microRNA 17/20 inhibits cellular invasion and tumor metastasis in breast cancer by heterotypic signaling

Abstract

microRNAs are thought to regulate tumor progression and invasion via direct interaction with target genes within cells. Here the microRNA17/20 cluster is shown to govern cellular migration and invasion of nearby cells via heterotypic secreted signals. microRNA17/20 abundance is reduced in highly invasive breast cancer cell lines and node-positive breast cancer specimens. Cell-conditioned medium from microRNA17/20–overexpressing noninvasive breast cancer cell MCF7 was sufficient to inhibit MDA-MB-231 cell migration and invasion through inhibiting secretion of a subset of cytokines, and suppressing plasminogen activation via inhibition of the secreted plasminogen activators (cytokeratin 8 and α-enolase). microRNA17/20 directly repressed IL-8 by targeting its 3′ UTR, and inhibited cytokeratin 8 via the cell cycle control protein cyclin D1. At variance with prior studies, these results demonstrated a unique mechanism of how the altered microRNA17/20 expression regulates cellular secretion and tumor microenvironment to control migration and invasion of neighboring cells in breast cancer. These findings not only reveal an antiinvasive function of miR-17/20 in breast cancer, but also identify a heterotypic secreted signal that mediates the microRNA regulation of tumor metastasis.

MicroRNAs (miRNAs) are 21- to 22-nucleotide molecules that regulate cellular phenotype through altering the stability or translational efficiency of targeted mRNAs (1). Important associations between miRNA gene expression and human cancer have implicated miRNA in human tumorigenesis (2–4). miRNA encoding genes are frequently located at fragile sites, common breakpoints, or minimal regions of amplification (5). The human miR-17/20 cluster, located on chromosome 13q31, undergoes loss of heterozygosity in several different cancers, including breast cancer. In the human B cell line P493-6, miR-17/20 inhibits cell growth via suppression of E2F1 expression (6). In mice, transgenic mice overexpressing miR-17 alone showed overall tissue growth retardation, smaller organs, and greatly reduced hematopoietic cell lineages (7). Analysis of miRNAs regulated in cyclin D1 transgenic mammary tumors and reciprocally regulated in cyclin D1–knockout mice identified the miR-17/20 cluster as an important regulatory switch in mammary tumor growth. miR-17/20 inhibited breast cancer tumor growth through repression of cyclin D1 expression via the 3′ UTR binding site (8). Recent evidence implicates miRNA in breast cancer metastasis by inhibiting target genes. miR-335 inhibited human breast cancer cell metastasis via repression of SOX4, a regulator of progenitor cell development and migration (9). miR-10b promotes cell migration in vitro and initiates breast tumor invasion in vivo by targeting gene HOXD10 (10). miR-200 family prevented epithelial to mesenchymal transition of breast cancer cells by repressing ZEB1 and SIP1 (11). However, the role of the miR-17/20 cluster in regulation of breast cancer metastasis remains unknown.

Tumor cell migration and invasion is in part dependent upon plasminogen activity (12), which is regulated by urokinase-type plasminogen activator (uPA), plasminogen activator inhibitor (PAI)–1, PAI-2, and binding to the secreted proteins cytokeratin 8 (CK8) (13) and α-enolase (α-ENO) (14). At variance with prior studies in which miRNA regulate migration and invasion via genes within the tumor epithelial cells (15), we demonstrate that the miR-17/20 cluster mediates breast cancer cell migration and invasion via a heterotypic secreted signal. miRNA expression within breast cancer epithelial cells conveys an antimetastatic phenotype that can be transmitted to other cell types.

Results and Discussion

Reduced miR-17/20 Abundance in Metastatic Breast Cancer.

To test the possibility that miR-17/20 is related to human breast cancer metastasis, human breast cancer specimens and human breast cancer cell lines were analyzed. miR-17/20 expression in lymph node–positive human mammary gland primary tumors and lymph node–negative human mammary gland primary tumors were examined with quantitative real-time PCR (Fig.1A) and confirmed by Northern blot. The miR-17/20 abundance in each tumor sample was normalized to that of the matching normal mammary gland sample from the same patient. Comparison between human mammary gland primary tumors with lymph nodes positive and negative indicated a significant inverse correlation between the miR-17–5p expression level and nodal metastatic status of the breast cancer (Fig.1A). A further analysis of miR-17/20 abundance in three highly invasive human breast cancer cell lines (MDA-MB-231, Hs578T, and SKBR3) and four noninvasive human breast cancer cell lines (BT474, MCF7, MDA-MB-468, and T-47D) were compared by miRNA Northern blot hybridization (Fig.1B). The quantitative analysis showed an overall higher expression of both miR-17–5p and miR-20a in the noninvasive breast cancer cell lines than that in highly invasive lines (Fig. S1).

Fig. 1.

miR-17/20–conditioned medium inhibits breast cancer cell migration and invasiveness. (A) Quantitative real-time PCR analysis of miR-17–5p abundance in human mammary gland primary tumors with lymph nodes positive (n = 10) or negative (n = 10). The matching normal breast tissue sample from the same patient was used for normalization. **P < 0.01. (B) Northern blot of miR-17–5p and miR-20a in highly invasive (MDA-MB-231, Hs578T, and SKBR3) and weakly invasive (BT474, MCF7, MDA-MB-468, and T-47D) human breast cancer cell lines as indicated. tRNAs served as loading control. (C) Three-dimensional collagen gel invasion assays. MDA-MB-231 cells were assayed for cellular invasiveness, and shown by fluorescence. Cells were cultured with conditioned medium from MCF7 cells either expressing a control vector (Ctrl CM) or transduced with the miR-17/20 cluster (miRNA CM). The relative distance of cell invasion was quantitated. CM, conditioned medium. (D) Boyden chamber assays. MDA-MB-231 cells were incubated either with control or miR-17/20–conditioned medium from MCF7 cells. Migrated cells were stained with crystal violet, and counted for quantitative analysis. (E) Wound healing assays. Cell culture plates attached with MDA-MB-231 cells were wounded and incubated with control or miR-17/20–conditioned medium. The wound healing was determined at the time points as indicated. Data are presented as mean ± SEM (n = 3). **P < 0.01.

miR-17/20 Conditioned Medium Suppressed Breast Cancer Cell Migration and Invasion.

To determine whether miR-17/20 reduces cellular migration and invasion directly, MDA-MB-231 and MCF-7 cells were examined. Both cell lines were transduced with the miR-17/20 cluster or vector control. The overexpression of miR-17 and miR-20a was shown in Fig. S2. Cell migration assays were performed on the miR-17/20 transduced cells. No significant difference was observed between the miR-17/20 overexpressing MDA-MB-231 cells and control cells.

The miR-17/20 conditioned cell medium was assayed for migration and invasion. Surprisingly, we found that the miR-17/20 conditioned medium from MCF7 cells reduced the invasion of MDA-MB-231 breast cancer cells in 3D collagen invasion assays (Fig. 1C) and reduced cell migration of MDA-MB-231 in Boyden chamber assays (Fig. 1D). Wound healing assays demonstrated the addition of conditioned medium from miR-17/20 transduced MCF7 breast cancer cells reduced the ability of MDA-MB-231 cells to migrate into the wound (Fig. 1E). We fractioned the miR-17/20 conditioned medium using a P10 column (protein cut off at 10 kDa molecular weight), and performed the cell migration assays. Both the high molecular weight fraction (proteins >10 kDa) and low molecular weight fraction (proteins <10 kDa) inhibited breast cancer cell migration, although the former had more inhibition than the latter, indicating multiple factors in the miR-17/20 conditioned medium are involved in the inhibition of cell migration and invasion. The miR-17/20 conditioned medium did not change the self-expression pattern of miR-17/20 in cells (Fig. S3). These observations suggest miR-17/20 regulates cellular secretion, alters cellular microenvironment, and regulates cell migration and invasion via a heterotypic signal.

miR-17/20 Reduces Cell Secretion.

To identify the secreted factors within the medium regulated by miR-17/20 transduction, MALDI-TOF MS analysis was conducted of miR-17/20 transduced MCF7 cell medium. Medium derived from vector transduced MCF7 cells served as control. The proteins recovered from the gel were subjected to in-gel tryptic digestion. Sequential MS/MS peptide sequencing of the excised bands at approximately 40 to 50 kDa identified the proteins α-ENO and CK8 (Fig. 2A and Figs. S4 and S5). CK8 and α-ENO are secreted by human breast cancer cells (13, 16). Western blot confirmed the reduction in α-ENO and CK8 protein abundance in miR-17/20 conditioned MCF7 cell medium (Fig. 2B). Several other protein bands showing different abundance in the miR-17/20–overexpressing MCF7 cell medium were analyzed by MS as well; unfortunately, the identities of these proteins were not validated yet.

Fig. 2.

miR-17/20 decreased cellular secretion. (A) Proteomic analysis of the miR-17/20 or control conditioned medium from MCF7 cells. SDS/PAGE–MS analysis of two bands that were differentially expressed in the miR-17/20–conditioned medium identified two proteins, Y1 and Y2. Y1 corresponds to α-ENO, and Y2 corresponds to CK8. (B) Western blot analysis of control or miR-17/20–conditioned medium for proteins as indicated. The abundance of α-ENO and CK8 decreased in the miR-17/20–conditioned medium. (C) Human cytokine antibody array analysis of the conditioned medium from either control or miR-17/20–transduced MCF7 cells indicated altered abundance of cytokines CXCL1, IL-10, IL-8, and NT-4. (D) Bioinformatic analysis demonstrated an overlap between predicted targets of miR-17/20 (n = 959) and known cellular secreted factors implicated in cellular migration (IL-8, VEGFA, LAMA3, and NTN1). The overlap between the predicted targets of miR-17/20 and the cytokines secreted in the miR-17/20–conditioned medium identified IL-8 as a target. ELISAs for IL-8 (E) or CXCL1 (F) confirmed their decreased abundance in the miR-17/20–conditioned medium. Data are presented as mean ± SEM (n = 5).

A human cytokine antibody array was used to examine further the miR-17/20 conditioned medium. Subtractive analysis identified secreted cytokines (CXCL1, IL-10, IL-8, and NT4) which were decreased in the miR-17/20 conditioned medium (Fig. 2C). This assay was repeated two times with independent medium samples. Only those spots on the array showing decreased level in the miR-17/20 conditioned medium of both experiments were chosen to follow up.

A bioinformatic screening analysis using miRNA TargetScan and Gene Ontology annotation tools identified four cell-secreted factors that are predicted targets of miR-17/20 and are known to be involved in the regulation of cell migration (VEGFA, NTN1, IL-8, and LAMA3; Fig. 2D). Quantitative analyses by ELISA confirmed that IL-8 and CXCL1 were reduced in miR-17/20–transduced breast cancer cell medium (Fig. 2 E and F).

To determine whether the decreased abundance of those cellular secreted factors in miR-17/20 conditioned medium governs cell migration and invasion, add-back experiments were conducted using transwell migration (Fig. 3A) and 3D invasion assays (Fig. 3B). IL-8, CK8, or CXCL1 either partially or fully reverted the anti-migratory and anti-invasive phenotypes of miR-17/20 conditioned medium.

Fig. 3.

miR-17/20 suppresses cell migration and invasion by altering cytokine secretion (IL-8, CK8, and CXCL1). (A) Boyden chamber assays of MDA-MB-231 cell migration or (B) 3D invasion assays examining the effects of miR-17/20–conditioned medium and the addition of specific cytokines. The addition of CK8 (100 ng/mL), IL-8 (50 ng/mL), or CXCL1 (50 ng/mL) partially or fully reversed the antimigratory and antiinvasive phenotype of miR-17/20–conditioned medium. Data are presented as mean ± SEM (n = 3).

miR-17/20 Inhibits Plasminogen Activation.

Tumor cell migration and invasion is in part dependent upon plasminogen activity (12). CK8 and α-ENO bind and activate plasminogen to promote cell migration and invasion (13, 14, 17). The reduction of CK8 and α-ENO abundance in the miR-17/20 conditioned medium led to a 40% to 50% reduction in plasminogen activity (Fig. 4A). As neither CK8 nor α-ENO contains a miR-17/20 binding site within their 3′ UTRs, we considered the effect may be indirect, and examined the validated targets of miR-17/20. Cyclin D1 is one of the downstream targets of miR-17/20 in MCF7 cells via the 3′ UTR binding (8). Cyclin D1 regulates expression of target genes through binding to target gene promoters in the context of local chromatin (18). To investigate the mechanism by which miR-17/20 reduced CK8 abundance, we considered the effect of miR-17/20 on CK8 may be indirect via cyclin D1. MCF7 cells were treated with cyclin D1 siRNA, the conditioned medium was collected, and functional assays were conducted. CK8 abundance was reduced in the medium from cyclin D1 siRNA treated MCF7 cells (Fig. 4B). α-ENO abundance did not show a change in cyclin D1 siRNA conditioned medium. Plasminogen activity was reduced in cyclin D1 siRNA–treated cell-conditioned medium (Fig. 4C), consistent with the reduction in CK8 abundance in the medium. These findings suggest the repression of cyclin D1 by miR-17/20 (8) reduces CK8 level in the conditioned medium, which results in the reduction of plasminogen activation.

Fig. 4.

miR-17/20 suppresses CK8 secretion and thereby plasminogen activity through cyclin D1. (A) Plasminogen activity in control and miR-17/20–conditioned MCF7 cell medium. Data are mean ± SEM (n = 4), **P < 0.01. CM, conditioned medium; Pg, plasminogen; PA, plasminogen activator; Pm, plasmin; VLK-pNA, Val-Leu-Lys-4-nitroanilide; NA, nitrophenolate anion. (B) Western blot of the cyclin D1 siRNA conditioned medium and cell lysates demonstrated reduction of cyclin D1 in MCF7 cells and decrease of CK8 secretion. Cyclin D1 siRNA did not affect α-ENO secretion. (C) Plasminogen activation was assessed in the conditioned medium from cyclin D1–deficient or control cells. Data are mean ± SEM (n = 3). **P < 0.01. (D) Western blot showed the cyclin D1 and CK8 abundance in breast cancer tumor samples and matching normal tissues. T, tumor sample; N, normal tissue sample.

Cyclin D1 and CK8 abundance were determined in breast cancer tumor samples as well as matched normal tissue from the same patient, as shown in Fig. 4D. Compared with normal tissues, the cyclin D1 and CK8 expression was increased and correlated positively in the breast tumor samples.

To determine the relative importance of plasminogen activity on cell migration, PAI-1 was added to the empty vector–transduced MCF7 cell–conditioned medium. PAI-1 inhibited MDA-MB-231 cell migration as miR-17/20 conditioned medium did (Fig. S6).

miR-17/20 Inhibits IL-8 Through 3′ UTR Binding.

IL-8 abundance was decreased in the miR17/20 conditioned medium. A highly conserved sequence complementary to the “seed sequence” of miR-17–5p and miR-20a was identified within the IL-8 3′ UTR (Fig. 5A). miR-17/20 repressed IL-8 luciferase reporter gene activity, and mutation of the miR-17/20 binding site of the IL-8 3′ UTR reduced the repression (Fig. 5 B and C).

Fig. 5.

miR-17/20 inhibits IL-8 expression through 3′ UTR interaction. (A) Schematic representation of the human IL-8 3′ UTR showing the highly conserved miR-17/20 binding site (in italics and boxed) between species. The “seed” sequence of miR-17/20 (nt 2–10) is indicated in blue. (B) The pGL3 reporter vectors carrying the wide type or mutated IL-8 3′ UTR are indicated. (C) Luciferase reporter assay showed the suppression of luciferase activity by IL-8 3′ UTR. In the miR-17/20–overexpressing MCF7 cells, the luciferase activity was suppressed more than that in control cells. But for the mutated construct in which the miR-17/20 binding site was mutated, the luciferase activity was significantly rescued compared with the WT construct in the miR-17/20 transduced cells. Values are equal to mean ± SEM (n = 6). **P < 0.01. (D) Schematic representation of the hypothetical molecular mechanisms by which miR-17/20 regulates breast cancer cellular migration and invasion. Pg, plasminogen; PA, plasminogen activator; IL-8R, IL-8 receptor; CXCR2, CXCL1 receptor.

The “fibroblastic system” is a broadly distributed cell surface–associated regulator of cellular migration and invasion. The accumulation of plasminogen and its activators uPA and tissue-type plasminogen activator at the cell surface augments their catalytic activities. Plasmin, generated from plasminogen, plays an important role in degradation of extracellular matrix, facilitating cell migration and invasion (19). α-ENO and CK8 bind to plasminogen to regulate plasminogen activation. α-ENO is a glycolytic enzyme that is a key 2-phospho-D-glycerate hydrolase in the cytoplasm of prokaryotes and eukaryotes. Plasminogen binds α-ENO via its C-terminal lysine residue (20), enhancing its activation and protecting activated plasmin against inhibition by α2-antiplasmin. α-ENO expression enhances plasminogen activation and promotes mammalian cell migration. Clinical evidence suggests increased plasminogen activity correlates with breast cancer invasiveness. The level of plasminogen activator, such as uPA, correlates with poor overall survival in breast cancer (12). In metastatic breast cancers, miR-17/20 expression is reduced. Reduced miR-17/20 expression increases α-ENO secretion and increases IL-8 and cyclin D1 abundance. Increased cyclin D1 induces CK8 secretion, which in turn activates plasminogen and promotes cancer cell migration and invasion, as increased α-ENO does. Thus, miR-17/20 regulates the migration and invasion of neighboring cells via heterotypic secreted signals as shown in Fig. 5D.

Materials and Methods

Human Breast Tumor Samples.

Human breast cancer specimens and matching normal breast tissue samples (frozen tissues) were provided by Peter McCue (Philadelphia, PA). All procedures were approved by the institutional review board of Thomas Jefferson University.

Vectors and Oligonucleotides.

pMSCV_puro vector was used to express miR-17/20 cluster as described previously (6, 8). WT and mutated IL-8 3′ UTR sequences were inserted into the XbaI–FseI site immediately downstream of the stop codon in the pGL3 control firefly luciferase reporter vector.

All DNA oligonucleotides were synthesized by Integrated DNA Technologies. The primer sequences, siRNA sequences, and miRNA probe sequences for Northern blot are available upon request.

Cell Culture, Retrovirus Infection, siRNA/Plasmid Transfection, and Luciferase Reporter Assay.

All cell types used in this study were cultured in DMEM containing penicillin and streptomycin (each 100 mg/L) and supplemented with 10% FBS. Retroviral production and infection methods were as described (21). The cellular transfection with plasmid DNA and luciferase reporter assay were described earlier (8).

miR-17/20–Conditioned Medium Preparation.

MCF-7 cells were transduced with the pMSCV_puro vector expressing miR-17/20 cluster or vector control as described previously (6, 8). The same number of miR-17/20–overexpressing MCF7 cells and control cells were plated at equal seeding densities. After attachment to the plates, cells were washed twice with PBS solution and overlaid with phenol red–free DMEM without serum and allowed to grow for another 24 h before collecting the medium (supernatant) from the cultures. The collected medium was centrifuged at 2,000 × g for 10 min and filtered through a 0.2-μm membrane to remove cellular debris and intact cells to yield conditioned medium.

MALDI MS.

MALDI mass spectra were recorded with a PerSeptive Voyager-DE STR MALDI time-of-flight mass spectrometer operated in the reflection mode. The measured peptide masses were used for database searching with ProFound algorithm.

Cellular Transwell Migration Assay.

Cellular transwell migration assay was performed as described previously (22).

Cellular 3D Invasion Assay.

The 3D invasion assay was adapted from Vial et al. (23).

Wound Healing Assay.

MDA-MB231 cells were counted and plated in equal numbers in 12-well tissue culture plates to achieve 90% confluence. Thereafter, a vertical wound was created using a 0.1-μL pipette tip. The wounded wells were covered with either miR-17/20–conditioned medium or control medium from MCF7 cells. Images were collected with a CCD camera at planned intervals and were digitized and stored using Metamorph 3.5 software (24). Images were captured at designated times to assess wound closure.

Cytokine Array Analysis.

Human cytokine antibody arrays were obtained from Raybiotech. The hybridization was performed following the instructions from the user manual. The intensities of signals were quantified using AlphaImager software.

ELISA.

The human IL-8 ELISA kit and human CXCL1/GROα ELISA kit were purchased from R&D Systems. The reagent preparation and assay procedure followed the manufacturer's instructions.

Plasminogen Activation Assay.

To assay the plasminogen activator activity in conditioned medium, a biochemical reaction was performed in 96-well plate at 37 °C for 1 to 2 h followed by absorbance readings at 405 and 570 nm. The reaction includes 150 μL of 20 mM Hepes buffer (pH 7.4), 5 μL of 10 mM Val-Leu-Lys-4-nitroanilide (Sigma), 0.1 U of human plasminogen (MP Biomedicals), and 20 μL of conditioned medium. The absorbance value for each well was corrected by subtracting the reading at 570 nm from the reading at 405 nm.

miRNA Northern Blot Analysis.

Northern blot analysis of miRNAs was performed as previously described (8).

miRNA Quantitative Real-Time PCR Analysis.

miRNA specific primer sets were purchased from Ambion. SYBR Green Master Mix was from Applied Biosystems. The 7900 HT Sequence Detection system (Applied Biosystems) was used for PCR. 5s rRNA was used for normalization.

Western Blot Analysis.

Cell-conditioned medium was concentrated using a 3K MWCO Amicon centrifugal filter (Millipore). The following antibodies were used for Western blotting: anti–cyclin D1 was from Neomarker; and anti-CK8 (sc-8020), anti-α-enolase (sc-100812), and anti–β-tubulin (sc-9104) were from Santa Cruz Biotechnology.

Statistical Analysis.

Data are presented as mean ± SEM. Student t test was used for analysis, and P values <0.05 were considered significant.