This Article Abstract Full Text (PDF) Supplementary Data Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Smid, M. Articles by Martens, J. W.M. Search for Related Content PubMed PubMed Citation Articles by Smid, M. Articles by Martens, J. W.M.

Subtypes of Breast Cancer Show Preferential Site of Relapse

¹ Department of Medical Oncology, Erasmus MC, Josephine Nefkens Institute; ² Cancer Genomics Centre, Rotterdam, the Netherlands; and ³ Veridex LLC, a Johnson & Johnson Company, San Diego, California

Requests for reprints: John W.M. Martens, Erasmus MC, Josephine Nefkens Institute, Department of Medical Oncology, Room Be400 P.O. Box 2040, 3000 CA Rotterdam, the Netherlands. Phone: 31-0-10-4088372; Fax: 31-0-10-4088377; E-mail: j.martens{at}erasmusmc.nl.

	Abstract

Top Abstract Introduction Materials and Methods Results Discussion References

 
We explored whether the five previously reported molecular subtypes in breast cancer show a preference for organ-specific relapse and searched for molecular pathways involved. The "intrinsic" gene list describing the subtypes was used to classify 344 primary breast tumors of lymph node–negative patients. Fisher exact tests were used to determine the association between a tumor subtype and a particular site of distant relapse in these patients who only received local treatment. Modulated genes and pathways were identified in the various groups using Significance Analysis of Microarrays and Global Testing. Bone relapse patients were most abundant in the luminal subtypes but were found less than expected in the basal subtype. The reverse was true for lung and brain relapse patients with the remark that absence of lung relapse was luminal A specific. Finally, a pleura relapse, although rare, was found almost exclusively in both luminal subtypes. Many differentially expressed genes were identified, of which several were in common in a subtype and the site to which the subtype preferentially relapsed. WNT signaling was up-regulated in the basal subtype and in brain-specific relapse, and down-modulated in the luminal B subtype and in bone-specific relapse. Focal adhesion was found up-regulated in the luminal A subtype but down-regulated in lung relapse. The five major molecular subtypes in breast cancer are evidently different with regard to their ability to metastasize to distant organ(s), and share biological features and pathways with their preferred distant metastatic site. [Cancer Res 2008;68(9):3108–14]

	Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

 
Molecular subtypes in breast cancer were first described by Perou et al. (1) who mapped the phenotypic diversity seen in breast cancer to a specific gene expression pattern. Later studies reported differences in prognosis (2, 3) and chemotherapy response (4) with respect to the subtypes in specific patient cohorts. These observations reinforce the hypothesis that the molecular gene expression patterns specific for a subtype have clinical relevance. Besides the above-mentioned clinical aspects (prognosis and response to therapy), the distant site to which a tumor preferentially metastasizes is of clinical and biological importance. Recently, microarray analyses identified gene expression profiles for bone (5–7) and lung (8) metastasis in breast cancer. One of the reported aspects is that the aggressiveness of a tumor, i.e., the ability to metastasize, is driven by a distinct set of genes, different from those involved in the capacity to home, survive, and proliferate in a particular organ (6, 8). This may tie in with the self-seeding theory (9) and the seed and soil theory of Paget (10). The latter theory proposes that specific organs are in some way predisposed targets for secondary growth. This may reflect the necessity for the primary tumor to express a certain genetic module to invade specific organs. The self-seeding theory offers the view that dislodged cancer cells may either reenter the primary tumor bed or otherwise colonize a distant organ, the latter cell possibly needing additional (genetic) properties. In this respect, we consider multiple affected organs as a first event after removal of the primary tumor as the outcome of the genetic capabilities of the primary tumor.

With the understanding that the reported molecular subtypes in breast cancer reflect phenotypic and genotypic variation of the biology of breast disease, we evaluated in this study expression data of 344 primary breast cancer patients; first to ascertain the presence of the molecular subtypes within our cohort, then to study if samples in a specific subtype showed organ-specific metastasis. If so, analyses were performed to establish if shared biology is present among the subtypes and the site of relapse.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

 
Patients. A group of 344 primary breast tumors was used in this study; all from patients with lymph node–negative breast cancer who did not receive any adjuvant systemic treatment. The study was approved by the Medical Ethics Committee of the Erasmus MC Rotterdam, the Netherlands (MEC-02.953) and was performed in accordance to the Code of Conduct of the Federation of Medical Scientific Societies in the Netherlands.⁴ Two hundred eighty-six patients were included from an earlier study (11) that was supplemented with an additional 58 estrogen receptor (ER)-negative patients (12). The mean age at time of surgery was 53 years (SD, 12); 221 patients were ER-positive and 123 patients were ER-negative; 198 patients were premenopausal and 146 were postmenopausal. T₁ tumors (2 cm) were present in 168 patients (49%), T₂ tumors (>2–5 cm) in 163 patients (47%), T_3/4 tumors (>5 cm) in 12 patients (3%), and 1 patient with unknown tumor stage. Of the complete group, 226 patients had no relapse during follow-up (median follow-up time of patients still alive was 101 mo; range, 61–171 mo), 81 patients had a distant relapse to a single organ, and 37 patients had multiple relapse sites (of which 30 patients had two organs affected at first presentation). Including the multiple sites of relapse in a single patient, the 118 relapse patients showed 165 relapses: 71 bone, 30 lung, 18 liver, 14 brain, 12 pleura, 5 skin, and 15 miscellaneous sites. Due to the low numbers, the latter two sites were not evaluated.

Microarray analysis. The expression data have been deposited in the National Center for Biotechnology Information/Gene Expression Omnibus database entries GSE2034 and GSE5327. In short, targets were hybridized to Affymetrix HG-U133A chips. Gene expression signals were calculated using Affymetrix GeneChip analysis software MAS 5.0. Global scaling was performed to bring the average signal intensity of a chip to a target of 600 before data analysis. For each probe set, intensities were thresholded at 30 and were then expressed relative to the geometric mean of that probe set and were 2-log transformed.

Bioinformatic analysis. The Genbank accession numbers of the original 496 subtype genes were matched to the Affymetrix probe-sets using Unigene cluster numbers. Some of the genes have multiple probe sets present. To ensure analysis of only the informative probe sets, the probe sets that did not vary across all samples were removed, leaving 66% of the probe sets, i.e., the most variable ones. To identify the molecular subtypes in our cohort, these top 66% variable probe sets were used to cluster 344 samples, using average linkage hierarchical clustering with correlation as a distance metric (13). The robustness of the clusters was ascertained as described (14) using the 10th percentile of the variance of log signals of all samples for the Gaussian noise, as recommended (14).

To identify differentially expressed genes and pathways, the samples were grouped according to the site of relapse, e.g., all bone relapse patients versus patients relapsing elsewhere. In a separate analysis, genes and pathways were identified in the subtype groups, e.g., basal-type samples versus the rest. Differentially expressed genes were obtained via a Significance Analysis of Microarrays (SAM) analysis (15). The top 33% variable genes (n = 7,353) were used for input, and 300 permutations of the data were performed to calculate the false discovery rate (FDR). Genes were selected with a 1.7-fold difference or more between groups that showed a FDR of 5% or less (for two analyses, the FDR was set at 25%; see Results). The overlap between the gene lists was statistically evaluated by calculating z-scores based on the hypergeometric distribution.

The Global Test program (16, 17) was used (version 4.2.0) to associate KEGG pathways (18) to the site of relapse and to the molecular subtypes. All P values were corrected for multiple testing and checked by resampling if an equally sized, randomly chosen group of genes is also significant (1,000 samplings). Pathways were considered of interest if the P value of the Global Test, after correcting for multiple testing, and the resampling P value were both below 0.05. Pathway P values mentioned in the text are two-sided P values corrected for multiple testing, except where stated otherwise. The contribution of individual genes in a pathway was evaluated using z-score calculated by the Global Test program. Genes with z-scores that are >1.96 were considered significant contributors to the pathway. R version 2.3.1⁵ was used to run the Global Test package.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

 
Subtyping
The intrinsic gene list (1) of the molecular subtypes was linked to the Unigene cluster number, which left 417 unique genes. These matched to 684 probe sets from the U133A chip (Supplementary Table S1). Based on identical Unigene number, these 684 probe sets represent 360 unique genes. To ascertain that these 360 genes still contain the information to distinguish the subtypes, the original data (1) were reclustered using this shorter gene list, consisting of only those intrinsic genes that are present on the Affymetrix U133A chip. Only one sample switched from not labeled in the original data to normal-like in the new clustering; all other samples remained in the same clusters (see Fig. 1 ). The top 66% variable probe sets were used to cluster 344 primary breast tumors (Fig. 2 ). In line with the initial reports (1, 2), clustering using the intrinsic genes revealed five main molecular subtypes. Evaluation of the expression patterns and in particular, the expression of ERBB2, ESR1, ADH1B, and KRT5 (see arrows; Fig. 2, left to right) were used to identify each clustering branch. The two luminal subtypes can be distinguished by the higher expression of the ER and its target genes (e.g., GATA3, TFF1, and NAT1). Among others, the high expression of ADH1B sets the luminal A subtype apart from the luminal B subtype. The erbb2 subtype was identified by the higher expression of the ERBB2 gene and its chromosomal neighboring genes. The basal subtype can be characterized by the low expression of ER and its target genes, plus the higher expression of the KRT5 and KRT17 genes. With four of the cluster groups identified, the last one is the normal-like subtype, which in addition, as described by Perou et al. (1), expresses ADH1B abundantly. Thus, samples were labeled (top to bottom) as luminal B (n = 68), luminal A (n = 93), which comprise the first major category, and erbb2 (n = 70), normal-like (n = 27), and basal subtype (n = 86) forming the second major category. The reproducibility of the clusters was studied using previously published methods (14), in which the R measure, which is the proportion of the time a sample pair stays in the same cluster after perturbation and reclustering, is calculated. Using 100 perturbations of the data, we obtained R values of 0.808 and 0.957 for the luminal A and B subtypes, respectively, and 0.982, 0.725, and 0.654, respectively for the basal, erbb2, and normal-like tumors.

View larger version (18K): [in this window] [in a new window] [Download PPT slide]   Figure 1. Clustering of original data set of Perou et al. (1) using the Affymetrix-matched, intrinsic gene list. The 360 unique genes of the intrinsic gene list that could be mapped to the Affymetrix platform were used to recluster the original data. The original data (top dendogram) was compared with the hierarchical clustering using this shorter gene list (bottom dendogram). Lines were drawn between the positions of the samples and colored according to the label used by Perou et al. (1); orange, basal subtype; red, erbb2 subtype; green, normal-like subtype; black, unlabeled; blue, luminal.

View larger version (118K): [in this window] [in a new window] [Download PPT slide]   Figure 2. Association of subtypes and site of relapse. Subtypes in 344 primary breast cancers. A, dendogram after hierarchical clustering. B, clustered expression data of the Affymetrix intrinsic genes; rows, tumor samples; columns, genes; red, relative high expression; blue, relative low expression. Arrows indicate from left to right the position of ERBB2, ESR1, ADH1B, and KRT5. C, subtypes (first column). Dark blue, luminal B; light blue, luminal A; red, erbb2; green, normal-like; orange, basal subtype. The remaining columns indicate the site of relapse of a patient; light blue, bone relapse; green, lung relapse; blue, liver relapse; maroon, brain relapse; orange, pleura relapse.

Pathway mapping. To evaluate whether the processes that govern the metastasis to a certain organ are also present in the subtypes, we mapped KEGG (18) pathways to all site of relapse and subtype groups using the Global Test package (16, 17). Although KEGG contains fewer entries than Gene Ontology (20, 21), in our view, it is more suited for our purpose because this curated database contains well-established pathways, whereas Gene Ontology groups genes with a similar function.

We tested the same groups of samples as were used in the SAM analysis. All pathways were corrected for multiple testing and verified by resampling (see Materials and Methods for details). We focused on pathways with a high percentage of influential genes, and found two pathways standing out (see Supplementary Table S3). "Focal adhesion" is found significant in the luminal A and normal-like subtype (both P < 0.0001) and in tumors from patients with a lung relapse (P = 0.001). We subsequently concentrated on analyzing focal adhesion genes in lung relapse patients and luminal A subtype (Fig. 3 ), as the normal-like subtype only harbors one lung relapse patient. The 10 significant genes (z-score, >1.96) of focal adhesion that are higher expressed in the lung relapse patients are associated with signaling and particularly with signaling toward cell survival. Of the 16 down-regulated genes in the lung relapse patients, 14 are found significantly higher expressed in the luminal A subtype group (Fig. 3, arrows). Six of these 14 genes are extracellular matrix genes.

View larger version (18K): [in this window] [in a new window] [Download PPT slide]   Figure 3. Pathway analysis plot: influential genes in focal adhesion. Genes annotated to the focal adhesion pathway in KEGG and their association with tumors from lung relapse patients and luminal A subtype samples. Each bar at the X-axis represents a gene in the focal adhesion pathway. The height of the bar indicates the contribution (influence) of each individual gene of the pathway with the clinical variable. Horizontal markers in a bar indicate one SD away from the reference point; two or more horizontal markers in a bar indicates a statistically significant association of the corresponding gene with the clinical end point. A, lung relapse patients versus patients with relapses at other sites. Red bars, genes higher expressed in tumors from lung relapse patients; green bars, genes higher expressed in tumors of patients with other relapse sites. B, luminal A subtype versus the other subtypes. Red bars, genes higher expressed in luminal A; green bars, genes higher expressed in the other subtypes. Arrows, genes significantly lower expressed in tumors of lung relapse patients but higher expressed in the luminal A subtype.

View larger version (20K): [in this window] [in a new window] [Download PPT slide]   Figure 4. Selected genes of WNT signaling. Important molecules of the WNT signaling pathway that are implicated in the basal and luminal B subtypes and in tumors from bone and brain relapse patients. Capped lines, inhibitory effect on the protein-protein interaction. The bottom graph depicts z-score values. These z-scores are a measure for the unlikelihood of the null hypothesis that the gene is not associated with the clinical attribute. Positive and negative z-values indicate genes significantly higher or lower expressed in the corresponding group, respectively. Orange bar, basal subtype; maroon, brain relapse patient; blue, luminal B subtype; light blue, bone relapse patient.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

The Affymetrix-matched, intrinsic gene list, although comprised of fewer genes, correctly classified the samples in the original data set. Therefore, we and others (27), show that the intrinsic gene list extracted by Perou and colleagues from a relatively small, heterogeneous patient group can be translated across microarray platforms, and as reported here, retains its value even in a large, homogenous (lymph node negative) sample cohort. Thus, the intrinsic gene list proves a robust indicator of the five major molecular subtypes present in breast cancer. Association of the subtypes with clinical data showed that the site of distant relapse was not randomly distributed across the subtypes. The large numbers of genes differentially expressed between the five molecular subtypes confirm that the underlying biology of the subtypes is indeed very different. Also, the many differentially expressed genes in the tumors that relapse to various sites support the view that specific biological processes are involved in organ-specific metastasis.

Taken all three types of analysis together, i.e., the number of overlapping genes, the function of differentially expressed genes, and the identified pathways, all show consistent results when analyzing the subtypes and the site of relapse, which suggests that shared biology has been unraveled. The most prominent results that emerged relate to the biology of brain, lung, and bone relapse. The WNT pathway was associated with patients relapsing to brain and bone, as well as with the basal and luminal B subtypes. The findings suggest that active WNT/β-catenin signaling contributes to basal breast tumors metastasizing to brain, whereas the absence of WNT/β-catenin signaling allows for luminal B–type tumors to metastasize to bone. However, the activation of β-catenin may not originate from WNT ligands, as we find inhibitors of WNT ligands and negative regulatory WNT receptors, i.e., FZD6 (28) overexpressed in the basal tumors relapsing to brain (Fig. 4). Overexpression of these negative regulators, however, could be the results of negative feedback due to activation of the WNT pathway (29). Taken together, these data suggest that a specific route of the WNT pathway may be active in basal tumors relapsing to brain. Interestingly, in brain tissue itself, a large body of evidence points to an important involvement of WNT/β-catenin signaling in the development and maintenance of normal brain tissue as well as in brain tumorigenesis [among others, reviewed by Fogarty et al. (30)]. Thus, the active WNT/β-catenin signaling by breast cancer cells metastasizing to brain could point to mimicry which, if proven, supports the view that the seed grows better in the soil it resembles (10). Along the same line, one could speculate that luminal B tumors, in which WNT/β-catenin signaling is down-regulated, thus, lack the specific genetic module that facilitates a brain relapse and thus relapse elsewhere.

For the lung relapse patients, we found that the focal adhesion signaling cascade is an important modulator of organ-specific relapse. Focal adhesions are specific types of large protein complexes through which the cytoskeleton of a cell connects to and communicates with the extracellular matrix. Of the focal adhesion genes that were annotated by KEGG (18), many are up-regulated in the luminal A subtype and down-regulated in tumors from patients who had a lung relapse. Because very few patients in the luminal A subtype had relapses to the lung, it seems the involved focal adhesion molecules impede a lung relapse. These observations, together with the high frequency of extracellular matrix genes that were found significantly differentially regulated, imply that luminal A–like tumor cells lack the ability to create a specific microenvironment surrounding the metastasizing cells, necessary for invading and proliferating in lung tissue. Similar findings were also reported by Minn et al. (8) who described many genes involved in the tumor cell microenvironment, which were found differentially expressed in lung metastasis in both a mouse model and primary breast tumors.

Finally, despite the fact that the patients with a relapse to the bone were not overrepresented in any of the subtypes, we find in agreement with earlier findings by us (5) and others (6), a clear association with ER status. This is also echoed in the list of differentially expressed genes in bone relapse samples, which holds many ER-target genes. Our findings in this cohort are further in line with the recently (5, 6) revealed involvement of transforming growth factor β and fibroblast growth factor (FGF) signaling, TFF proteins, IL11, and CTGF in bone relapse. Noteworthy is the fact that the genes up-regulated in the bone relapse patients of the erbb2 subtype are entirely different from those found in the luminal subtypes. This suggests that the erbb2 subtype may metastasize to bone via processes different from luminal tumors. Also, among the list of differentially expressed genes in erbb2 tumors that metastasize to bone are two members of the secretoglobin family, mammaglobin 1 and lipophilin B, which are both on 11q13, a common amplified region in breast cancer (which includes the genes cylin D1, FGF 3 and 4, and cortactin). Both mammaglobin and lipophilin B are implicated as important breast cancer markers (31–33) and are next to being up-regulated in bone relapse patients of the erbb2 subtype also significantly down-regulated in lung relapse patients. This could suggest that mammaglobin, lipophilin B, or possibly one or more other genes in the 11q13 amplicon modulate the ability of erbb2 tumors to relapse to bone instead of lung.

Although many gene expression studies on breast cancer are available, none provided all essential information to validate our results. Nevertheless, an interesting observation in this respect, is reported by Minn et al. (34). In this collaborative study on lung metastasis in breast cancer, the tumors that expressed the lung metastasis signature (developed by Minn et al.) were very often of the basal subtype, which agrees with our results. However, this observation was partly based on our own data set; thus, we do not regard this as a true independent validation.

In conclusion, the observations reported here indicate that the five major molecular subtypes in breast cancer are not only distinct with regard to primary tumor characteristics, tumor aggressiveness, and response to certain types of chemotherapy, they are also clearly different with regard to their ability to metastasize to distant organ(s). In depth data mining uncovered shared biology and gene expression patterns specific for a subtype and its preferred distant metastatic site. Many identified genes agree with the notion that the tumor microenvironment plays an important role in distant metastasis which, together with specific attractors such as chemokines (35, 36) and specific signaling pathways (e.g., WNT/β-catenin signaling and focal adhesion), are important determinants involved in homing, survival, and proliferation of the tumor cells in their new niche. The genetic modules involved in these three parts are balanced differently in the breast cancer subtypes and, thus, facilitate metastasis to different organs.

	Acknowledgments

 
Grant support: Veridex LLC for tissue processing and isolating RNA for microarray analysis as well as a research grant from the Netherlands Genomics Initiative/Netherlands Organization for Scientific Research. This organization had no role in study design, analysis of data, or preparation of the manuscript.

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.

We thank Michael Moorhouse of Erasmus MC, Rotterdam for providing us an altered version of a Perl-script, The American Women's Club of The Hague, and the Netherlands Genomics Initiative/Netherlands Organization for Scientific Research for the financial support.

	Footnotes

 
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

⁴ http://www.fmwv.nl

⁵ http://www.cran.org

Received 9/26/07. Revised 2/28/08. Accepted 2/28/08.

	References

Top Abstract Introduction Materials and Methods Results Discussion References

 

1. Perou CM, Sorlie T, Eisen MB, et al. Molecular portraits of human breast tumours. Nature 2000;406:747–52.[CrossRef][Medline]
2. Sorlie T, Perou CM, Tibshirani R, et al. Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci U S A 2001;98:10869–74.[Abstract/Free Full Text]
3. Sorlie T, Tibshirani R, Parker J, et al. Repeated observation of breast tumor subtypes in independent gene expression data sets. Proc Natl Acad Sci U S A 2003;100:8418–23.[Abstract/Free Full Text]
4. Rouzier R, Perou CM, Symmans WF, et al. Breast cancer molecular subtypes respond differently to preoperative chemotherapy. Clin Cancer Res 2005;11:5678–85.[Abstract/Free Full Text]
5. Smid M, Wang Y, Klijn JG, et al. Genes associated with breast cancer metastatic to bone. J Clin Oncol 2006;24:2261–7.[Abstract/Free Full Text]
6. Kang Y, Siegel PM, Shu W, et al. A multigenic program mediating breast cancer metastasis to bone. Cancer Cell 2003;3:537–49.[CrossRef][Medline]
7. Deckers M, van Dinther M, Buijs J, et al. The tumor suppressor Smad4 is required for transforming growth factor β-induced epithelial to mesenchymal transition and bone metastasis of breast cancer cells. Cancer Res 2006;66:2202–9.[Abstract/Free Full Text]
8. Minn AJ, Gupta GP, Siegel PM, et al. Genes that mediate breast cancer metastasis to lung. Nature 2005;436:518–24.[CrossRef][Medline]
9. Norton L, Massague J. Is cancer a disease of self-seeding? Nat Med 2006;12:875–8.[CrossRef][Medline]
10. Paget S. The distribution of secondary growths in cancer of the breast. 1889. Cancer Metastasis Rev 1989;8:98–101.[Medline]
11. Wang Y, Klijn JG, Zhang Y, et al. Gene-expression profiles to predict distant metastasis of lymph-node-negative primary breast cancer. Lancet 2005;365:671–9.[Medline]
12. Yu JX, Sieuwerts AM, Zhang Y, et al. Pathway analysis of gene signatures predicting metastasis of node-negative primary breast cancer. BMC Cancer 2007;7:182.[CrossRef][Medline]
13. Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci U S A 1998;95:14863–8.[Abstract/Free Full Text]
14. McShane LM, Radmacher MD, Freidlin B, Yu R, Li MC, Simon R. Methods for assessing reproducibility of clustering patterns observed in analyses of microarray data. Bioinformatics 2002;18:1462–9.[Abstract/Free Full Text]
15. Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A 2001;98:5116–21.[Abstract/Free Full Text]
16. Goeman JJ, Oosting J, Cleton-Jansen AM, Anninga JK, van Houwelingen HC. Testing association of a pathway with survival using gene expression data. Bioinformatics 2005;21:1950–7.[Abstract/Free Full Text]
17. Goeman JJ, van de Geer SA, de Kort F, van Houwelingen HC. A global test for groups of genes: testing association with a clinical outcome. Bioinformatics 2004;20:93–9.[Abstract/Free Full Text]
18. Kanehisa M, Goto S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res 2000;28:27–30.[Abstract/Free Full Text]
19. Holm S. A simple sequentially rejective multiple test procedure. Scand J Stat 1979;6:65–70.
20. Ashburner M, Ball CA, Blake JA, et al. The Gene Ontology Consortium. Gene ontology: tool for the unification of biology. Nat Genet 2000;25:25–9.[CrossRef][Medline]
21. Harris MA, Clark J, Ireland A, et al. The Gene Ontology (GO) database and informatics resource. Nucleic Acids Res 2004;32:D258–61.[Abstract/Free Full Text]
22. Chang HY, Sneddon JB, Alizadeh AA, et al. Gene expression signature of fibroblast serum response predicts human cancer progression: similarities between tumors and wounds. PLoS Biol 2004;2:E7.[CrossRef][Medline]
23. West RB, Nuyten DS, Subramanian S, et al. Determination of stromal signatures in breast carcinoma. PLoS Biol 2005;3:e187.[CrossRef][Medline]
24. Chi JT, Wang Z, Nuyten DS, et al. Gene expression programs in response to hypoxia: cell type specificity and prognostic significance in human cancers. PLoS Med 2006;3:e47.[CrossRef][Medline]
25. Winter SC, Buffa FM, Silva P, et al. Relation of a hypoxia metagene derived from head and neck cancer to prognosis of multiple cancers. Cancer Res 2007;67:3441–9.[Abstract/Free Full Text]
26. Hiraga T, Kizaka-Kondoh S, Hirota K, Hiraoka M, Yoneda T. Hypoxia and hypoxia-inducible factor-1 expression enhance osteolytic bone metastases of breast cancer. Cancer Res 2007;67:4157–63.[Abstract/Free Full Text]
27. Hu Z, Fan C, Oh DS, et al. The molecular portraits of breast tumors are conserved across microarray platforms. BMC Genomics 2006;7:96.[CrossRef][Medline]
28. Golan T, Yaniv A, Bafico A, Liu G, Gazit A. The human Frizzled 6 (HFz6) acts as a negative regulator of the canonical Wnt. β-catenin signaling cascade. J Biol Chem 2004;279:14879–88.[Abstract/Free Full Text]
29. Zirn B, Samans B, Wittmann S, et al. Target genes of the WNT/β-catenin pathway in Wilms tumors. Genes Chromosomes Cancer 2006;45:565–74.[CrossRef][Medline]
30. Fogarty MP, Kessler JD, Wechsler-Reya RJ. Morphing into cancer: the role of developmental signaling pathways in brain tumor formation. J Neurobiol 2005;64:458–75.[CrossRef][Medline]
31. Ferrucci PF, Rabascio C, Mazzetta C, et al. Mammaglobin expression in leukapheresis products is a predictive marker of poor prognosis in women with high-risk breast cancer. Clin Cancer Res 2004;10:6039–46.[Abstract/Free Full Text]
32. Span PN, Waanders E, Manders P, et al. Mammaglobin is associated with low-grade, steroid receptor-positive breast tumors from postmenopausal patients, and has independent prognostic value for relapse-free survival time. J Clin Oncol 2004;22:691–8.[Abstract/Free Full Text]
33. Zafrakas M, Petschke B, Donner A, et al. Expression analysis of mammaglobin A (SCGB2A2) and lipophilin B (SCGB1D2) in more than 300 human tumors and matching normal tissues reveals their co-expression in gynecologic malignancies. BMC Cancer 2006;6:88.[CrossRef][Medline]
34. Minn AJ, Gupta GP, Padua D, et al. Lung metastasis genes couple breast tumor size and metastatic spread. Proc Natl Acad Sci U S A 2007;104:6740–5.[Abstract/Free Full Text]
35. Burger JA, Kipps TJ. CXCR4: a key receptor in the crosstalk between tumor cells and their microenvironment. Blood 2006;107:1761–7.[Abstract/Free Full Text]
36. Muller A, Homey B, Soto H, et al. Involvement of chemokine receptors in breast cancer metastasis. Nature 2001;410:50–6.[CrossRef][Medline]

This article has been cited by other articles:

				T. A. DiMeo, K. Anderson, P. Phadke, C. Feng, C. M. Perou, S. Naber, and C. Kuperwasser A Novel Lung Metastasis Signature Links Wnt Signaling with Cancer Cell Self-Renewal and Epithelial-Mesenchymal Transition in Basal-like Breast Cancer Cancer Res., July 1, 2009; 69(13): 5364 - 5373. [Abstract] [Full Text] [PDF]

				A. Moreno-Aspitia and E. A. Perez Treatment Options for Breast Cancer Resistant to Anthracycline and Taxane Mayo Clin. Proc., June 1, 2009; 84(6): 533 - 545. [Abstract] [Full Text] [PDF]

				S. Dawood, K. Broglio, F. J. Esteva, W. Yang, S.-W. Kau, R. Islam, C. Albarracin, T. K. Yu, M. Green, G. N. Hortobagyi, et al. Survival among women with triple receptor-negative breast cancer and brain metastases Ann. Onc., April 1, 2009; 20(4): 621 - 627. [Abstract] [Full Text] [PDF]

				A. M. Sieuwerts, J. Kraan, J. Bolt, P. van der Spoel, F. Elstrodt, M. Schutte, J. W. M. Martens, J.-W. Gratama, S. Sleijfer, and J. A. Foekens Anti-Epithelial Cell Adhesion Molecule Antibodies and the Detection of Circulating Normal-Like Breast Tumor Cells J Natl Cancer Inst, January 7, 2009; 101(1): 61 - 66. [Abstract] [Full Text] [PDF]

				J. Lin, N. R. Cook, C. Albert, E. Zaharris, J. M. Gaziano, M. Van Denburgh, J. E. Buring, and J. E. Manson Vitamins C and E and Beta Carotene Supplementation and Cancer Risk: A Randomized Controlled Trial J Natl Cancer Inst, January 7, 2009; 101(1): 14 - 23. [Abstract] [Full Text] [PDF]

				J. A. Foekens, A. M. Sieuwerts, M. Smid, M. P. Look, V. de Weerd, A. W. M. Boersma, J. G. M. Klijn, E. A. C. Wiemer, and J. W. M. Martens Four miRNAs associated with aggressiveness of lymph node-negative, estrogen receptor-positive human breast cancer PNAS, September 2, 2008; 105(35): 13021 - 13026. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) Supplementary Data Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Smid, M. Articles by Martens, J. W.M. Search for Related Content PubMed PubMed Citation Articles by Smid, M. Articles by Martens, J. W.M.