Research ArticleCancer

The lncRNA H19 mediates breast cancer cell plasticity during EMT and MET plasticity by differentially sponging miR-200b/c and let-7b

+ See all authors and affiliations

Sci. Signal.  13 Jun 2017:
Vol. 10, Issue 483, eaak9557
DOI: 10.1126/scisignal.aak9557

A noncoding mediator of metastasis

To metastasize, cancer cells undergo dynamic shifts in phenotype called epithelial-to-mesenchymal transition (EMT) and its reverse process (MET). These phenotypic shifts are controlled in part by microRNAs (miRNAs), which, by binding to target transcripts, suppress the abundance of various proteins. By isolating tumor cells from the primary site, the circulation, and metastatic nodules in the lung in a mouse model of metastatic breast cancer, Zhou et al. found that the long noncoding RNA (lncRNA) H19 acted as a “sponge” for miRNAs to promote an epithelial or mesenchymal switch in tumor cells. In epithelial-like tumor cells in the primary and metastatic sites, H19 sequestered miR-200b/c, which ultimately inhibited the migration-related protein ARF. In mesenchymal-like, disseminated cells in the circulation, H19 sequestered let-7b, ultimately activating ARF. H19 abundance was greater in metastatic than in nonmetastatic human tumors. These findings reveal a previously unknown mediator of the EMT/MET phenomenon in metastasis.


Metastasis is a multistep process by which tumor cells disseminate from their primary site and form secondary tumors at a distant site. The pathophysiological course of metastasis is mediated by the dynamic plasticity of cancer cells, which enables them to shift between epithelial and mesenchymal phenotypes through a transcriptionally regulated program termed epithelial-to-mesenchymal transition (EMT) and its reverse process, mesenchymal-to-epithelial transition (MET). Using a mouse model of spontaneous metastatic breast cancer, we investigated the molecular mediators of metastatic competence within a heterogeneous primary tumor and how these cells then manipulated their epithelial-mesenchymal plasticity during the metastatic process. We isolated cells from the primary mammary tumor, the circulation, and metastatic lesions in the lung in TA2 mice and found that the long noncoding RNA (lncRNA) H19 mediated EMT and MET by differentially acting as a sponge for the microRNAs miR-200b/c and let-7b. We found that this ability enabled H19 to modulate the expression of the microRNA targets Git2 and Cyth3, respectively, which encode regulators of the RAS superfamily member adenosine 5′-diphosphate (ADP) ribosylation factor (ARF), a guanosine triphosphatase (GTPase) that promotes cell migration associated with EMT and disseminating tumor cells. Decreasing the abundance of H19 or manipulating that of members in its axis prevented metastasis from grafts in syngeneic mice. Abundance of H19, GIT2, and CYTH3 in patient samples further suggests that H19 might be exploited as a biomarker for metastatic cells within breast tumors and perhaps as a therapeutic target to prevent metastasis.


Metastasis, which causes more than 90% of cancer-related deaths, is a multistage process during which malignant cells spread from the primary tumor into distant organs (1). The vast majority of tumors are carcinoma [meaning (the tumors are) derived from an epithelium]. Although these tumors can release large numbers of cancer cells into the circulation, only a small proportion of these epithelial-derived cells survive the migration process to infiltrate distant organs, and even fewer cells successfully form clinically relevant metastases. Successful metastatic cells must acquire the ability to invade the tumor stroma, to intravasate and extravasate the vascular endothelium, and, most importantly, to survive and thrive in a new tissue environment (2).

The ability to adopt phenotypic changes, including changes between epithelial and mesenchymal phenotypes, helps carcinoma cells through metastatic progression bottlenecks. Cytokine signals induce the reactivation of developmental epithelial-to-mesenchymal transition (EMT) programs that are implicated in the metastatic process (3). EMT is proposed to provide cancer cells with several prometastatic traits, including stemness, motility, and resistance to chemotherapy (4, 5). One argument that has been raised against a role for EMT in cancer progression is that metastatic tumors examined histologically often exhibit an epithelial-like phenotype and resemble the primary tumor (68). Increasing evidence that the reverse process, mesenchymal-to-epithelial transition (MET), is vital for the successful metastatic colonization of a secondary organ has been published recently (912). These studies support a reversible EMT-MET model for metastatic tumor cells, which undergo EMT to intravasate into blood capillaries at the primary tumor site and to extravasate into the distant organ but then revert to an epithelial phenotype to grow in the secondary site and become a clinically relevant, detectable mass (1315).

It is increasingly evident that the pathophysiological course of metastasis is not dependent solely on epithelial or mesenchymal phenotypes; it is also dependent on the ability of carcinoma cells to flexibly and dynamically transition between these two states, adapting their metastatic behavior to current needs (1620). However, an unknown question in the field is whether cancer cells dynamically switch between epithelial and mesenchymal phenotypes as a result of gradual accumulation of changes that confer an advantage on the cell, allowing it to thrive under different conditions, or whether metastatic “seed cells” already exist from the onset of primary tumor formation. The classical view of tumor progression, based on the clonal evolutionary theory of cancer (21), postulated that metastatic ability is conferred by rare random mutations in primary tumor cells that then become clonally expanded after selection at secondary organ sites (22). This conclusion, based on microarray analysis, suggested that clonal heterogeneity within primary tumors endows these different cells with distinct metastatic capabilities (2325). Thus, metastatic competence can emerge with selection among preexisting heterogeneous cancer cells in a population without the need for new mutations (23, 26, 27). Given that the inherent heterogeneous subpopulations within primary tumors are a source for the selection of metastatic cancer cells, we sought to identify the innate characteristics of metastatic cancer cells within the primary tumor to determine whether metastatic competence requires the ability to flexibly shift between EMT and MET.

Thus, we investigated the intrinsic properties of metastatic cancer cells using a set of otherwise isogenic tumor cell populations that are able to complete distinct steps of metastasis when implanted into the mammary glands of TA1 mice. We identified the long noncoding RNA (lncRNA) H19 as being essential for tumor metastasis and have characterized the molecular actions of H19 that may be required for its involvement in human breast cancer metastasis.


H19 is essential for tumor metastasis

To understand the mechanisms by which primary tumor–derived subpopulations differ in their metastatic capabilities, we performed single-cell cloning using serial dilution to establish a series of cell lines from a single mouse mammary tumor that arose spontaneously without chemical stimulus in a wild-type TA2 mouse (2831). Although these cell lines form primary tumors within a month with equivalent kinetics, they differ substantially in their metastatic potential (table S1). The behavior of these tumor lines reflects their origin from distinct subpopulations within the same primary tumor having distinct metastatic potency. We therefore divided these cell lines into three distinct groups, nonmetastatic group N, weakly metastatic group W, and highly metastatic group H (Fig. 1A). We hypothesized that the defined metastatic properties exhibited by each of these subpopulations result from alterations in the expression of specific genes. Accordingly, we compared the gene expression profiles of these three groups to dissect the specific genetic and epigenetic changes associated with their respective metastatic abilities. We compared the transcription profile of these tumor cell lines pairwise and assigned differentially expressed genes (changes greater than 2.5-fold) to class X (group W versus group N), class Y (group H versus group W), and class Z (group H versus group N) (Fig. 1A). Class X comprised 27 up-regulated and 14 down-regulated genes, class Y comprised 23 up-regulated and 19 down-regulated genes, and class Z comprised 29 up-regulated and 25 down-regulated genes (data file S1).

Fig. 1 H19 was essential for cancer metastasis.

(A) The classification for pair comparison between cell clones with differently metastatic potency (group N, nonmetastatic clones; group W, weakly metastatic clones; group H, highly metastatic clones). (B) qRT-PCR analysis of H19 abundance in 16 heterogeneous subclones isolated from the primary tumor of one TA2 mice with spontaneous breast cancer and normalized to that in clone 1. (C) qRT-PCR assessment of H19 abundance in metastatic 168FARN, 4TO7, and 4T1 cells and nonmetastatic 67NR cells and normalized to that in mouse mammary epithelial Scp2 cells. (D) qRT-PCR assessment of H19 abundance in TA2-C13 and TA2-C47 cells infected with lentiviral short hairpin RNA (shRNA) control or H19 shRNA and normalized to U6 small nuclear RNA (snRNA). (E) IVIS bioluminescence imaging of mice (n = 3 each) or extracted tissue from syngeneic TA1 mice 4 weeks after orthotopic injection of TA2-C13 and TA2-C47 cells transfected with luciferase and H19 shRNA or control shRNA. (F) The circulating tumor cells (CTCs) and metastatic lesions were calculated from syngeneic TA1 mice 4 weeks after orthotopic injection of TA2-C13 and TA2-C47 cells transfected with luciferase and H19 shRNA or control shRNA. (G) qRT-PCR assessment of H19 abundance of primary tumors from 48 primary breast cancer (PBC) patients and of metastases from 60 metastatic breast cancer (MBC) patients and normalized to U6 snRNA (Sh-Ctr, shRNA control; Sh-H19, H19 shRNA). Data are means ± SD from three independent experiments. **P < 0.01, ***P < 0.001.

Among the identified genes, H19 transcript stood out as an attractive candidate because of its high expression in highly metastatic and weakly metastatic cell lines, but not in nonmetastatic cell lines (data file S1). H19 is a maternally imprinted oncofetal gene that does not code for a protein but transcribes an lncRNA. Quantitative real time PCR (qRT-PCR) analysis confirmed that H19 transcript was expressed 5- to 15-fold higher in class W and H tumor cells compared to nonmetastatic class N cells (Fig. 1B). Consistently, a previous microarray analysis showed that H19 transcript was the most strongly up-regulated gene in metastatic 4T1 cell–derived primary tumors compared to nonmetastatic 67NR cell–derived primary tumors (5). 4T1 and 67NR cells, together with 168FARN and 4TO7 cells, are four breast cancer cell lines derived from a single mammary tumor that arose spontaneously in a wild-type BALB/c mouse. Our quantitative real-time PCR showed that H19 transcript is also expressed in 168FARN, 4TO7, and 4T1 cells, but not in 67NR cells or in mouse mammary epithelial line Scp2 cells (Fig. 1C). These observations led us to pursue H19 as an attractive candidate for involvement in metastasis.

To determine whether H19 plays a causal role in tumor metastasis, we tested whether inhibition of H19 expression affects metastatic ability. To do so, we used a lentiviral shRNA system to stably knock down H19 abundance in highly metastatic clone 13 (TA2-C13) and weakly metastatic clone 47 (TA2-C47) cells. H19 shRNA significantly reduced the expression of H19 transcript in TA2-C13 and TA2-C47 cells, compared with shRNA control (Fig. 1D). The knocked-down cells were also stably transfected with a firefly luciferase gene, and 4 weeks after implantation into the mammary glands of syngeneic TA1 mice, metastasis was examined by bioluminescence imaging. In contrast to controls, when H19 was knocked down in the weakly metastatic TA2-C47 or highly metastatic TA2-C13 cells before injection in mice, we did not detect tumor cells in the peripheral blood or secondary organs (Fig. 1, E and F).

To assess human relevance, H19 abundance was investigated in primary tumors from 48 PBC and 60 MBC patients and in normal breast tissue from 132 donors (sample descriptions are in Materials and Methods and table S2). The median H19 expression in MBC samples was about 2.5-fold higher than that in PBC and normal samples (Fig. 1G). Together, these data suggest that the lncRNA H19 may be essential for tumor metastasis and that the lncRNA H19 expression correlates with metastasis in human breast cancers.

H19 sponges miR-200b/c

Because the lncRNA H19 was highly expressed in both weakly and highly metastatic cancer cells and proved to be critical for tumor metastasis, we attempted to determine the specific steps of the metastatic process to which H19 contributes. To do so, we labeled metastatic murine TA2-C13 cells with green fluorescent protein (GFP) and implanted them into the mammary fat pads of syngeneic TA1 mice. Cells from the primary tumor, those circulating in the peripheral blood, or those that had metastasized to the lung were sorted by fluorescence-activated cell sorting (FACS) and called C13-PT, C13-PB, and C13-LM, respectively (Fig. 2A). Microarray analysis revealed that the ArfGAP [adenosine 5′-diphosphate (ADP) ribosylation factor (ARF) guanosine triphosphatase (GTPase)–activating protein (GAP)] GIT2, which we previously identified as a regulator of EMT (32), was expressed only in C13-PT and C13-LM cells (Fig. 2B). Western blots showed that C13-PT and C13-LM cells had a greater abundance of the epithelial marker E-cadherin and lower abundance of mesenchymal markers N-cadherin and vimentin compared to C13-PB cells (fig. S1, A and B). Consistent with the mRNA expression profiling, Git2 expression was greater in C13-PT and C13-LM cells than in C13-PB cells (fig. S1C). These results suggest that GIT2 expression may be dynamically regulated in tumor cells between the primary, disseminated, and metastatic sites.

Fig. 2 Git2 expression was conditionally targeted by miR-200b/c.

(A) Schematic outlining the origin of C13-PT, C13-PB, and C13-LM cells. (B) Gene cluster by microarray analysis of C13-PT, C13-PB, and C13-LM cells using the genes encoding GIT2, CYTH3, E-cadherin, N-cadherin, and vimentin. For each gene, the average signal among all cell samples was used as the baseline (onefold). Changes from baseline are represented by color intensity, with up-regulation shown as red and down-regulation shown as green. (C and D) qRT-PCR analysis of miR-200b (C) and miR-200c (D) expression in nonmetastatic TA2-C7 cells, weakly metastatic TA2-C47 cells, and highly metastatic TA2-C13 cells and in TA2-C13–derived cells and normalized to U6 snRNA. (E and F) qRT-PCR assessment of miR-200b (E) or miR-200c (F) abundance in C13-PT, C13-PB, or C13-LM cells transfected with the antagomirs of miR-200b, miR-200c, or their antagomir controls and normalized to U6 snRNA. (G and H) Western blotting analysis of GIT2 in C13-PT, C13-PB, or C13-LM cells transfected with the antagomirs of miR-200b (G), miR-200c (H), or their antagomir controls. Densitometry was normalized to that of glyceraldehyde-3-phosphate dehydrogenase (GAPDH). E-cad, E-cadherin; Vim, vimentin; PT, primary tumor; PB, peripheral blood; LM, lung metastasis. Data are means ± SD from three independent experiments. **P < 0.01, ***P < 0.001.

To explain how GIT2 expression might be controlled dynamically during metastasis, we considered the involvement of microRNAs (miRNAs). In particular, we hypothesized that GIT2 abundance might be inhibited by miRNAs in C13-PB cells but that this inhibition was abrogated in C13-LM cells. We found that mouse Git2 gene was a predicted target of miR-200b/c (fig. S1, D and E), which are central players in the maintenance of the epithelial phenotype (33, 34). Abundance of miR-200b and miR-200c was increased in the highly metastatic TA2-C13 cells, as well as C13-PT, C13-PB, and C13-LM cells, compared to weakly metastatic TA2-C47 cells and nonmetastatic clone 7 (TA2-C7) cells, which all share the same primary tumor origin (Fig. 2, C and D). When miR-200b and miR-200c were inhibited by targeted antagomirs (Fig. 2, E and F), the abundance of GIT2 protein in C13-PT and C13-LM cells was unaffected, but that in C13-PB cells was significantly increased (Fig. 2, G and H). This supports the idea that Git2 may be a target of miR-200b/c, but that the functional effects of this targeting are somehow cell-specific (or, rather, context-specific).

To explore how Git2 transcripts are not down-regulated by miR-200b/c in C13-PT and C13-LM cells, we turned to lncRNAs. A growing body of evidence indicates that lncRNAs may act as decoys to sequester or sponge miRNAs and hence modulate miRNA downstream targets (35, 36); thus, we posited that the lncRNA H19 might sponge miR-200b/c to inhibit their functions. The abundance of the lncRNA H19 was substantially increased in both highly and weakly metastatic cells compared with nonmetastatic cells (Fig. 1B). We performed qRT-PCR for the lncRNA H19, confirming its expression in TA2-C47, TA2-C13, C13-PT, C13-PB, and C13-LM cells (fig. S2A). RNA immunoprecipitation (RIP) assays using antibodies against mouse Ago2 (Fig. 3A), which is a core component of the RNA-induced silencing complex (RISC) (37), further supported the sponging hypothesis. Ago2 antibody precipitated Ago2 protein-RNA complexes from C13-PT, C13-PB, and C13-LM cell lysates (fig. S2B), and we found that endogenous H19 was preferentially enriched in Ago2 RIPs compared to control immunoglobulin G (IgG) antibody RIPs (Fig. 3B). Moreover, Ago2 RIP samples from C13-PT and C13-LM cells were significantly enriched for endogenous miR-200b/c compared to those from C13-PB cells (Fig. 3C), from which we infer that H19 and miR-200b/c were in the same Ago2 complex in C13-PT and C13-LM cells. Additionally, bioinformatics tools (38) revealed potential binding sites in mouse H19 for mouse miR-200b and miR-200c: miR-200b/c shared 14-mer conserved H19 target sequences (340–353), wherein two predicted miR-200b sites start at nucleotide position 339 and three predicted miR-200c sites start at nucleotide position 338 (data files S2 and S3).

Fig. 3 H19 sponged miR-200b/c.

(A) Schematic outlining the Ago2 RIP strategy to validate endogenous miRNA:H19 binding. (B) qRT-PCR assessment of H19 in C13-PT, C13-PB, and C13-LM cells that were pulled down by AGO2 or negative control IgG and normalized to U6 snRNA. (C) qRT-PCR detection of miR-200b/c endogenously associated with H19 in C13-PT, C13-PB, and C13-LM cells that were pulled down by AGO2 or negative control IgG and normalized to U6 snRNA. (D) The luciferase activity assessment in TA2-C7 cells transfected with sensor (miR-200b/c 4×, psiCHECK2-miR-200b/c 4×), together with 0, 20, 40, 80, or 160 ng of sponge plasmid pH19 or pH19mut1. (E) Western blotting detection of GIT2 in C13-PT, C13-PB, and C13-LM cells transfected with #2 H19 small interfering RNA (siRNA) or siRNA control and normalized to GAPDH. (F) Western blotting detection of guanosine 5′-triphosphate (GTP)–bound ARF6 that were pulled down by glutathione S-transferase (GST)–GGA3–protein binding domain (PBD) beads in C13-PB, C13-PT, and C13-LM cells and normalized to total ARF6. (G) Endogenous CYTH3 in C13-PT, C13-PB, and C13-LM cells was detected by Western blotting and quantified relative to GAPDH as a loading control (siCtr, siRNA control; siH19, H19 siRNA). Data are means ± SD from three independent experiments. **P < 0.01, ***P < 0.001.

To further explore whether H19 might act as a “sponge” to sequester miR-200b/c, we transfected TA2-C7 cells, which have a low abundance of endogenous H19 (Fig. 1B), with the plasmids psiCHECK2-miR-200b/c 4× (encoding miR-200b/c binding sites; the “sensor” in this experiment) and various amounts of pH19 (which expresses full-length mouse H19; the “sponge” in this experiment). The relative luciferase activity increased in response to pH19 in a dose-dependent manner (Fig. 3D), suggesting that ectopically expressed H19 specifically sequestered endogenous miR-200b/c, thereby preventing it from inhibiting luciferase expression. Expression of a mutant H19 (pH19mut1), in which predicted miR-200b/c interaction sites were mutated, did not inhibit luciferase expression (Fig. 3D), confirming that miR-200b/c binding sites are required for this effect. Luciferase reporter assays in which a luciferase reporter of the 3′ untranslated region (3′UTR) of Git2 (Git2 pMirTarget) was cotransfected into TA2-C7 cells with miR-200b/c mimics or their controls. TA2-C7 cells cotransfected with miR-200b/c mimics significantly decreased the activity of luciferase constructs carrying the sequence of Git2 3′UTR (fig. S5A). The effects of H19 on the Git2 luciferase reporter were also analyzed with full-length mouse H19 (pH19) and its mutant derivative (pH19mut1). The relative activity of luciferase was rescued in pH19 cotransfected cells but not in pH19mut1 cotransfected cells (fig. S5A). Moreover, we knocked down the abundance of H19 with siRNA (fig. S2, C and D) and monitored the effect on the abundance of miR-200b/c and GIT2 abundance. H19 had no effect on the abundance of miR-200b/c (fig. S5B) but reduced the abundance of GIT2 in C13-PT and C13-LM cells (Fig. 3E). However, GIT2 abundance in C13-PB cells was not rescued because they were with miR-200b/c antagomirs (Figs. 2, G and H, and 3F). Together, these data suggest that H19 physically associates with miR-200b/c to function as a competing endogenous RNA (ceRNA) for miR-200b/c in C13-PT and C13-LM cells.

H19 sponges let-7b

As stated, the data thus far suggested that H19 acts as a ceRNA or sponge for miR-200b/c miRNA in tumor cells within the primary tumor and in lung metastases, but somehow, miR-200b/c remained functional in targeting Git2 in tumor cells in the circulation. GIT2 belongs to the family of GAPs for the small GTPase ARF, called ArfGAPs (39, 40). ARF proteins, which regulate vesicular trafficking and actin remodeling (and thus are implicated in cell migration), cycle between their active GTP-bound and inactive guanosine diphosphate (GDP)–bound conformations. ArfGAPs promote hydrolysis of bound GTP and induce binding of GDP to inactivate the ARF protein (41), whereas ARF guanine nucleotide exchange factors (ArfGEFs) release bound GDP and facilitate binding of GTP to activate the ARF protein (42). Because we found that H19 inhibits miR-200b/c targeting of Git2, we examined the functional effects on ARF activity in C13-PT, C13-PB, and C13-LM cells. We found that increased amounts of GTP-bound ARF proteins were precipitated by the immobilized ARF-specific effector GGA3 (Golgi-associated, γ adaptin ear containing, ARF-binding protein 3) (43) in C13-PB cells relative to C13-PT or C13-LM cells (Fig. 3F). Because GTP binding to ARF is facilitated by ArfGEFs, we thus hypothesized that an ArfGEF is functional in C13-PB cells.

Among 10 known mouse ArfGEFs, only the transcript for cytohesin 3 (CYTH3; also known as GRP1 or ARNO3) (data file S4) was predicted by TargetScan to be targeted by the miRNA let-7 (fig. S2, E and F), and let-7 is reportedly bound and sequestered (sponged) by H19 (44). CYTH3 protein was more abundant in C13-PB cells than in C13-PT and C13-LM cells (Fig. 3G), as was GTP-bound ARF6 (Fig. 3F), but the opposite (less in C13-PB cells) was observed for the abundance of GIT2 (fig. S1C). Although the amount of seven members of the let-7 family was relatively similar in subclones of the three initial cell groups (nonmetastatic TA2-C7, weakly metastatic TA2-C47, and highly metastatic TA2-C13 and TA2-C55 cells) and in C13-PT, C13-PB, and C13-LM cells (fig. S3A), selectively decreasing the amount of let-7b with an antagomir increased the abundance of CYTH3 in C13-PT and C13-LM cells but not in C13-PB cells (Fig. 4A and fig. S3, B and C), suggesting that CYTH3 may be a target of let-7b in C13-PT and C13-LM cells. In addition, bioinformatics tools (38) revealed potential binding sites for mature mouse let-7b in mouse H19 (data file S3). Ago2 RIP experiments showed that endogenous let-7b and H19 were both pulled down in the Ago2 complex in C13-PB cells, but not in C13-PT and C13-LM cells (Figs. 3B and 4B). Furthermore, luciferase reporter assays demonstrated that the relative luciferase activity of TA2-C7 cells transfected with psiCHECK2-let-7b 4× (the sensor) increased in response to pH19 (the sponge) in a dose-dependent manner, whereas the sensor was unaffected by pH19mut2, in which predicted let-7b interaction sites were mutated (Fig. 4C). Furthermore, H19 knockdown with siRNA had no significant effect on the abundance of let-7b (fig. S5C) or CYTH3 in C13-PT or C13-LM cells but decreased the abundance of CYTH3 in C13-PB cells (Fig. 4D), indicating that H19 competed with let-7b’s ability to bind and down-regulate CYTH3 transcripts. Together, these data demonstrated that the lncRNA H19 sponged miR-200b/c in C13-PT and C13-LM cells but sponged let-7b in C13-PB cells, mediating the regulation of GIT2 and CYTH3 abundance, respectively.

Fig. 4 H19 sponged let-7b.

(A) Western blotting detection of CYTH3 protein expression of C13-PT, C13-PB, and C13-LM cells transfected with let-7b antagomir or antagomir control and normalized to GAPDH. (B) qRT-PCR detection of let-7b endogenously associated with H19 in C13-PT, C13-PB, and C13-LM cells that were pulled down by AGO2 or negative control IgG and normalized to U6 snRNA. (C) The luciferase activity assessment in TA2-C7 cells transfected with sensor (let-7b 4×, psiCHECK2-let-7b 4×), together with 0, 20, 40, 80, or 160 ng of sponge plasmid pH19 or pH19mut2. (D) Western blotting detection of CYTH3 in C13-PT, C13-PB, and C13-LM cells transfected with #2 H19 siRNA or siRNA control and normalized to GAPDH. (E) Number of CTCs in syngeneic TA1 mice 4 weeks after orthotopic injection of C13-PT, C13-PB, and C13-LM cells infected by CYTH3 shRNA or control shRNA. (F) Bioluminescence (Biolum), hematoxylin and eosin (H&E) staining, and immunohistochemistry staining with luciferase antibody (Anti-Luc) of lung metastatic lesions of TA1 mice (n = 3) 4 weeks after orthotopic injection of C13-PT cells infected with Git2 shRNA or control shRNA. (G) Quantification of bioluminescence of region of interest (ROI) for lung metastatic lesions of TA1 mice (n = 3) 4 weeks after orthotopic or tail vein injection of C13-PT cells infected with Git2 shRNA or control shRNA. (H) Immunostaining for CTCs in peripheral blood samples from MBC patients. CTCs are defined as CK+/CD45/DAPI+ cells with cytokeratin 8 (CK8) (red), CD45 (green), and 4′,6-diamidino-2-phenylindole (DAPI) (blue). Scale bars, 10 μm. NC, negative control; BF, before filtrating; AF, after filtrating; siCtr, siRNA control; siH19, H19 siRNA. Data are means ± SD from three independent experiments. **P < 0.01, ***P < 0.001.

Context-specific sponging by H19 mediates epithelial-mesenchymal plasticity

Given the relationship between GIT2, CYTH3, and EMT (32, 40), we next investigated the role of GIT2 and CYTH3 on ARF activity and epithelial or mesenchymal state in C13-PT, C13-PB, and C13-LM cells. Our data show that the knocking down of GIT2 increased the amount of active, GTP-bound ARF6 (fig. S4A) and reduced the abundance of the epithelial marker E-cadherin (fig. S4B) in C13-PT and C13-LM cells, whereas CYTH3 knockdown decreased the amount of active ARF6 (fig. S4C) and increased E-cadherin abundance (fig. S4D) in C13-PB cells.

We then examined the functional effect of potentially selective sponging of miR-200b/c and let-7b by the lncRNA H19. The miR-200b/c antagomir increased E-cadherin abundance in C13-PB cells (fig. S4E), whereas the let-7b antagomir reduced E-cadherin abundance in C13-PT and C13-LM cells (fig. S4F). Most importantly, forced expression of a miR-200b/c–resistant mutant of the Git2 construct decreased the amount of active ARF6 and increased the amount of E-cadherin in C13-PB cells (fig. S5, D and E), indicating that the endogenous Git2 transcript in C13-PB cells was targeted by miR-200b/c and thus could not inhibit translation of Arf6. On the other hand, expressing let-7b–resistant CYTH3 mutants increased the amount of active ARF6 and decreased the amount of E-cadherin in C13-PT and C13-LM cells (fig. S5, F and G). Furthermore, knocking down H19 reduced E-cadherin and increased vimentin abundance in C13-PT and C13-LM cells, whereas the opposite was observed in C13-PB cells (fig. S4, G and H). These data reveal that GIT2 is essential for maintaining the epithelial state, whereas CYTH3 is critical for inducing a mesenchymal phenotype, and that the abundance of each—and hence EMT/MET dynamics (at least in metastatic murine TA2-C13 breast cancer cells)—is regulated by context-specific sponging of miR-200b/c and let-7b by the lncRNA H19.

CYTH3 and GIT2 are involved in metastatic initiation and metastatic colonization, respectively

To relate these findings to cancer metastasis, we tested the involvement of CYTH3 and GIT2 in tumorigenic and metastatic behaviors, namely, invasion, proliferation, and colony-forming ability. GIT2 knockdown significantly promoted the invasive ability of C13-PT and C13-LM cells, whereas CYTH3 knockdown significantly suppressed the invasive ability of C13-PB cells (fig. S4, I to L) in Transwell assays. Furthermore, GIT2 knockdown reduced the colony-forming ability and proliferation of C13-PT and C13-LM cells (fig. S4, M, O, and Q), whereas CYTH3 knockdown increased the colony-forming ability and proliferation of C13-PB cells (fig. S4, N, P, and R) as assessed by soft agar and 5-bromo-2′-deoxyuridine (BrdU) incorporation assays.

We then examined the necessity of GIT2 and CYTH3 for successful metastasis in two mouse models. C13-PT, C13-PB, and C13-LM cells were injected into syngeneic TA1 mice either orthotopically into a mammary fat pad or into the tail vein. Cells from orthotopically initiated tumors failed to intravasate when CYTH3 was knocked down (Fig. 4E). Cells from either orthotopically initiated tumors or tail vein injections failed to form macrometastasis in the lung when GIT2 was knocked down (Fig. 4, F and G). These data thus far demonstrate that CYTH3 promoted metastatic initiation (presumably EMT) but that GIT2 promoted metastatic colonization (presumably MET).

Last, we evaluated the involvement of GIT2 and CYTH3 in human breast tumor metastasis by correlating the abundance of each with pathological phenotypes in clinical breast tumor samples. We obtained primary breast tumor specimens, paired CTCs, and metastatic samples from 13 MBC patients. CTCs were isolated by an optimized filtration method and verified by immunostaining as CK+/CD45/DAPI+ cells (Fig. 4H). In agreement with our findings that GIT2 and CYTH3 were dynamically expressed in primary tumor, circulating, and metastatic cancer cells (that is, C13-PT, C13-PB, and C13-LM cells; Fig. 3G and fig. S1C), GIT2 expression was down-regulated and CYTH3 expression was up-regulated in CTCs (as assessed by single-cell qRT-PCR) compared to each in primary tumors and metastatic nodules from MBC patients (Fig. 5, A and B). Together, our findings demonstrated that the lncRNA H19 critically contributes to metastatic progression by sponging miR-200b/c or let-7b during distinct phases of the metastatic process.

Fig. 5 ArfGAP GIT2 and ArfGEF CYTH3 are involved in metastatic initiation or colonization.

(A and B) Western blotting for GIT2 (A) and CYTH3 (B) in the primary breast tumor specimens, peripheral blood samples, and metastatic samples from the same patient of 13 MBC patients. (C and D) A possible mechanism explains the sequential metastatic capability of nonmetastatic TA2-C7, weakly metastatic TA2-C47, and highly metastatic TA2-C13 cells. (E) Schemes describing the mechanism of H19 regulating the step acquirement of the competence to flexibly shift EMT and MET during metastasis.


Metastasis is a complex multistep process in which cells from a primary tumor spread to distant organs. Cancer cells must exit the primary tumor to invade through the surrounding tissue, intravasate into a blood or lymphatic vessel, be carried to a distant site, extravasate from the vessel into a foreign tissue, and proliferate to reestablish a tumor mass in the new organ environment (45). The vast majority of disseminated cancer cells fail to survive and proliferate after invading a foreign tissue (46). However, studies of clonal cell lines derived from late-stage human carcinomas (47) have provided direct evidence that individual cancer cells within a tumor differ in their metastatic capability, including cells that are not even metastatic, confirming the heterogeneity shown in preclinical studies with murine (22) and human tumors (48). Still, the intrinsic mechanisms governing the ability of the metastatic subpopulation within the heterogeneous primary tumor to successfully metastasize to secondary organs are incompletely understood.

Here, we reported that the lncRNA H19 differentially sponged miR-200b/c and let-7b in a cell context–specific manner, thereby controlling the abundance of the ArfGAP GIT2 or the ArfGEF CYTH3 and the reversible shifts between epithelial and mesenchymal states in tumor cells (Fig. 5E). These seemingly contradictory functions of H19 were essential to successful metastasis and were differentially harnessed by a tumor cell depending on its state within the process of tumor metastasis.

Our results also highlight a central role for ARF6 activity in regulating distinct aspects of metastasis (4951). GIT2 is an ArfGAP protein that reduces ARF6 activity (52). Disseminated tumor cells had reduced GIT2 abundance and increased GTP-bound ARF6 compared to each in cells derived from primary tumor or cells from metastatic lesions. CYTH3 is an ArfGEF that increases ARF6 activity (53). Conversely, disseminated tumor cells had greater abundance of CYTH3 and GTP-bound (active) ARF6 than cells from primary tumor or cells from metastatic lesions. Thus, regulation of ARF6 might be the final common mediator of the regulatory network composed of the lncRNA H19, miR-200b/c and let-7b, and GIT2 and CYTH3.

The cell lines used in this study with distinct metastatic potency were derived from a single mammary tumor that arose spontaneously in a TA2 mouse. These cells reflect the intrinsic cellular heterogeneity within a primary tumor and their ability to engage into distinct steps in the metastatic process. All metastatic and nonmetastatic cell lines expressed high amounts of let-7b, whereas only metastatic cells expressed miR-200b/c and nonmetastatic TA2-C7 cells expressed relatively low amounts of H19, suggesting that H19 might be a factor for metastasis. The lack of expression of H19 in nonmetastatic cells may account for their inability to metastasize. Although weakly metastatic cells expressed H19 and let-7b miRNA, these cells failed to colonize secondary sites because they had no miR-200b/c. Accordingly, we propose that in highly metastatic cells, high expression of H19 results in selective sponging of miR-200b/c and let-7b, conferring on these cells the ability to accomplish each of the EMT-MET transitions required for successful metastasis, compared to the weakly metastatic or nonmetastatic cell lines (Fig. 5, C and D).

The miR-200 family members, including miR-200c, directly target and repress Zeb1 expression, and their frequent loss with concomitant increase in ZEB1 abundance promotes EMT by down-regulating E-cadherin abundance (54). Controversially, our data showed that miR-200b/c was highly abundant in cells from the primary tumor, peripheral circulation, and metastatic lesions, whereas the abundance of both E-cadherin and GIT2 were decreased in cells from the peripheral circulation. Although the Zeb1 transcript was a possible target of miR-200b/c, the Git2 transcript was apparently targeted by miR-200b/c (fig. S1, D and E). In addition, we previously reported that loss of GIT2 induced the expression of Zeb1 (32). Therefore, we speculate that miR-200b/c might directly target Git2 instead of Zeb1 in disseminated tumor cells and thus decrease the abundance of GIT2, promote the abundance of ZEB1, and induce EMT. Consistently, the miR-200b/c antagomir restored both GIT2 and E-cadherin abundance in disseminated tumor cells. On the other hand, the amount of E-cadherin was rescued by inhibiting either H19 or miR-200b only in disseminated tumor cells, suggesting that miR-200b/c works by targeting Git2 in circulating cancer cells due to sponging by H19.

Immunohistochemistry staining showed that E-cadherin was highly abundant in tumors of mice injected with cells isolated from the circulation or from metastatic lesions. Additionally, in these tumors, the abundance of both GIT2 and CYTH3 was similar to that in “parental” cells (those in the original primary tumor), indicating that the cells in the circulation and metastases can flexibly reverse their phenotypes in response to the new microenvironment through intrinsic regulation by lncRNA (H19)–mediated selective sponging.

Our results demonstrate that the lncRNA H19 is more abundant in metastatic cancer cells than in nonmetastatic cancer cells or normal mammary epithelial cells and that the H19 transcript is more abundant in cells in the peripheral blood from patients with MBC than from those with PBC. It remains to be determined how and why the lncRNA H19 is induced in metastatic cancer cells. Nevertheless, our findings suggest that the lncRNA H19 might be a biomarker for metastatic breast cancer and a therapeutic target to prevent metastasis.


shRNAs, siRNAs, miRNAs, and inhibitors

H19 SMARTvector Lentiviral shRNA, GIT2 SMARTvector Lentiviral shRNA, CYTH3 SMARTvector Lentiviral shRNA, and shRNA control were purchased from GE Healthcare Dharmacon Inc. (data file S5 for sequence information). Three unique 27-mer siRNA duplexes of mouse H19-specific siRNAs (#SR402967) were purchased from OriGene Technologies (sequence information in data file S5). The mouse let-7b miRNA mimics hsa-let-7b (stem-loop accession number: MI0000063), miR-200b miRNA mimics mmu-miR-200b (stem-loop accession number: MI0000342), and miR-200c miRNA mimics mmu-miR-200c (stem-loop accession number: MI0000650) were purchased from Ambion/Life Technologies. The mouse let-7b antagomir (#MmiR-AN0004-AM02), miR-200b antagomir (#MmiR-AN0300-SN20), miR-200c antagomir (#MmiR-AN0302-AM02), and control antagomir (#CmiR-AN0001-SN) were purchased from GeneCopoeia.


GFP (#30127), psiCHECK2-let-7 4× (#20930), and luciferase (#17186) plasmids were obtained from Addgene. H19 mouse open reading frame clone (#MC207056), GIT2 mouse complementary DNA (cDNA) clone (#MC220430), CYTH3 mouse cDNA clone (#MC209242), and GIT2 pMirTarget luciferase reporter (#SC209774) were purchased from OriGene Technologies.

To make pH19mut1 in which miR200b/c binding sites were mutated and pH19mut2 in which let-7b binding sites were mutated, PCR was carried out using pCMV6-Kan/Neo-pH19 vector as a template with primers (data file S5). The resulting PCR fragment was ligated to pCMV6-Kan/Neo opened with Hind III and Xho I. To make psiCHECK2-miR-200b/c 4× and psiCHECK2-let-7b 4×, an annealed oligonucleotide fragment containing copies of miR-200b/c or let-7b binding sites (data file S5) was inserted into psiCHECK2-let-7 4× opened with Xho I and Not I. To make Git2 constructs with mutant for miR-200b/c binding sites and CYTH3 constructs with mutant for let-7b binding sites, PCR was carried out using GIT2 mouse cDNA or CYTH3 mouse cDNA as a template with the primers (data file S5). The resulting PCR fragment was ligated to pCMV6-Kan/Neo (carrying GIT2 or CYTH3 cDNA) opened with Bam HI and Nco I to replace the corresponding miRNA binding sites.

Microarray analysis

Total RNA samples from 16 subclones (group N, group P, and group F) or C13-PT, C13-PB, and C13-LM cells were isolated using the RNA isolation kit (Qiagen) according to the manufacturer’s protocol. Total RNA (100 ng) was reverse-transcribed to produce cDNA/mRNA hybrids, which were used as a template to create double-stranded cDNA with a unique DNA/RNA heteroduplex at one end. The cDNA was then amplified via SPIA (single primer isothermal amplification), which produces single-stranded anti-sense DNA. Post-SPIA modification generates sense target cDNA that was fragmented, biotin-labeled, and hybridized to Affymetrix Mouse Gene 1.0 ST arrays for 18 hours at 45°C while rotating at 60 rpm (Affymetrix). Arrays were then washed and developed using the FS450_0007 fluidics protocol and scanned using an Affymetrix 3000 7G scanner.

Fluorescence-activated cell sorting

To generate C13-PT and C13-LM cell cultures, primary tumor tissue and lung metastatic nodules from mice orthotopically injected with GFP-labeled TA2-C13 cells were dissected and crushed in RPMI 1640 medium and passed through a 100-μm nylon mesh. The filtrate was centrifuged at 50g for 5 min, and the supernatant was collected. To generate C13-PB cell cultures, peripheral blood samples from mice bearing GFP-labeled TA2-C13 cells were depleted of red blood cells using lysis solution [100 mM NH4Cl (pH 7.4)] for not more than 5 min. Cells were suspended in a 40% Percoll (GE Healthcare) solution and overlaid on a 70% Percoll solution. After centrifugation at 800g for 20 min, the interphase was collected and washed twice with phosphate-buffered saline (PBS). GFP-positive cells were sorted on a FACSAria II (BD Biosciences) cell sorter equipped with a 488-nm laser directly into the sorting buffer. To culture cells after FACS, collected GFP-positive cells were mixed in 50% medium/50% serum to maximize cell viability and plated in one well of a chamber slide and grown until confluent. Cells were passaged and sorted over several cycles to obtain stable GFP-labeled C13-PT, C13-PB, and C13-LM cells.

Ago2 RIP assay

To determine whether the lncRNA H19, miR-200b/c, and let-7b are associated with the RISC, we performed RNA pull-down assay using Ago2 antibody to precipitate the RISC and detected H19 or miRNAs from the pellet using qRT-PCR. Briefly, C13-PT, C13-PB, and C13-LM cells were rinsed with cold PBS and fixed with 1% formaldehyde for 10 min. After centrifugation, cell pellets were collected and resuspended in NP-40 lysis buffer supplemented with 1 mM phenylmethylsulfonyl fluoride (PMSF), 1 mM dithiothreitol, 1% protease inhibitor cocktail (Sigma-Aldrich), and RNase inhibitor (200 U/ml) (Life Technologies). The cell lysates were stored at −80°C before use. The supernatant from cell lysates was collected by high-speed centrifugation. To generate antibody-coated beads, protein G Sepharose 4 Fast Flow bead slurry (GE Healthcare) was rinsed with NT2 buffer (50 mM tris-HCl, 150 mM NaCl, 1 mM MgCl2, and 0.5% NP-40) and incubated with antibody against Ago2 (Abcam). Nonimmune mouse IgG (Sigma-Aldrich) was used as negative control. For RIP, the supernatant was incubated with the antibody-coated Sepharose beads overnight. The beads were rinsed with cold NT2 buffer, followed by incubation with proteinase K (10 mg/ml) (Sigma-Aldrich). The RNAs bound to Ago2 antibody were purified with the RNeasy Mini Kit (Qiagen) and were used for qRT-PCR.

Animal studies

TA1 and TA2 mouse strains were bred by the Animal Center of Tianjin Medical University (31). Six-week-old female TA1 mice were used for orthotopic mammary fat pad or tail vein injection metastasis model assays. For bioluminescence tracking of cells in vivo, mice were anesthetized with isoflurane (RWD Life Science), intraperitoneally injected with d-luciferin (PerkinElmer), and imaged with an IVIS Lumina II system (Caliper Life Sciences). Signal quantification was performed with Living Image Software (PerkinElmer). All mouse experiments were conducted in accordance with a protocol approved by the Institutional Animal Care and Use Committee at Lishui University.

RNA extraction and qRT-PCR

Total RNA was isolated by TRIzol Reagent (Invitrogen) following the manufacturer’s instructions. After RNA extraction, RNA samples were reverse-transcribed using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). The FastStart Universal SYBR Green Master mix (Roche) was used for the qRT-PCR. The relative fold changes of candidate genes were analyzed by using the 2−ΔΔCT method. Results were normalized to U6 snRNA as indicated.

Isolation of CTCs by size

Patient peripheral blood samples (6 ml; anticoagulated with EDTA) were collected after discarding the first 2 ml to avoid potential skin cell contamination by normal epithelial cells. Next, a filtration method was applied to isolate CTCs using a calibrated membrane with 8-μm-diameter pores (Millipore). The filtration system consisted of a filtration tube containing the membrane (SurExam), a manifold vacuum plate with valve settings (SurExam), an E-Z 96 vacuum manifold (Omega), and a vacuum pump (Auto Science). Erythrocytes were removed using red blood cell lysis buffer [154 mM NH4Cl, 10 mM KHCO3, and 0.1 mM EDTA (all from Sigma-Aldrich) in deionized water], and then the remaining cells were resuspended in PBS (Sigma-Aldrich) containing 4% formaldehyde (Sigma-Aldrich) for 5 min before filtration. After the cell suspension was transferred to the filtration tube, the pump valve was switched on to reach at least 0.08 MPa; the manifold vacuum plate valve was then switched on, and filtration began. Isolated CTCs were further confirmed by immunostaining analysis and used for single-cell qRT-PCR with the Single Cell-to-CT qRT-PCR kit (Thermo Fisher).

Patient eligibility

This study included 48 PBC patients and 60 MBC patients treated at Zhejiang Cancer Hospital from August 2014 to August 2015. Patients meeting all of the following requirements were eligible for enrollment: (i) a diagnosis of invasive breast cancer confirmed by histology, (ii) no treatment before diagnosis, and (iii) provision of voluntary written informed consent. Human specimens included primary tumor tissue and adjacent normal tissue (beyond tumor margins) from PBC and MBC patients for analysis of the expression of the lncRNA H19. Of the 60 MBC patients, primary tumor, paired peripheral blood samples, and distant metastasis were all collected from 13 patients and assessed (blinded to the patient/tumor status) by a pathologist for the abundance of GIT2 and CYTH3.

Immunostaining analysis

The peripheral blood of MBC patients was collected and fixed with 10% formaldehyde for 10 min, blocked with 10% fetal bovine serum (FBS), and permeabilized with 0.5% Triton X-100 for 20 min before immunostaining. The samples were incubated in the dark with an antibody cocktail containing anti–CK8-PE (1:10), anti–CD45–fluorescein isothiocyanate (1:10) and DAPI (1:100) for 30 min at 25°C. The smears were analyzed under a fluorescence microscope using a 20× objective. CTCs are defined as CK+CD45DAPI+ cells.

Generation of knockdown cells

Git2 shRNA, Cyth3 shRNA, H19 shRNA, or shRNA control lentiviral system vectors were transfected into 293T cells. The lentiviral particles were packaged and collected following the manufacturer’s guideline and were mixed with 1 ml of polybrene (10 mg/ml) (#H9268, Sigma-Aldrich) and diluted into 1 ml of Dulbecco’s modified Eagle’s medium (DMEM) without FBS. The diluted medium with lentiviral particles was added to cells seeded into 24-well plates. Twenty-four to 48 hours later, the medium was replaced by DMEM with 10% FBS and 1% penicillin-streptomycin and followed by positive selection using puromycin (2.5 mg/ml) (Invitrogen). The positive clones were further confirmed by Western blotting (for CYTH3 and GIT2) or qRT-PCR (for H19), and the one with the best knockdown efficiency was chosen in our experiments.

ARF activity assay

ARF activity was analyzed by ARF6-GTP pull-down assay following the manufacturer’s protocol (Pierce Biotechnology). Briefly, C13-PT, C13-PB, and C13-LM cells were washed with PBS at 4°C and harvested into 500 ml of lysis buffer [200 mM NaCl, 50 mM tris-HCl (pH 7.5), 10 mM MgCl2, 1% Triton X-100, 0.1% SDS, 0.5% deoxycholate, 5% glycerol, 1 mM PMSF, 9 nM pepstatin, 9 nM antipain, 10 nM leupeptin, and 10 nM chymostatin]. Cell extracts were incubated with 30 μg of GST-GGA3 or GST alone immobilized on glutathione-Sepharose for 1 hour at 4°C. The pellets were washed three times with lysis buffer. Bound proteins were eluted by 30 μl of elution buffer. GTP-bound ARF6 was identified by immunoblotting with an ARF6-specific monoclonal antibody.

Colony formation assays

Stable GIT2 knockdown C13-PT, C13-PB, and C13-LM cells and their corresponding control cells (1 × 103 cells each) were seeded into a six-well plate. After culture for 12 days, cells were fixed with 70% ethanol and subsequently stained by 0.2% cresyl violet solution. The images were captured, and the colonies were counted by ImageJ software (U.S. National Institutes of Health).

In vitro invasion and BrdU incorporation assays

The invasive ability of stable GIT2 knockdown C13-PT, C13-PB, and C13-LM cells and their corresponding control cells was measured using 24-well Transwell plates (8-μm pore size; Corning) as previously described (55). Cells (5 × 104) were suspended in 200 μl of DMEM and added into the upper chamber with a noncoated membrane. The chamber inserts were coated with Matrigel (BD Biosciences) at 1:7 dilution. For the proliferation assay, cells were seeded into 96-well plates at 5000 cells per well for 24 hours and assessed using a BrdU chemiluminescent cell proliferation enzyme-linked immunosorbent assay (Roche) as previously described (56).


Tumor specimens from lung metastatic nodules of mice bearing C13-PT Git2 shRNA or shRNA control cell tumors were fixed in 4% paraformaldehyde and then embedded in paraffin. Paraffin blocks were cut into 5-μm-thick sections and stained with H&E or luciferase antibody.

Western blotting

Cell lysates were lysed in RIPA buffer (Sigma-Aldrich) with Complete Protease Inhibitor Cocktail (Roche). Cell lysates were stored at −20°C before use. Cellular proteins were separated by SDS–polyacrylamide gel electrophoresis using a 5% stacking gel and 10% running gel. The molecular weight of candidate proteins was interpolated using Pre-Stained SeeBlue rainbow marker (Invitrogen) loaded in parallel. The membranes were probed with the following antibodies: E-cadherin (Cell Signaling Technology), vimentin (Santa Cruz Biotechnology), Git2 (Santa Cruz Biotechnology), CYTH3 (Zymed Laboratories), and GAPDH (Sigma-Aldrich). Chemiluminescence was induced by SuperSignal West Dura Substrate (Pierce), and the images were acquired by a LAS-3000 (Fujifilm Life Sciences) and quantified by Multi Gauge software (Fujifilm Life Sciences). Data are presented as means ± SD, based on at least three independent repeats.

Bioinformatics analysis

The miRNA binding sites on H19 were predicted using the web-based program RNAhybrid (38). The predicted miRNA let-7 targets were searched from TargetScan ( For the mouse lncRNA H19, the sequence was downloaded from an lncRNA database ( For mature mouse miR-200b/c and let-7b miRNAs, sequences were downloaded from the miRBase miRNA database ( The sequences and predicted binding results are summarized in data files S2 and S3.


Statistical data analysis was performed using SPSS 17.0. Data are presented as means ± SD, based on at least three independent repeats. Comparisons between two groups were conducted using two-tailed Student’s t test, and differences were considered to be statistically significant when the P value is less than 0.05.


Fig. S1. Git2 transcript is a predicted target of miR-200b/c.

Fig. S2. Cyth3 transcript is a predicted target of let-7b.

Fig. S3. The functional effect of let-7 assessed with a targeted antagomir.

Fig. S4. The role of GIT2 and CYTH3 as analyzed by RIP and in vitro assays.

Fig. S5. Luciferase assay results.

Table S1. Metastases resulting from orthotopic injection of cells from distinct clones.

Table S2. Clinical characteristics of MBC patients.

Data file S1. Microarray results.

Data file S2. miRNA and H19 sequences.

Data file S3. miRNA target information.

Data file S4. ArfGEF target information.

Data file S5. siRNA, shRNA, and PCR primer sequences.


Acknowledgments: We are grateful to F. Miller (Wayne State University) for mouse breast cancer cell lines 67NR, 168FARN, 4TO7, and 4T1. We thank J.-P. Thiery (National University of Singapore) and R. T. Premont (Duke University) for pre-reviewing our manuscript. Funding: W.Z. was supported by the National Natural Science Foundation of China (81572879). D.X. was supported by the National Natural Science Foundation of China (3152010390 and 81230058). Author contributions: W.Z. designed and interpreted the experiments and wrote the manuscript. X.-l.Y., J.X., M.G.-C., Z.-Y.F., L.-Y.L., G.-H.G., Q.L., and Y.-H.Q. conducted the experiments and analyzed the data. D.X. provided input on the manuscript. All authors provided comments and critically read the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: The microarray data are deposited at Dryad, doi:10.5061/dryad.t5pv8.
View Abstract

Navigate This Article