Control of EVI-1 Oncogene Expression in Metastatic Breast Cancer Cells Through Microrna Mir-22

Oncogene (2011) 30, 1290–1301 & 2011 Macmillan Publishers Limited All rights reserved 0950-9232/11 www.nature.com/onc ORIGINAL ARTICLE Control of EVI-1 oncogene expression in metastatic breast cancer cells through microRNA miR-22

JB Patel1,8, HN Appaiah1,8, RM Burnett1, P Bhat-Nakshatri1, G Wang2, R Mehta3, S Badve3, MJ Thomson4, S Hammond5, P Steeg6, Y Liu2 and H Nakshatri1,7

1Department of Surgery, Indiana University School of Medicine, Indianapolis, IN, USA; 2Department of Medicine, Indiana University School of Medicine, Indianapolis, IN, USA; 3Department of Pathology, Indiana University School of Medicine, Indianapolis, IN, USA; 4Department of Cancer Biology, Vanderbilt-Ingram Cancer Center, Vanderbilt University Medical School, Nashville, TN, USA; 5Department of Cell and Developmental Biology, Lineberger Comprehensive Cancer Center, University of North Carolina, Chapel Hill, NC, USA; 6Laboratory of Molecular Pharmacology, Center for Cancer Research National Cancer Institute, Bethesda, MD, USA and 7Department of Biochemistry and Molecular Biology, Indiana University School of Medicine, Indianapolis, IN, USA

Metastasis in breast cancer carries a disproportionately Keywords: breast cancer; metastasis; microRNA; worse prognosis than localized primary disease. To miR-22; let-7 identify microRNAs (miRNA) involved in metastasis, the expression of 254 miRNAs was measured across the following cell lines using microarray analysis: MDA-MB- 231 breast cancer cells, cells that grew as a tumor in the Introduction mammary fat pad of nude mice (TMD-231), metastatic disease to the lungs (LMD-231), bone (BMD-231) and Breast cancer metastasis is the dominant mechanism adrenal gland (ADMD-231). A brain-seeking variant of underlying treatment failure and mortality (Steeg, 2006). this cell line (231-BR) was used additionally in validation Metastatic cancer cells have been shown to carry a studies. Twenty miRNAs were upregulated and seven were distinct biological profile from that of primary tumor downregulated in metastatic cancer cells compared with cells by displaying characteristics such as enhanced TMD-231 cells. The expression of the tumor suppressor survival, motility, invasiveness and resistance to che- miRNAs let-7 and miR-22 was consistently downregulated motherapy (Kang et al., 2003; Gupta et al., 2005). in metastatic cancer cells. These metastatic cells ex- Investigating the extent to which metastatic cancer cells pressed higher levels of putative/proven miR-22 target differ from primary tumor cells with respect to gene oncogenes ERBB3, CDC25C and EVI-1. Introduction of expression is a newly developing field. At the center of miR-22 into cancer cells reduced the levels of ERBB3 and these investigations are microRNAs (miRNAs). These EVI-1 as well as phospho-AKT, an EVI-1 downstream small non-coding RNAs are now thought to be one of target. The miR-22 primary transcript is located in the the major regulators of gene expression (Nicoloso et al., 50-untranslated region of an open reading frame C17orf91, 2009). and the promoter/enhancer of C17orf91 drives miR-22 The role of miRNAs in cancer pathogenesis is tumor expression. We observed elevated C17orf91 expression in subtype specific. Recent reports describe miRNA non-basal subtype compared with basal subtype breast expression patterns in breast cancer in relation to cancers. In contrast, elevated expression of EVI-1 was hormone receptor status (Mattie et al., 2006; Blenkiron observed in basal subtype and was associated with poor et al., 2007). For example, the expression of let-7 family outcome in estrogen receptor-negative breast cancer members is lower in HER-2(ERBB2)-positive breast patients. These results suggest that metastatic cancer cancers compared with other cancer types. In total, 43 cells increase specific oncogenic signaling proteins through miRNAs are expressed at a higher level in HER-2- downregulation of miRNAs. Identifying such metastasis- positve breast cancers compared with HER-2-negative specific oncogenic pathways may help to manipulate breast cancers. Similarly, 43 miRNAs are expressed at tumor behavior and aid in the design of more effective higher levels in estrogen receptor a (ERa)-positive targeted therapies. breast cancers compared with ERa-negative breast Oncogene (2011) 30, 1290–1301; doi:10.1038/onc.2010.510; cancers. Luminal type A breast cancers, which express published online 8 November 2010 ERa and correspond to a good prognostic subgroup, show elevated expression of miR-10a, miR-10b, miR-21, miR-126*, miR-130a, miR-31, miR-382, miR-152, miR- Correspondence: Dr H Nakshatri, Department of Surgery, 100, miR-99a, miR-30-a-3p, miR-30a-5p, miR-224, Biochemistry, Molecular Biology, Indiana University School of Medicine, miR-214, let-7a, let-7b, let-7c, let-7f and miR-342. 980 West Walnut Street C218E, Indianapolis, IN 46202, USA. Estrogen regulates the expression of several of these E-mail: [email protected] 8These authors contributed equally to this work. miRNAs (Bhat-Nakshatri et al., 2009; Castellano et al., Received 6 May 2010; revised and accepted 30 September 2010; 2009; Maillot et al., 2009). Basal type of breast cancers, published online 8 November 2010 which are hormone receptor negative and correspond to Breast cancer metastasis JB Patel et al 1291 a significantly worse prognostic subgroup, express that showed altered expression was miR-22, which was higher levels of miR-150, miR-142-3p, miR-142-5p, downregulated in the metastatic clones. We describe a miR-148a, miR-106a, miR-106b, miR-18a, miR-93, pattern of metastatic site-specific differential protein miR-155, miR-25, miR-187 and miR-135b. Interest- expression of predicted/proven targets of miR-22 and ingly, HER-2-positive tumors show a similar miRNA describe how reduced levels of miR-22 in metastatic cells expression profile to basal type tumors with the contribute to activation of oncogenic pathways. exceptions of miR-106a, miR-18a, miR-93, miR-155 and miR-135b. However, none of these studies compared the miRNA expression pattern between primary and metastatic tumors. Utilizing the highly aggressive Results and basal type MDA-MB-231 (called MD-231 here- after) breast cancer cell line as a metastasis model, we MiRNA expression compared the miRNA expression profiles of parental MiRNA expression profiling of the five cell lines MD-231 cells, primary tumor cells (TMD-231) as well as including parental MD-231, primary tumor-derived metastatic disease from the brain (231-BR), lung (LMD- TMD-231, bone metastatic BMD-231, lung metastatic 231), adrenal (ADMD-231) and bone (BMD-231). LMD-231 and adrenal metastatic ADMD-231 revealed Using microarray analysis, we found differential expres- significant differential expression between different cell sion of miRNAs between metastatic cell lines compared types (Figure 1a). The miRNA profile was similar with the primary tumor. Most notable among miRNAs between duplicate experiments (Figure 1b). Metastatic

Samples

MD MD-231 MD

TMD TMD-231 MD

BMD

BMD BMD-231

BMD

ADMD ADMD-231 ADMD

BMD BMD-231 BMD

LMD

LMD LMD-231

LMD

-1.9 0 1.9 -1.9 0 +1.9

10 ADMD-231 7.3 BMD-231 LMD-231 4.6 MD-231 TMD-231 1.9

-0.8

-3.5

-6.2 PC #2 13.7% -8.9

-11.6

-14.3

-17 -18.8 -15.6 -12.4 -9.2 -6 -2.8 0.4 3.6 6.8 10 PC #1 19.8% Figure 1 MicroRNA expression profile in parental (MD-231), tumor (TMD-231), bone metastatic (BMD-231), lung metastatic (LMD-231) and adrenal metastatic (ADMD-231) variants of MDA-MB-231 cells. (a) Heat map showing the expression pattern of microRNAs in different cell types. (b) Principal component analysis (PCA) of microRNA expression profiles in different cell types. Each circle represents a cell type sample plotted based on their first three components (x, y and z axis in the plot). Samples are colored according to the cell types. The figure clearly shows the similarity in microRNA expression between duplicate samples.

Oncogene Breast cancer metastasis JB Patel et al 1292 Table 1 MicroRNAs differentially expressed in metastatic cells Table 2 MicroRNAs differentially expressed in TMD-231 and compared with parental (MD-231) and primary tumor (TMD-231) metastatic variants compared with parental cells cells MicroRNA name Fold change P-value MicroRNA name Fold change P-value MicroRNAs elevated in tumor and metastatic cells MicroRNAs elevated in metastatic cells hsa-miR-216 1.868754 0.025069 hsa-miR-138 1.569841 0.013959 hsa-miR-328 1.577275 0.025139 hsa-miR-200a 1.650099 0.014243 hsa-miR-30a-3p 1.48373 0.025259 hsa-miR-101 1.519873 0.020937 hsa-miR-424 4.311446 0.00569 hsa-miR-10a 1.243558 0.021184 hsa-miR-331 1.679677 0.005939 hsa-miR-26b 1.178239 0.021506 hsa-miR-302c* 1.617726 0.025732 hsa-miR-149 1.418833 0.022479 hsa-miR-324-3p 1.429711 0.027095 hsa-miR-129 1.439338 0.026166 hsa-miR-129 1.591343 0.029512 hsa-miR-218 1.605859 0.026291 hsa-miR-346 1.654502 0.030471 hsa-miR-324-3p 1.313153 0.031767 hsa-miR-202* 1.768106 0.032251 hsa-miR-212 1.572052 0.032277 hsa-miR-199b 2.004496 0.033587 hsa-miR-30a-3p 1.347055 0.032524 hsa-miR-19b 1.307697 0.039029 hsa-miR-204 1.569835 0.033425 hsa-miR-765 1.405249 0.039775 hsa-miR-202* 1.549166 0.034841 hsa-miR-138 1.67829 0.046795 hsa-miR-199b 1.7066 0.035982 hsa-miR-212 1.740812 0.049538 hsa-miR-220 1.407553 0.036564 hsa-miR-560 1.835449 0.009438 hsa-miR-19b 1.231013 0.038968 hsa-miR-101 1.754769 0.012318 hsa-miR-216 1.583479 0.040803 hsa-miR-220 1.621864 0.013034 hsa-miR-346 1.456886 0.041714 hsa-miR-326 1.579604 0.018446 hsa-miR-560 1.511125 0.042804 hsa-miR-149 1.578043 0.020005 hsa-miR-328 1.388716 0.047731 hsa-miR-200a 1.876417 0.020234 hsa-miR-218 1.850163 0.024632 MicroRNAs reduced in metastatic cells hsa-miR-21 0.599389 0.003418 MicroRNAs reduced in tumor and metastatic cells hsa-miR-634 0.720565 0.006854 hsa-miR-22 0.497554 0.003643 hsa-miR-181d 0.749118 0.011775 hsa-miR-195 0.535825 0.00686 hsa-miR-574 0.626508 0.018176 hsa-miR-574 0.54643 0.017916 hsa-miR-181a 0.66919 0.036095 hsa-miR-21 0.580566 0.049115 hsa-miR-195 0.657854 0.041401 hsa-miR-363* 0.661028 0.047242 hsa-miR-30d 0.823367 0.04561 hsa-miR-26a 0.661568 0.013301 hsa-miR-768-5p 0.731164 0.00507 hsa-miR-181d 0.677433 0.005183

clones displayed elevated levels of 20 miRNAs and reduced levels of 7 miRNAs compared with MD-231 or TMD-231 cells (Table 1). Selection of cells through Table 3 MicroRNAs differentially expressed in LMD-231 compared mammary fat pad injection alone appears to change the with other cell types miRNAs expression pattern as 22 miRNAs were MicroRNA name Fold change P-value elevated and 8 miRNAs were downregulated in TMD- 231 cells and its metastatic variants compared with MicroRNAs elevated in LMD-231 compared with other cells parental MD-231 cells (Table 2). Of these, miR-22 hsa-miR-565 1.277767 0.01041 B hsa-miR-210 1.35664 0.026345 showed the most significant repression of 50% in hsa-miR-28 1.225833 0.029151 metastatic clones compared with parental MD-231 cells. hsa-miR-487b 1.419737 0.030153 LMD-231 clones displayed a unique expression pattern in which four let-7 family members were downregulated MicroRNAs reduced in LMD-231 compared with other cell types hsa-let-7b 0.716811 0.031371 compared with the other clones (Table 3). Very few hsa-let-7a 0.661869 0.031699 changes specific to BMD-231 and ADMD-231 were hsa-let-7d 0.555842 0.0404 observed. hsa-let-7f 0.585863 0.01298 hsa-miR-181a 0.604505 0.046016 hsa-miR-16 0.591411 0.046521 hsa-miR-23b 0.601896 0.018737 Differential expression of let-7f and miR-22 in primary hsa-miR-23a 0.605637 0.020189 and metastatic cells hsa-miR-17-5p 0.681955 0.034475 It was recognized recently that the measurement of miRNA expression results obtained using various methods including miRNA profiling and quantitative (Gril et al., 2008). As 231-BR cells originated from a reverse transcription–PCR (qRT–PCR) were not different laboratory, parental cells used to generate these always compatible (Koshiol et al., 2010). Taking these cells from that same laboratory (MD-231P) were used observations into consideration, we performed qRT– for side-by-side comparison. PCR using primers specific to mature let-7f and miR-22 Let-7f expression was similarly low in metastatic across our cell lines to further establish their differential clones compared with MD-231 and TMD-231 expression in various cell lines (Figure 2). This analysis (Figure 2a). We are not sure why the level of this also included the brain metastatic variant 231-BR cells miRNA was elevated in TMD-231 compared with the

Oncogene Breast cancer metastasis JB Patel et al 1293

Figure 2 Let-7f, miR-22 and miR-424 expression pattern in various cell types. Mature microRNAs were measured by quantitative reverse transcription polymerase chain reaction (qRT–PCR) and were normalized to RNU66 or 5S RNA. (a) Let-7f expression in different cell types with RNU66 as normalization control. Right panel shows expression of Let-7f in MD-231P and its brain metastatic variant (231-BR) using 5S RNA as a normalization control. Average and standard error of the mean are shown. (b) miR-22 expression in various cell types measured as in (a). (c) miR-424 expression in MD-231 and TMD-231 cells with RNU66 as a normalization control. Statistically significant difference in expression (P ¼ 0.003 as indicated by an asterisk) was observed only between these two cell types. parental cell line. Differences between the variants of a miRgen, PCDHA13 (procadherin 13), PTCH1 (patched particular metastatic clone were not statistically sig- 1, a tumor suppressor in hedgehog pathway) and RARb nificant, which was discordant with the microarray (retinoic acid receptor b, a tumor suppressor) are its results (Table 3). Nonetheless, the results clearly showed predicted targets. In qRT–PCR validation assays, reduced expression of let-7f in metastatic clones although there was considerable experimental variation compared with parental and/or primary tumor-derived in expression in metastatic cell lines, a statistically clones with minimal difference among metastatic clones. significant increase was observed only in TMD-231 cells MD-231 and TMD-231 showed a modest difference compared with MD-231 cells with RNU66 as a normal- in the expression of miR-22 with metastatic clones ization control (Figure 2c and data not shown). With 5S showing even further reduction in expression of this RNA as a normalization control, a 3.3-fold increase in miRNA (Pp0.0005). Expression was least in 231-BR expression (P ¼ 0.01) was observed in ADMD-231 cells followed by BMD-231 (Figure 2b). The expression of compared with MD-231 cells (data not shown). miR-22 was lower in LMD-231 and BMD-231 cells compared with MD-231 cells with U6 or 5S as normalization controls (data not shown). Expression levels of predicted miR-22 targets in primary MiR-424 displayed 4.3-fold elevated expression in tumor and metastatic cells TMD-231 and metastatic clones compared with MD- Based on targetscan and miRGen predictions, more 231 cells (Table 2). MiR-424 is located on the X- than 200 genes are putative targets of miR-22. Genes chromosome and the corresponding region is activated relevant to breast cancer include AKT3, ERBB3, in sporadic basal breast cancers as well as in BRCA1 SATB2, HDAC4, CDC25C and SIRT1. Additionally, mutant cancers suggesting its relevance to these aggres- miR-22 has been shown to target the transcription sive types of cancer (Richardson et al., 2006). As per factors SP-1, EVI-1 oncogene, ERBB3 and ERa

Oncogene Breast cancer metastasis JB Patel et al 1294 (Sun et al., 2008; Pandey and Picard, 2009; Nagaraja clones, particularly in LMD-231 and ADMD-231 cells, et al., 2010). miR-22-mediated regulation of ERa is not compared with MD-231 cells (Figure 3a). Similarly, relevant to this study as cells used in this study are ERa- EVI-1 expression showed negative correlation with miR- negative. First, we determined whether miR-22 expres- 22. In particular, 231-BR cells showed remarkably sion levels showed any correlation with the levels of its higher levels of EVI-1 compared with MD-231P putative/proven target genes. ERBB3 protein, but not cells. EGFR, although not a direct target of miR-22, AKT3 and CBL proteins, was elevated in metastatic but is functionally linked to ERBB3, is differentially

Figure 3 Expression pattern of potential miR-22 targets in various cell types. (a) Western blot analysis shows expression of EGFR, ERBB3, EVI-1, CBL, CDC25C, AKT3 and SP-1 in various cell types. Three distinct isoforms of EVI-1 (MDA/EVI-1, EVI-1 and EVI- 1d are indicated. (b) EGFR transcripts levels are elevated in tumorigenic and metastatic cells compared with parental cells. EGFR transcripts were measured by qRT–PCR. (c) Generation of miR-22 overexpressing MD-231P cells. qRT–PCR was used to measure miR-22 expression in vector control (MD-231P-pQXIN) and miR-22 overexpressing cells (MD-231-miR-22). (d) Expression levels of miR-22 targets in MD-231P-pQXIN and MD-231-miR-22 cells. Western blotting was used to measure the expression levels of ERBB3 and EVI-1. (e) Basal phospho-AKT levels in MD-231P, 231-BR, MD-231P-pQXIN and MD-231P-miR-22 cells under serum starved (24 h) or normal growth media condition. (f) Migration assay. Scratch wound migration assay shows differences in migration capacity of MD-231P-pQXIN and MD-231-miR-22 cells (left panel). Cells were plated in regular media and migration was monitored for 48 h after creating a wound. Cell migration as measured by quantitative CyoSelect also shows differences in migration capacity of two cell types (right panel, P ¼ 0.004 as indicated by an asterisk).

Oncogene Breast cancer metastasis JB Patel et al 1295 expressed; both protein and transcript levels were expression with outcome in breast cancer patients elevated in TMD-231 and all metastatic clones com- (Gyorffy et al., 2010). Elevated C17orf91 expression pared with parental MD-231 cells (Figures 3a and b). correlated with significantly higher relapse-free survival Both SP-1 and CDC25C showed variable expression (Figure 4a). Metastasis-free survival also showed a among different cell lines; however, their expression similar trend, but did not reach statistical significance. did not negatively correlate with miR-22 expression When truly prognostic data sets (no systemic therapy) in all cell types. Between MD-231p and 231-BR were used in the analysis, elevated C17orf91 expression cells, CDC25C expression was higher in 231-BR cells correlated with poor relapse-free survival (data not (Figure 3a). shown). Whether these contrasting observations are The degree of miRNA-mediated target downregula- linked to the therapeutic response is unknown. tion often tends to be quantitatively modest and is We used additional microarray data sets in Gene evident mostly upon overexpression (Inui et al., 2010). Expression Omnibus to determine correlation between To investigate whether miR-22 directly regulates the C17orf91 and EVI-1 expression. C17orf91 expression expression of above genes, we generated MD-231P cells was lowest in basal breast cancers compared with non- overexpressing miR-22 (MD-231-miR-22) (Figure 3c). basal cancers or normal breast cells (Figure 4b). In The expression levels of EVI-1 and ERBB3, but not contrast, EVI-1 expression was elevated in basal breast AKT3, SP-1 and SP-3, were substantially lower in MD- cancers compared with non-basal cancers or normal 231-miR-22 cells compared with MD-231P cells with breast. C17orf91 (miR-22) expression appeared to be vector alone (MD-231P-pQXIN) (Figure 3d and data negatively regulated by hypoxia-inducible transcription not shown). factors as MCF-7 cells exposed to hypoxia expressed EVI-1 expression is linked to PI3Kinase-dependent lower levels of C17orf91 compared with the same cells activation of AKT and miR-22 overexpression is exposed to hypoxia under HIF1a, HIF2a or both expected to reduce AKT activation (Liu et al., 2006; depleted conditions (Figure 4c). In the same set of Nagaraja et al., 2010). We examined this possibility by experiments, EVI-1 expression under hypoxic conditions measuring phospho-AKT levels in MD-231P-pQXIN was lower in cells depleted of HIF2a and both HIF1a and MD-231P-miR-22 cells. Basal phospho-AKT levels and HIF2a compared with cells depleted of HIF1a or were substantially lower in MD-231P-miR-22 cells treated with control siRNA (Figure 4c). compared with MD-231P-pQXIN cells (Figure 3e). We also note that 231-BR cells, which express lower levels of EVI-1 expression in ERa-negative breast cancers is miR-22, but higher levels of EVI-1 compared with MD- associated with poor outcome 231P cells, contain higher levels of pAKT (Figure 3e). The results of our in vitro studies showing clear negative MD-231P-pQXIN cells were more migratory compared correlation between miR-22 and EVI-1 expression in with MD-231P-miR-22 cells in scratch wound assay, metastatic breast cancer cells prompted additional which may be linked to differential levels of pAKT in evaluation of EVI-1 for its relevance in breast cancer these cells (Figure 3f). Additionally, a quantitative prognosis. In ERa-negative subgroup of patients, who migration assay revealed B25% lower migration of often develop metastatic disease, higher expression of MD-231P-miR-22 cells compared with MD-231P- EVI-1 in tumors was associated with reduced relapse- pQXIN cells over a 24-h period (Figure 3f). Taken free, metastasis-free and overall survival (Figure 5a). In together, these results reveal specific changes in the ERa-positive subgroup, EVI-1 expression did not show expression levels of oncogenic proteins in metastatic any correlation with patient outcome (Figure 5b). When variants compared with primary tumor cells in our all patients were considered, elevated EVI-1 expression model system. Some of these changes may be linked to was associated with reduced metastasis-free but not differential miRNA expression. overall survival. For unknown reason, higher EVI-1 expression appeared to correlate with improved relapse- Elevated expression levels of C17orf91, which harbor free survival (Figure 5c). When the analysis was miR-22, correlate with better outcome, luminal phenotype restricted to truly prognostic data sets, elevated EVI-1 and lower ERBB3 and EVI-1 levels in primary tumors expression was associated with poor distant metastasis- Previous studies with limited breast cancer samples have free (P ¼ 0.031) but not recurrence-free or overall shown lower miR-22 expression in ERBB2-positive survival. Overall, our results showed an association breast cancers compared with ERBB2-negative tumors between reduced miR-22/C17orf91 levels, elevated and higher expression in ER and PR tumors þ þ EVI-1 and poor patient outcome. compared with ERÀ and PRÀ tumors, respectively; however, its expression status in basal or triple negative breast cancer is unknown (Mattie et al., 2006). We took advantage of the fact that miR-22 is located in the Discussion 50-untranslated region of a transcript called C17orf91 on chromosome 17 and utilizes the enhancer/promoter There is an explosion of data emerging from labora- region of this gene for transcription to indirectly tories worldwide describing the pivotal role of miRNAs determine the expression levels of miR-22 in published in tumorigenesis. From their discovery in 1997 to the gene expression databases (Chang et al., 2008). We used present, over 700 such molecules have been identified the recently reported online tool to correlate C17orf91 (O’Day and Lal, 2010). These non-coding RNAs have

Oncogene Breast cancer metastasis JB Patel et al 1296 Recurrence free SurvivalOverall Survival Metastasis-free Survival 1.0 1.0 1.0 HR = 0.81 (0.69 - 0.94) HR = 0.82 (0.54 - 1.23) HR = 0.76 (0.56 - 1.01) Iogrank P = 0.0047 Iogrank P = 0.33 Iogrank P = 0.06 0.8 0.8 0.8

0.6 0.6 0.6

0.4 0.4 0.4 Probability Probability Probability

0.2 0.2 0.2 low low low 0.0 high 0.0 high 0.0 high 0 5 10 15 20 0 5 10 15 20 25 0 5 10 15 20 25 Time (in years) Time (in years) Time (in years) Numbers at risk Numbers at risk Numbers at risk 796 455 144 36 1 180 150 76 39 2 0383 263138 49 3 0 796 507 206 11 0 177 146 52 17 1 0 383 294 117 20 00

C17orf91 (mir-22) EVI-1 9 6 p=0.0001 2.35 39

2.30 8 2.25 7 2.20 6 2.15 Expression levels 5 2.10 36 4 2.05 p=0.013 Normal Non-Basal BRCA1 Basal Normal Non-Basal BRCA1 Basal

C17orf91 (miR-22) EVI-1 600 180

500 160

400 140

300 120 Expression levels 200 100

p=0.0001 100 p=0.0001 80 Control HIF1α HIF2α Both Control HIF1α HIF2α Both depleted depleted depleted depleted depleted depleted Figure 4 Relationship between C17orf91 (miR-22) and EVI-1 expression and breast cancer outcome. (a) Kaplan–Meier survival analysis showing signiﬁcantly improved recurrence-free survival of patients with tumors that express higher levels of C17orf91, which harbors miR-22 (left panel). Overall survival (center panel) and distant metastasis-free survival (right panel) are shown; only metastasis-free survival showed a trend of positive correlation with elevated C17orf21 expression. Number of patients with higher expression (grey) and lower expression (black) are indicated along with number of patients at risk at speciﬁc time points. All data sets irrespective of systemic therapy were included in the analysis. (b) Basal breast cancers express lower levels of C17orf91 compared with normal breast and non-basal breast cancers (left panel); the same cancer type expresses higher levels of EVI-1, an miR-22 target (right panel). (c) Hypoxia-inducible factors regulate C17orf91 and EVI-1 expression in opposite ways. The expression levels of C17orf91 were higher in MCF-7 breast cancer cells depleted of hypoxia-inducible factors HIF1a, HIF2a or both (left panel). In contrast, HIF2a- depleted or HIF1a and HIF2a-depleted cells show lower EVI-1 expression (right panel).

been linked to all aspects of cancer from initiation, altered expression of genes such as SOX4, neuropilin 1 proliferation, motility and invasion to metastasis (Hurst and semaphoring 3C, which control progenitor cell et al., 2009; Inui et al., 2010). They are thought to development and migration (Hoser et al., 2007). Studies function much like proto-oncogenes and tumor sup- on miR-126 show the context-speciﬁc actions in pressor genes, whereby alterations in their expression tumorigenesis. In breast cancer, loss of miR-126 leads lead to aberrant cellular behavior. Examples of tumor to increased cell proliferation without effects on suppressor miRNAs include let-7, miR-335, miR-126 motility, whereas in non-small cell lung cancers, loss of and miR-206 (Tavazoie et al., 2008). The downregula- miR-126 leads to increased motility without effects on tion of these miRNAs had many effects, but the most proliferation (Nicoloso et al., 2009). Other such predominant function is increasing the metastatic miRNAs include miR-29c involvement in modulating potential via upregulation of genes involved in motility the tumor microenvironment and miR-146 modulation and proliferation. Speciﬁcally, miR-335 loss leads to of amoeboid motility (Friedl and Wolf, 2003;

Oncogene Breast cancer metastasis JB Patel et al 1297

Figure 5 Relationship between EVI-1 expression and breast cancer outcome. (a) Kaplan–Meier curve showing signiﬁcantly worse recurrence-free, overall and distant metastasis-free survival of patients with ERa-negative tumors expressing higher levels of EV-1. (b) Survival analysis of patients with ERa-positive tumors expressing high and low levels of EVI-1. (c) Survival analysis of all patients with tumors expressing high or lower levels of EVI-1.

Ramaswamy et al., 2003). Several other miRNAs have metastasis is well known (O’Day and Lal, 2010); been described to have a function in metastasis, but not therefore, we did not place much emphasis on this primary tumor growth (Hurst et al., 2009). miR-31 is miRNA. However, miR-22 is a relatively new player and one among them, which is downregulated in metastatic has been linked to cellular metabolism and chronic lesions compared with primary tumors; however, we did inﬂammatory diseases such as osteoarthritis (Iliopoulos not observe a change in miR-31 expression between et al., 2008). Normal mammary stem cells express high primary tumors and metastatic cells because the cell levels of miR-22 and little-to-no let-7; however, during lines used in our study did not express this miRNA progressive differentiation, the expression of miR-22 (Valastyan et al., 2009). decreases and expression of let-7 family members is From our studies, we found a strong correlation induced (Ibarra et al., 2007). These results suggest a role between decreased expression of let-7f and miR-22 and for miR-22 in maintaining a stem cell phenotype in the breast cancer metastasis. From their downregulation, we normal mammary gland (Ibarra et al., 2007). However, can infer that these miRNAs have a metastasis it appears that the role of miR-22 is different in a suppressor function; loss of their expression leads to metastasis setting. increased expression of their downstream oncogenic miR-22 targets the extracellular environment via products. The role of let-7f in tumorigenesis and matrix metalloproteinases and interleukins and thus

Oncogene Breast cancer metastasis JB Patel et al 1298 has a function in inflammatory conditions (Iliopoulos expression arrays from primary breast tumor samples et al., 2008). Putative downstream targets of miR-22 suggest a role for this protein in breast cancer include ERBB3, CDC25C, EVI-1, SIRT1, AKT3, metastasis. As EVI-1 is a transcription factor capable SATB2, HDAC4, CBL, peroxisome proliferator acti- of causing epigenetic changes through recruitment of vated receptor a and BMP7. Peroxisome proliferator histone methyltransferases (Goyama et al., 2010), a activated receptor a and BMP7 have been validated in modest alteration in its expression levels may be other studies as targets of miR-22 (Iliopoulos et al., sufficient to cause a major shift in gene expression 2008); however, we did not find them to be targets of programs, helping cancer cells to adapt to sites of miR-22 in breast cancer cells (data not shown). ERBB3 metastasis. and EVI-1 have recently been shown to be targets of While most of our studies were confined to miR-22, miR-22 in clear cell ovarian cancer (Nagaraja et al., we also observed several other miRNAs, whose altered 2010). Overexpression of miR-22 reduced the levels of expression may potentially contribute to metastasis. One ERBB3 and EVI-1 in breast cancer cells confirming that example is miR-200a. Reduced expression of miR-200a they both were targets of miR-22. Collectively, it in primary tumors is essential for cancer cells to acquire appears that potent oncogenes linked to breast cancer invasive phenotype through epithelial-to-mesenchymal progression are the targets of miR-22. Although it is transition (Gibbons et al., 2009), whereas its over- extremely unlikely that only miR-22 regulates the levels expression in mammary tumor cell lines enhances of these proteins in metastatic cancer cells compared mesenchymal-to-epithelial transition and increases with parental tumor cells, several organ-specific variants macrometastasis (Dykxhoorn et al., 2009). Consistent expressed higher levels of miR-22 targets compared with with this possibility, we observed elevated miR-200a in primary tumor cells. For example, ERBB3 levels were metastatic cancer cells compared with parental cells or higher in the lung and adrenal metastatic clones TMD-231 cells. miR-17-5P has been suggested as a compared with MD-231 cells. ERBB3, EVI-1 and tumor suppressor in breast cancer; its expression was CDC25C levels were higher in 231-BR cells compared lower in lung metastatic cancer cells compared with with MD-231P cells. All metastatic cancer cells ex- other cell types (O’Day and Lal, 2010). The protumori- pressed higher levels of EGFR both at the transcript and genic and prometastatic miRNA miR-19b was upregu- protein levels; therefore, EGFR:ERBB3 heterodimers lated in metastatic cells (Table 2), whereas the may be predominant in metastatic cancer cells compared antimetastatic miR-16 was downregulated in lung with tumors cells. metastatic cells (Table 3) (Hurst et al., 2009). However, ERBB3 lacks kinase activity and, therefore, it received there were few exceptions: metastatic cells expressed little attention until recently. It is now believed to have a elevated levels of miR-101, which has been described as significant function in conferring resistance to growth tumor suppressor in other cancer types (Varambally factor receptor tyrosine kinase inhibitors (Sergina et al., et al., 2008). In summary, we have described results that 2007). Whether metastatic cancer cells are more resistant suggest a role for miRNAs in changing the protein to these inhibitors compared with primary tumors cells expression pattern in cancer cells depending on the site remains to be determined. PI3Kinase/AKT is one of the of metastasis. Further studies are required to investigate major downstream targets of ERBB3 (Campbell et al., the impact of manipulating individual miRNAs on 2010). Interestingly, EVI-1 also activates this pathway; organ-specific metastasis and sensitivity to specific activation of PI3 kinase/AKT is linked to drug therapies. resistance (Liu et al., 2006). It is likely that two redundant pathways of PI3Kinase/AKT activation in metastatic cancer cells make these cells highly resistant to chemotherapeutic drugs. Indeed, in our in vitro Materials and methods studies, the metastatic variants were resistant to doxorubicin compared with parental cells (data not Cell lines shown). However, miR-22 overexpression alone did not MDA-MB-231, TMD-231, BMD-231, LMD-231, ADMD-231 and 231-BR cells were maintained in MEM media supple- change sensitivity of MD-231P cells to doxorubicin or mented with 10% fetal bovine serum, 10 nM insulin and docetaxel, suggesting the involvement of additional penicillin/streptomycin. Our metastasis model involved injec- signaling networks in resistance to chemotherapy (data tion of parental MDA-MB-231 into the mammary fat pad of not shown). nude mice. After B8 weeks time, the mice were killed, the In addition to its role in PI3Kinase/AKT activation, primary tumors were harvested and select distant sites EVI-1 is a major epigenetic regulator of gene expression including bone, lung, adrenal and brain were examined for and mainly targets genes associated with stemness evidence of metastatic disease. Using this model system, only phenotype (Kumano and Kurokawa, 2010). EVI-1 lung metastatic (LMD-231) cells could be obtained (Helbig induces PBX1, one of the major stemness-associated et al., 2003). Therefore, a slight variation on the previous gene (Shimabe et al., 2009). Therefore, reduced miR-22 model system was implemented. TMD-231 cells were injected via intracardiac route and after 8 weeks, the mice were killed, expression in metastatic cells may lead to elevated and BMD-231, ADMD-231 and 231-BR metastatic cell lines expression of stemness-associated genes. EVI-1 has were successfully generated. Two variants of LMD-231 were previously been linked to myeloid leukemia and used for microarray analysis, one that had metastasized from myelodysplastic syndrome (Wieser, 2007). Our studies the mammary fat pad to the lungs and the other via in breast cancer cell lines, as well as the analysis of gene intracardiac injection. Similarly, three clones of BMD-231

Oncogene Breast cancer metastasis JB Patel et al 1299 were used for microarray analysis—one isolated from the lined sequences are cloning restriction enzyme sites) and mandible, one from the vertebral column and the other from cloned into BamH1-EcoR1 sites of the bicistronic retrovirus the tibia; all generated via intracardiac injection. Typically, vector pcQXIN. Retrovirus packaging, transduction and metastatic lesions or whole organs were minced and incubated selection of cells overexpressing miR-22 have been described in 50 ml media containing collagenase IV (0.72 mg/ml) and previously (Chua et al., 2007). Cell migration assay was hyaluronidase (0.2 mg/ml) for 3 h. Cells were washed in PBS performed using CytoSelect 24-well Cell Migration assay kit as and plated in regular media. After 2–3 days, cells were per instruction from manufacturers (Cell Biolabs, Inc., San incubated in serum-free media for 2 days to reduce the number Diego, CA, USA). of fibroblasts. Additionally, flow cytometry with EpCAM confirmed that 495% of cells are cancer cells (data not Western blot analysis shown). As 231-BR cells were developed in a different Whole cell lysates were prepared in RIPA buffer and western laboratory, the expression pattern of miRNAs and their blotting was performed as previously described (Bhat-Nak- targets in 231-BR was compared with parental cells main- shatri et al., 2004). The primary antibodies used were c-Myc, tained in that laboratory (MD-231P) (Gril et al., 2008). EGFR, TLR4, ERBB2, ERBB3 (Santa Cruz Biotechnology, Santa Cruz, CA), AKT3, EVI-1 (Cell Signaling, Danvers, MA, miRNA expression analysis USA) and b-actin (Sigma Chemical Co., St Louis, MO, USA). The small RNA fraction from all cell lines was prepared using the mirVana miRNA isolation kit (Ambion Inc., Austin, TX, USA). Hybridization to miRNA arrays has been described Statistical analysis previously (Yu et al., 2007). Assays were carried out in All experiments were performed at least in duplicate. The duplicate with RNA from MD-231, TMD-231, ADMD-231, results of qRT–PCR were analyzed using GraphPad software two independent clones of LMD-231 and three independent (Graphpad.com). Analysis of variance was used to determine clones of BMD-231. As all of the probes were printed twice on the P-values between mean measurements. A P-value of o0.05 the microarray in two independent blocks, four measurements was deemed statistically significant. Error bars on all were available for every miRNA in each experimental histograms represent the standard error of the mean. condition. Signal intensity of tRNA-Thr was used for normalization between samples. Differentially expressed miRNAs Expression analysis of primary tumors were identified using a linear model that describes the Relationship between the expression levels of C17orf91, relationship between cell type and probe blocks. Changes in ERBB3 and EVI-1 in different breast cancer subtypes as well let-7f and miR-22 expression were verified in three independent as in MCF-7 cells depleted of hypoxia-inducible factors was RNA preparations by qRT–PCR using TaqMan miRNA determined using the expression array data sets in Gene assays designed to detect and accurately quantify mature Expression Omnibus (Elvidge et al., 2006). Relationship miRNAs (Applied Biosystems, Foster City, CA, USA). between EVI-1 expression and breast cancer outcome was Primers specific to miR-22 (catalog #000398), miR-424 determined using the gene expression data of 22 277 genes in (catalog #000604), let-7f (catalog #000382) and RNU66 1809 breast cancer patients (Gyorffy et al., 2010). All or only (catalog number 001002) were purchased from Applied true prognostic data sets (no systemic therapy) were included Biosystems. Primers specific for 5S (catalog #201509) were in the analysis. These data sets are described in Gene purchased from Exiqon (Vedbaek, Denmark). Target genes Expression Omnibus accession numbers GSE11121 (Schmidt of differentially expressed miRNAs were predicted using et al., 2008), GSE7390 (Desmedt et al., 2007) and select data TargetScan and miRgen programs (Lewis et al., 2005). from GSE3494, GSE2990 and GSE2034 (Miller et al., 2005; Wang et al., 2005; Sotiriou et al., 2006). C17orf91 expression RT–PCR and quantitative real-time RT–PCR was analyzed using the affymetrix probe set 214696_at Independent samples of total RNA were prepared using the (excellent probe set), whereas EVI-1 (MECOM) was measured RNAeasy kit (Qiagen, Valencia, CA, USA). First strand using the probe set 221884_at (intermediate probe set). cDNA was synthesized using random hexamers and super- script II reverse transcriptase (Invitrogen, Carlsbad, CA, Conflict of interest USA). To quantify the level of EGFR mRNA, qRT–PCR was performed using the SyBr green mix according to the The authors declare no conflict of interest. manufacturer’s protocol (Applied Biosystems). Expression of b-actin was used as an internal control. Sequences of primers used will be provided upon request. Acknowledgements

Generation of miR-22 overexpressing cell line and This work is supported by Indiana University Simon Cancer migration assay Center Pilot grant and by Komen for Cure grant Genomic DNA harboring mature miR-22 sequences were BCTR0601111 to HN. JBP and RMB are supported by ampliﬁed by PCR using the primers 50-GGGGGATCCCT National Institutes of Health Training Grants T32 DK07519 GGGGCAGGACCCT-30 and GGGGAATTCAACGTATC and T32 CA111198, respectively. HN is Marian J Morrison ATCCACCC-30 (chromosome 17:1563850–1564149, under- Professor of Breast Cancer Research.

References

Bhat-Nakshatri P, Campbell RA, Patel NM, Newton TR, King AJ, Bhat-Nakshatri P, Wang G, Collins NR, Thomson MJ, Geistlinger Marshall MS et al. (2004). Tumour necrosis factor and PI3-kinase TR, Carroll JS et al. (2009). Estradiol-regulated microRNAs control oestrogen receptor alpha protein level and its transrepres- control estradiol response in breast cancer cells. Nucleic Acids Res sion function. Br J Cancer 90: 853–859. 37: 4850–4861.

Oncogene Breast cancer metastasis JB Patel et al 1300 Blenkiron C, Goldstein LD, Thorne NP, Spiteri I, Chin SF, Dunning Inui M, Martello G, Piccolo S. (2010). MicroRNA control of signal MJ et al. (2007). MicroRNA expression profiling of human breast transduction. Nat Rev Mol Cell Biol 11: 252–263. cancer identifies new markers of tumour subtype. Genome Biol 8: Kang Y, Siegel PM, Shu W, Drobnjak M, Kakonen SM, Cordon- R214. Cardo C et al. (2003). A multigenic program mediating breast Campbell MR, Amin D, Moasser MM. (2010). HER3 comes of age: cancer metastasis to bone. Cancer Cell 3: 537–549. new insights into its functions and role in signaling, tumor biology, Koshiol J, Wang E, Zhao Y, Marincola F, Landi MT. (2010). and cancer therapy. Clin Cancer Res 16: 1373–1383. Strengths and limitations of laboratory procedures for microRNA Castellano L, Giamas G, Jacob J, Coombes RC, Lucchesi W, detection. Cancer Epidemiol Biomarkers Prev 19: 907–911. Thiruchelvam P et al. (2009). The estrogen receptor-alpha-induced Kumano K, Kurokawa M. (2010). The role of Runx1/AML1 and Evi- microRNA signature regulates itself and its transcriptional 1 in the regulation of hematopoietic stem cells. J Cell Physiol 222: response. Proc Natl Acad Sci USA 106: 15732–15737. 282–285. Chang TC, Yu D, Lee YS, Wentzel EA, Arking DE, West KM et al. Lewis BP, Burge CB, Bartel DP. (2005). Conserved seed pairing, often (2008). Widespread microRNA repression by Myc contributes to flanked by adenosines, indicates that thousands of human genes are tumorigenesis. Nat Genet 40: 43–50. microRNA targets. Cell 120: 15–20. Chua HL, Bhat-Nakshatri P, Clare SE, Morimiya A, Badve S, Liu Y, Chen L, Ko TC, Fields AP, Thompson EA. (2006). Nakshatri H. (2007). NF-kappaB represses E-cadherin expression Evi1 is a survival factor which conveys resistance to both and enhances epithelial to mesenchymal transition of mammary TGFbeta- and taxol-mediated cell death via PI3K/AKT. Oncogene epithelial cells: potential involvement of ZEB-1 and ZEB-2. 25: 3565–3575. Oncogene 26: 711–724. Maillot G, Lacroix-Triki M, Pierredon S, Gratadou L, Schmidt S, Desmedt C, Piette F, Loi S, Wang Y, Lallemand F, Haibe-Kains B Benes V et al. (2009). Widespread estrogen-dependent repression of et al. (2007). Strong time dependence of the 76-gene prognostic micrornas involved in breast tumor cell growth. Cancer Res 69: signature for node-negative breast cancer patients in the TRANS- 8332–8340. BIG multicenter independent validation series. Clin Cancer Res 13: Mattie MD, Benz CC, Bowers J, Sensinger K, Wong L, Scott GK et al. 3207–3214. (2006). Optimized high-throughput microRNA expression profiling Dykxhoorn DM, Wu Y, Xie H, Yu F, Lal A, Petrocca F et al. (2009). provides novel biomarker assessment of clinical prostate and breast miR-200 enhances mouse breast cancer cell colonization to form cancer biopsies. Mol Cancer 5: 24. distant metastases. PLoS One 4: e7181. Miller LD, Smeds J, George J, Vega VB, Vergara L, Ploner A et al. Elvidge GP, Glenny L, Appelhoff RJ, Ratcliffe PJ, Ragoussis J, (2005). An expression signature for p53 status in human breast Gleadle JM. (2006). Concordant regulation of gene expression by cancer predicts mutation status, transcriptional effects, and patient hypoxia and 2-oxoglutarate-dependent dioxygenase inhibition: the survival. Proc Natl Acad Sci USA 102: 13550–13555. role of HIF-1alpha, HIF-2alpha, and other pathways. J Biol Chem Nagaraja AK, Creighton CJ, Yu Z, Zhu H, Gunaratne PH, Reid JG 281: 15215–15226. et al. (2010). A link between mir-100 and FRAP1/mTOR in clear Friedl P, Wolf K. (2003). Tumour-cell invasion and migration: cell ovarian cancer. Mol Endocrinol 24: 447–463. diversity and escape mechanisms. Nat Rev Cancer 3: 362–374. Nicoloso MS, Spizzo R, Shimizu M, Rossi S, Calin GA. (2009). Gibbons DL, Lin W, Creighton CJ, Rizvi ZH, Gregory PA, Goodall MicroRNAs—the micro steering wheel of tumour metastases. Nat GJ et al. (2009). Contextual extracellular cues promote tumor cell Rev Cancer 9: 293–302. EMT and metastasis by regulating miR-200 family expression. O’Day E, Lal A. (2010). MicroRNAs and their target gene networks in Genes Dev 23: 2140–2151. breast cancer. Breast Cancer Res 12: 201. Goyama S, Nitta E, Yoshino T, Kako S, Watanabe-Okochi N, Pandey DP, Picard D. (2009). miR-22 inhibits estrogen signaling by Shimabe M et al. (2010). EVI-1 interacts with histone methyl- directly targeting the estrogen receptor alpha mRNA. Mol Cell Biol transferases SUV39H1 and G9a for transcriptional repression and 29: 3783–3790. bone marrow immortalization. Leukemia 24: 81–88. Ramaswamy S, Ross KN, Lander ES, Golub TR. (2003). A molecular Gril B, Palmieri D, Bronder JL, Herring JM, Vega-Valle E, signature of metastasis in primary solid tumors. Nat Genet 33: Feigenbaum L et al. (2008). Effect of lapatinib on the outgrowth 49–54. of metastatic breast cancer cells to the brain. J Natl Cancer Inst 100: Richardson AL, Wang ZC, De Nicolo A, Lu X, Brown M, Miron A 1092–1103. et al. (2006). X chromosomal abnormalities in basal-like human Gupta GP, Minn AJ, Kang Y, Siegel PM, Serganova I, Cordon-Cardo breast cancer. Cancer Cell 9: 121–132. C et al. (2005). Identifying site-specific metastasis genes and Schmidt M, Bohm D, von Torne C, Steiner E, Puhl A, Pilch H et al. functions. Cold Spring Harb Symp Quant Biol 70: 149–158. (2008). The humoral immune system has a key prognostic impact in Gyorffy B, Lanczky A, Eklund AC, Denkert C, Budczies J, Li Q et al. node-negative breast cancer. Cancer Res 68: 5405–5413. (2010). An online survival analysis tool to rapidly assess the effect of Sergina NV, Rausch M, Wang D, Blair J, Hann B, Shokat KM et al. 22 277 genes on breast cancer prognosis using microarray data of (2007). Escape from HER-family tyrosine kinase inhibitor therapy 1809 patients. Breast Cancer Res Treat 123: 725–731. by the kinase-inactive HER3. Nature 445: 437–441. Helbig G, Christopherson II KW, Bhat-Nakshatri P, Kumar S, Shimabe M, Goyama S, Watanabe-Okochi N, Yoshimi A, Ichikawa Kishimoto H, Miller KD et al. (2003). NF-kappaB promotes breast M, Imai Y et al. (2009). Pbx1 is a downstream target of Evi-1 in cancer cell migration and metastasis by inducing the expression of hematopoietic stem/progenitors and leukemic cells. Oncogene 28: the chemokine receptor CXCR4. J Biol Chem 278: 21631–21638. 4364–4374. Hoser M, Baader SL, Bosl MR, Ihmer A, Wegner M, Sock E. (2007). Sotiriou C, Wirapati P, Loi S, Harris A, Fox S, Smeds J et al. (2006). Prolonged glial expression of Sox4 in the CNS leads to architectural Gene expression profiling in breast cancer: understanding the cerebellar defects and ataxia. J Neurosci 27: 5495–5505. molecular basis of histologic grade to improve prognosis. J Natl Hurst DR, Edmonds MD, Welch DR. (2009). Metastamir: the field Cancer Inst 98: 262–272. of metastasis-regulatory microRNA is spreading. Cancer Res 69: Steeg PS. (2006). Tumor metastasis: mechanistic insights and clinical 7495–7498. challenges. Nat Med 12: 895–904. Ibarra I, Erlich Y, Muthuswamy SK, Sachidanandam R, Hannon GJ. Sun M, Estrov Z, Ji Y, Coombes KR, Harris DH, Kurzrock R. (2008). (2007). A role for microRNAs in maintenance of mouse mammary Curcumin (diferuloylmethane) alters the expression profiles of epithelial progenitor cells. Genes Dev 21: 3238–3243. microRNAs in human pancreatic cancer cells. Mol Cancer Ther 7: Iliopoulos D, Malizos KN, Oikonomou P, Tsezou A. (2008). 464–473. Integrative microRNA and proteomic approaches identify novel Tavazoie SF, Alarcon C, Oskarsson T, Padua D, Wang Q, Bos PD osteoarthritis genes and their collaborative metabolic and inflam- et al. (2008). Endogenous human microRNAs that suppress breast matory networks. PLoS One 3: e3740. cancer metastasis. Nature 451: 147–152.

Oncogene Breast cancer metastasis JB Patel et al 1301 Valastyan S, Reinhardt F, Benaich N, Calogrias D, Szasz AM, Wang metastasis of lymph-node-negative primary breast cancer. Lancet ZC et al. (2009). A pleiotropically acting microRNA, miR-31, 365: 671–679. inhibits breast cancer metastasis. Cell 137: 1032–1046. Wieser R. (2007). The oncogene and developmental regulator EVI1: Varambally S, Cao Q, Mani RS, Shankar S, Wang X, Ateeq B et al. expression, biochemical properties, and biological functions. Gene (2008). Genomic loss of microRNA-101 leads to overexpression of 396: 346–357. histone methyltransferase EZH2 in cancer. Science 322: 1695–1699. Yu J, Vodyanik MA, Smuga-Otto K, Antosiewicz-Bourget J, Frane Wang Y, Klijn JG, Zhang Y, Sieuwerts AM, Look MP, JL, Tian S et al. (2007). Induced pluripotent stem cell lines derived Yang F et al. (2005). Gene-expression proﬁles to predict distant from human somatic cells. Science 318: 1917–1920.

Oncogene