MicroRNAs (miRNAs) are single-stranded, noncoding RNAs that are important in many biological processes1,2. Although the oncogenic and tumour-suppressive functions of several miRNAs have been characterized, the role of miRNAs in mediating tumour metastasis was addressed only recently3 and still remains largely unexplored4,5. To identify potential metastasis-promoting miRNAs, we set up a genetic screen using a non-metastatic, human breast tumour cell line that was transduced with a miRNA-expression library and subjected to a trans-well migration assay. We found that human miR-373 and miR-520c stimulated cancer cell migration and invasion in vitro and in vivo, and that certain cancer cell lines depend on endogenous miR-373 activity to migrate efficiently. Mechanistically, the migration phenotype of miR-373 and miR-520c can be explained by suppression of CD44. We found significant upregulation of miR-373 in clinical breast cancer metastasis samples that correlated inversely with CD44 expression. Taken together, our findings indicate that miRNAs are involved in tumour migration and invasion, and implicate miR-373 and miR-520c as metastasis-promoting miRNAs.
Metastasis is a process whereby cancer cells spread from a primary site and form tumours at distant sites6,7. It occurs through a specific series of steps, starting with local invasion, followed by entrance of cancer cells into the blood stream (intravasation), survival in the circulation, exit from blood vessels (extravasation), initiation and maintenance of micro-metastases at distant sites and finally, vascularization of the resulting tumours8. Cell motility is an essential feature of the metastatic process8, and the identification and characterization of molecules that control cell motility is critical to our understanding of cancer dissemination. One set of molecules that may be involved in this process are miRNAs.
Mature miRNAs are single-stranded RNAs consisting of 21–24 nucleotides that are generated from sequential processing of primary miRNA transcripts by Drosha and Dicer9,10. In mammals, mature miRNAs are integrated into an RNA-inducing silencing complex (RISC) and associate with 3′untranslated regions (3′UTR) of specific target messenger RNAs (mRNAs) to suppress translation and occasionally also induce their degradation by mRNA decay11,12,13. Computational and biological analyses estimate that each miRNA controls tens or hundreds of gene targets, whereas all the miRNAs in the human genome may regulate up to 30% of human genes and the majority of genetic pathways14,15.
To identify miRNAs that have the capacity to induce tumour migration and invasion, we used a forward genetic screen (Fig. 1a). It is well documented that the MCF-7 human breast cancer cell line has a non-migratory and non-metastatic phenotype that can be reversed by the introduction of genes such as N-cadherin or human growth hormone gene16,17. We transduced MCF-7 cells with approximately 450 individual miRNA vectors (miR-Vecs) from our miRNA-expression library (miR-Lib) by retroviral infection18, and subjected these cells to a trans-well cell migration assay to identify miRNAs that stimulate cell migration. The abundance of each miR-Vec in the migratory population was compared with its abundance in the total cell population, as described previously18. Enrichment of miR-373, miR-520c and miR-520e vectors was observed in the migratory population (Fig. 1b and data not shown). Subsequently, the miRNA inserts that incorporated into the genome of these migrated cells were retrieved and sequenced, and using PCR, the recurrent presence of these miRNAs in the migratory cells was confirmed (Fig. 1b).
Previously, we identified miR-373 as a potential oncogene in testicular germ-cell tumours18. It suppresses the oncogene-induced p53 pathway and cooperates with oncogenic RAS to promote cellular transformation, in part through direct inhibition of the tumour suppressor LATS2. However, miR-373 has not been implicated in tumour metastasis; miR-520c was discovered recently by sequencing of cellular miRNAs19 but has not yet been functionally characterized.
To determine whether they are capable of promoting metastasis, miR-373 and miR-520c were introduced into MCF-7 cells to generate stable, polyclonal cell populations that express these miRNAs (Fig. S1). Enforced expression of either miR-373 or miR-520c, but not a control miRNA or a mutant of miR-373, in MCF-7 cells produced a potent migratory phenotype (Fig. 1c) and an invasive phenotype through Matrigel (Fig. 1d). No effects of miR-373 or miR-520c on MCF-7 cell proliferation or cell cycle distribution were observed (data not shown). These results indicate that miR-373 and miR-520c promote tumour cell migration and invasion in vitro.
MCF-7 cells contain wild-type and functional p53 but the expression of LATS2 was undetectable (data not shown). As miR-373 suppresses an oncogene-induced p53 pathway and promotes cellular transformation, partly by inhibiting LATS218, we tested whether either LATS2 or p53 are involved in blocking cell migration and invasion. Stable MCF-7 cells containing retroviral vectors expressing short hairpin (sh) RNAs that suppress either LATS2 or p5318,20 were generated. In migration and invasion assays, we found that the behaviour of LATS2kd and p53kd cells resembled that of control cells and cells that express a non-target control shRNA (Fig. 1c, d). Thus, although LATS2 is hypermethylated in many human breast cancer samples, particularly aggressive forms21, our data suggest that downregulation of LATS2 or p53 is unlikely to be involved in the metastasis process mediated by miR-373 and miR-520c.
To further assess the contribution of miR-373 and miR-520c to cell migration, we screened several cancer cell lines for their endogenous expression. We found that the human breast cancer cell line MDA-MB-435, prostate cancer cell line DU-145 and human colon cancer cell line HCT-15 expressed endogenous miR-373 and displayed a high capacity to migrate in a trans-well cell migration assay (Figs 1e, f, g). However, no expression of miR-520c was detected in these cells (data not shown). Next, to examine the contribution of endogenous miR-373 to the migratory phenotype of these cells, MDA-MB-435 and HCT-15 cells were incubated with antagomiR-373, an inhibitory antisense molecule for miR-373. Addition of increasing amounts of antagomiR-373 to the culture medium caused a potent and specific inhibition of miR-373 activity (Supplementary Information, Fig. S2). Introduction of antagomiR-373, but not a mutated antagomiR-373 or antagomiR-17 (a closely related miRNA), significantly reduced the migration capacity of both cell types by more than 70% (Fig. 1f, g). The effect of antagomiR-373 on cell migration was concentration-dependent and correlated closely with its ability to counteract miR-373 activity (compare Fig. 1f, g with Supplementary Information, Fig. S2). These observations demonstrate the requirement of endogenous miR-373 expression for motility of some human cancer cells.
Next, the migration and invasion-promoting activity of miR-373 and miR-520c was measured in vivo. Luciferase-tagged MCF-7 cells expressing miR-373, miR-520c or control, were transplanted into severe combined immunodeficiency (SCID) mice via tail-vein injection. Metastatic nodules developed in the skull or spine and pleura or lungs in a period of six to eight weeks after injection (Fig. 2a). Histological analysis showed osteolytic metastases infiltrating into medullary bone spaces and destroying calcified bone tissue. In the case of upper maxillary metastases, there was invasion of periodontal tissues and occasional cementolysis (Fig. 2b). The metastatic phenotype of inoculated MCF-7 cells that expressed miR-373 and miR-520c was similar to osteolytic lesions observed in human breast cancer bone metastasis. As expected, metastasis did not occur with intravenous injection of control MCF-7 cells or MCF-7 cells expressing mutant miR-373 (Fig. 2a and data not shown). Together these results demonstrate that miR-373 and miR-520c are capable of promoting tumour invasiveness in vivo.
Seed sequences of miR-373 and miR-520c are similar, indicating that they may function through the same pathway and share many of their target mRNAs15. Each of these miRNAs also has the capacity to generate an additional miRNA from the complementary strand of the respective pre-miRNA (miR-373* and miR-526, respectively). Although we cannot exclude the possibility that miR-373* and miR-526 contribute to induction of the metastasis phenotype of miR-373 and miR-520c, respectively, only very low levels of these pre-miRNAs were detected in transduced cells; however, the inhibitory effect of antagomiR-373 on migration suggests that miR-373 and miR-520c are essential components in the induction of the invasive phenotype.
To identify the mechanism of action of miR-373 and miR-520c, an expression array analysis was performed on MCF-7 cells that stably expressed either miR-373, miR-520c or a control vector (Supplementary Information, Table S1). The target prediction program TargetScan was used to search for predicted direct target genes of miR-373 and miR-520c from the top 50 genes that were downregulated in both miR-373 and miR-520c-expressing MCF-7 cells when compared with control cells. Nine genes (ADCY1, TFF1, CD44, NFKBIZ, MAP3K8, GNA14, ETV5, RAPGEFL1 and BNIP3L) were predicted to have at least one potential binding site at their 3′UTRs for miR-373/520c family members. Their reduction in miR-373- and 520c-expressing cells was verified using quantitative RT-PCR (Fig. 3a and data not shown). Among these nine genes, enhanced tumour progression caused by reduced expression has previously been reported only for TFF1, CD44 and BNIP3L22,23,24. To test whether CD44, TFF1 and BNIP3L are direct targets of miR-373 and miR-520c, we constructed reporter plasmids containing the 3′UTR of these three genes downstream of a luciferase gene. Co-transfection experiments showed that the introduction of either miR-373 or miR-520c markedly suppressed the expression of a luciferase gene containing the 3′UTR of CD44 (Fig. 3b) but did not affect luciferase genes containing the 3′UTR of TFF1 or BNIP3L (data not shown), indicating that CD44, which encodes a cell surface receptor for hyaluronan25, is a potential direct target of miR-373 and miR-520c. Increased expression of CD44s, the most widely expressed isoform of CD44, correlates with overall survival of breast cancer patients26,27. Moreover, CD44s is expressed at a lower level in invasive micropapillary carcinoma, which is highly metastatic and has poor clinical outcome, compared with tubular carcinoma, a type of breast cancer that rarely metastasizes28. CD44 has also been identified as a metastasis suppressor in prostate cancer and colon cancer29,30,31. These results are consistent with the observation in an animal model that loss of CD44 was associated with induction of breast cancer metastasis to the lung, whereas tumour onset and size were unaffected32.
To investigate CD44, we initially verified changes in its protein expression level by immunoblot analysis using an anti-human-CD44 antibody (Hermes-3) that recognizes all CD44 isoform variants33,34. Indeed, expression of all CD44 isoforms was lower in MCF-7 cells that expressed miR-373 and miR-520c (Fig. 3c), compared with control cells. To test the possibility that CD44 suppresses cell migration and invasion, a shRNA-expressing vector was introduced into MCF-7 cells to reduce CD44 expression to a similar level to that seen in miR-373 and miR-520c cells (Fig. 3c, d). Loss of CD44 conferred a migratory and invasive phenotype in MCF-7 cells (Fig. 3e, f). To examine this phenotype in vivo, luciferase-tagged MCF-7 cells expressing a CD44 knockdown shRNA or a control non-target shRNA, were transplanted into SCID mice via tail vein injection. As found in experiments where mice were injected with miR-373- or miR-520c-expressing cells, bone metastases in skull and metastasis nodules in pleura or lung developed six to eight weeks after injection of CD44 shRNA. In contrast, mice injected with the control cells did not develop metastases (Fig. 3g). To examine the importance of CD44 in the context of miR-373- and miR-520c-induced migration, CD44s without its 3′UTR was expressed ectopically. (Supplementary Information, Fig. S3). Maintenance of CD44s levels significantly reduced the number of migrated MCF-7 cells that express miR-373 or miR-520c (Fig. 3h). These results support the notion that suppression of CD44 is required for the migratory phenotype of MCF-7 cells expressing miR-373 and miR-520c. However, it is likely that other targets of miR-373 and miR-520c may also participate in this process.
To further confirm that CD44 is a specific target of miR-373 and miR-520c, we obtained the luciferase signal of CD44 3′UTR reporter plasmid in the presence of antagomiR-373 in MCF-7 cells. The addition of antagomiR-373, but not a mutated antagomiR-373 or antagomiR-17, to the transfected cells reversed the inhibitory effect of miR-373 on the 3′UTR of CD44 (Fig. 4a). Similar results were obtained in HCT-15 cells, which express miR-373 endogenously (Fig. 4b), further supporting the specificity of the genetic interaction between miR-373 and CD44. To examine whether this interaction is direct, we looked for putative miR-373 target sites in the 3′UTR of CD44. Two sites at the 3′ end of CD44-3′UTR (∼70 base pairs (bp) and ∼140 bp from the end of 3′UTR, predicted by target-prediction algorithms TargetScan and RNA22; Fig. 4c) were deleted in the 3′UTR reporter plasmid. The luciferase mRNA was stable in both deletion constructs (as demonstrated by quantitative RT-PCR of luciferase mRNA in MCF-7 cells transfected with either wild-type CD44-3′UTR luciferase reporter or deletion reporter; see Supplementary Information, Fig. S4). In transient transfection assays, deletion of both sites completely abrogated the inhibition of the luciferase signal by miR-373 (Fig. 4c). Together, these results suggest that miR-373 and miR-520c promote cell migration and invasion, at least in part, by limiting CD44 expression directly.
It is likely that other target genes of miR-373 and miR-520c may also contribute to metastasis. Some genes with reduced expression in the microarray analysis (Supplementary Information, Table S1), such as S100 family members, may be targeted either indirectly and contribute to tumour metastasis, or directly by an unknown mechanism. Furthermore, some classical metastasis genes involved in the progression of invasive phenotypes were not found using microarray analysis. Possible explanations for this are that (1) miRNAs may function on multiple genes acting downstream of the known metastasis genes or (2) miRNAs function mainly by suppressing translation of their target genes, which only occasionally is associated with mRNA decay. Therefore, only a subset of relevant targets can be identified on the basis of mRNA expression analysis. To explore the full impact of a miRNA, genome-wide proteomic studies should be performed; however, methods for such studies are unavailable at present.
Finally, we examined the expression of miR-373 in primary and metastatic clinical samples. Expression analysis of miR-373 in 11 pairs of primary breast cancer and corresponding lymph-node metastases, each from the same patient, showed higher miR-373 levels in lymph-node metastases compared with primary tumours (Fig. 5a). The P value of the Wilcoxon paired rank sum test was below 0.01, indicating a significant association between the expression level of miR-373 and the ability of tumours to metastasize to lymph nodes. Notably, the expression level of miR-373 in some lymph-node tumour samples was in the range of several human cancer cell lines (Fig. 5b). As the expression of miR-373 in MDA-MB-435 cells and HCT-15 is required for their migratory phenotype (Fig. 1f, g), our results indicate that the increase in miR-373 levels observed in lymph node specimens of breast cancer is potentially relevant for the metastatic phenotype. To further investigate this issue, we examined a collection of (non-matched) 34 lymph-node-positive and 38 lymph-node-negative human primary breast tumours. We found that the mean expression of miR-373 in the patients with lymph node metastases was higher than the mean expression in patients without metastases (Fig. 5c, P = 0.0041 versus 0.0014, respectively). The P-value of the two-sample t-test assuming unequal variances was 0.0146. Interestingly, tumours with high miR-373 expression invariably showed an invasive phenotype (logistic regression, P = 0.0019) (Fig. 5d).
We further investigated the expression of CD44 in these tumour samples by quantitative RT-PCR. Mean expression of CD44 was significantly lower in the group of patients with lymph node metastases compared with those without lymph node metastases (Wilcoxon two-sample test, P = 0.0261, Fig. 5e). Regression analysis showed a significant inverse relationship between the likelihood of having lymph node metastases and CD44 expression in the tumours (logistic regression, P = 0.0347; Fig. 5f). More importantly, there was a statistically significant Spearman correlation that characterized an inverse relationship between the expression of miR-373 and CD44 (Spearman correlation = −0.27; P = 0.0196, Fig. 5g). Collectively, these results further support the involvement of miR-373 in tumour progression in humans. However, although the expression of miR-373 in metastatic tumours is significantly higher than that in non-metastatic tumours, miR-373 on its own is probably not a strong enough biomarker.
Herein, we have shown that miR-373 and miR-520c can stimulate tumour cell migration and invasion, at least in part through direct suppression of CD44. Both miR-373 and miR-520c belong to a large primate-specific miRNA family that shares similar seed sequence. Recently, it was reported that miR-93, another member of this family, is upregulated in basal-like breast cancer samples, associated with oestrogen receptor (ER)-negative status, and positively correlated with higher tumour grade35. Whether other miRNA family members can also stimulate tumour migration and invasion remains to be investigated. Our experiments documented the higher expression of miR-373 in lymph node-positive breast cancer specimens. Future studies should be conducted to understand the mechanism of this upregulation. We believe that cell lines, such as MDA-MB-435, DU145 and HCT-15, which express miR-373 and show a migratory phenotype, may be instrumental in this regard. Suppression of cell migration by an anti-miR-373 oligonucleotide indicates that such a strategy may serve as a basis for the development of therapies against metastasis.
MCF-7 and EcoPack II cells were cultured in Dulbecco's modified Eagle's medium (DMEM, 41966, Invitrogen) supplemented with 10% fetal calf serum (FCS) and antibiotics (complete medium). Retrovirus was made by polyethyleneimine (PEI)-transfection of EcoPAck II cells. The pMSCV–miR constructs were made as described previously18. All miRNA transfection and virus collection steps were carried out on a Hamilton ML STAR (Hamilton Bonaduz). Protocols were developed at the Netherlands Cancer Institute using Hamilton STAR Software 3.2. The methods were completely automated.
Genomic DNA was isolated from MCF-7 cells with a DNeasy Tissue kit (Qiagen). The inserts were recovered by PCR using primers specific for the pMSCV vector. The PCR product was labelled using ULS-Cy3 or Cy5 (Kreatech) and hybridized to the miR-Array according to manufacturer's instructions. Microarray analysis: total RNA was isolated from MCF-7 control cells, MCF-7 cells expressing miR-373 and MCF-7 cells expressing miR-520c. Amplification and hybridization were performed according to the manufacturer's protocol (Illumina). Illumina human V6 array was used for gene expression analysis. The raw data of the spot density was extracted from Illumina BeadStudio software and deposited on the Gene Expression Omnibus (GEO) database (accession number GSE9742). Sample clustering analysis and raw data filtering (P < 0.05) were performed. Quantile normalization was performed on the filtering data, followed by one-way analysis of variance (ANOVA) to identify significant genes.
Migration and invasion assays.
In vitro cell migration assays were performed as described previously36 using Trans-well chambers (8 μM pore size; Costar). Cells were allowed to grow to subconfluency (∼75–80%) and were serum-starved for 24 h. After detachment with trypsin, cells were washed with PBS, resuspended in serum-free medium and 250 μl cell suspension (2 × 105 cells ml−1) was added to the upper chamber. Complete medium was added to the bottom wells of the chambers. For the screen, after 12 h the cells that had not migrated were removed from the upper face of the filters using cotton swabs, and the cells that had migrated were collected with trypsin and subjected to another two rounds of enrichment. The migratory cells were then collected for the identification of miRNA. To determine the number of migratory cells, the lower surfaces of the filters were fixed with 5% glutaraldehyde solution and stained with 0.5% solution of Toluidine Blue in 2% sodium carbonate. Images of three different ×10 fields were captured from each membrane and the number of migratory cells was counted. The mean of triplicate assays for each experimental condition was used. Similar inserts coated with Matrigel were used to determine invasive potential in the invasion assay.
Isolation of RNA, reverse transcription and real-time PCR quantification.
Total RNA was extracted from frozen primary and metastasis tissues using Trizol total RNA isolation reagent (Invitrogen), according to the manufacturer's instructions. cDNA was synthesized from total RNA using gene-specific primers or random hexamers with TaqMan MicroRNA and High Capacity cDNA Reverse Transcription Kit (Applied Biosystems), according to the manufacturer's instructions. The reactions were incubated in a thermal cycler for 30 min at 16 °C, 30 min at 42 °C, 5 min at 85 °C and then held at 4 °C. Real-time PCR was performed using an Applied Biosystems 7500 Fast Real Time PCR system with miR-373- and miR-520c-specific primers (Kit 4378073, Kit 4373253) and TaqMan Universal PCR Master Mix, no AmpErase UNG (Applied Biosystems). To determine the level of CD44 expression, gene primers were designed using Primer Express v3.0 Software and real- time PCR was performed with SYBR Green Jumpstart Taq ReadyMix (Sigma). The reactions were incubated in a 96-well plate at 95 °C for 10 min followed by 40 cycles of 15 s at 95 °C and 1 min at 60 °C. The relative expression level was calculated from a relative standard curve obtained by using log dilutions of cDNA containing the gene or miRNA of interest. The average of two independent analyses for each gene and sample was calculated and was normalized to the endogenous reference control gene GAPDH.
AntagomiR-373 (CAGGGACACCCCAAAAUCGAAGCACUUCCCAGU), AntagomiR-17-5p (CAGGGACUACCUGCACUGUAAGCACUUUGCCAGU) and mutant-AntagomiR-373 (CAGGGACACCCCAAAAUCGAACGCGUUCCCAGU) were purchased from Dharmacon. These antagomiRs were incubated with cells for 48 h before cell migration assay. For luciferase assays, cells were incubated with these antagomiRs 10 h post-transfection and luciferase activity was measured 48 h later.
The 3′UTR of human CD44 was amplified using PCR and cloned into a pLuc-4 vector to generate pLuc–CD44–3′UTR. This construct (2 ng) was co-transfected into MCF-7 cells in 96-well plates together with 200 ng of control plasmid or plasmids expressing miR-373 or miR-520c and Renilla plasmid (0.2 ng). Luciferase activity was measured 48 h after transfection using the Dual-luciferase reporter assay system (Promega).
Lentivirus shRNA gene transfection and transduction.
To establish CD44 knockdown cell lines we used a vector-based shRNA technique. CD44 shRNA was purchased from Sigma. Recombinant lentiviruses were produced by co-transfecting subconfluent human embryonic kidney (HEK) 293T cells with the CD44 shRNA lentivirus expression plasmid and packaging plasmids (pMDLg/pRRE and pRSV-Rev) using Fugene6 as a transfection reagent. HEK 293T cells were cultured in DMEM supplemented with 10% FCS, 1% penicillin/streptomycin, in a 37 °C incubator with 5% CO2. Infectious lentiviruses were collected 48 h after transfection. The supernatant was centrifuged to remove cell debris and filtered through 0.45 μm filters (Millipore). MCF-7 cells were transduced with the lentivirus containing CD44 shRNA. The CD44 expression knockdown efficiency was determined by quantitative PCR.
Validation of metastasis-promoting activity of miRNAs in an animal model.
The MCF-7 human breast cancer cell line stably expressing the firefly luciferase gene with the respective miRNA were maintained in DMEM supplemented with 10% FCS at 37 °C with 5% CO2. A 17β-oestradiol pellet (1.7 mg per pellet, 90-day release time; Innovative Research of America) was implanted subcutaneously in the dorsal interscapular region of female SCID mice (6–7 weeks old) with the aid of a 10-gauge precision trochar two days before intravenous injection of MCF-7 Luc cells or cells with the respective miRNA (4 × 106 cells 0.2 ml−1 PBS). Mice were imaged using a IVIS 200 Imaging System (Xenogen) starting 3 weeks after implantation. Fifteen minutes prior to in vivo imaging, animals were anaesthetized with 1–2% isoflurane and injected intraperitoneally with D-luciferin (150 mg kg−1 in PBS). The experiments were performed using 5 or 6 mice per group and repeated 2–3 times. Animals were euthanized before dissection and tissues were subsequently fixed in 10% formalin (Sigma) and prepared for standard histological examination.
Breast cancer specimens were collected at the time of surgery from previously untreated patients. Samples were snap-frozen immediately and stored at −80 °C. Total RNA was isolated from frozen tissues with Trizol reagent (Invitrogen). Approval to collect specimens was granted by the local Institutional Review Board. Specimens were processed using procedures approved by the Health Insurance Portability and Accountability Act.
Note: Supplementary Information is available on the Nature Cell Biology website.
We would like to thank Janet Price (The University of Texas, MD Anderson Cancer Center, Houston, TX) for providing the MDA-MB-435 cell line, Ron Kerkhoven and Mike Heimerikx (Netherlands Cancer Institute, Amsterdam) for assistance in using the array facility, Roderick Beijersbergen (Netherlands Cancer Institute, Amsterdam) for establishing the high-throughput-screening facility and Louise Showe, Celia Chang and Wenhai Horng (The Wistar Institute) for microarray analysis. Q.H. is supported by Breast Cancer Alliance, Pardee Foundation, V Foundation and Commonwealth Universal Research Enhancement Program, Pennsylvania Department of Health. R.A. is supported by the Dutch Cancer Society (KWF), the European Young Investigator Award (EURYI), the Dr Josef Steiner Cancer Research Foundation and the EMBO Young Investigator Program. G.C. and L.Z. are supported by the Netherlands Cancer Institute Ovarian Cancer Research fund and the American Cancer Society.
About this article
Pathology & Oncology Research (2019)