Epigenetic reprogramming of epithelial mesenchymal transition in triple negative breast cancer cells with DNA methyltransferase and histone deacetylase inhibitors

Triple negative human breast epithelial/cancer cell lines MCF10A, MCF10F, trMCF, bsMCF, bsMCF-luciferase (bsMCF-luc), XtMCF, and LmMCF cell lines were cultured as described in the previous studies [10, 23]. Other breast cancer cell lines used in the study are: MCF7, T47D, BT474, SK-BR-3, BT-549, MDA-MB-468, MDA-MB-231, Hs578t, HCC1954, Sum149pt, and Sum159pt. The media used for these cells are described in Additional file 1. The subtype information of these cell lines are shown in Additional file 2: Table S1. All cell lines used for this study were authenticated or certified, and used in less than ten passages after recovery.

Three DNA methyltransferase inhibitors (DNMTi): 5-Azacytidine (AZA), Decitabine (DAC), and Guadecitabine/SGI-110 (SGI), and six histone deacetylase inhibitors (HDACi): JNJ-26481585 (JNJ), Vorinostat (SAHA), Entinostat (MS275), SB939, LBH589, and Tubastatin A HCL (Tub), were used to treat cells. Two forms of SGI-110 were used, SGI-110 powder for injection was used for RTCA assay and in vivo study, and SGI-110 salt was used for all other in vitro studies. The information and preparation of these drugs are shown in Additional file 3: Table S2.

Cells were plated in 96-well plates for MTT assay (see Additional file 1), or in plastic flasks or dishes for in vitro phenotypic studies. Cells were plated on day 1, treated with drug at a dose around IC50 on day 2, cultured for additional 96 h; then cells were trypsinized, examined the viability by trypan blue staining, and the viable cells were used for phenotypic studies unless specified.

Viable cells after treatment with the dose around IC50 were used for wound healing assay as previously described [10] and Matrigel invasion assay (see Additional file 1). Wound healing assay was presented as percent wound closed, and invasion assay was presented as percent of control.

Cell migration and invasion were also measured by RTCA (real-time cell analyzer) assay using xCELLigence system (ACEA Biosciences INC, San Diego, CA). Cells were treated with the dose around IC50 (refers to high dose) or a dose that inhibits cell growth by around 20% (refers to low dose), 96 h after treatment, cells were trypsinized, and the viable cells were seeded in RTCA plate for the assay (Additional file 1). As at the late time point, the impedance microelectrodes of the RTCA plate not only measure cell migration and invasion from the upper chamber to the lower chamber but also the cell proliferation on the underside of the membrane, we chose a time point in the slope of the cell index curves based on cell doubling time (20 to 30 h for these cell lines) to present data by bar graphs.

3D culture was done as previously described [10]. Drug-treated or control cells were used for 3D culture. At the end of the examination period, images were acquired, the number of ductal structures or masses formed in the collagen was quantified and classified.

Drug-treated or control cells were used for colony formation assay as described [10]. The colony formation was monitored under microscope. At the end of the examination period, cells were stained and the images were acquired. The number and the size of colonies were quantified.

Immunofluorescence staining was performed using antibodies vimentin, EpCAM{Abbiotech}, EpCAM(VU1D9), EpCAM(E144), or cleaved caspase 3. The information of the antibodies is shown in Additional file 4: Table S3. The details of staining are described in Additional file 1.

Primary antibodies E-cadherin, vimentin, SLUG, EpCAM(VU1D9), TCF4, p53, EZH2, ZEB1, H3K27me3, Histone H3, Beta actin, and GAPDH were used for WB (Additional file 1). The information of the antibodies is shown in Additional file 4: Table S3.

Single cell suspension from drug- or vehicle-treated cells was plated in ultra-low attachment 6-well plate at a density of 50,000 cells/well in culture media supplemented with 10 ng/ml EGF, 20 ng/ml bFGF, and 1 × B27. Four days after culture, images of tumorspheres were acquired using NIKON ECLIPSE TS100 microscope. The number of spheres were counted and graphed.

Animal studies were conducted in the laboratory animal facility at Fox Chase Cancer Center (FCCC) using protocol approved by the IACUC committee. Eight to nine weeks old female CB17/SCID or NOD/SCID mice from FCCC animal facility were used. 5 × 10⁴ viable XtMCF or LmMCF cells in 100 μl of PBS were mixed with 100 μl of Matrigel (BD Biosciences) and injected subcutaneously to the inguinal mammary fat pad. The treatment for CB17/SCID mice started from three days post cell injection, mice were randomly divided into four groups with 10 mice per group: (1) vehicle control; (2) SGI (1.5 mg/kg, Monday and Wednesday for XtMCF model; Monday, Wednesday, and Friday for LmMCF model; s.c. injection); (3) MS275 (0.5 mg/kg, Thursday; i.p. injection); (4) SGI + MS275. The treatment for NOD/SCID mice started from two weeks post cell injection when a xenograft was palpable. Mice were assigned into four groups: (1) vehicle control; (2) SGI (2.0 mg/kg, Monday, Wednesday, Friday; s.c. injection; two weeks); (3) MS275 (1.0 mg/kg, Thursday; i.p.; three weeks); (4) SGI + MS275. Tumor size was measured with caliper and calculated using the formula ½*length*width² twice a week.

The lung metastatic study was conducted by injecting 8 × 10⁴ viable LmMCF cells into the tail vein. Six days post cell injection, mice were randomly divided into two groups: (1) control; (2) MS275 (0.5 mg/kg, Thursday; i.p. injeciton). Mice were treated for three weeks, sacrificed 25 days post cell injection. Lungs were fixed in Bouin’s solution to visualize the metastases, and then processed and embedded in paraffin for histology examination.

Paraffin-embedded mouse lung sections at 4 μm thickness were stained with anti-human vimentin antibody (BioGenex, Fremont, CA) to evaluate lung metastases. Staining was performed following the standard protocol using i6000 Autostainer (BioGenex). Super Sensitive™ Polymer-HRP Detection System (BioGenex) was used to detect the immunostaining.

To determine the effect of DNMTi and HDACi on cell migration, wound healing assay was performed. As shown in Fig. 1c and d, SGI but not DAC significantly inhibited migration of bsMCF cells (Wound closed: 66.9% ± 6.95% for control vs. 50.1% ± 4.2% for SGI over 16 h). Four out of six HDACi significantly inhibited bsMCF cells migration. Among them, MS275 showed the greatest inhibition (Wound closed: 92.3% ± 7.6% for control vs. 55.2% ± 2.6% for MS275, 72.9% ± 4.5% for SB939, 77.9% ± 5.8% for LBH, 82.9% ± 3.7% for Tub over 18 h) (Fig. 1d, e).

We further examined whether the treatment affects cell invasion. Both DAC and SGI significantly suppressed bsMCF cell invasion by 56%. SAHA, MS275, and Tub did not have much effect. Interestingly, although the growth of bsMCF cells was significantly inhibited by JNJ, SB939, and LBH at very low doses treatment, the three drugs significantly increased cell invasion by 64% (JNJ), 43% (SB939), and 88% (LBH) (Fig. 1f), indicating that the inhibition of cell growth should not be used as the only parameter for selecting the drug for cancer treatment. Drugs like JNJ, SB939, and LBH can promote cell invasion while inhibiting cell growth, thus may accelerate the progression of the disease.

The 3D cell culture is a technique that more closely resembles the cell growth behavior in vivo. When cultured in 3D collagen matrix, MCF10F cells form branching ductal structures. Tumorigenic bsMCF cell line usually does not form ductal structures; instead, it forms solid mass displaying protrusions and shows disseminated growth of individual cells [10]. The number of solid masses may indicate the potential of tumor cells to form a structure mimicking the primary tumor with cell-cell contact, whereas the protrusions resemble cell invasion, and the disseminated cell growth exhibits the mesenchymal feature of the tumor cells. Treatment of JNJ, MS275, SB939, LBH, and Tub significantly increased the number of masses (Fig. 2a-c), suggesting the treatment increases cell-cell contact. The masses were then classified based upon the display of protrusions. DAC, SGI, and MS275 significantly increased type 1 masses, which showed no protrusions on the outer surface of the mass (Fig. 2d-f). SGI is superior to DAC, and MS275 is superior to other HDACi with regarding to the formation of type 1 masses, implying cells treated with SGI or MS275 are less invasive. It is noteworthy that DAC, SGI, and MS275 treatment reduced the disseminated growth of the individual cells (Fig. 2a and Additional file 6: Figure S2a), suggesting the reversal of mesenchymal feature by these treatments.

Anchorage independent growth of cells is usually correlated to the tumorigenicity of transformed cells. We next assessed the effect of drug treatment on the colony formation ability of bsMCF cells in methylcellulose matrix. DAC, SGI, SAHA, MS275, SB939, and Tub significantly reduced the colonies number formed by bsMCF cells. DAC and SGI were comparable in the extent of inhibition. MS275 was the most potent one among six HDACi in decreasing the colonies number (Fig. 2g-i; Additional file 6: Figure S2b).

The treatment of DNMTi in combination with HDACi in cancers has been shown more effective than the treatment of single agent in some studies [24, 25]. We evaluated the effect of combined treatment using bsMCF cells. Treatment of SGI combined with SAHA (SGI + SAHA) or with MS275 (SGI + MS275) induced a greater inhibition in cell proliferation compared to single agent (Fig. 3a). In addition, colony formation assay also showed SGI + SAHA and SGI + MS275 formed fewer colonies compared to SAHA or MS275 treatment alone (Fig. 3b). SGI + MS275 is better than SGI + SAHA in inhibiting both cell proliferation and colony formation, thus SGI + MS275 was chosen for further study. The effect of the combined treatment on cell migration and invasion was evaluated by RTCA assay at two different doses (Additional file 7: Figure S3). A significant decrease of cell migration was observed by SGI + MS275 but not by single agent at low dose treatment. At high dose, SGI + MS275 induced a significant decrease in both migration and invasion compared to control or single agent (Fig. 3c).

When examined cell growth in collagen matrix, the treatment of SGI + MS275 formed more masses than SGI treatment alone, although there was no difference in the type of masses between these two groups. By contrast, there was no significant difference in the number of masses between the SGI + MS275 and MS275 treatment, whereas SGI + MS275 increased the proportion of masses with smooth outer surface than MS275 treatment alone (Fig. 3d), indicating the combined treatment is better than single agent in certain characteristics.

As floating cells were observed when treated with MS275 or SGI + MS275, cell apoptosis was evaluated by staining with cleaved caspase 3 on cytospun slides, The result showed SGI + MS275 induced significant apoptosis than single agent. In addition, the apoptosis was more pronounced in bsMCF cells than in trMCF cells (Fig. 3e, f).

The above finding prompted us to investigate whether these treatments have in vivo antitumor activity. For this purpose, we decided to use two highly tumorigenic and metastatic TNBC cell lines XtMCF and LmMCF [10]. We first evaluated the in vitro antitumor activity. Compared to parental bsMCF-luc cells, the proliferation of XtMCF and LmMCF cells was more sensitive to SGI or MS275 (Additional file 8: Figure S4). The combined treatment significantly reduced cell viability compared to single agent (Fig. 4a). Concentrations lower than IC50, as indicated in Fig. 4b, were then used for in vitro phenotypic studies. The colony formation assay (Additional file 9: Figure S5) showed that SGI and SGI + MS275 induced a marked decrease of colony number in all three cell lines. Whereas MS275 decreased colony number in bsMCF-luc and XtMCF cells only. SGI + MS275 reduced colony number compared to MS275 alone in bsMCF-luc and LmMCF cell lines.

RTCA assay was used to evaluate the effect of the treatment on migration and invasion. The doses and the cell index plots were shown in Additional file 10: Figure S6. For XtMCF cells, SGI treatment significantly inhibited cell migration (Fig. 4c) but not invasion. MS275 did not have much effect at any dose tested. Strikingly, SGI + MS275 significantly inhibited both migration and invasion at both low and high dose (Fig. 4c, d). LmMCF are not invasive without adding additional growth factors in in vitro study. With basic FGF as a chemoattractant, only high dose MS275 showed inhibition in cell migration and invasion for LmMCF cell line, whereas SGI-110 or the SGI + MS275 promoted both migration and invasion at low or high dose, one explanation could be the treatments changed the expression of the growth factor receptor and that might have influence on migration or invasion (Fig. 4e, f).

We observed bsMCF-luc, XtMCF, and LmMCF cells have undergone EMT and are able to generate tumorspheres [10]. Tumorsphere formation assay showed SGI treated cells formed tumorspheres at the beginning in all three cell lines, however, the tumorspheres started to dissociate after four days culture, this phenomenon was more striking in bsMCF-luc cell line than in other two cell lines. Treatment of MS275 alone did not significantly affect the formation of tumorspheres. SGI + MS275 resulted in dissociation of almost all tumorspheres formed by bsMCF-luc cells, reduced mammospheres formed by XtMCF compared to MS275 alone, and significantly reduced the number of tumorspheres formed by LmMCF cells compared to single agent (Fig. 4g, h).

The above results show that the combined treatment was better than single agent in majority of the parameters evaluated, while for the effects on certain aspects such as colony formation or cancer stemness, the combined treatment was better than one of the two agents only in certain cell lines, suggesting that different pathways might be involved in different cell lines, and patient selection is very important for a treatment to exert its best effect in the clinics.

We next evaluated the effect of the treatments on the in vivo growth of TNBC cells. All XtMCF xenograft tumors in CB17/SCID mice continued growing during the course of therapy; however, the growth of the majority of tumors from treated animals was slower than tumors of control group in the week following the termination of therapy and thereafter. Mice were sacrificed 10 days after treatment stopped, there was a 54% decrease in the median tumor weight in MS275 treated group, and a 52% decrease in SGI + MS275 group (Fig. 5a, b). We also evaluated the effect of the treatments using NOD/SCID mice, with higher doses but a short treatment period. By this schedule, SGI + MS275 treatment reduced tumor weight by 33% compared to control, although SGI or MS275 alone didn’t show much effect on the tumor growth (Additional file 11: Figure S7a-c). Using LmMCF cells and CB17/SCID mice model, all treated groups showed the trend of decrease in tumor weight but it didn’t reach statistical difference (Additional file 11: Figure S7d, e). These results suggest that epigenetic drugs can exert a significant anti-tumor effect on TNBC in vivo when appropriate treatment schedule is used.

The lung metastatic study was conducted using LmMCF cell line, 80,000 cells were injected into the tail vein. MS275 treatment reduced the tumor size on lungs surface compared to control shown by Bouin’s solution fixation (Fig. 5c). The IHC staining to human vimentin on lung section showed the presence of tumor foci in the lungs (Fig. 5d), quantification of the lung areas affected by the metastases showed MS275 reduced lung metastases by 2 fold compared to control (1.9% ± 0.9% for MS275 vs. 3.8% ± 1.5% for control) (Fig. 5e). This result was consistent to the in vitro study that showed MS275 inhibits the migration and invasion of LmMCF cells.

To investigate the mechanisms underlying the reversal of mesenchymal-like features induced by the treatment of DNMTi and HDACi, Western blot analyses were performed to assess the expression of EMT markers. All treatments induced up-regulation of E-cadherin in bsMCF cells, without significant change of vimentin and SLUG expression (Fig. 6a).

The change of E-cadherin partially explained the reversal of bsMCF mesenchymal phenotype, but it could not address why the treatments of JNJ, SB939, and LBH promotes bsMCF cell invasion while up-regulating E-cadherin. Our previous study suggested that the cleavage of EGF-like domain and the nuclear translocation of EpCAM are associated with EMT process [10]. Therefore, we examined the expression of EpCAM with antibodies targeting different epitopes (Additional file 12: Figure S8a, b). The staining using antibody EpCAM(VU1D9) targeting N-terminal domain showed there was no detectable EpCAM in bsMCF cells. However, using antibody EpCAM{Abbiotech} which recognizes thyroglobulin repeat-like domain and a part of cysteine poor region, we were able to observe the expression of EpCAM in these cells. The quantification of immunofluorescence showed that the nuclear EpCAM intensity was reduced in bsMCF cells treated with DAC, SGI, SAHA, MS275, and Tub, in contrast it was increased in bsMCF cells treated with JNJ and LBH (Fig. 6b, c, and Additional file 13: Figure S9a), suggesting the elevated nuclear EpCAM level may be related to the increased cell invasion in JNJ and LBH treated bsMCF cells. EpCAM nuclear translocation is closely related to the WNT signaling, the intracellular domain of EpCAM can form a nuclear protein complex with beta-catenin, leading to gene transcription [26, 27]. Therefore, TCF4, the main effector of WNT signaling, was examined. All treatments reduced the expression of TCF4, and the change was more pronounced with the treatment of MS275 and SB939 (Fig. 6a). These results suggest both EpCAM and WNT signaling may contribute to the reversal of mesenchymal phenotype of bsMCF cells, but the crosstalk between these two pathways may be affected by the particular drug been used.

We next determined whether the combined treatment of SGI with MS275 would be superior to the single agents in regulating these signaling. Figure 7a shows E-cadherin was significantly up-regulated by the combined treatment, whereas it was not detectable in control cells, and was just slightly observed in MS275 treated cells (Fig. 7a). Similarly, the full length EpCAM showed by the reactivity to the antibody EpCAM(VU1D9), was significantly increased in cells treated with the combination (Fig. 7a), and the fluorescence staining further proved the presence of full length EpCAM in these cells (Fig. 7b and Additional file 13: Figure S9b), indicating the combined treatment protects EpCAM from being cleaved. Consistently, TCF4 was significantly reduced in cells treated with the combination when compared with the single agent. These findings suggest that SGI and MS275 have a synergistic effect on up-regulating E-cadherin, and this effect may be partially mediated by inhibiting EpCAM cleavage and WNT signaling.

It has been reported that SAHA and MS275 suppress mutant p53 expression in TNBC and pancreatic cancer [28, 29]. TP53 mutations are the most frequent genetic alterations in breast cancers, especially in basal-like cancers [30]. The mutant p53 can induce ZEB1 and EZH2 expression thus promote EMT [31, 32]. We performed cDNA sequencing of exon 5–8 (central DNA binding domain) of TP53 gene, and found the tumorigenic cell lines bsMCF, bsMCF-luc, XtMCF, and LmMCF all harbor one copy of mutation in codon 241 (TCC in wild type vs. CCC in mutant), in contrast, the MCF10F and trMCF which present epithelial phenotype do not have this mutation. Study of TP53 mutation database showed that the mutation in codon 241 is also observed in bladder carcinoma, brain tumor, breast cancer, and other cancer types. Western blotting showed p53 protein was significantly increased in trMCF cells and tumorigenic cell lines. Consistent to the mutation of TP53, EZH2 was increased in all transformed cells, and ZEB1 was up-regulated in cells that underwent EMT (Fig. 7c). Treatment of SGI did not change the expression of p53, whereas MS275 treatment significantly reduced p53 in all cell lines. Combined treatment of SGI with MS275 resulted in further suppression in p53 expression in XtMCF and LmMCF cells compared to single agent. Similarly, EZH2 and ZEB1 expression were inhibited by MS275 alone or MS275 combined with SGI (Fig. 7d, e). Our results revealed that MS275, or the combination of MS275 with SGI down-regulates mutant p53 expression and its downstream signaling in TNBC cells.

EZH2 is a functional enzymatic component of the polycomb repressive complex2. The primary function of EZH2 is to methylate lysine-27 on histone H3 (H3K27me3). However, EZH2 and H3K27me3 show an inverse correlation across breast cancer subtypes, the EZH2 expression is the highest and the H3K27me3 is the least in TNBC tumors compared to other subtypes [33]. We were interested in determining whether the H3K27me3 level was also altered during the EMT and whether it can be regulated by SGI and MS275 treatment. As shown in Fig. 7c, the global H3K27me3 was lower in mesenchymal-like tumorigenic cells compared to epithelial MCF10F and trMCF cells. SGI or MS275 treatment resulted in an elevated H3K27me3 in XtMCF and LmMCF cells, the combined treatment synergistically increased H3K27me3 in all cell lines, especially in XtMCF (Fig. 7d, e). This result indicates the reversal of EMT may also be mediated by the regulation of epigenetic marker H3K27me3 that is associated with the transcription repression.

To determine whether the growth inhibitory effects of SGI, MS275, and the combination also apply to other breast cancer cell lines, we tested the treatments in a panel of commercial available luminal, HER2 type, and TNBC cell lines. The MCF series cell lines were used for comparison. The growth of luminal and HER2 type cell lines was not sensitive to SGI (Fig. 8a, middle panel). Among the six tested TNBC cell lines (Fig. 8a, right panel), MDA-MB-468 was extremely sensitive to SGI. (Fig. 8a). All tested cell lines showed a dose dependent response upon MS275 treatment. The dose around IC50 was then used to do combined treatment. For the cell lines with the IC50 above 1000 nM, cells were treated with the agent at 1000 nM. Although luminal and HER2 type cell lines did not respond very well to SGI alone at concentrations below 1000 nM, combined treatment showed greater inhibition in cell growth than single agent. The MCF series and all other TNBC cell lines showed significantly inhibitory effect upon the combined treatment, with the most sensitive in MDA-MB-468 and Sum149pt cell lines (Fig. 8b).

To verify our findings that SGI combined with MS275 induced reversal of EMT through different pathways in TNBC cells, two mesenchymal-like cell lines MDA-MB-231 and Sum159pt were used for further studies, Western blotting showed that MS275 significantly induced E-cadherin and full length EpCAM, and reduced TCF4 in MDA-MB-231 cells. Combined treatment induced higher expression of E-cadherin and full length EpCAM than single agent alone in MDA-MB-231. The restoration of full length EpCAM was not observed in Sum159pt cells although the combined treatment showed significant up-regulation of E-cadherin and inhibition of TCF4 (Fig. 8c). Both MDA-MB-231 and Sum159pt cells have TP53 missense mutation [34]. Western blotting showed there was a notable decrease of p53 in both cell lines after MS275 and the combined treatment. EZH2 was also reduced by MS275 and the combined treatment in MDA-MB-231, but not in Sum159pt cells (Fig. 8c). There was no significant change in ZEB1 expression in both cell lines with any of the treatments. H3K27me3 was increased by MS275, with greater increase by the combined treatment in both cell lines. These results suggest that the growth inhibition and the reversal of EMT by the combined treatment of SGI with MS275 are not limited to our MCF series model, but also apply to other TNBC cells presenting EMT features, nevertheless the involved pathways might be dependent on the cell lines.