Molecular Cancer Research Targeting the PI3-Kinase Pathway in Cancer Bridging the Lab and the Clinic in Cancer Medicine
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Meeting Abstracts Online
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Creighton, C. J.
Right arrow Articles by Hanash, S. M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Creighton, C. J.
Right arrow Articles by Hanash, S. M.
Molecular Cancer Research 3:119-129 (2005)
© 2005 American Association for Cancer Research

Cancer Genes and Genomics

Analysis of Tumor-Host Interactions by Gene Expression Profiling of Lung Adenocarcinoma Xenografts Identifies Genes Involved in Tumor Formation

Chad J. Creighton1, Jennifer L. Bromberg-White5, David E. Misek2, David J. Monsma5, Frank Brichory2, Rork Kuick2, Thomas J. Giordano3, Weimin Gao2, Gilbert S. Omenn4, Craig P. Webb5 and Samir M. Hanash2

1 Bioinformatics Program, Departments of 2 Pediatrics and Communicable Diseases, 3 Pathology, 4 Internal Medicine, Human Genetics, and Public Health, University of Michigan, Ann Arbor, Michigan, and 5 Laboratory of Tumor Metastasis and Angiogenesis, Van Andel Research Institute, Grand Rapids, Michigan

Requests for reprints: Chad J. Creighton, Department of Pathology, University of Michigan, 4237 Med Sci I, 1150 West Medical Center Drive, Ann Arbor, MI 48109. Phone: 734-763-5823; Fax: 734-763-6476. E-mail: ccreight{at}umich.edu


   Abstract
Top
 Notes
 Abstract
Introduction
 Results
 Discussion
 Materials and Methods
 Acknowledgements
 References
 
Tumor cell lines are relied on extensively for cancer investigations, yet cultured cells in an in vitro environment differ considerably in behavior compared with those of the same cancer cells that proliferate and form tumors in vivo. To uncover gene expression changes related to tumor formation, gene expression profiles of human lung adenocarcinoma (A549) cells grown as lung tumors in immune-compromised mice were compared with profiles of the same cells grown in vitro. Additionally, profiles of uninvolved adjacent mouse tissue were determined. A profound interplay between cancer cells and the host was shown that affected a complex protein interaction network involving processes of extracellular interaction, growth factor signaling, hemostasis, immune response, and transcriptional regulation. Growth in vivo of A549 cells, which carry an activating k-ras mutation, induced changes in gene expression that corresponded highly to a pattern characteristic of human lung tumors with k-ras mutation. Cytokines interleukin-4, interleukin-6, and IFN-{gamma} each induced distinct in vitro genomic responses in cancer cells that emulated many of the changes in gene expression observed in vivo. Genes that were both selectively induced in vivo and overexpressed in human lung adenocarcinoma tumors included CSPG2, which has not been associated previously with tumor formation. Knockdown in A549 of CSPG2 by RNA interference significantly inhibited tumor growth in vivo but not in vitro. Thus, analysis of tumor xenografts by gene expression profiling has the potential for identifying genes involved in tumor development that may not be expressed in cancer cells grown in vitro.

Key Words: bioinformatics • microarray • interleukins • IFN-{gamma} • k-ras


   Introduction
Top
 Notes
 Abstract
 Introduction
Results
 Discussion
 Materials and Methods
 Acknowledgements
 References
 
It has become increasingly clear that cancer is a complex biological system affected by numerous factors and pathways that dampen or enhance the effects of genetic alterations (1, 2). In particular, the microenvironment has a significant effect on tumor development and progression (3). Tumor cell lines have been considered an important resource for cancer investigations because of their limited cellular complexity compared with human tumors and the ease with which they can be grown and manipulated under controlled conditions. However, as the in vitro microenvironment is substantially less complex than the environment in which cancer cells proliferate in vivo, the conclusions reached from cancer cell line investigations, although informative, may not be directly applicable to the natural setting in which human tumors arise, progress, and are treated. Given that the consequences of particular genetic alterations in cancer may be highly context dependent and governed by the spatial and temporal involvement of multiple interacting components, identification of gene expression differences between in vitro proliferation and in vivo tumor growth potentially may uncover genes that play an important role in tumor formation.

Global RNA expression profiling of xenograft tumors from human-derived cells grown in immune-compromised mice makes it possible to determine the extent to which cancer cells modify their expression profile in response to the local microenvironment. Gene expression profiling using oligonucleotide arrays represents an excellent means to study such tumor-host relationships in this setting because the potential background due to cross-species hybridization is minimal (4, 5). In this study, we have investigated the transcriptional changes in human A549 lung adenocarcinoma cells grown orthotopically in the lung of mice to better mimic the environment of human lung tumors. We have also examined potential responses of the host tissue to the implantation of cancer cells using a mouse oligonucleotide array to measure the expression of mouse genes in lung tissue adjacent to tumor. To gain further insight into the contributions of various factors in the microenvironment of lung tumors, we compared transcriptional profiles of cells grown in vitro under the influence of various cytokines with profiles of orthotopically implanted tumors and human lung adenocarcinomas. Furthermore, using small interfering RNAs, we silenced CSPG2, a gene arising in our analysis, to determine its effect on tumor growth in vivo.


   Results
Top
 Notes
 Abstract
 Introduction
 Results
Discussion
 Materials and Methods
 Acknowledgements
 References
 
Global Differences in Expression of Human Genes between Xenograft Tumors and Cell Cultures Indicate Extracellular Interactions Arising from Cancer Cells In vivo
Using Affymetrix (Santa Clara, CA) U133A oligonucleotide arrays (composed of 22,283 probe sets designed to measure human mRNAs), we compared profiles of A549 orthotopically implanted and grown as xenograft tumors (Fig. 1A) with profiles obtained from the cell line grown in culture. Principal components analysis of the expression data showed xenograft tumor profiles clustering separately from cell culture profiles (Fig. 2). Between the A549 xenografts and cell cultures, 2,865 probe sets (2,260 distinct genes) differed significantly at P < 0.05 with a fold change >1.8 in either direction (1,138 genes being higher in tumors and 1,122 genes being higher in culture), with 245 expected by chance (SD = 113). Criteria for selecting a statistically significant set of genes are a balance between false negatives and false positives. Here, we initially used a looser criteria yielding a larger gene set, with more information and fewer false negatives, for subsequent gene ontology (GO) term and pathway analysis.



View larger version (64K):
[in this window]
[in a new window]
  		FIGURE 1. Tissues used for gene expression profiling. A. Representative adenocarcinoma derived from A549 orthotopic xenograft lung tumor (profiled using Affymetrix U133A human oligonucleotide arrays). Whereas the tumor is poorly differentiated in areas, gland formation is morphologically appreciated. B. Representative adjacent nonneoplastic lung tissue (profiled using 430A mouse arrays) showing expansion of the interstitial spaces by an inflammatory infiltrate. The underlying architecture of the lung is preserved. C. Normal lung tissue from healthy mouse controls (profiled using 430A mouse arrays) showing normal alveolar parenchyma without interstitial expansion and inflammation.

 


View larger version (19K):
[in this window]
[in a new window]
  		FIGURE 2. A. Principal components analysis plot of U133A (human) gene expression profiles from A549 cell cultures and xenografts. Principal components were extracted using all 22,283 probe sets. The first two principal components are shown. B. Principal components analysis plot of 430A (mouse) gene expression profiles from normal mouse lung and diseased (tumor-involved) mouse lung.

 
Based on profiling of normal mouse lung tissue, the contribution of mRNA from mouse-derived tissue to the tumor profile as assessed with human probe sets was minimal (Fig. 3). From a comparison of the individual probe set intensities of the mouse lung profiles with those of the culture profiles, <5% of the genes with fold changes >1.8 in tumors relative to cultured cells could be attributed to mouse tissue alone rather than to the (human) neoplastic cells. The cellular composition and gene expression patterns of normal lung tissue are expected to be very different from those of tumor-involved lung tissue, making it difficult to determine specifically which U133A probe sets may be particularly susceptible to cross-contamination by mouse mRNAs. However, the control profile experiments indicate that the expected error rate due to cross-species contamination is comparable with the expected false-positive rate using the t test due to multiple testing.



View larger version (28K):
[in this window]
[in a new window]
  		FIGURE 3. Results of profiling mouse tissue on Affymetrix U133A arrays designed for human mRNAs. Distribution of fold changes in expression of normal mouse lung tissue over human A549 cell cultures, each profiled using the Hu133A array, along with the fold changes of A549 xenograft tumors over A549 cultures. From histologic analysis, the amount of mouse tissue in a xenograft tumor was established between 10% and 20%. Results shown are for probe sets that gave hybridization intensities >50 for A549 cultures.

 
We searched for significantly enriched (i.e., overrepresented) GO annotation terms associated with the genes that differed between xenografts and cell cultures for clues as to the processes that may contribute to the coordinate expression of these genes. Classification of 1,138 human genes with up-regulated expression in xenograft tumors relative to in vitro cultures revealed 28 significant GO terms (P < 0.005) associated with numerous processes of tumor-host interactions. Simulations using 1,138 randomly selected genes gave an average of only 3.3 (SD = 2.9) significant terms; thus, obtaining 28 such terms is not likely due to chance. Representative terms that characterized this set of genes include "extracellular" (95 genes; P = 5.6 x 10–6), "immune response" (82 genes; P = 1.5 x 10–6), "regulation of cell proliferation" (35 genes; P = 8.5 x 10–5), "complement activation" (10 genes; P = 0.00019), "serine-type endopeptidase activity" (19 genes; P = 0.00065), "trypsin activity" (14 genes; P = 0.0012), "induction of apoptosis" (17 genes; P = 0.0020), and "class I MHC antigen" (4 genes; P = 0.00086). In contrast, 85 significant GO terms were associated with the set of 1,122 genes expressed at higher levels in culture. Notable among these are the terms "cell cycle" (80 genes; P = 1.5 x 10–10), "cell proliferation" (102 genes; P = 8.5 x 10–7), and "mitochondrion" (60 genes; P = 7.1 x 10–7), indicative of increased cell division and metabolism in cells grown in the culture environment.

A potential difference in conditions between growth of cells as tumors and growth as in vitro cultures relates to oxygen availability. Hypoxia-inducible factor-1{alpha} is a key regulator of the hypoxic response. Of 75 genes known to be induced by hypoxia-inducible factor-1{alpha} (6), 17 genes, a highly significant number (P = 0.0003, Fisher's exact test), occurred among the set of 1,138 genes with higher expression in A549 tumors relative to cultures, suggesting a role for hypoxia in modulating gene expression in cancer cells in vivo. Although hypoxia-inducible factor-1{alpha} itself was not up-regulated at the mRNA level in A549 xenografts, its expression is regulated at the translational level in response to various signaling pathways, such as ERBB2 and Ras (6).

Using the publicly available Novartis atlas of gene expression profiles from 45 different mouse-derived normal tissues (7), we ranked the set of genes represented in the atlas by their expression in lung tissues over other tissues. We calculated the significance of overlap between the top A549 xenograft genes (when compared with cell cultures) and the top Novartis lung tissue-enriched genes. Similar analyses were carried out using genes enriched in each of the other 44 tissues in the Novartis atlas (Fig. 4A and B). Tissue-enriched genes that showed significant correspondence (i.e., overlap) with A549 tumor genes were representative of epithelial-derived tissues, including lung (P < 1 x 10–7 for the top 500 genes for each comparison). A similar pattern of lung-enriched genes was observed (P < 0.002) using an independent profile data set of A549 s.c. tumors (5), which were not implanted near the lung. In contrast, genes expressed in s.c. xenograft tumors derived from brain cells (5) had a high correspondence with cortex-enriched genes (P < 1 x 10–5). These results are indicative of acquisition of organ-specific expression in vivo relative to in vitro expression patterns.



View larger version (50K):
[in this window]
[in a new window]
  		FIGURE 4. A. Correspondence of xenograft tumor-associated genes (genes higher in A549 orthotopic tumors over culture and genes expressed higher in tumor-bearing mouse lung relative to control lung) with tissue-enriched genes (mRNAs) as defined using the Novartis mouse gene expression atlas of 45 different normal tissues. Genes represented in the Novartis atlas were ranked by expression in the given tissue over the other normal tissues. For each given tissue, the top 500 expressed genes for the given tissue type that were also represented by their human orthologues in the A549 data set were compared with the top 500 A549 tumor genes for significance of overlap. The same type of comparison was made between the top 500 Novartis genes enriched in the given tissue and the top 500 diseased (tumor-bearing) mouse lung genes. Arrow, nominal significance level at P < 0.05. For selected tissue and xenograft gene set comparisons, the actual number of genes in the shared overlap is given. The A549 tumor profile data set shares some 3,800 genes with the Novartis atlas, whereas the mouse lung data set shares some 4,900 genes. B. Correspondence of A549 tumor genes with lung tissue-enriched genes (mRNAs) as well as with prostate-enriched genes and placenta-enriched genes. Of the genes shared between the Novartis and A549 data sets, the top 1 to 1,000 genes from each comparison were examined for significance of overlap. C. Correspondence of diseased mouse lung genes with both bone marrow–enriched genes and lymph node–enriched genes.

 
Global Differences in Expression of Mouse Genes between Lung Tissue Adjacent to Tumors and Healthy Control Lung Tissue Indicate Host Tissue Responses
Mouse genes that are differentially expressed in stromal tissue relative to normal tissue may be indicative of host responses to the tumor that include recruitment of infiltrating cells, such as fibroblasts and inflammatory cells. Using Affymetrix mouse 430A arrays (composed of 22,626 mouse probe sets), we compared lung tissue from healthy mouse controls (Fig. 1C) with lung tissue immediately adjacent to A549 xenografts (Fig. 1B). Between the diseased and healthy lung tissues, 2,500 probe sets corresponding to 2,098 named genes differed significantly at P < 0.05 with a fold change >1.5 in either direction (expected 385 by chance; SD = 235), with 939 genes being expressed at higher levels in diseased lung and 1,159 genes expressed higher in healthy lung. Less than 10% of the differentially expressed genes were estimated to be attributed to infiltration of mouse lung tissue samples with A549 cells, based on hybridization of mouse arrays with human cDNA. Mouse genes that were lower in diseased lung tissue compared with healthy lung tissue include a significant number involved in the "mitochondrion" (31 genes in the 1,159 gene set had this GO term; P = 6.9 x 10–9) and "oxidoreductase activity" (56 genes; P = 3.8 x 10–4). Enriched GO terms (P < 0.01) found in the set of 939 genes expressed at higher levels in response to tumor formation include "cell adhesion" (35 genes), "metal ion binding activity" (60 genes), "defense/immunity protein activity" (12 genes), "cell-matrix adhesion" (8 genes), "response to stress" (6 genes), and "complement activity" (5 genes). Many of the tumor-induced mouse genes are expressed at relatively high levels in bone marrow and lymphoid tissue relative to other normal tissues (Fig. 4A and C), indicative of infiltration with lymphoid and myeloid elements as part of a host immune response.

Evidence for a Vast Protein Interaction Network Involving Both Human and Mouse Genes
To determine whether the human and mouse genes that are overexpressed in xenograft tumor development encode proteins that may interact with each other, we queried the Human Protein Reference Database (HPRD) catalogue of >10,000 known physical protein-protein interactions involving >3,800 genes represented on the U133A array (8). We sought to define a protein interaction network that encompassed the following: (a) protein-protein interactions resulting from human genes that were both up-regulated in A549 xenograft tumors compared with cultures; (b) genes encoding proteins with a significant number of interactions with other proteins encoded by A549 tumor genes; and (c) putative interactions between human proteins (derived from the tumor) and mouse proteins (derived from stromal tissue). In the set of 1,138 human genes overexpressed in xenograft lung tumors (P < 0.05; fold change >1.8), 143 HPRD interactions were represented (both members being included in the 1,138 gene set). Simulation testing (using randomly selected gene sets) yielded 77 interactions on average (SD = 15), indicating a high enrichment for interactions within the set of 143 occurring in vivo within xenograft tumors.

We next searched for "active hubs," which we defined as proteins with at least four interactions with proteins represented in the set of 1,138 human xenograft tumor genes, where the number of interactions was statistically significant (P < 0.05, Fisher's exact test) compared with the interactions shared by the active hub with genes in the entire U133A data set. These active hubs may represent nodes of high activity in the biological network. By this definition, a gene may not necessarily be significantly expressed to be an active hub; in this way, proteins that are active at the cellular signaling level, through post-translational modification (e.g., phosphorylation) but not at the transcriptome level, may be implicated in the network. We found 44 active hubs represented in the 1,138 gene set (expected 12, SD = 6.2). Of these 44 hubs, 12 showed significant expression (P < 0.05) in A549 tumors over in vitro culture, such as STAT1 and insulin-like growth factor-I (IGF1), and 7 showed significant expression in stromal tissue, such as (murine) EGF and plasminogen urokinase activator receptor (PLAUR).

We also searched for evidence of interactions occurring in vivo between human proteins expressed in xenograft tumors and mouse proteins expressed in the surrounding host tissue. We converted the set of 939 mouse genes overexpressed in diseased lung tissue over healthy controls (P < 0.05; fold change >1.5) to their human orthologues (791 genes). Of the set of 1,138 human xenograft tumor genes and the set of 791 diseased mouse lung genes, we selected 245 and 252 genes, respectively, that were annotated by GO as located outside of the cell (see Materials and Methods). Following removal of 32 genes from the mouse set that were shared by the human set, we found 33 HPRD protein-protein interactions in which one member was from the human set and the other member was from the mouse set. Simulation testing (using randomly selected gene sets of 245 human genes from the 2,388 extracellular genes represented on the Hu133A array and 220 mouse genes from the 2,817 extracellular human orthologue genes on the 430A array) yielded only 18 interactions on average (SD = 5.3).

Figure 5 graphically displays the HPRD interactions that were linked to the xenograft data as described above, with nodes representing genes and edges connecting any two genes for which the corresponding proteins are thought to physically interact. One subnetwork of interactions consists of several constituents of the extracellular matrix (both human and mouse derived); these interactions are understood to occur outside of the cell. Another subnetwork involves several genes with roles in hemostasis, neovascularization, and angiogenesis, including human-derived vascular endothelial growth factor (VEGF), plasminogen urokinase activator (PLAU), and fibrinogen and mouse-derived PLAUR and vascular-epithelium cadherin (CDH5). Other interactions involve human-derived cell adhesion molecules, including E-cadherin (CDH1), junction plakoglobin (JUP), and ß-catenin (CTNNB1) as well as murine polycystin-1 (PKD1); these interactions are thought to occur at the cell surface. Another set of interactions represents a growth factor signaling cascade initiated at the cell surface by murine EGF interacting with (human-derived) growth factor receptors ERBB2 (HER-2/neu) and ERBB3, from which the signal is transduced intracellularly into the neoplastic cell. Within the nucleus of the cancer cell, an extensive transcription regulation network involves several genes, including JUN, JUND, FOS, TOP2A, STAT1, and CTNNB1.



View larger version (96K):
[in this window]
[in a new window]
  		FIGURE 5. Graphical representation of protein-protein interactions involving both human genes overexpressed in A549 xenograft lung tumors and human orthologues of mouse genes overexpressed in lung tissue immediately adjacent to tumor. Nodes, genes; line between two nodes, the corresponding proteins can physically interact. Red, human genes overexpressed in xenograft tumors over culture (P < 0.05); yellow, human gene orthologues of mouse genes overexpressed in diseased lung over normal lung (P < 0.05); white, genes that were not significantly expressed. Active hubs, or genes connected to a significant number (P < 0.05) of other genes overexpressed in xenograft tumors, appear enlarged over the other nodes. Except for active hubs (fold change >1), human genes and mouse genes are expressed with fold changes >1.8 and 1.5, respectively, for the given comparison. The borders of some nodes are colored to reflect GO term annotation. Colored edges (other than gray) represent a common GO term annotation shared by both of the connected genes. Genes that showed higher expression (P < 0.05) in lung tumors with k-ras mutation relative to k-ras wild-type tumors (+). Genes that showed induction (fold change >1.8) in A549 cells under treatment of IFN-{gamma}, IL-4, or IL-6 appear in bold, in italics, and with asterisks, respectively. CSPG2 was targeted for further functional studies (Fig. 7).

 


View larger version (74K):
[in this window]
[in a new window]
  		FIGURE 7. CSPG2 has an essential role in A549 tumor growth in vivo. A. Expression values across various data sets (human lung, A549, and mouse lung) for genes that (a) are overexpressed in A549 orthotopic tumors over culture (P < 0.05, fold change >1.8) and (b) show higher expression (P < 0.01) in human lung tumors over normal lung (using the data set from ref. 13). Red, high expression; blue, low expression (relative to centroid mean), with color intensity indicating the degree of expression difference. Of the 109 genes that met these selection criteria, selected genes shown here also appear in the interaction network of Fig. 4. Genes marked with asterisk were also up-regulated in A549 s.c. tumors. B and C. Effect of RNA interference knockdown of CSPG2 and scrambled sequence (control) on growth of A549 tumor cells implanted s.c. into the back of athymic nude female mice (n = 5). B. Kaplan-Meier curves for time to tumor detection. C. Mean tumor volume; bars, SE. After day 23, CSPG2 group differs significantly in tumor volume from the control group (P < 0.05, two-sided t test).

 
A549 Tumors over In vitro Cultured A549 Cells Exhibit a Signature Profile for Human Adenocarcinoma Tumors with k-ras Mutation
The protein interaction network associated with A549 tumor formation (Fig. 5) includes a large number of interactions surrounding ERBB2 and EGF. Relative to other cell lines, A549 cells do not overexpress ERBB2 in vitro (9). Thus, a link between A549 and the ERBB2 pathway would not have been uncovered based on in vitro expression data. ERBB2 signaling occurs via the Ras-mitogen-activated protein kinase pathway leading to up-regulation of cyclin D1 via the JUN and FOS heterodimers (both genes being represented in the network) and cell cycle progression (10). A549 cells have an activating mutation in k-ras (homozygous at codon 12) that leads to overamplification of the EGF signal (11). We asked whether the in vivo–induced gene expression patterns in A549 were reflective in part of its k-ras genotype and whether these patterns had any correspondence with what may be found in human tumors. We relied on two independent public data sets of profiles of human lung adenocarcinomas for which the k-ras mutation status was known (12, 13). In both data sets, we found a high correspondence between the genes most highly expressed in lung tumors with k-ras mutation relative to k-ras wild-type tumors and the genes most highly expressed in A549 orthotopic xenografts over cultured cells (Fig. 6A). Although this correspondence was well within statistical significance in both data sets, differences in the actual Fisher's exact test P between the lung data sets should not be stressed, as they arise in part from the different platforms used (HuGeneFL versus U95Av2). Interestingly, in an independent xenograft tumor profile data set (5), a similar pattern of correspondence was observed for A549 s.c. tumors (Fig. 6B) but not for U118 (glial) cell-derived tumors (Fig. 6C), which are wild-type for k-ras. The protein interaction network of Fig. 5 indicates which of the genes represented were also found up-regulated in lung tumors with k-ras mutation in either of the data sets examined.



View larger version (59K):
[in this window]
[in a new window]
  		FIGURE 6. A. Correspondence of genes higher in A549 orthotopic tumors over A549 cultures with (a) genes showing higher expression in patient lung adenocarcinomas with k-ras mutation over lung tumors with wild-type k-ras genes (Lung k-ras mut) in both the Michigan data set (MI; ref. 12) and the Massachusetts data set (MA; ref. 13) and (b) genes showing induction in vitro under treatment by various cytokines. Genes in each comparison were ranked by the two-sided t test. The entire set of genes commonly represented between the orthotopic (U133A) data set and the other data sets was first determined (U133A versus HuGeneFL, 4,648 genes; U133A versus U95Av2, 7,788 genes). The top 1 to 1,000 A549 tumor genes were compared with the same number of top genes from the other data sets. The Fisher's exact test was used to determine whether the number of top genes shared significantly exceeds the number expected by chance. Arrow, nominal significance level at P < 0.05. B. Significance of the overlap between the top A549 s.c. tumor genes and the top genes from the other comparisons mentioned in A. S.c. tumor data are from the study in ref. 5. C. Significance of the overlap between the top U118 s.c. tumor genes and the top genes from the other comparisons. D. Significance of the overlap between the top lung k-ras mutant genes (Michigan data set) and the top genes from the other comparisons.

 
Responses of A549 to Cytokine Treatments In vitro Are Encoded in A549 Gene Expression Patterns In vivo
The tumor microenvironment contains various cytokines that play a role in tumor behavior. The protein interaction network of Fig. 5 includes several cytokine-related genes, such as interleukin (IL)-4 (IL-4), IL-4 receptor (IL-4R), IL-1{alpha} (IL-1A), IL-6 signal transducer or gp130 (IL-6ST), IL-11, and IL-13RA1. We hypothesized that cytokines contribute to the differential gene expression patterns between A549 tumors and in vitro cultures. We profiled A549 cells under separate treatments with growth stimulatory cytokines IL-4 and IL-6 and the inhibitory cytokine IFN-{gamma}. A high correspondence was found (Fig. 6A and B) between the genes most highly induced by a given cytokine and the top A549 tumor genes (induced in vivo over in vitro in both orthotopically implanted and s.c. tumors). Significantly enriched GO terms (P < 0.05) represented in the genes that showed induction (fold change >1.8) in A549 cells under the various cytokine treatments (IFN-{gamma}, 228 genes; IL-6, 230 genes; IL-4, 193 genes) reflect many of the significant terms found for A549 in vivo–induced genes, including "extracellular" (IFN-{gamma}, 31 genes; IL-4, 26 genes), "complement activity" (IFN-{gamma}, 5 genes; IL-6, 6 genes; IL-4, 4 genes), "immune response" (IFN-{gamma}, 45 genes; P < 1 x 10–9; IL-6, 26 genes), and "cell adhesion" (IL-4, 15 genes).

The protein interaction network of Fig. 5 indicates which of the genes represented were also induced in vitro by cytokines. IL-6 signal transduction through IL-6ST is mediated by the Ras-mitogen-activated protein kinase pathway (14). We see a high correspondence existing in our data (Fig. 6D) between genes overexpressed in lung tumors with k-ras mutation and IL-6-induced genes. IL-6ST (overexpressed in k-ras mutant lung tumors) is an active hub in our interaction network (with interactions to ERBB2 and ERBB3) but shows significant down-regulation at the mRNA level in vivo over in vitro (fold change <0.6), which may offer an example of interactions between proteins for which the corresponding transcripts are not coexpressed.

IFN-{gamma} inhibits cell growth by initiating cell cycle arrest and inducing apoptosis. IFN-{gamma} regulates gene transcription through activation of STAT1 via Janus-activated kinase (15). STAT1 is an active hub in our network and shows induction in A549 both in vivo and by IFN-{gamma} stimulation in vitro. As expected (15), elements of the class I MHC (HLA-A and HLA-E), which presents antigens to immune cells to facilitate immune surveillance, also show induction by IFN-{gamma} (Fig. 5). IFN-{gamma} and IL-4 exert opposite effects on the regulation of immune responses, with IFN-{gamma} repressing IL-4 expression (16) and IL-4 antagonizing macrophage activities induced or enhanced by IFN-{gamma} (17). Activation of the Ras-mitogen-activated protein kinase pathway in T-helper type 2 cells has been shown to alter IL-4R function directly (18), although our data show a correspondence between the mutant k-ras expression signature and IL-4-induced genes (Fig. 6D). No correspondence is evident between k-ras signaling and IFN-{gamma}-induced gene expression (Fig. 6D), which may be expected as the k-ras and IFN-{gamma} pathways have contrasting effects.

Up-Regulated Expression In vivo of Genes with Essential Roles in Tumor Formation
Genomic profiling of xenograft tumors may reveal critical genes and pathways on which the cancer cells rely in in vivo over in vitro growth conditions. A set of 109 genes were overexpressed in A549 orthotopic tumors over culture (P < 0.05; fold change >1.8) and showed higher expression (P < 0.01) in human lung tumors over normal lung in the public data set from ref. 13. The intersection of our xenograft data with human tumor data would allow us to better select for genes essential for tumor formation that are also relevant to human disease progression. Almost none of these 109 genes showed any increase in tumor stromal (mouse) tissue relative to normal (mouse) lung. This is a further indication that these genes are derived from A549 cells rather than stroma cells. Many genes well known for having essential roles in tumor growth were found in this list, including ERBB2 and VEGF. Of the 109 genes, 24 (of 74 represented in the s.c. tumor data set) were also up-regulated (P < 0.05) in A549 s.c. tumors over culture (Fig. 7A). Three of these genes (XBP1, COL1A1, and CSPG2) were also found to be similarly up-regulated in both (a) U118 brain s.c. xenograft tumors over culture (P < 0.05) and (b) human brain tumors over normal tissues (P < 0.01) in the public data set from ref. 19, suggesting a role for these genes in tumor formation in multiple cancer types. For further investigation, we selected CSPG2, given the limited available data regarding its role in tumor formation.

Versican, the protein encoded by CSPG2 (represented in the Fig. 4 interaction network), belongs to the family of hyaluronan-binding proteoglycans and has roles in cell adhesion, proliferation, migration, and extracellular matrix assembly. Expression of CSPG2 is associated with a proliferative cell phenotype and is often found in tissues exhibiting elevated proliferation, such as in development and in the stroma of a variety of tumors (20). Versican accumulation in human prostatic fibroblast cultures was shown to be enhanced by prostate cancer cell-derived transforming growth factor ß1, which observation has suggested to some that the host stromal cells and not the tumor cells secrete versican (21). Our xenograft profile data showed no significant increase in CSPG2 mRNA expression in mouse lung tissue adjacent to A549 tumors over uninvolved lung tissue but significant induction in tumor cells (Fig. 7A).

We sought to show an essential role for A549-derived CSPG2 in in vivo tumor growth. We carried out a knockdown of mRNA expression of CSPG2 in A549 using a replication-competent avian retroviral vector method of delivery of short hairpin RNA (shRNA). This method of RNA interference has a stable effect lasting for >3 weeks with >75% knockdown (ref. 22; data not shown) and specific to the A549 cells over the mouse host cells. There was no observable effect on in vitro growth for knockdown of CSPG2 compared with knockdown using a scrambled shRNA sequence (data not shown). The shRNA-treated cells were then implanted s.c. into nude mice. CSPG2 knockdown cells formed detectable s.c. tumors significantly later (P = 0.005) than control cells (Fig. 7B), resulting in smaller tumor growth in vivo (Fig. 7C), showing a role for tumor-derived (rather than stromal-derived) CSPG2 as a promoter of tumor growth.


   Discussion
Top
 Notes
 Abstract
 Introduction
 Results
 Discussion
Materials and Methods
 Acknowledgements
 References
 
Several findings have emerged from our gene expression profiling of A549 xenograft tumors, including (a) the set of genes induced in vivo over in vitro in A549 cells represents an extensive protein interaction network spanning a wide range of processes and pathways, including tumor-host interactions, cell differentiation, cell adhesion, hemostasis, growth factor–induced signal transduction, and gene transcription; (b) the in vivo–induced genomic response of A549 cells reflects its genotype in part, particularly its k-ras mutation status, as the A549 in vivo gene expression signature corresponds highly to a k-ras mutation signature found in human lung tumors; (c) treatment of A549 cells in vitro with various cytokines identified in the microenvironment of the orthotopic lung tumors results in a genomic response in cultured A549 cells that emulates the in vivo response in part, demonstrating how cells in vitro can be perturbed in specific ways to activate gene expression programs associated with tumor formation; and (d) CSPG2, a gene induced in vivo in A549 and overexpressed in human lung tumors, has an essential role in tumor growth in vivo but not in vitro.

Our studies show an intricate molecular interplay between the tumor and host microenvironments that likely regulates important aspects of tumor biology. Concerns regarding the potential confounding effect of mouse tissue when profiling xenograft tumors using human arrays to determine (human) cancer cell expression patterns is alleviated by our findings of tissue-specific, genotype-specific, and cytokine-induced responses of cells to the microenvironment. We may expect that different cell lines respond differently to the in vivo microenvironment. Consequently, xenograft tumor profiling studies on a larger scale, as for example with the entire NCI-60 set of cell lines, for which in vitro expression data are available (23), may be highly informative with respect to the extent to which these cell lines when grown in vivo may exhibit different properties compared with their in vitro growth. Although cell lines grown in vitro remain a powerful model for studying cancer, our study indicates that novel therapeutic targets or biomarkers may be found by superseding in vitro models and looking initially at in vivo models.


   Materials and Methods
Top
 Notes
 Abstract
 Introduction
 Results
 Discussion
 Materials and Methods
Acknowledgements
 References
 
Cell Lines and Culture Conditions
The A549 lung adenocarcinoma cell line was cultured at 37°C in a 6% CO2-humidified incubator in DMEM supplemented with 10% fetal bovine serum, 100 units/mL penicillin, and 100 units/mL streptomycin. The cells were passaged weekly on reaching confluence. A549 cells were incubated with or without IL-6 (10 ng/mg), IFN-{gamma} (100 ng/mL), or IL-4 (20 ng/mL) for 24 hours (37°C in 5% CO2) in DMEM without FCS.

Nude Mouse Orthotopic Xenograft Model
Six-week-old athymic nude female mice were obtained from National Cancer Institute (Frederick, MD); all animal care was provided in accordance with the procedures outlined in the Guide for the Care and Use of Laboratory Animals (National Academy of Sciences, 1996). Surgeries were done under sterile conditions and surgical instruments were steam sterilized before surgery. Mice were anesthetized with 2% isofluorane by inhalation. Following cleaning with 70% alcohol, a right posterolateral thoracic incision was made and the thoracic wall was exposed by blunt dissection of the muscles. The tip of a 0.5 inch, 27 gauge needle was advanced ~3 mm under visual control through the translucent pleura at the third intercostal space at the dorsal midaxillary line into the pleura cavity, where the A549 cells (106/20 µL PBS) containing 1 mg/mL Matrigel (BD Biosciences, Bedford, MA) were injected. Mice were euthanized at 7 weeks postimplant. Tumor tissue and lungs were harvested, snap frozen in chilled isopentane, and stored at –80°C. Tumors ranged in size between 0.5 and 10 mm at 7 weeks.

Tissue Processing and RNA Extraction
Frozen mouse tissue was embedded in OCT freezing medium (Miles Scientific, Naperville, IL) before cyrosectioning. Sections (5 µm thick) were prepared, stained by routine H&E, and evaluated by light microscopy. Areas of relatively pure tumor (Fig. 1A) as well as areas immediately adjacent to tumor (Fig. 1B) and normal lung tissue (Fig. 1C) were selected for RNA extraction and removed from the OCT block with a pointed scalpel. Tissues were homogenized in the presence of Trizol reagent (Life Technologies, Inc., Gaithersburg, MD). Total RNA was purified according to manufacturer's protocol and further purified by acid-phenol extraction and RNeasy spin columns (Qiagen, Valencia, CA). RNA quality was assessed by agarose gel electrophoresis in the presence of ethidium bromide.

Gene Expression Profiling
Four A549 orthotopic xenograft lung tumors obtained from different mice and four different A549 cell culture samples were each profiled using Affymetrix human U133A oligonucleotide arrays. From histologic analysis, the amount of mouse tissue in a xenograft tumor sample was established between 10% and 20%. Two samples of normal mouse lung tissue were also profiled with the U133A array (designed for human mRNAs) to assess the amount of hybridization that could be attributable to mouse mRNA in a xenograft tumor sample. Using Affymetrix mouse 430A arrays, four samples of lung tissue adjacent to tumor obtained from different xenograft mice and three samples of normal lung tissue from healthy mouse controls were each profiled. The amount of A549 cells in a diseased mouse lung sample was <20%. Two samples of A549 cell cultures were also profiled with the 430A array to assess the amount of hybridization that could be attributable to human mRNA in a diseased mouse lung sample. Two different cultures of A549 cells grown under each of the various cytokine treatments described above were profiled using Affymetrix HuGeneFL arrays. Preparation of mRNA, hybridization of the arrays, and computation of probe set intensities were as described previously (5). The exogenous probe set controls on the U133A array (probe sets that give constant hybridization from sample to sample) were used to determine scaling factors for comparing the U133A mouse lung profiles with the U133A A549 xenograft tumor and culture profiles. In the same way, the 430A A549 profiles were compared with the 430A mouse lung profiles. Complete data sets and analysis results are available at http://dot.ped.med.umich.edu:2000/pub/xeno/net/index.htm.

Statistical Analysis
Affymetrix intensity values <50 units were set to 50. As criteria for determining significant differences in mean gene mRNA expression levels between groups of samples, two-sample t tests were done on log-transformed data. For sets of genes selected as differing between two sets of samples, permutation testing estimated the number of genes that would be obtained for the same selection criteria by chance. For a given set of significantly expressed genes, a search was made within the set for significantly enriched GO annotation terms (5). Monte Carlo simulations were used to estimate the number of significant GO terms that could arise in a randomly selected set of genes of the same size as the given gene set (based on 1,000 tests).

Gene Expression Meta-analysis
Given two different comparisons for differential gene expression using independent data sets, we sought to determine the significance of the overlap shared between genes significant in one comparison and genes significant in the other. Genes in each data set were ranked based on their significance of expression either by the two-sided t test (xenograft tumor, human lung tumor, and cytokine treatment; see Results) or by the rank-sum test (Novartis atlas). In the Novartis mouse expression atlas (U74A), expression values <20 for any of the profiles of the tissue type under consideration were excluded from the analysis for that tissue. For two given ranked gene lists, the common represented gene population was determined. In the case where the two data sets considered were generated using the same Affymetrix array, all probe sets were used in the analysis. In different gene expression platforms being used between different data sets, a mapping between the two was made using the gene names; if a gene was represented more than once on a given platform, then the highest ranking for the gene was used. For a given number (ranging from 1 to 1,000) of top genes from each ranked list, the significance of shared gene overlap was determined by a one-sided Fisher's exact test.

Protein Interaction Network Analysis
A catalogue of >10,000 physical human protein-protein interactions was obtained from the HPRD (http://www.hprd.org, downloaded February 2004), each interaction being inferred previously by a manual search of the literature (8). Human orthologues of mouse genes were obtained from Homologene (ftp://ftp.ncbi.nih.gov/pub/HomoloGene/). For a given set of genes, the set of interactions for which both proteins were represented was obtained and Monte Carlo simulation testing assessed the number of interactions expected from a randomly selected gene set (based on 1,000 tests). We defined an extracellular gene as either a gene with GO annotation "extracellular," "cell adhesion," or "cell adhesion molecule activity" or a gene with GO annotation "membrane" but not "cytoplasm." Given two distinct sets of extracellular genes, the significance of the number of binary HPRD interactions in which genes from each set were represented was also assessed by Monte Carlo, the number of interactions between two randomly selected gene sets being tabulated 1,000 times. The statistical significance of "active hubs" (described in Results) for a given gene set were calculated by one-sided Fisher's exact tests, with Monte Carlo being used to determine the number of active hubs that may arise in a randomly selected set of genes (based on 1,000 tests). Graphical visualization of a set of interactions of interest was generated using the Pajek software package (http://vlado.fmf.uni-lj.si/pub/networks/pajek/).

RNA Interference–Mediated Inhibition of In vivo Tumor Growth
To inhibit the expression of specific genes of interest, an avian-specific, replication-competent, Gateway-compatible retroviral vector capable of expressing shRNA was used (22). Viral vectors were developed for RNA interference of a specific gene, whereby the shRNA was expressed under the control of the H1 promoter. shRNA sequences were chosen based on standard RNA interference guidelines, and all sequences were BLAST searched to ensure no homology to other genes. Viral stocks were generated and titers were determined as described in ref. 22. A stable clonal line from A549, A549-TVA, was created to express the avian retroviral receptor TVA, allowing cell-specific targeted infection of tumor cells.

PAGE-Northern blots were done to determine the delivery of shRNA from virally infected cells. Specific gene knockdown was verified by Northern blot analysis of infected A549-TVA cells in vitro. Different groups of five nude mice were implanted s.c. with 105 A549-TVA cells infected with virus expressing shRNA for a given gene. A control group of mice were implanted with A549-TVA infected with virus expressing a scrambled shRNA sequence.


   Acknowledgements
Top
 Notes
 Abstract
 Introduction
 Results
 Discussion
 Materials and Methods
 Acknowledgements
References
 
We thank Michelle Lizyness, Angela Smith, and Barbara Lamb for excellent technical expertise regarding tissue processing, RNA extraction, and Affymetrix profiling.


   Notes
Top
 Notes
Abstract
 Introduction
 Results
 Discussion
 Materials and Methods
 Acknowledgements
 References
 
Grant support: Michigan Life Sciences Corridor grant MEDC-238 and National Cancer Institute institutional training grant CA09676 (C.J. Creighton).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Received 11/15/04; revised 12/31/04; accepted 2/ 8/05.


   References
Top
 Notes
 Abstract
 Introduction
 Results
 Discussion
 Materials and Methods
 Acknowledgements
 References
 

  1. Hanahan D, Weinberg RA. The hallmarks of cancer. Cell 2000;100:57–70.[CrossRef][Medline]
  2. Webb CP, Vande Woude GF. Genes that regulate metastasis and angiogenesis. J Neurooncol 2000;50:71–87.[CrossRef][Medline]
  3. Bissell MJ, Radisky D. Putting tumours in context. Nat Rev Cancer 2001;1:46–54.[CrossRef][Medline]
  4. Clark EA, Golub TR, Lander ES, Hynes RO. Genomic analysis of metastasis reveals an essential role for RhoC. Nature 2000;406:532–5.[CrossRef][Medline]
  5. Creighton C, Kuick R, Misek DE, et al. Profiling of pathway-specific changes in gene expression following growth of human cancer cell lines transplanted into mice. Genome Biol 2003;4:R46.[CrossRef][Medline]
  6. Harris AL. Hypoxia—a key regulatory factor in tumour growth. Nat Rev Cancer 2002;2:38–47.[CrossRef][Medline]
  7. Su AI, Cooke MP, Ching KA, et al. Large-scale analysis of the human and mouse transcriptomes. Proc Natl Acad Sci U S A 2002;99:4465–70.[Abstract/Free Full Text]
  8. Peri S, Navarro JD, Amanchy R, et al. Development of Human Protein Reference Database as an initial platform for approaching systems biology in humans. Genome Res 2003;13:2363–71.[Abstract/Free Full Text]
  9. Bunn PA Jr, Helfrich B, Soriano AF, et al. Expression of Her-2/neu in human lung cancer cell lines by immunohistochemistry and fluorescence in situ hybridization and its relationship to in vitro cytotoxicity by trastuzumab and chemotherapeutic agents. Clin Cancer Res 2001;7:3239–50.[Abstract/Free Full Text]
  10. Maruta H, Burgess AW. Regulation of the Ras signalling network. Bioessays 1994;16:489–96.[CrossRef][Medline]
  11. Jiang Y, Cui L, Yie TA, Rom WN, Cheng H, Tchou-Wong KM. Inhibition of anchorage-independent growth and lung metastasis of A549 lung carcinoma cells by I{kappa}Bß. Oncogene 2001;20:2254–63.[CrossRef][Medline]
  12. Beer DG, Kardia SL, Huang CC, et al. Gene-expression profiles predict survival of patients with lung adenocarcinoma. Nat Med 2002;8:816–24.[Medline]
  13. Bhattacharjee A, Richards WG, Staunton J, et al. Classification of human lung carcinomas by mRNA expression profiling reveals distinct adenocarcinoma subclasses. Proc Natl Acad Sci U S A 2001;98:13790–5.[Abstract/Free Full Text]
  14. Naka T, Nishimoto N, Kishimoto T. The paradigm of IL-6: from basic science to medicine. Arthritis Res 2002;4 Suppl 3:S233–42.
  15. Schroder K, Hertzog PJ, Ravasi T, Hume DA. Interferon-{gamma}: an overview of signals, mechanisms and functions. J Leukoc Biol 2004;75:163–89.[Abstract/Free Full Text]
  16. Elser B, Lohoff M, Kock S, et al. IFN-{gamma} represses IL-4 expression via IRF-1 and IRF-2. Immunity 2002;17:703–12.[CrossRef][Medline]
  17. Paludan SR. Interleukin-4 and interferon-{gamma}: the quintessence of a mutual antagonistic relationship. Scand J Immunol 1998;48:459–68.[CrossRef][Medline]
  18. Yamashita M, Kimura M, Kubo M, et al. T cell antigen receptor-mediated activation of the Ras/mitogen-activated protein kinase pathway controls interleukin 4 receptor function and type-2 helper T cell differentiation. Proc Natl Acad Sci U S A 1999;96:1024–9.[Abstract/Free Full Text]
  19. Rickman DS, Bobek MP, Misek DE, et al. Distinctive molecular profiles of high-grade and low-grade gliomas based on oligonucleotide microarray analysis. Cancer Res 2001;61:6885–91.[Abstract/Free Full Text]
  20. Wight TN. Versican: a versatile extracellular matrix proteoglycan in cell biology. Curr Opin Cell Biol 2002;14:617–23.[CrossRef][Medline]
  21. Sakko AJ, Ricciardelli C, Mayne K, Tilley WD, Lebaron RG, Horsfall DJ. Versican accumulation in human prostatic fibroblast cultures is enhanced by prostate cancer cell-derived transforming growth factor ß1. Cancer Res 2001;61:926–30.
  22. Bromberg-White JL, Webb CP, Patacsil VS, Miranti CK, Williams BO, Holmen SL. Delivery of short hairpin RNA sequences by using a replication-competent avian retroviral vector. J Virol 2004;78:4914–6.[Abstract/Free Full Text]
  23. Scherf U, Ross DT, Waltham M, et al. A gene expression database for the molecular pharmacology of cancer. Nat Genet 2000;24:236–44.[CrossRef][Medline]



This article has been cited by other articles:


Home page
 Mol. Cell. ProteomicsHome page
E. Bjorling, C. Lindskog, P. Oksvold, J. Linne, C. Kampf, S. Hober, M. Uhlen, and F. Ponten
A Web-based Tool for in Silico Biomarker Discovery Based on Tissue-specific Protein Profiles in Normal and Cancer Tissues
Mol. Cell. Proteomics, May 1, 2008; 7(5): 825 - 844.
[Abstract] [Full Text] [PDF]

Home page
 Clin. Cancer Res.Home page
J. S. Fridman, E. Caulder, M. Hansbury, X. Liu, G. Yang, Q. Wang, Y. Lo, B.-B. Zhou, M. Pan, S. M. Thomas, et al.
Selective Inhibition of ADAM Metalloproteases as a Novel Approach for Modulating ErbB Pathways in Cancer
Clin. Cancer Res., March 15, 2007; 13(6): 1892 - 1902.
[Abstract] [Full Text] [PDF]

Home page
 Endocr Relat CancerHome page
C L Wilson, A H Sims, A Howell, C J Miller, and R B Clarke
Effects of oestrogen on gene expression in epithelium and stroma of normal human breast tissue.
Endocr. Relat. Cancer, June 1, 2006; 13(2): 617 - 628.
[Abstract] [Full Text] [PDF]

Home page
 Cancer Res.Home page
C. J. Creighton, A. M. Hilger, S. Murthy, J. M. Rae, A. M. Chinnaiyan, and D. El-Ashry
Activation of Mitogen-Activated Protein Kinase in Estrogen Receptor {alpha}-Positive Breast Cancer Cells In vitro Induces an In vivo Molecular Phenotype of Estrogen Receptor {alpha}-Negative Human Breast Tumors.
Cancer Res., April 1, 2006; 66(7): 3903 - 3911.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Creighton, C. J.
Right arrow Articles by Hanash, S. M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Creighton, C. J.
Right arrow Articles by Hanash, S. M.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Meeting Abstracts Online