identification of novel metastasis suppressor signaling pathways for breast cancer

6

Click here to load reader

Upload: marsha-rich

Post on 01-Oct-2016

217 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: Identification of novel metastasis suppressor signaling pathways for breast cancer

© 2012 Landes Bioscience.

Do not distribute.

review

Cell Cycle 11:13, 2452-2457; July 1, 2012; © 2012 Landes Bioscience

2452 Cell Cycle volume 11 issue 13

Cancer Phenotypes Result from Context-Dependent Signaling Pathways

The application of genomics and high-throughput DNA sequenc-ing technologies to cancer has produced a formidable amount of data that highlights the complexity and heterogeneity between cancer genomes. These studies showed that the genetic origins of most cancers are complicated and unclear. A wide variety of rare mutations cause cancer in any one tumor type, but muta-tions in the same genetic pathways seem to occur in multiple tumor types. Thus, while any individual gene may or may not be mutated, it appears that a limited number of signaling pathways are deregulated in cancer.1-3

*Correspondence to: Andy J. Minn and Marsha Rich Rosner; Email: [email protected] and [email protected]: 04/26/12; Accepted: 05/03/12http://dx.doi.org/10.4161/cc.20624

Cancer lethality is mainly caused by metastasis. Therefore, understanding the nature of the genes involved in this process has become a priority. Given the heterogeneity of mutations in cancer cells, considerable focus has been directed toward characterizing metastasis genes in the context of relevant signaling pathways rather than treating genes as independent and equal entities. One signaling cascade implicated in the regulation of cell growth, invasion and metastasis is the MAP kinase pathway. raf kinase inhibitory protein (rKiP) functions as an inhibitor of the MAP kinase pathway and is a metastasis suppressor in different cancer models. By utilizing statistical analysis of clinical data integrated with experimental validation, we recently identified components of the rKiP signaling pathway relevant to breast cancer metastasis. Using the rKiP pathway as an example, we show how prior biological knowledge can be efficiently combined with genome-wide patient data to identify gene regulatory mechanisms that control metastasis.

Identification of novel metastasis suppressor signaling pathways for breast cancer

Andy J. Minn,1,* elena Bevilacqua,2 Jieun Yun3 and Marsha rich rosner2,*

1Department of radiation Oncology and Abramson Family Cancer research institute; Perelman School of Medicine; University of Pennsylvania; Philadelphia, PA USA; 2Ben May Department for Cancer research; University of Chicago; Chicago, iL USA; 3Bioevaluation Center; Korea research institute of Bioscience and Biotechnology;

Cheongwon, Chungbuk Korea

Keywords: breast cancer, gene signature, metastasis, invasion, Raf kinase inhibitory protein, signaling pathway, gene set enrichment analysis, gene set analysis, random forests

Abbreviations: RKIP, Raf kinase inhibitory protein; GSEA, gene set enrichment analysis; GSA, gene set analysis; ORA, over-representation analysis; RF, random forests; GRN, genetic regulatory networks; RPMS, RKIP pathway metastasis signature;

HMGA2, high mobility group AT-Hook 2; BACH1, BTB and CNC homology 1; Let-7-TG, let-7 target gene; BrCa, breast cancer

For this reason there has been considerable focus on charac-terizing tumor genes in the context of signaling pathways rather than as single agents. Signal transduction cascades maintain cellular robustness in at least three discrete ways: (1) fidelity of signal transmission; (2) adaptability to changes in the signal-ing environment; and (3) specificity of signaling. In order to understand the regulatory relationship between genes we need to consider the signaling environment—what external signals are being communicated by the microenvironment, and what inter-nal signals are activated within the cell. The identification of key regulators of signaling pathways activated during tumorigenesis must thus take into account the nature of the specific tissues and the cells of interest and also reflect the discrete cellular signaling environments. Together, the signaling pathways and the cellular context determine the biological state of a cell.

Metastasis Gene Signatures Controlled by Specific Signaling Pathways

In most cases, lethality in cancer patients is caused by metastasis rather than the primary tumor. Understanding the nature of the genes involved in the regulation of metastatic spread is thus of major importance for improving cancer survival. However, like most phenotypes, metastasis is characterized by many distinct bio-logical states that are driven by equally discrete but incompletely characterized pathways. Ignoring the underlying biological het-erogeneity can prevent the discovery of the leading regulators of metastasis in each specific context. For example breast cancer is comprised of multiple molecular subtypes. In the past, attempts to discover metastasis genes have utilized a top-down approach that groups all breast cancers together and exhaustively searches for genes associated with metastasis. This approach has largely failed to identify mediators of distant spread.

In recent years, novel approaches driven by experimental biol-ogy have been developed, and they have led to the definition of signatures for organ-specific metastasis.4,5 Many of the genes that comprise these metastasis signatures have been shown to drive specific steps in the metastatic cascade, and multiple genes from

Page 2: Identification of novel metastasis suppressor signaling pathways for breast cancer

© 2012 Landes Bioscience.

Do not distribute.

www.landesbioscience.com Cell Cycle 2453

review review

using simple correlation coefficients to construct gene-gene con-nections; however, such methods are not well suited for nonlin-ear relationships or the dissection of higher-order interactions. Recently, ensemble regression tree-based methods have been shown to perform very well in inferring regulatory networks.18 These multivariate methods do not assume linear relationships between genes and evaluate all gene-gene interactions in the net-work or pathway. The strength of the association between genes is measured by a variable importance score, which can be used to rank which genes may be regulating each other.

Regardless of the statistical method, signal-to-noise due to the high dimensionality of genomics data is a major barrier to accurate deduction. One powerful strategy to overcome this issue is to incorporate prior biological knowledge into network and pathway modeling as filters and integrators.19 Filters are used to decrease noise by removing variables that are unrelated and/or biologically uninteresting to the hypothesis at hand. For exam-ple, if one is interested in discovering which of several hundred genes are most likely to be controlled by a microRNA-mediated signaling pathway, removing all genes not predicted to be a target of microRNA would be a useful filter. Integrators increase statis-tical power by aggregating signals together. In this way, signals that are individually weak can together yield a larger effect and be more robust. GSEA takes advantage of this concept and com-bines gene-level statistics into a pathway-level statistic in order to calculate whether a group of genes, rather than the individual genes, are significantly associated with a phenotype or outcome. Another example of an integrator is a meta-gene. Meta-genes combine the individual expression of a group of genes into a sin-gle value. Thus, both filters and integrators can be an effective way to deal with signal-to-noise issues inherent in high-dimen-sional data.

Connecting Signaling Pathways to Gene Signatures

In order to discover novel signaling pathways that regulate metas-tasis, we utilized both experimental data and many of the com-putational strategies described above. The experimental data provided partial pathway information and were used as prior knowledge to define filters that focused the search space. Then, a form of GSEA called gene set analysis (GSA)20 was combined with a tree-based method called random forests (RF).21 In the first step, GSA was used to determine if a metastasis regulator likely controls a particular pathway that is represented by a set of genes. In the second step, RF was used to test the individual components of the pathway in order to determine which genes likely form a gene-gene interaction with the regulator. To ensure robust signals as the pathway is being dissected, we employed integrators in the form of meta-genes. Moreover, to make the analysis relevant to disease, we utilized gene expression data from primary human breast cancer. Finally, experimental validation using breast cancer cell lines provided confirmation and mecha-nistic insight into the predicted gene-gene interactions. Below, we describe how we used this integrative approach to identify a new signaling pathway that controls breast cancer metastasis.

these signatures are often required for successful colonization of distant organs. For example, a lung metastasis signature for breast cancer5 contains genes that promote extravasation,6 tumor initiation,7 maintenance of a metastatic niches8 and immune-mediated survival signals.9 Other genes in the signature seem to provide redundancy for the same function. Thus, metastasis gene signatures are comprised of multiple mediators that serve comple-mentary and sometimes redundant functions that together drive metastatic spread.

Recent evidence suggests that a limited number of signaling pathways rather than a collection of independent regulators may control metastasis gene signatures. Several studies have revealed that a limited number of transcription factors or chromatin modifiers10 and microRNAs11 or other non-coding RNAs12 may broadly regulate many genes that comprise metastasis signa-tures.13-16 These observations suggest that metastasis genes are co-regulated, and a few definable signaling pathways may control many genes within a single metastasis signature.

In summary, metastasis is a major cancer-related problem. Recently identified gene signatures not only identify genes that control multiple steps in the metastatic spread, but may also cap-ture biological states associated with context-dependent metas-tasis signaling pathways. The signaling pathways that control multiple components of a metastasis signature would be an effec-tive way to control distant spread of tumor cells. However, devel-oping methods to identify the signaling pathways is a significant challenge.

Strategies and Challenges to Computationally Predict Cancer Pathways

Common approaches to identify the pathways that character-ize a gene expression signature or any other list of differentially expressed genes include over-representation analysis (ORA) and gene set enrichment analysis (GSEA).17 Both analyses are used to investigate whether the identified gene group (i.e., the sig-nature) is statistically enriched in genes from predefined gene sets, which are assembled by manually curated pathways or gene ontologies. In ORA, over-representation is determined by using counting statistics similar to Chi square analysis. In contrast, GSEA takes all genes in a predefined gene set and calculates a gene-level statistic for each gene. Next, a pathway-level statis-tic is derived, by combining the gene-level statistics, and used to determine whether the pathway represented by the gene set is significantly enriched. In this way, GSEA considers coordi-nate expression changes of all genes in a pathway and has better statistical power. Although popular and useful, both ORA and GSEA suffer from the limitation that pathway topology, or how genes are connected to one another to transmit signals, is not accounted for.

The principal goal of creating genetic regulatory networks (GRNs) using gene expression data are to predict how genes are connected by determining statistical relationships. Since pathways can be considered subsets of GRNs, similar principles can be adapted to pathway analysis. These approaches include

Page 3: Identification of novel metastasis suppressor signaling pathways for breast cancer

© 2012 Landes Bioscience.

Do not distribute.

2454 Cell Cycle volume 11 issue 13

been previously described4 using the breast cancer cell lines we used in the RKIP study, we reasoned that RKIP and let-7 would control some of the BMS genes. Using microRNA target predic-tion databases, it appeared that BMS genes were unlikely to be direct targets of let-7, suggesting, rather, that a let-7 target gene could act as a regulator of downstream BMS genes. Based on the prior knowledge gained from this partial pathway, we employed a strategy to computationally predict the missing signaling path-ways components downstream of RKIP and let-7 that converge on critical BMS genes (summarized in Fig. 1).

Since RKIP acts as a metastasis suppressor through positive regulation of mature let-7 microRNA expression, we hypothe-sized that high levels of RKIP expression would induce let-7 and, in turn, decrease the expression of let-7 target genes. In order to test the importance of this regulatory pathway in primary human tumors and to identify relevant let-7 targets, we used the filter strategy to assemble a set of 38 high-confidence let-7 target genes. As a first step, we utilized a data set of 443 breast tumors (BrCa443) to show that expression of RKIP negatively associates with expression of the let-7 targets using GSA. As a second step, we used RF to measure which individual let-7 target genes may be most likely regulated by RKIP and let-7. This analysis revealed that two transcription regulators had particularly high variable importance scores, which is indicative of possible gene-gene inter-action. The first, high mobility group AT-Hook 2 (HMGA2), is a chromatin remodeling factor and known let-7 target.36 Elevated HMGA2 expression has been observed in a variety of cancers and correlated with cancer progression, identifying it as a potential contributor to the disease. The second gene of interest, BTB and

RKIP Suppress Invasion and Metastasis in Multiple Cancer

Models

Genes that mediate the initial forma-tion and progression of tumors may not be specifically the same genes that medi-ate metastatic spread, even if they can be considered prerequisites to metastasis. Metastatic spread requires loss or gain of function of an additional set of genes. These “metastasis genes” specifically enable or inhibit one or multiple steps of the process that lead to metastasis forma-tion: invasion, intravasation, circulation, extravasation and colonization of the sec-ondary site.

The MAP kinase signaling pathway is a highly conserved signaling cascade that has been implicated in the regulation of cell growth, differentiation, invasion and metastasis. A key regulator of the MAPK pathway is Raf kinase inhibitory protein (RKIP), which has been shown to act as a metastasis suppressor in prostate and breast cancer models among others.22,23 Recent studies suggest that RKIP plays a key role in maintaining checkpoint control and cellular homeo-stasis for MAP kinase, G-protein-coupled receptor and NFκB signaling cascades. RKIP binds to Raf, inhibiting MAPK signal-ing; phosphorylation of RKIP at S153 releases RKIP from Raf-124,25 but potentiates RKIP inhibition of GRK2, a kinase that downregulates G protein-coupled receptors.26 RKIP is also an inhibitor of NFκB activation.27 Loss of RKIP leads to suppres-sion of the spindle checkpoint, chromosomal aberrations and genomic instability.28 RKIP expression is decreased in many solid tumors, including breast, prostate, colorectal and hepatocellular carcinoma, and RKIP is prognostic for survival of prostate and colon tumor patients.29-34 Consistent with this correlation, RKIP has been implicated as a suppressor of lung metastasis of prostate tumor cells.32

We recently identified a signaling pathway by which RKIP inhibits breast cancer invasion, intravasation and bone metastasis in an orthotopic murine model.22 The mechanism involves inhi-bition of MAPK, leading to suppression of LIN28 and enhanced processing of the microRNA let-7. Specifically, inhibition of the Raf/MEK/MAP kinase cascade by RKIP leads to inhibition of Myc activation. Myc is a transcriptional activator of LIN28, which, in turn, inhibits let-7 maturation.

Using RKIP to Identify a Novel Metastasis Suppressor Signaling Cascade for Breast Cancer

As a follow-up to our initial RKIP study, we sought to identify a signaling pathway that explains how RKIP and let-7 regulate metastasis.35 Because a bone metastasis signature (BMS) had

Figure 1. Schematic representation of our strategy to identify the signaling pathway compo-nents downstream of rKiP that lead to BMS genes.

Page 4: Identification of novel metastasis suppressor signaling pathways for breast cancer

© 2012 Landes Bioscience.

Do not distribute.

www.landesbioscience.com Cell Cycle 2455

that include MMP1, OPN and CXCR4. All of these predictions were statistically validated using a second independent data set of 871 breast tumors (BrCa871) prior to experimental validation.

Experimental Validation of the RKIP Signaling Pathway

Despite independent statistical validation of the predictions, the critical part of our study was to provide in vitro and in vivo vali-dation. Using multiple breast cancer cell lines, we demonstrated that RKIP positively regulates let-7 expression and negatively regulates HMGA2 and BACH1 expression. We experimentally tested the relationship of RKIP and let-7 to a subset of BMS genes (MMP1, CXCR4 and OPN) and found that both RKIP and let-7 negatively regulate expression of these three BMS genes in several breast cancer cell lines. Ectopic expression of RKIP in MDA-MB-231-derived 1833 and MDA-MB-436 cells, which are aggressive triple-negative breast tumor lines, functionally suppressed invasion and metastasis in culture and in mouse xenografts, respectively. Conversely, loss of RKIP decreased let-7 expression and potentiated invasion and metastasis by MDA-MB-435 cells. The suppressive effect of RKIP on invasion and metastasis can be rescued by overexpression of the three BMS genes, which act more potently when expressed in concert than when expressed individually.

We also validated the functional role of BACH1 and HMGA2 as regulators of the BMS genes and promoters of invasion and metastasis. We showed that BACH1 transcriptionally activates MMP1; HMGA2 induces CXCR4; and both BACH1 and HMGA2 stimulate OPN. Suppression of BACH1 and HMGA2, either together or separately, decreased invasion in vitro and intravasation and metastasis in vivo. By overexpressing the three BMS genes MMP1, CXCR4 and OPN, we effectively reversed the decrease in invasion and metastasis that we observed follow-ing HMGA2 and/or BACH1 depletion. These results establish BACH1 and HMGA2 as key mediators of the RKIP signaling pathway via a let-7 mechanism and demonstrate the functional impact of BACH1 and HMGA2 on expression of the BMS genes, MMP1, CXCR4 and OPN (see Fig. 2 for signaling pathway). Finally the components of the RKIP signaling cascade we iden-tified function in a coordinated manner and specifically define tumor-stroma interactions in the context of breast cancer metas-tasis to the bone microenvironment.37

The RKIP Signaling Pathway Identifies Patients at Risk for Metastasis

As described above, we used clinical gene expression data com-bined with experimental validation with the aim of generating a novel signaling cascade relevant to disease. Therefore, to investi-gate the clinical significance of our findings, we also generated a gene signature to identify a subset of patient tumors that express our RKIP-directed metastasis pathway. The RKIP cascade we validated experimentally was a seven-gene pathway. To convert this pathway into a gene signature, which we call the RKIP path-way metastasis signature (RPMS), we included RKIP, the let-7

CNC homology 1 (BACH1), is a transcription factor and novel let-7 target. The role of BACH1 in the context of cancer is largely unknown.

Since BACH1 is a novel let-7 target gene with largely unknown function, we focused on its potential role in the pathway. As a transcription factor, it may regulate hundreds of target genes. Therefore, we repeated our two-step GSA/RF approach to first test if BACH1 target genes are associated with RKIP and let-7. As a modification to the GSA step, we introduced an integrator strategy for let-7 activity. As modulation by microRNAs tends to result in small but significant changes in target gene expression, signal robustness was increased using a meta-gene for the let-7 targets (LET7-TG). The ability of this integrator for let-7 to accu-rately function as a surrogate for let-7 activity was validated using direct probes for let-7e and let-7g. Using GSA, we confirmed that BACH1 target genes positively associated with LET7-TG, imply-ing a negative regulation by let-7. Next, we applied the meta-gene integrator for BACH1 (BACH1-TG) and found that it strongly associated with the BMS genes. In the RF step, matrix metallo-peptidase 1 (MMP1) was identified as the top BACH1 target gene and, thus, the most likely involved in propagating the pathway toward the BMS. Consistent with this, MMP1 is known to be a key BMS gene.4 Finally, the two-step GSA/RF approach also pre-dicted other BMS genes on the RKIP pathway, including osteo-pontin (OPN) and chemokine (C-X-C motif) receptor 4 (CXR4).

In total, our approach utilized GSA to test for enrichment of groups of genes controlled by a regulator, and RF to rank individ-ual genes most likely to biologically interact with upstream regu-lators. Prior knowledge, based on experimental results, partially elucidated the pathway and was used to create filters and integra-tors. Importantly, this knowledge also allowed us to initiate the computational approach using RKIP and to direct it toward the BMS. As a result, the analysis suggested that RKIP inhibits metas-tasis by regulating, via let-7, HMGA2 and the novel let-7 target BACH1. In turn, HMGA2 and BACH1 controls key BMS genes

Figure 2. rKiP signaling pathway regulates breast cancer metastasis to the bone. Lght blue rectangles are the components of the pathway that are also part of the rPMS. Light blue ovals are the meta-genes that were added to the signaling pathway to make the rPMS.

Page 5: Identification of novel metastasis suppressor signaling pathways for breast cancer

© 2012 Landes Bioscience.

Do not distribute.

2456 Cell Cycle volume 11 issue 13

Because the RPMS was computationally predicted and experi-mentally validated, we believe that this serves as an example of how primary cancer expression data can be utilized to infer sig-naling pathway components. In the case of the RPMS, we chose to focus on a single signaling pathway. An alternative possibility is to use gene expression data from patient tumors to determine how an individual signature is regulated by two different mas-ter regulators to control a common outcome such as tumor size. Other possible extensions of our approach or similar approaches would include predicting entire biological networks. However, for this to be successful and generalizable, numerous statistical challenges would have to be addressed, including how to select thresholds for inclusion of likely gene-gene interactions.

In conclusion, we would like to stress that the relationships we define with this approach are important in the context of a spe-cific signaling environment and may not be the only way that a specific component of the pathway is regulated. We showed here that the RKIP signaling pathway regulates breast cancer metasta-sis; however, some of the genes we identified are likely important in other signaling contexts as well. Although the RPMS identifies a limited subset of breast cancer patients at high risk for metasta-sis, it also identifies genes that are relevant as potential targets for this population. Thus, the use of signatures for disease-relevant signaling pathways can simultaneously serve as biomarkers and targets for therapy.

Acknowledgements

The work described here was supported by NIH grants NS33858 and CA112310 to M.R.R. A.J.M. is supported by the Department of Defense grant W81XWH-09-1-0339 and the W.W. Smith Charitable Trust. We thank Eva Eves and Casey Frankenberger for helpful comments.

meta-gene, the BACH1 meta-gene, HMGA2, MMP1, CXCR4 and OPN (Fig. 2).

Using the RPMS, we were able to identify a subset (~5%) of a heterogeneous patient population that express above median levels for all of these markers (except RKIP which would be below median) and have significantly lower survival rates. Interestingly, each of the RPMS genes alone fail or only poorly stratify patients based on metastasis-free survival; only the com-bination of the genes comprising the RPMS effectively separates patients through synergistic rather than additive interactions that presumably reflect their mechanistic relationships. In total, these results argue that the RPMS can identify an aggressive subset of breast cancer patients with tumors that may utilize the RKIP/let-7 metastasis pathway to drive distant spread.

Significance

With 200,000 new cases of breast cancer per year in the United States, stringent application of the RPMS allows us to identify at least 10,000 patients with tumors expressing this pathway. Identification using RPMS could define novel therapeutic targets as well as lead to more focused clinical trials for these patient pop-ulations. Moreover, pathway-based signatures such as the RPMS can be used to understand the complexity of the underlying bio-logical processes and increase our ability to interpret the mean-ing of certain metastasis-specific characteristics in these patients. Understanding pathway deregulation in breast cancer cell lines and corresponding tumors should help to predict drug sensitivity. Thus, although many gene signatures can predict metastatic risk, the RPMS is unique, because it is based on building mechanistic relationships between genes that serve as markers for particular signaling environments within tumors.

References1. Bozic I, Antal T, Ohtsuki H, Carter H, Kim D, Chen

S, et al. Accumulation of driver and passenger muta-tions during tumor progression. Proc Natl Acad Sci USA 2010; 107:18545-50; PMID:20876136; http://dx.doi.org/10.1073/pnas.1010978107.

2. Jones S, Zhang X, Parsons DW, Lin JC, Leary RJ, Angenendt P, et al. Core signaling pathways in human pancreatic cancers revealed by global genomic analyses. Science 2008; 321:1801-6; PMID:18772397; http://dx.doi.org/10.1126/science.1164368.

3. Wood LD, Parsons DW, Jones S, Lin J, Sjöblom T, Leary RJ, et al. The genomic landscapes of human breast and colorectal cancers. Science 2007; 318:1108-13; PMID:17932254; http://dx.doi.org/10.1126/sci-ence.1145720.

4. Kang Y, Siegel PM, Shu W, Drobnjak M, Kakonen SM, Cordón-Cardo C, et al. A multigenic program mediating breast cancer metastasis to bone. Cancer Cell 2003; 3:537-49; PMID:12842083; http://dx.doi.org/10.1016/S1535-6108(03)00132-6.

5. Minn AJ, Gupta GP, Siegel PM, Bos PD, Shu W, Giri DD, et al. Genes that mediate breast cancer metastasis to lung. Nature 2005; 436:518-24; PMID:16049480; http://dx.doi.org/10.1038/nature03799.

6. Gupta GP, Nguyen DX, Chiang AC, Bos PD, Kim JY, Nadal C, et al. Mediators of vascular remodelling co-opted for sequential steps in lung metastasis. Nature 2007; 446:765-70; PMID:17429393; http://dx.doi.org/10.1038/nature05760.

7. Gupta GP, Perk J, Acharyya S, de Candia P, Mittal V, Todorova-Manova K, et al. ID genes mediate tumor reinitiation during breast cancer lung metas-tasis. Proc Natl Acad Sci USA 2007; 104:19506-11; PMID:18048329; http://dx.doi.org/10.1073/pnas.0709185104.

8. Oskarsson T, Acharyya S, Zhang XH, Vanharanta S, Tavazoie SF, Morris PG, et al. Breast cancer cells produce tenascin C as a metastatic niche compo-nent to colonize the lungs. Nat Med 2011; 17:867-74; PMID:21706029; http://dx.doi.org/10.1038/nm.2379.

9. Chen Q, Zhang XH, Massagué J. Macrophage bind-ing to receptor VCAM-1 transmits survival signals in breast cancer cells that invade the lungs. Cancer Cell 2011; 20:538-49; PMID:22014578; http://dx.doi.org/10.1016/j.ccr.2011.08.025.

10. Han HJ, Russo J, Kohwi Y, Kohwi-Shigematsu T. SATB1 reprogrammes gene expression to pro-mote breast tumour growth and metastasis. Nature 2008; 452:187-93; PMID:18337816; http://dx.doi.org/10.1038/nature06781.

11. Tavazoie SF, Alarcón C, Oskarsson T, Padua D, Wang Q, Bos PD, et al. Endogenous human microR-NAs that suppress breast cancer metastasis. Nature 2008; 451:147-52; PMID:18185580; http://dx.doi.org/10.1038/nature06487.

12. Gupta RA, Shah N, Wang KC, Kim J, Horlings HM, Wong DJ, et al. Long non-coding RNA HOTAIR reprograms chromatin state to promote cancer metas-tasis. Nature 2010; 464:1071-6; PMID:20393566; http://dx.doi.org/10.1038/nature08975.

13. Ma L, Weinberg RA. Micromanagers of malignancy: role of microRNAs in regulating metastasis. Trends Genet 2008; 24:448-56; PMID:18674843; http://dx.doi.org/10.1016/j.tig.2008.06.004.

14. Nguyen DX, Chiang AC, Zhang XH, Kim JY, Kris MG, Ladanyi M, et al. WNT/TCF signaling through LEF1 and HOXB9 mediates lung adenocarcinoma metastasis. Cell 2009; 138:51-62; PMID:19576624; http://dx.doi.org/10.1016/j.cell.2009.04.030.

15. Valastyan S, Reinhardt F, Benaich N, Calogrias D, Szász AM, Wang ZC, et al. A pleiotropically acting microRNA, miR-31, inhibits breast cancer metastasis. Cell 2009; 137:1032-46; PMID:19524507; http://dx.doi.org/10.1016/j.cell.2009.03.047.

16. Zhang XH, Wang Q, Gerald W, Hudis CA, Norton L, Smid M, et al. Latent bone metastasis in breast cancer tied to Src-dependent survival signals. Cancer Cell 2009; 16:67-78; PMID:19573813; http://dx.doi.org/10.1016/j.ccr.2009.05.017.

17. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci USA 2005; 102:15545-50; PMID:16199517; http://dx.doi.org/10.1073/pnas.0506580102.

18. Huynh-Thu VA, Irrthum A, Wehenkel L, Geurts P. Inferring regulatory networks from expression data using tree-based methods. PLoS ONE 2010; 5:12776; PMID:20927193; http://dx.doi.org/10.1371/journal.pone.0012776.

Page 6: Identification of novel metastasis suppressor signaling pathways for breast cancer

© 2012 Landes Bioscience.

Do not distribute.

www.landesbioscience.com Cell Cycle 2457

33. Hagan S, Al-Mulla F, Mallon E, Oien K, Ferrier R, Gusterson B, et al. Reduction of Raf-1 kinase inhibitor protein expression correlates with breast cancer metastasis. Clin Cancer Res 2005; 11:7392-7; PMID:16243812; http://dx.doi.org/10.1158/1078-0432.CCR-05-0283.

34. Schuierer MM, Bataille F, Hagan S, Kolch W, Bosserhoff AK. Reduction in Raf kinase inhibi-tor protein expression is associated with increased Ras-extracellular signal-regulated kinase signaling in melanoma cell lines. Cancer Res 2004; 64:5186-92; PMID:15289323; http://dx.doi.org/10.1158/0008-5472.CAN-03-3861.

35. Yun J, Frankenberger CA, Kuo WL, Boelens MC, Eves EM, Cheng N, et al. Signalling pathway for RKIP and Let-7 regulates and predicts metastatic breast cancer. EMBO J 2011; 30:4500-14; PMID:21873975; http://dx.doi.org/10.1038/emboj.2011.312.

36. Park SM, Shell S, Radjabi AR, Schickel R, Feig C, Boyerinas B, et al. Let-7 prevents early cancer progression by suppressing expression of the embry-onic gene HMGA2. Cell Cycle 2007; 6:2585-90; PMID:17957144; http://dx.doi.org/10.4161/cc.6.21.4845.

37. Bevilacqua E, Frankenberger CA, Rosner MR. RKIP Suppresses Breast Cancer Metastasis to the Bone by Regulating Stroma-Associated Genes. Int J Breast Cancer 2012; 2012:124704; PMID:22482058; http://dx.doi.org/10.1155/2012/124704.

27. Yeung KC, Rose DW, Dhillon AS, Yaros D, Gustafsson M, Chatterjee D, et al. Raf kinase inhibitor protein interacts with NFkappaB-inducing kinase and TAK1 and inhibits NFkappaB activation. Mol Cell Biol 2001; 21:7207-17; PMID:11585904; http://dx.doi.org/10.1128/MCB.21.21.7207-17.2001.

28. Eves EM, Shapiro P, Naik K, Klein UR, Trakul N, Rosner MR. Raf kinase inhibitory protein regulates aurora B kinase and the spindle checkpoint. Mol Cell 2006; 23:561-74; PMID:16916643; http://dx.doi.org/10.1016/j.molcel.2006.07.015.

29. Akaishi J, Onda M, Asaka S, Okamoto J, Miyamoto S, Nagahama M, et al. Growth-suppressive function of phosphatidylethanolamine-binding protein in anaplas-tic thyroid cancer. Anticancer Res 2006; 26:4437-42; PMID:17201166.

30. Al-Mulla F, Hagan S, Behbehani AI, Bitar MS, George SS, Going JJ, et al. Raf kinase inhibitor protein expres-sion in a survival analysis of colorectal cancer patients. J Clin Oncol 2006; 24:5672-9; PMID:17179102; http://dx.doi.org/10.1200/JCO.2006.07.5499.

31. Fu Z, Kitagawa Y, Shen R, Shah R, Mehra R, Rhodes D, et al. Metastasis suppressor gene Raf kinase inhibitor protein (RKIP) is a novel prognostic marker in prostate cancer. Prostate 2006; 66:248-56; PMID:16175585; http://dx.doi.org/10.1002/pros.20319.

32. Fu Z, Smith PC, Zhang L, Rubin MA, Dunn RL, Yao Z, et al. Effects of raf kinase inhibitor protein expres-sion on suppression of prostate cancer metastasis. J Natl Cancer Inst 2003; 95:878-89; PMID:12813171; http://dx.doi.org/10.1093/jnci/95.12.878.

19. Ideker T, Dutkowski J, Hood L. Boosting signal-to-noise in complex biology: prior knowledge is power. Cell 2011; 144:860-3; PMID:21414478; http://dx.doi.org/10.1016/j.cell.2011.03.007.

20. Efron B, Tibshirani R. On testing the significance of sets of genes. Annals of Applied Statistics 2007; 1:107-29; http://dx.doi.org/10.1214/07-AOAS101.

21. Breiman L. Random Forests. Mach Learn 2001; 45:5-32; http://dx.doi.org/10.1023/A:1010933404324.

22. Dangi-Garimella S, Yun J, Eves EM, Newman M, Erkeland SJ, Hammond SM, et al. Raf kinase inhibi-tory protein suppresses a metastasis signalling cascade involving LIN28 and let-7. EMBO J 2009; 28:347-58; PMID:19153603; http://dx.doi.org/10.1038/emboj.2008.294.

23. Granovsky AE, Rosner MR. Raf kinase inhibitory protein: a signal transduction modulator and metastasis suppressor. Cell Res 2008; 18:452-7; PMID:18379591; http://dx.doi.org/10.1038/cr.2008.43.

24. Corbit KC, Trakul N, Eves EM, Diaz B, Marshall M, Rosner MR. Activation of Raf-1 signaling by protein kinase C through a mechanism involving Raf kinase inhibitory protein. J Biol Chem 2003; 278:13061-8; PMID:12551925; http://dx.doi.org/10.1074/jbc.M210015200.

25. Trakul N, Menard RE, Schade GR, Qian Z, Rosner MR. Raf kinase inhibitory protein regulates Raf-1 but not B-Raf kinase activation. J Biol Chem 2005; 280:24931-40; PMID:15886202; http://dx.doi.org/10.1074/jbc.M413929200.

26. Lorenz K, Lohse MJ, Quitterer U. Protein kinase C switches the Raf kinase inhibitor from Raf-1 to GRK-2. Nature 2003; 426:574-9; PMID:14654844; http://dx.doi.org/10.1038/nature02158.