Each BAL sample is closely grouped in either tumor- measurements of the exposed films. bearing (T; red) or control (C; green) branches of the tree diagram with only sample C36 misidentified (magenta asterisk). On inspection of the formalin-fixed lungs from this mouse, it clearly Validation of increased protein levels of the Field genes in had developed a spontaneous tumor, something the A/J strain is the tumor microenvironment. The protein level of five different known to do as it ages (17, 42), suggesting why it was more closely Field genes (Ctsd, Ctsz, Chi3l1, Spp1, and Sirpa) was verified by grouped with the BAL samples from treated mice. The same BAL Western blot analysis of normal and adjacent A/J lung homoge- datasets were clustered with supervision using the independently nates (Fig. 2). In each case, the protein levels were higher in lung derived Field probeIDs (49 probeIDs; Fig. 4B) producing a similar homogenates from adjacent tissue from urethane-treated A/J mice degree of accuracy as the BAL-derived classifiers. Using the compared with untreated normal controls. Our previous work on prediction algorithm within BRB-ArrayTools, the Field probeIDs Ptgs1 (COX1), another Field gene, had increased protein in lung correctly predicted 94% to 100% of the BAL samples using a variety homogenates from adjacent tissue (41). of statistical approaches (compound covariate, nearest neighbors, Immunohistochemical analyses for the same five proteins was nearest centroid, and support vector machine; ref. 33). Interestingly, carried out to determine if the up-regulated expression observed in the Field gene and BAL-derived classifiers had only three genes lung homogenates can be attributed to specific cell types within in common: Acp5, Lrg1, and the expressed sequence tag (EST) the tumor microenvironment (Fig. 3). Generally, the overall inten- 1100001G20Rik (Fig. 4C). Although the majority of the Field genes sity of immunohistochemical staining increased in macrophages at were not selected as BAL classifiers, most had significant absolute 42 weeks after urethane injection, along with their number and size signal intensities (>64 arbitrary units) and nonoverlapping signal in the adjacent tissue. Ctsd was present in all macrophages and intensity distributions in the BAL microarray datasets (Supple- RBC, as well as occasional Clara cells in all groups. Ctsz was not mentary Fig. S7). found in RBC or Clara cells, and was not expressed in all macro- The BAL samples were correctly classified using the Field Effect phages, but increased macrophage expression was seen in 42-week genes (derived from lung tissue samples), although the majority of urethane-treated mouse lung. Chi3l1 protein was highly expressed these genes did not overlap with the BAL-specific classifiers. For in the large macrophages found in 42-week urethane-treated this reason, we verified the BAL microarray signal intensities for mouse lung but present at lower levels in Type 2 and Clara cells. a number of different Field genes by quantitative real-time PCR Spp1 was found in macrophages and this expression did not (qRT-PCR) analysis (Fig. 5A and B). Theoretically, the Field gene depend on prior urethane-treatment, but its overall level increased classifiers could give accurate predictions of the BAL samples yet in macrophages at 42 weeks after urethane treatment. Sirpa was have opposite expression levels (e.g., up-regulated Field genes are expressed in Clara cells and macrophages in urethane-treated mice down-regulated in BAL RNA). The mRNA levels for the 11 genes but was absent in these cell types in age-matched controls. Ptgs1 tested were significantly up-regulated in BAL cells from treated was previously shown to be expressed in Clara and Type 2 cells, animals compared with control BAL cells, showing 1.5- to 21-fold and macrophages in adjacent tissue (41). At lower magnification increases (P V 0.01). The qRT-PCR results further support the www.aacrjournals.org 39 Cancer Res 2008; 68: (1). January 1, 2008

Downloaded from cancerres.aacrjournals.org on September 25, 2021. © 2008 American Association for Cancer Research. Cancer Research in silico assessment (Supplementary Figs. S4 and S5) that the Field disease, when it is most curable by surgical resection and Effect signature is attributable to macrophages because 95% of the chemotherapy. The controversial use of spiral computed tomog- recovered cells in BAL are macrophages. In addition, Western blot raphy in asymptomatic smokers (7, 8) also suffers from high analysis showed increased protein levels in cell-free BAL fluid for monetary cost and relatively poor rate of new patient identification the same proteins measured in lung homogenates (Fig. 5C). This during the extensive follow-up period. An alternative approach suggests that macrophages infiltrating the tumor microenviron- uses laser-induced fluorescence endoscope bronchoscopy to ment have a unique Field gene expression signature reflected in localize early stage bronchial lesions (43). Although screening both the mRNAs of the BAL cells and the secreted proteins in BAL studies continue to show the survival advantage of early diagnosis, of urethane-treated A/J mice. no strategy other than direct biopsy exists for the evaluation of peripheral nodules. Our studies were designed to investigate the gene expression signature of the lung tumor microenvironment Discussion and evaluate the utility of BAL expression profiling. As with many cardiovascular diseases, lung cancer risk can easily In our previous work, we presented evidence using a microarray be assigned to a specific patient population, with the overwhelming approach that the murine A/J mouse-urethane system is an risk factors being cigarette smoke inhalation and environmental excellent animal model of human adenocarcinoma due to the high exposures (1). Unlike cardiovascular diseases, there are few degree of conservation in the altered gene expression patterns clinically valid methods for identifying individuals with early stage due to the cancer (12). We have now extended the study of the

Figure 3. Immunohistochemical analysis of Field Effect proteins in fixed lung tissues. Formalin-fixed lung tissues from normal (42-wk untreated controls)and urethane-treated (24 or 42 wk after injection)A/J mice were stained for immunohistochemistry analyses of five different Field gene products (Ctsd, Ctsz, Chi3l1, Spp1, and Sirpa). Colored arrows, pointing toward representative cells or locations, show cell type identifications. Intensity of the brownish gray staining was generally highest in 42-wk urethane-treated animals. Twenty-four–week untreated animals were similar to 42-wk normal controls (not shown). Magnification bar, 40 Am.

Cancer Res 2008; 68: (1). January 1, 2008 40 www.aacrjournals.org

remodeling (cathepsins D, K, S, and Z; proteinase inhibitors cystatin B and a1-antitrypsin; and integrin a X), macrophage markers (CD68 and macrophage-expressed gene 1), and cellular immune functions (Ly75; Ccl6; CD200; colony-stimulating factor 2 receptor, h 1, and low affinity (granulocyte-macrophage); suppressor of cytokine signaling 3; and signal-regulatory protein-a). In addition, Ptgs1 was found up-regulated within the Field genes, a well-characterized marker for the role of inflammation in cancer (12, 35). A number of recent reports have identified macrophage infiltration into the lung tumor microenvironment as an important consequence of multiple signaling pathways, including C-C motif and angiogenesis pathways (19, 21, 24). Our data argue strongly for the infiltration of a specific macrophage phenotype into the lung tumor microenvironment. These macrophages produce a gene expression signature (Field Effect genes) attributed to their presence in the tumor microenvironment, perhaps due to functional differences from the normally resident lung macrophages. The ‘‘strength’’ of the Field genes was shown by their predictive abilities in correctly classifying cells recovered by saline BAL of urethane or control A/J mouse lungs. In addition, a number of secreted proteins, predicted to be overexpressed with in the tumor microenvironment, were increased in cell-free BAL fluid. Because there is a high degree of similarity between the A/J-urethane model and human lung adenocarcinoma, we speculate that the Field genes could have predictive value in analyzing cells and/or fluid from BAL of cigarette-smoking patients. Ideally, a unique signature found by gene expression or proteomic analysis will be identified in high-risk asymptomatic individuals with early stage lung cancer. The influence of immune cell infiltration on metastatic potential and patient survival has been examined in patients with colorectal cancer (44, 45). Instead of using a microarray approach, large-scale flow cytometry (48 CD markers and 10surface markers/receptors) and qRT-PCR measurements (20genes) were used to create a panel of cellular immune markers. Differing from our lung cancer results, their studies focused on the adaptive immune response, specifically the effector T cells, which were important in colorectal cancer, rather than inflammatory or immunosuppressive molecules. They found that T-cell markers of migration, activation, and differentiation were increased in colorectal tumors without signs of metastatic invasion. The density, type, and location of immune Figure 4. Field genes correctly classify microarray datasets from cells cells within the colorectal tumors, gave a better predictor of patient recovered by BAL. A, RNAs from cells in BAL fluid from urethane-treated or age-matched normal controls were analyzed using MOE430 2.0 microarrays survival than the current histologic methods to stage their cancers (Affymetrix). The datasets were analyzed using a two-way comparison and was reproduced in two additional patient populations. [tumor-bearing (T)versus control ( C)], generating a set of BAL-specific classifier The tumor microenvironment is clearly complex in terms of cell probeID lists (168 probeIDs; P < 0.0001). The microarray datasets are displayed after supervised clustering using the BAL-specific probeIDs with color-coding types within the milieu, regulation of signaling pathways, and bars for control (green)or tumor-bearing ( red)animals. Sample C36, from a tissue remodeling. In selected cases, potential novel therapeutic control animal, nearest neighbors treated BAL cells; the formalin-fixed lung of targets have been identified by analysis of the tumor microenvi- C36 was examined and found to have a spontaneous tumor (magenta asterisks). B, microarray datasets from BAL fluid cells are displayed after supervised ronment. Strikingly, the single largest functional group within the clustering using the Field Effect probeIDs (n49). The Field Effect probeIDs lung cancer Field genes is proteases (cathepsins D, K, S, and Z) and correctly identify the origin of the BAL cells, from control or urethane-treated a animals, as well as the BAL-specific classifiers. Color coding and labeling are as their inhibitors (cystatin B and 1-antitrypsin). The role of the in A. C, Venn diagram representing the overlap of the BAL-specific classifiers cathepsin protease family in pancreatic cancer has been studied and Field Effect genes showing the Field genes are distinct (46 of 49; 94%) using both transgenic mouse models and pharmacologic agents from the BAL classifiers. Only Acp5, Lrg1, and the EST 1100001G20Rik are shared between the two classifier lists. (46, 47). Their initial findings, using microarray analysis of the RIP1-Tag2 murine pancreatic cancer model, showed higher expression of a number of cathepsins in tumor tissue. In addition, A/J-urethane model to ask if the tumor microenvironment deve- cathepsins are up-regulated in human pancreatic cancer and in a lops a unique gene expression signature within the histologically number of other tumor types (48, 49). Small molecule cathepsin normal tissue (adjacent to tumor) compared with untreated age- inhibitors and specific cathepsin gene knockouts caused decreased matched controls (normal). The 48 Field probeIDs represents pancreatic tumor formation, impaired angiogenesis, and lowered 46 different genes with broadly functional groupings in matrix metastatic potential. Although cathepsins are typically lysosomal, www.aacrjournals.org 41 Cancer Res 2008; 68: (1). January 1, 2008

Downloaded from cancerres.aacrjournals.org on September 25, 2021. © 2008 American Association for Cancer Research. Cancer Research in pancreatic cancer, they are found extracellularly at the interface A/J mice, suggesting these proteases are also present in the between normal and tumor cells, implicating them in extracellular extracellular environment of lung adenocarcinoma. matrix remodeling and angiogenesis. We found elevated protein In summary, many studies suggest that secreted chemotactic and levels for cathepsins D and S in BAL fluid from urethane-treated differentiating factors affect TAMs in the tumor microenvironment

Figure 5. Validation of increased expression of Field Effect genes in BAL. A, mRNA levels for 11 different Field genes were determined using qRT-PCR with primer/probe pairs. DCt values were measured relative to h-actin as the endogenous control mRNA in BAL from 5 normal and 11 treated animals. Smaller DCt values indicate higher expression relative to h-actin in the BAL samples. Fold increase (tumor BAL/control BAL)was calculated by the DDCt with the t test P value of <0.02 in all cases. Columns, mean; bars, the SE. B, the protein levels for five different Field genes (Ctsd, Ctsz, Chi3l1, Spp1, and Sirpa)were determined by Western blot analysis of cell-free BAL fluid. In each case, the protein levels were higher in the BAL fluid from treated animals compared with the controls. The expression of Ctsd, Chi3l1, and Spp1 was significantly increased (P < 0.05) as determined by densitometry measurements of the exposed films.

Cancer Res 2008; 68: (1). January 1, 2008 42 www.aacrjournals.org

(19, 23, 50). TAMs have a distinct phenotype that promotes genes as human classifiers using BAL cells and fluid from control inflammation, angiogenesis, and matrix remodeling, while sup- and lung cancer patients. In addition, this work suggests new pressing the adaptive immune response. Our microarray analysis approaches in murine preclinical studies, investigating the rela- has identified a unique gene expression signature (Field genes) tionship between the cathepsin family members and lung cancer from TAMs in the tumor microenvironment that is distinct from using pharmacologic agents and/or transgenic mouse models. normal tissue. The Field genes identified in TAMs were able to correctly classify BAL-derived cells from urethane-treated or Acknowledgments control animals, which is one step further removed from direct Received 3/19/2007; revised 8/7/2007; accepted 10/8/2007. measurement of lung biopsy tissue. Overexpression of several Grant support: FAMRI Clinical Innovator Award (R.S. Stearman), R01 HL 72340 proteins from the Field gene list was validated in both lung homo- (M.W. Geraci), R01 CA 96133 (A.M. Malkison and M.W. Geraci), R01 CA33497 (A.M. genates and cell-free BAL fluid. Four different cathepsins were Malkinson), and P50CA 58187 (M.W. Geraci). The costs of publication of this article were defrayed in part by the payment of page found as Field genes suggesting their potential role(s) in lung charges. This article must therefore be hereby marked advertisement in accordance tumorigenesis. Future studies could examine the utility of the Field with 18 U.S.C. Section 1734 solely to indicate this fact.

References 19. Strieter RM, Belperio JA, Burdick MD, Sharma S, hydroxytoluene (BHT), causes chronic inflammation in Dubinett SM, Keane MP. CXC chemokines: angiogenesis, promotion-sensitive BALB/cByJ mice but not in 1. American Cancer Society. Cancer Facts and Figures immunoangiostasis, and metastases in lung cancer. Ann promotion-resistant CXB4 mice. Toxicology 2001;169: 2007. Available from: http://www.cancer.org/docroot/ N Y Acad Sci 2004;1028:351–60. 1–15. stt/stt_0.asp. 20. Yamamoto Y, Gaynor RB. InB kinases: key regu- 36. Kakiuchi S, Daigo Y, Ishikawa N, et al. Prediction of 2. Ruano-Ravina A, Figueiras A, Barros-Dios JM. Lung lators of the NF-nB pathway. Trends Biochem Sci sensitivity of advanced non-small cell lung cancers to cancer and related risk factors: an update of the 2004;29:72–9. gefitinib (Iressa, ZD1839). Hum Mol Genet 2004;13: literature. Public Health 2003;117:149–56. 21. Chen JJ, Lin YC, Yao PL, et al. Tumor-associated 3029–43. 3. Ruano-Ravina A, Figueiras A, Barreiro-Carracedo MA, macrophages: the double-edged sword in cancer 37. von Eyben FE. Epidermal growth factor receptor Barros-Dios J. Occupation and smoking as risk factors progression. J Clin Oncol 2005;23:953–64. inhibition and non-small cell lung cancer. Crit Rev Clin for lung cancer: a population-based case-control study. 22. Redente EF, Orlicky DJ, Bouchard RJ, Malkinson AM. Lab Sci 2006;43:291–323. Am J Ind Med 2003;43:149–55. Tumor signaling to the bone marrow changes the 38. Stearman RS, Keith RL, Dwyer-Nield L, et al. 4. Neuberger JS, Mahnken JD, Mayo MS, Field RW. Risk phenotype of monocytes and pulmonary macrophages Expression analysis in lung cancer: the search for factors for lung cancer in Iowa women: implications for during urethane-induced primary lung tumorigenesis in biomarkers. Am J Respir Cell Mol Biol 2004;31:S39–44. prevention. Cancer Detect Prev 2006;30:158–67. A/J mice. Am J Pathol 2007;170:693–708. 39. Su AI, Cooke MP, Ching KA, et al. Large-scale analysis 5. Mountain CF. Revisions in the international system for 23. Sica A, Schioppa T, Mantovani A, Allavena P. of the human and mouse transcriptomes. Proc Natl staging lung cancer. Chest 1997;111:1710–7. Tumour-associated macrophages are a distinct M2 Acad Sci U S A 2002;99:4465–70. 6. Mountain CF. Staging classification of lung cancer. A polarised population promoting tumour progression: 40. Du X, Tang Y, Xu H, et al. Genomic profiles for human critical evaluation. Clin Chest Med 2002;23:103–21. potential targets of anti-cancer therapy. Eur J Cancer peripheral blood T cells, B cells, natural killer cells, 7. Bach PB, Jett JR, Pastorino U, Tockman MS, Swensen 2006;42:717–27. monocytes, and polymorphonuclear cells: comparisons SJ, Begg CB. Computed tomography screening and lung 24. Mantovani A, Schioppa T, Porta C, Allavena P, Sica A. to ischemic stroke, migraine, and Tourette syndrome. cancer outcomes. JAMA 2007;297:953–61. Role of tumor-associated macrophages in tumor Genomics 2006;87:693–703. 8. Henschke CI, Yankelevitz DF, Libby DM, Pasmantier progression and invasion. Cancer Metastasis Rev 2006; 41. Bauer AK, Dwyer-Nield LD, Malkinson AM. High MW, Smith JP, Miettinen OS. Survival of patients with 25:315–22. cyclooxygenase 1 (COX-1) and cyclooxygenase 2 (COX-2) stage I lung cancer detected on CT screening. N Engl J 25. Strieter RM, Belperio JA, Phillips RJ, Keane MP. CXC contents in mouse lung tumors. Carcinogenesis 2000;21: Med 2006;355:1763–71. chemokines in angiogenesis of cancer. Semin Cancer 543–50. 9. Beer DG, Kardia SL, Huang CC, et al. Gene-expression Biol 2004;14:195–200. 42. Shimkin MB, Stoner GD. Lung tumors in mice: profiles predict survival of patients with lung adenocar- 26. Pouniotis DS, Plebanski M, Apostolopoulos V, application to carcinogenesis bioassay. Adv Cancer Res cinoma. Nat Med 2002;8:816–24. McDonald CF. Alveolar macrophage function is altered 1975;21:1–58. 10. Hayes DN, Monti S, Parmigiani G, et al. Gene in patients with lung cancer. Clin Exp Immunol 2006; 43. Hirsch FR, Franklin WA, Gazdar AF, Bunn PA, Jr. expression profiling reveals reproducible human lung 143:363–72. Early detection of lung cancer: clinical perspectives of adenocarcinoma subtypes in multiple independent 27. Clark JC, Tichelaar JW, Wert SE, et al. FGF-10disrupts recent advances in biology and radiology. Clin Cancer patient cohorts. J Clin Oncol 2006;24:5079–90. lung morphogenesis and causes pulmonary adenomas Res 2001;7:5–22. 11. Meyerson M, Franklin WA, Kelley MJ. Molecular in vivo. Am J Physiol Lung Cell Mol Physiol 2001;280: 44. Galon J, Costes A, Sanchez-Cabo F, et al. Type, classification and molecular genetics of human lung L705–15. density, and location of immune cells within human cancers. Semin Oncol 2004;31:4–19. 28. Malkinson AM. Role of inflammation in mouse lung colorectal tumors predict clinical outcome. Science 12. Stearman RS, Dwyer-Nield L, Zerbe L, et al. Analysis tumorigenesis: a review. Exp Lung Res 2005;31:57–82. 2006;313:1960–4. of orthologous gene expression between human pulmo- 29. Mantovani A, Sica A, Sozzani S, Allavena P, Vecchi A, 45. Pages F, Berger A, Camus M, et al. Effector memory T nary adenocarcinoma and a carcinogen-induced murine Locati M. The chemokine system in diverse forms of cells, early metastasis, and survival in colorectal cancer. model. Am J Pathol 2005;167:1763–75. macrophage activation and polarization. Trends Immu- N Engl J Med 2005;353:2654–66. 13. Buyse M, Loi S, van’t Veer L, et al. Validation and nol 2004;25:677–86. 46. Gocheva V, Zeng W, Ke D, et al. Distinct roles for clinical utility of a 70-gene prognostic signature for 30. Lewis CE, Pollard JW. Distinct role of macrophages in cysteine cathepsin genes in multistage tumorigenesis. women with node-negative breast cancer. J Natl Cancer different tumor microenvironments. Cancer Res 2006;66: Genes Dev 2006;20:543–56. Inst 2006;98:1183–92. 605–12. 47. Joyce JA, Baruch A, Chehade K, et al. Cathepsin 14. Chen HY, Yu SL, Chen CH, et al. A five-gene signature 31. Barrett T, Suzek TO, Troup DB, et al. NCBI GEO: cysteine proteases are effectors of invasive growth and and clinical outcome in non-small-cell lung cancer. N mining millions of expression profiles-database and angiogenesis during multistage tumorigenesis. Cancer Engl J Med 2007;356:11–20. tools. Nucleic Acids Res 2005;33:D562–6. Cell 2004;5:443–53. 15. Lu Y, Lemon W, Liu PY, et al. A gene expression 32. Liu G, Loraine AE, Shigeta R, et al. NetAffx: 48. Roshy S, Sloane BF, Moin K. Pericellular cathepsin B signature predicts survival of patients with stage I non- Affymetrix probesets and annotations. Nucleic Acids and malignant progression. Cancer Metastasis Rev 2003; small cell lung cancer. PLoS Med 2006;3:e467. Res 2003;31:82–6. 22:271–86. 16. Potti A, Mukherjee S, Petersen R, et al. A genomic 33. McShane LM, Radmacher MD, Freidlin B, Yu R, Li 49. Turk V, Turk B, Guncar G, Turk D, Kos J. Lysosomal strategy to refine prognosis in early-stage non-small-cell MC, Simon R. Methods for assessing reproducibility of cathepsins: structure, role in antigen processing and lung cancer. N Engl J Med 2006;355:570–80. clustering patterns observed in analyses of microarray presentation, and cancer. Adv Enzyme Regul 2002;42: 17. Malkinson AM. Primary lung tumors in mice as an data. Bioinformatics 2002;18:1462–9. 285–303. aid for understanding, preventing, and treating human 34. Eisen MB, Spellman PT, Brown PO, Botstein D. 50. Stout RD, Jiang C, Matta B, Tietzel I, Watkins SK, adenocarcinoma of the lung. Lung Cancer 2001;32:265–79. Cluster analysis and display of genome-wide expression Suttles J. Macrophages sequentially change their 18. Aggarwal BB, Takada Y, Shishodia S, et al. Nuclear patterns. Proc Natl Acad Sci U S A 1998;95:14863–8. functional phenotype in response to changes in transcription factor NF-nB: role in biology and medi- 35. Bauer AK, Dwyer-Nield LD, Hankin JA, Murphy RC, microenvironmental influences. J Immunol 2005;175: cine. Indian J Exp Biol 2004;42:341–53. Malkinson AM. The lung tumor promoter, butylated 342–9.

www.aacrjournals.org 43 Cancer Res 2008; 68: (1). January 1, 2008

Downloaded from cancerres.aacrjournals.org on September 25, 2021. © 2008 American Association for Cancer Research. A Macrophage Gene Expression Signature Defines a Field Effect in the Lung Tumor Microenvironment

Robert S. Stearman, Lori Dwyer-Nield, Michael C. Grady, et al.

Cancer Res 2008;68:34-43.

Updated version Access the most recent version of this article at: http://cancerres.aacrjournals.org/content/68/1/34

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2007/12/27/68.1.34.DC1

Cited articles This article cites 49 articles, 9 of which you can access for free at: http://cancerres.aacrjournals.org/content/68/1/34.full#ref-list-1

Citing articles This article has been cited by 5 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/68/1/34.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/68/1/34. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.