Paracrine and epigenetic control of CAF-induced metastasis: the role of HOTAIR stimulated by TGF-ß1 secretion

The human breast cancer cell lines MDA-MB-231 and MCF-7 were purchased from the American Type Culture Collection (ATCC). The CAFs were isolated and cultured as described previously [23]. CAFs were acquired from four invasive breast cancer patients who underwent a mastectomy at the Tianjin Medical University Cancer Institute and Hospital (TMUCIH), and the use of specimens was approved by the Institutional Review Board of TMUCIH. CAFs were obtained from tissues that had been cut into small pieces and digested with collagenase type I (1 mg/mL; Sigma) and hyaluronidase (125 units/mL; Sigma) for 6 h in DMEM without FBS at 37 °C. After filtering the undigested tissues, the stromal fraction was centrifuged at 1000 rpm for 5 min. The cells were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) or RMPI Medium 1640 supplemented with 10% fetal bovine serum (FBS), 1% penicillin and 1% streptomycin at 37 °C in a 5% CO₂ humidified incubator.

The profiles of cytokines secreted by CAFs were detected in the culture supernatants using a Human Cytokine Array (RayBiotech, Guangzhou, China) according to the manufacturer’s instructions. The cytokines with distinct differences in expression were screened out.

The culture medium was removed and used to assess the level of extracellular TGF-β1 with a TGF-β1 enzyme-linked immunosorbent assay (ELISA, abcam) according to the manufacturer’s instructions. The results are expressed in ng/mL.

The total protein was extracted with cold radio-immunoprecipitation buffer containing protease inhibitor and phosphatase inhibitor. The cell lysates were separated using 8-10% SDS–PAGE and transferred onto polyvinylidene difluoride (PVDF) membranes. Antibodies against human CDK5, CDK5RAP1, H3K27me3, EZH2 (1:1000 dilutions, Cell Signaling Technology), Egr-1, E-cadherin, vimentin, β-catenin (1:1000 dilutions, Abcam), and β-actin (1:4000 dilutions, Santa Cruz) were used as primary antibodies. Rabbit or mouse IgG antibody coupled with horseradish peroxidase (Santa Cruz) was used as a secondary antibody. The antibody-labeled protein bands on membranes were detected with a G-BOX iChemi XT instrument (Syngene).

After various treatments, cells were seeded on sterile coverslips and cultured for 48 h. The cells were then fixed with 4% paraformaldehyde for 15 min and permeabilized with 0.25% Triton-X 100 for 10 min, followed by blocking with 3% BSA for 1 h. Immunofluorescence staining was conducted with antibodies against β-catenin, E-cadherin and vimentin (1:100 dilutions, Abcam). The cells were washed with phosphate-buffered saline (PBS) and incubated with Alexa Fluor 488 or Alexa Fluor 546 (Life Technologies) secondary antibodies. To detect the formation of stress fibers, cells were stained with Alexa Fluor 568-phalloidin. Nuclei were stained using DAPI, and the cells were visualized using FV-1000 laser scanning confocal microscopes. Cells from three independent experiments in which at least 300 cells were counted were quantified.

Total RNA was extracted using TRIzol reagent (Life Technologies) according to the standard protocol. HOTAIR cDNA was obtained from total RNA using the Fast Quant RT Kit (TIANGEN). Real-time PCR was performed using Super Real Pre Mix Plus SYBR Green (TIANGEN) on an ABI-7500 Real-time PCR machine (Life Technologies).The following thermocycling protocol was employed: 95 °C for 30s, 60 °C for 30 s and 72 °C for 30 s for 40 cycles. GAPDH was used as the internal control. Data are presented as fold changes (2^−ΔΔCt) and were analyzed using the Opticon Monitor Analysis Software V 2.02 (MJ Research).

Approximately 2 × 10⁵ MDA-MB-231 cells or 3 × 10⁵ MCF-7 cells were plated in six-well plates after different treatments. A linear scratch/wound was made on the cell monolayer with a sterile pipette. Photomicrographs of live cells were taken at 40× magnification, and the distance migrated was observed after 48 h or 72 h.

The Matrigel invasion assay was performed in 24-well Transwell culture plates. Briefly, 40 μL of Matrigel (1 mg/mL, BD) was applied to 8 μm polycarbonate membrane filters. Approximately 3 × 10⁴ MDA-MB-231 cells or 5 × 10⁴ MCF-7 cells were resuspended and then seeded in 24-well Transwell plates containing FBS-free medium in the upper chamber and complete growth medium supplemented with 10% FBS in the lower chamber for 24 h or 48 h at 37 °C. Non-invading cells were removed from the upper surfaces of the invasion membranes, and the cells on the lower surface were stained with hematoxylin. The average number of cells per field was determined by counting the cells in six random fields per well. Cells were counted in four separate fields in three independent experiments.

The expression of HOTAIR in paraffin-embedded sections and cells of different conditions seeded on the sterile coverslips for 48 h was examined in situ hybridization. First, we deparaffinized and rehydrated the samples. Before digesting samples with proteinase K, we used 4% paraformaldehyde to fix the paraffin-embedded sections and DEPC water to fix the cell samples. We then hybridized samples with a 5’Cy3-labeled modified HOTAIR probe (6 ng/μl) overnight at 37 °C. After completing the hybridization, the samples were successively washed twice with solution buffer (2×, 1×, 0.5×, chloride sodium citrate buffer) at 37 °C for 10 min each. Nuclei were stained using DAPI, and the cells were visualized using an FV-1000 laser scanning confocal microscope.

The EZ-ChIP kit (Millipore) was used to perform the ChIP assays. Briefly, 1% formaldehyde was used to fix cells, and 0.125 M glycine was then used to neutralize the formaldehyde. The cells were then lysed in lysis buffer with SDS and protease inhibitors. Soluble chromatins of 200-1000 bp were collected after sonication and pre-cleared in 1:10 dilution buffer, followed by incubation with IgG, H3K27me3 and EZH2 anti-bodies on a rotating platform overnight at 4 °C. After washing the immunocomplexes captured by protein G-Sepharose beads, the bound DNA fragments were eluted for real-time qPCR analysis using ChIP primer.

The upstream 2.5 kb promoter regions of HOTAIR were cloned into a luciferase reporter vector (GV238 vector), and the mutations of predicted SMADs binding sites were also cloned into the same luciferase reporter vector. Cells were co-transfected with SMAD plasmid GV219 vector and HOTAIR plasmid (wild type or mutant) for 24 h, and the promoter activities were examined with a luciferase assay.

BALB/c nude mice aged 4-6 weeks were purchased from the Animal Center at the Cancer Institute at Chinese Academy of Medical Science (Beijing, China). MDA-MB-231 cells transfected with Lenti-HOTAIR or Lenti NC and CAFs were injected into the mammary fat pads of each nude mouse at a ratio of 1:3. The mice were imaged for luciferase activity with an IVIS imaging system once per week.

The paraffin-embedded tissue sections were stained with hematoxylin and eosin (H&E) and subjected to an immunohistochemistry analysis. After blocking with 3% H₂O₂ and non-immune rabbit serum, the sections were incubated with primary antibodies (1:200 dilutions) overnight at 4 °C, followed by a streptavidin-biotinylated secondary antibody for 1 h at 37 °C. The chromogenic substrate was developed with 3, 3′-diaminobenzidine, andhematoxylin was used for counter staining. The results were visualized using a light microscope.

SPSS 16.0 (IBM, USA) was used for all calculations. All values are presented as the mean ± SD. Statistical comparisons between the two groups were made using Student’s t-test, and differences among groups were assessed with a two-way ANOVA, followed by Dunnett’s test. Significance was set to P < 0.05.*P < 0.05, **P < 0.01.

During breast cancer progression, CAFs may supply appropriate signals that may promote the development of aggressive phenotypes in carcinoma cells and establish a complex scenario, which culminates in metastasis [26]. To identify the cytokines secreted by CAFs that enhance tumorigenic ability, the cytokine profiles of cell supernatants from MCF-7, MDA-MB-231 and treatment with conditioned media of CAFs (CAF-CM) were analyzed using a RayBiotech Human Cytokine Antibody Array. The top four cytokines secreted in the supernatant of CAFs were TGF-β1, IGFBP-4, IGF-1 and PDGF-AA (Fig. 1a, b). Furthermore, TGF-β1 was the biggest increased cytokine in both CAF-CM treated MCF-7 (Fig. 1c, d) as well as in MDA-MB-231 cells, compared to control cells (Fig.1e, f).

To gain a further insight of the role of CAFs in the clinical setting, the distribution of TGF-ß1 in clinic samples was performed (Additional file 1: Figure S1). It’s well documented that CAFs strongly expressed α-SMA, whereas were negative for the epithelial marker cytokeratin. Immunofluorescent staining with antibodies against α-SMA, cytokeratin-7 and TGF-ß1 showed that the red fluorescence of TGF-ß1 colocalized with green fluorescence of α-SMA, but not with cytokeratin-7 in both invasive breast carcinoma and ductal carcinoma in situ. More important, much more and stronger fluorescence of TGF-ß1 was detected in invasive breast carcinoma, compared with human ductal carcinoma in situ. Collectively, these results indicated that TGF-ß1may be secreted by CAFs.

Due to the changes of quantity and quality in CAFs from different patients, we extracted CAFs from four invasive breast carcinomas patients to reduce the original difference in this study (Additional file 1: Figure S2). MDA-MB-231 and MCF-7 cells were treated with conditioned media of CAF1#, CAF2#, CAF3# and CAF4#, respectively. ELISA further confirmed all of the four CAFs populations induced significant increased TGF-β1 level in the supernatants from CAF-CM-treated cells (mean value: 0.05 ± 0.01 ng/ml vs 6.34 ± 0.243 ng/ml in MCF-7 cells; 0.06 ± 0.01 ng/ml vs 4.10 ± 0.14 ng/ml in MDA-MB-231 cells) (Fig. 2a, Additional file 1: Figure S3). We then further evaluated the role of TGF-β1 in CAFs induced EMT of cancer cells. We found that TGF-β1 treatment achieved similar enhanced the migratory activity of cancer cell, compared with CAF-CM treated cells (Fig. 2b). Additionally, the small-molecule inhibitors of TGF-β1, SB451332 (SB) or Pirfenidone (PFD) treatment abrogated the CAF-CM induced promotion of cell migration. Moreover, all of the four CAFs populations were competent in enhancing the cell invasive ability, while the number of invading cells was deceased 2- to 3.5-fold in SB or PFD treated cells, as indicated by a Transwell assay (Fig. 2c, Additional file 1: Figure S4). Moreover, SB and PDF treatment markedly rescued the induction of vimentin and β-catenin expression as well as the repression of E-cadherin in CAF-CM-treated cells (Fig. 2d). Immunofluorescent images showed that SB or PFD treatment decreased the nuclear signals of ß-catenin in MDA-MB-231 and MCF-7 cells, suggesting that the nuclear translocation to the cytoplasm was promoted (Fig. 2e). Furthermore, the remodeling and polarization of F-actin was observed after CAF stimulation, whereas the formation of a stress fiber pattern was abolished after SB or PFD treatment. These results identified TGF-β1 as a crucial mediator of the interaction between CAFs and breast cancer cells and TGF-β1 was necessary for CAFs to induce EMT of cancer cells.

The epigenetic regulation of lncRNA triggered invasion and metastasis in breast cancer. To study the epigenetic mechanism by which CAFs induce EMT in cancer cells, the expression of several lncRNAs were measured by RT-PCR (Fig. 3a). Specifically, CAF-CM resulted in the largest increases in HOTAIR expression, 3.5- and 8.7-fold increases inMCF-7 and MDA-MB-231 cells, respectively, compared to the control cells. Similar to CAF-CM, recombinant TGF-β1 also increased the levels of HOTAIR (Fig. 3b). However, the CAF-CM mediated activation of HOTAIR was attenuated after SB or PFD treatment. FISH images revealed that SB or PFD repressed the nuclear distribution of HOTAIR induced by CAF-CM (Fig. 3c), and Western blotting analysis indicated that the protein levels of EZH2 and tri-methylated H3K27 were notably increased (Fig. 3d). Moreover, the induction of E-cadherin and repression of vimentin and β-catenin was observed in HOTAIR knock-down cells (Fig. 3e), and the depletion of HOTAIR decreased the number of invading cells, as indicated by a Transwell assay (Fig. 3f, Additional file 1: Figure S5). Based on these data, we concluded that CAFs promoted metastasis by activating HOTAIR expression in cancer cells. However, the mechanisms by which CAFs activate HOTAIR remain to be addressed.

TGF-β is known to assemble a receptor complex that activates Smads thereby assembling multi-subunit complexes that regulate transcription [27]. Western blotting analysis indicated that the levels of phosphorylated of Smad2 protein (p-SMAD2), p-SMAD3 and p-SMAD4were significantly elevated in CAF-CM and TGF-β1-treated cancer cells, suggesting that CAFs activated the TGF-β1/SMAD signaling pathway (Fig. 4a). However, the repression of p-SMAD2, p-SMAD3 and p-SMAD4 was detected in cells treated with SB or PFD. Moreover, RT-PCR showed a 50-60% reduction in HOTAIR expression in cells transfected with SMAD2, SMAD3 or SMAD4 siRNA (Fig. 4b). To further elucidate the mechanisms underlying CAF-induced HOTAIR activation, we used ChIP technology to identify the genomic locations bound by SMADs in cancer cells before and after co-culture with CAF-CM. The ChIP results indicated that CAFs stimulate the direct binding of SMAD2, 3, and 4 to the promoter region of HOTAIR (Fig. 4c), and a promoter analysis using JASPAR and PROMO indicated several candidate SMAD binding sites in the promoter region of HOTAIR (Fig. 4d). The relative profile score threshold set as 80% in JASPAR and the maximum matrix dissimilarity rate set as 15% in PROMO [28, 29]. Accordingly, the deletion of predicted binding sites (Mut-1 and Mut-2) within the HOTAIR promoter region decreased SMAD2/3/4 induced luciferase activity in MCF-7 and MDA-MB-231 cells, whereas the deletion of the remaining mutant binding sites did not result in significant differences compared with control group (Fig. 4e). Therefore, we concluded that SMADs directly bind the promoter site of HOTAIR located between nucleotides -386 and -398, -440 and -452 (the genome coordinates: chr12: 53975342-53975354, chr12: 53975396-53975408), and consequently induced the transactivation of HOTAIR.

Based on these results, we elucidated that CAFs transactivated HOTAIR via the secretion TGF-β1. Furthermore, we identified HOTAIR as a direct transcriptional target of SMAD2/3/4.

We previously reported that the active form of CDK5 kinase was responsible for tumor metastasis. Consistent with our previous study, CDK5 protein expression was significantly increased in recombinant TGF-β1- or CAF-CM treated cells, whereas the levels of EGR-1 and CDK5RAP1 were greatly decreased (Fig. 5a). These findings suggest that the activation of CDK5 was essential for CAF-induced EMT. To investigate the epigenetic events that were involved in the activation of CDK5 in breast cancer, we first examined the histone methylation status using ChIP assays (Fig. 5b). Profiling histone methylation marks on the CDK5RAP1 and EGR-1 promoter by qChIP showed that H3K27me3 is highly enriched at the promoter of both genes in CAF-CM treated MDA-MB-231 and MCF-7 cells compared with the control cells. Whereas the depletion of HOTAIR rescued the effect of CAFs, the H3K27me3 levels at the CDK5RAP1 and EGR-1 promoter were remarkably reduced. Furthermore, the recruitment of EZH2 to the CDK5RAP1 and EGR-1 promoter was measured, which indicated that the level of PRC2 components at the CDK5RAP1 and EGR-1 promoter was significantly decreased (Fig. 5c). Based on these data, we demonstrated that the activation of CDK5 was required for CAF-mediated EMT. Specifically, CAFs stimulated CDK5 expression by H3K27 trimethylation on the promoter of CDK5RAP1 and EGR-1. CAFs transactivated HOTAIR expression to induce EMT by targeting CDK5 signaling.

To study the epigenetic mechanism of CAF-induced tumor growth and metastasis in vivo, an MDA-MB-231 orthotopic tumor transplantation model was established in nude mice. Bioluminescence imaging was monitored weekly and indicated that all of the four CAF populations remarkably promoted primary tumor growth, whereas shHOTAIR treatment markedly reduced tumor volume (Fig. 6a-c, Additional file 1: Figure S6). Most importantly, CAFs resulted in dramatic lung metastasis with extensive tumor foci compared to control cells, as determined by quantitative bioluminescence images (Fig. 6d). However, this metastatic phenomenon was attenuated when HOTAIR was knocked down, and H&E staining further confirmed these results (Fig. 6e). Compared to the control group, CAFs enhanced the nuclear staining of HOTAIR, as revealed by FISH, which further confirmed our in vitro study (Fig. 6f). Moreover, immunohistochemistry staining showed the downregulation of nuclear β-catenin and reductions in p-SMAD2-, p-SMAD3-, EZH2- and H3K27me3- positive cells in the shHOTAIR group (Fig. 6g). Taken together, these findings indicate that shHOTAIR treatment efficiently blocked the crosstalk between CAFs and tumor cells, thereby impairing CAF-induced EMT. Collectively, these experiments supported the notion that the secretion TGF-β1 by CAFs activated the TGF-β1/SMAD pathway in breast cancer cells, resulting in the up-regulation of HOTAIR transcription and histone modification of the CDK5 signaling pathway, thereby promoting breast cancer progression and metastasis.

To gain a further insight of the role of CAFs in the clinical setting, we collected 59 breast carcinoma samples and examined markers of CAFs FAP-α and α-SMA by immunohistochemistry. We found that the expression levels of FAP-α and α-SMA were significantly up-regulated in invasive breast carcinoma (97.5%, 38 in 39), compared with human ductal carcinoma in situ (DCIS, 33.3%, 7 of 20), suggesting that CAFs facilitated tumor metastasis in patients with breast cancer (Additional file 1: Figure S7). Compared to DCIS, HOTAIR expression and nuclear staining were higher in metastatic breast carcinoma, as revealed by FISH images, which indicated that the upregulation of HOTAIR strongly correlated with tumor metastasis (Fig. 7a). Furthermore, the expression levels of EZH2 and CDK5 exhibited similar patterns, which further confirmed our in vitro and in vivo studies (Fig. 7b).