Envisioning metastasis as a transdifferentiation phenomenon clarifies discordant results on cancer

Cardiac departments are missing in cancer institutes

The Holy Grail of the anti-cancer recipe has been sought by researchers in exotic organisms resistant to cancer; but it is in fact not necessary to look as far, as the anti-cancer secret seems to be detained by one of our own organs: the heart. Heart is resistant to cancer despite its relatively high number of cells (300–400 g), compared for example to pancreas (70–100 g), that is highly subject to cancer. The rare tumours found in the heart are either of exogenous origin or derive from the connective tissue (myxomas), but never from cardiomyocytes. To understand this strange cellular specificity, one is tempted to look for genes specifically expressed in these cells and preventing their tumorigenesis. This inquiry readily leads to transcription factors specialised in the synthesis of contractile machineries, including myocardin, specific of cardiomyocytes and smooth muscle cells (SMC) [19,32,35,68,70] and myocardin-related factors, constitutively nuclear and strongly enriched in these cells, but sometimes expressed at lower levels in other cell types where they are subject to nucleocytoplasmic shuttling. The experimental disruption of myocardin leads to the de-differentiation and malignant transformation of smooth muscular cells [38] and conversely, myocardin is an ingredient of a cocktail of factors capable of reprogramming cells towards a cardiac fate [43]. These observations are in radical contrast with the activities of MRTFs reported in cancer.

Common modes of action but opposite effects of myocardin-related transcription factors in cancer and myocytes

Cardiomyocytes and SMCs overexpress actin isoforms and actin-related genes which are governed by serum-responsive elements (SRE), originally identified in the promoter of the immediate early gene c-fos and then extended to muscle-specific elements (named CArG). The selection of the subsets of SRE-dependent genes related to either division or motility, is partly dictated by the competing association with a common surface of SRF [77], with different coactivators [59], namely TCF for division and MRTFs for cytoskeletal dynamics [8,71]. The result of this competition depends on the mode of cellular adhesion [72]. In contractile cells, the poorly efficient SRF transcription factor is strengthened by fixation of MRTFs which work as coactivators because they are endowed with potent transcription activation domains while devoid of an own DNA-binding domain. Some members of the partially redundant MRTF family are specific of cardiomyocytes, such as its founder member Myocardin [68]. Other ones such as MRTF-A/MKL1 are abundant in muscle cells [58,69], but are also found at lower levels in other cell types. MRTFs are constitutively nuclear and active in muscular cells, whereas their nuclear translocation and activity are controlled by stimuli in other cell types [49]. Heart-specific myocardin is not regulatable by actin-mediated nuclear shuttling in spite of the presence of RPEL motifs in its amino-acid sequence [11]. At first glance, the roles of MRTFs appear strikingly different in the different cell types. In epithelial cells, MRTF-A seems to be involved in EMT and cancer progression [30,56] and consistently a factor inhibiting the transcriptional activity of MRTF-A, SCAI, is a suppressor of cancer [2], whereas in SMCs, MRTFs are clearly involved in differentiation [32]. The main interpretation suggested here to reconcile these opposite effects is that these factors are basically involved in the synthesis of actin-related contractile machineries (actin isoforms, myosin light chains, gelsolin, FHL2, etc.). These machineries participate to some extent to the mechanical phases of mitosis, but they are necessary in much larger amounts for the contractile processes of migration and thereby metastasis. Conversely, in the context of differentiated SMC and cardiomyocytes, these molecules participate to a normal differentiation program reinforcing a differentiated state.

MCF10A cells expressing or not the transforming oncogene src

The MCF10A-Er-Src cell line has been engineered from the parental MCF10A cell line to contain an inducible viral oncogene v-src oncogene [23]. This potent oncogene found in the Rous-sarcoma virus has the well-established capacity to transform the cell through deregulated phosphorylation of many target proteins from the cytoskeleton such as vinculin [16]. Upon addition of the inducer of the ER-src construct (the ER ligand 4OHT), certain cells switch from an epithelial-like to individualized and stellate morphology. This transition is accompanied by series of changes. As shown in Fig. 1, the epithelioid shapes are disrupted and the cells appear more individualized, with a disorganization and a dramatic drop of the expression of E-cadherin and an apparent shrinking of nuclei (top of panel A). This transformation is associated with a profound reorganization of actin with an increase of filamentous actin (Fig. 1A bottom), and a significant increase of the mark H3K9me3 (Fig. 1C). The transcriptional coactivator MRTF-A involved in the expression of actin and contractile machineries also increased globally, with a tendency of nuclear enrichment in certain dish areas where the cells present the most altered morphology (Panel D). Interestingly, these effects are to some extent recovered without induction of ER-src, by merely adjusting the density culture (panel E), consistent with the established influence of cell–cell contacts on SRF/MRTF signalling [3]. This effect of density, in conjunction with serum, has been examined in another widely used cellular tool for studying breast cancer, MCF7 cells.

MCF7 cells

When cultured at confluence, MCF7 mammary cells retain a relatively differentiated epithelial phenotype, which can be disrupted at low density. In our experimental conditions with epithelial cells, the proportion of proliferating and migrating cells can be adjusted experimentally in non-confluent cell cultures (Fig. 2A). Serum is a crude mixture containing some nutrients and, in variable amounts depending on the batches, factors capable of inducing proliferation (such as EGF) and cytoskeletal dynamics (such as TGFβ ). As a consequence, its global effect is complex and should be redefined for every cell type. The drastic effect of serum addition on nuclear translocation is well established in fibroblastic cells [39], but less spectacular in epithelial cells. We first examined the behaviour of MCF7 cells with low or high serum in cell spreading assays after release form confluence (Ibidi^® chambers). The cells initially densely packed are suddenly allowed to evade a dense area and different markers are examined in the spreading area (Fig. 2B). Low serum promotes cell spreading and migration, with a marked decrease of the fraction of cells with a strong nuclear staining of PCNA. In contrast, in presence of high serum, the cells actively proliferate and colonize free areas by increase of cell number instead of net moving. Interestingly, nuclear ERα staining decreased in all conditions in peripheral cells, proliferating as well as migrating (not shown), which indicates that in both cases, they lost their original epithelial differentiation state. Strikingly, the nuclear enrichment of MRTF-A was observed only in migrating cells whereas it remains predominantly cytoplasmic in proliferating cells. This result agrees with the admitted mode of nuclear translocation of MRTF-A, anchored in the cytoplasm by globular non-polymerized actin and released in the nucleus following actin polymerization [41]. In parallel, dramatic changes are observed at the chromatin level. Migrating cells with nuclear MRTF-A were strongly enriched in trimethylated H3K9 whereas acetylated H3K9 strongly increased in proliferative cells. Though simple, this experiment is illuminating with respect to the dual behaviour of cancer cells. H3K9 trimethylation only reflects the differentiation-like nature of migration but is not a primary marker of cancer since it is not increased in proliferative cells. Conversely, H3K9 hyperacetylation and PCNA accumulation are observed only in proliferating cells in which MRTF-A is less enriched in the nucleus. The nuclear enrichment of MRTF-A in presence of low serum is confirmed in homogeneous culture conditions at low density (Fig. 3A). This result seems discordant with the initial reports on nuclear translocation of MRTF-A [39]. We verified that the propensity of MRTF-A to translocate in the nucleus is much higher in highly aggressive cells in which MRTF-A is predominantly nuclear, such as the human breast cancer cells MDA, and the epithelioid Hela cells (data not shown). Hence, using the moderately transformed MCF7 cells, the combination of density and serum allows to separate proliferation and migration. Serum has been shown to induce both programs in NIH 3T3 cells fibroblastic [8], but it would be interesting to precise if they are indeed activated in the same individual cells. Our experiments suggest that these two programs coexist in the same population but are mutually exclusive at the level of individual cells. In our conditions, serum addition stimulates proliferation without triggering massive nuclear accumulation of MRTF-A in MCF7 cells and conversely, low serum favours the nuclear accumulation of MRTF-A.

The selective involvement of MRTF-A in the migratory phenotype can be further visualized by the co-detection of markers at the single cell level. The nuclear fraction of MRTF-A differs from one cell to another within the same cell culture, but the stronger enrichment of nuclear MRTF-A is systematically associated to the exclusion of traditional proliferation markers (Fig. 3B) such as cyclin D1 (cells in G1 phase), proliferating cell nuclear antigen PCNA (cells in S phase) and, at the level of chromatin, acetylated lysine 9 and phosphorylated serine 10 (cells in M phase) of histone H3. Conversely the cytostatic cyclin dependent kinase inhibitor P27 (cells in G0 phase) is increased in MRTF-A positive cells. The transfected cells in which MRTF-A is strongly nuclear correspond precisely to those containing the higher total cellular content of this protein, suggesting the existence of a simultaneous control of MRTF-A translocation and stabilization. As shown in Fig. 3C and quantified in Fig. 3D, the loss of the proliferation marker PCNA is predominant in the cells which acquire the higher contractile phenotype depicted by the peripheral accumulation and organization of the focal adhesion protein FHL2.

To further assess the mutual exclusion between MRTF-A activity and proliferation, we took advantage of a MCF7 cell line described in [10]. This cell line contains a transgene encoding a mutant version of MRTF-A lacking its N-terminal domain called RPEL and responsible for the cytoplasmic anchoring of MRTF-A with globular actin. This mutant version is constitutively active (CA) and nuclear, capable of binding to SRF and activating the expression of SRF-MRTF-A-activated genes. As shown in Fig. 3, the cell cycle is clearly downregulated in these cells as revealed by the traditional proliferation markers Ki-67 at the protein as well as mRNA level. Conversely the mRNA of the cell cycle arrest marker P21 was upregulated (not shown). We conclude that the proliferation of MCF7 cells is stimulated by mitogenic factors of serum and conversely migration is favoured in presence of low serum. The cell cycle is inhibited in the cells where MRTF-A/SRF is most strongly activated (Fig. 3B), consistent with the cytostatic effect of MRTF-A reported in certain cells [74].

Attractor switching

Some cells can occasionally leave this intensely biosynthetic stem-like state and commit to differentiate in various cell types which can then be selected by environmental conditions. In particular, a myocytic-like phenotype could underlie metastasis, providing selectable advantages to the cancer cell population. Although migration and proliferation are mutually incompatible at a given moment for a given cell, they coexist in different cells cultured in the same conditions and also in endogenous tumors. They can also coexit in the same cell, but at different times, because migration alone is likely to be not sustainable for individual cells if not periodically interrupted by phases of cellular division. These phases may alternate with a reorganization of chromatin, consistent with the plasticity and reversibility of H3K9me3 heterochromatin [14]. In this picture, chromatin changes are expected to mainly rely on dynamic protein modifications but to be not further locked by DNA methylation, in agreement with the absence of DNA methylation measured during EMT [34], thereby facilitating bidirectional switches between the two cancer attractors drawn in Fig. 3. By nature, Waddington epigenetics is reversible, so that individual cells in the tumoral population can occasionally switch from an attractor to another one, which could contribute to the phenotypic heterogeneity observed in tumors and in cell cultures. The conversions between these attractors could be stochastic or deterministic. Deterministic parameters cover a wide variety of chemical and physical microenvironmental conditions such as the number of neighbouring cells, concentration gradients of extracellular ligands or serum composition and concentration. The transitions can also result from negative feedback-driven oscillations or simply be stochastic, that is to say apparently random and underlain by hidden microscopic causes. The fundamental reversibility of Waddington epigenetics suggests that in theory, the stem-like cells of Fig. 4 could re-differentiate into epithelial cells or virtually any other cell type. This possibility seems to be not forbidden, as suggested by certain experiments of reprogramming of nuclei from maligant cells [18], but it is however likely to be prevented in most cases by no-return mutations with a ratchet effect, or mutations distorting the epigenetic landscapes [20].

The proliferative and migrating states are mutually exclusive but are expected to be close together [5]. The proximity of proliferation and migration is suggested by the dual involvement of SRF in the transcription of mitotic and contractile gene repertoires [8] and to some extent, by the involvement of MRTF-A in proliferation. At the cell population level, the src oncogene as well as serum can increase both the mitotic index and actin cytoskeletal structures, suggesting that they can induce the two facets of cancer (Fig. 4), which generally coexist in the cell population, in variable extents depending on the cancer grade. These two aspects are also gathered together in EMT, showing that EMT corresponds more to a dual cellular competence than to an integral cellular activity strictly speaking. In a single cell, only one of these programs can be activated at a time, as suggested but the exclusion of proliferative markers in the cells in which MRTF-A the most strongly activated (Fig. 3B).

Attractor disjoining and stability

The plasticity of EMT does not forbid the clear separation of the attractors which are individualized by specific self-stabilizing biochemical circuits [44]. The proliferative state is relatively stable as it corresponds to an archaic attractor reminiscent of unicellular organisms and maintained in stem cells present in most tissues [21,25]. The contractile phenotype is also strengthened by many stabilizing positive feedbacks. On the one hand, MRTF-A activities are controlled by actin dynamics and on the other hand, actin genes are key transcriptional targets of MRTF-A, in a typical feedback [55] that can enforce the migratory phenotype in cancer cells. Another example of self-sustained loop stabilizing the migratory phenotype involves the adaptor protein FHL2, closely associated to metastasis [66]. FHL2 is strongly upregulated at the transcriptional level by MRTF-A ([57] and Fig. 3), and in turn, MRTF-A is strongly stabilized by FHL2 [17].

Cell lines and treatments

MCF7 were routinely cultured in DMEM (Dulbecco’ Modified Eagle’ Medium, Invitrogen) supplemented with 10% FBS (biowest) and antibiotics (Invitrogen) at 37 ^∘ C in 5% CO2. MCF7 cells stably expressing MKL1 or MKL1-ΔN200 constructs were established using the T-Rex subclones system [10]. Expression of the proteins of interest was induced by a 48 hours treatment of the MCF7 subclones with tetracyclin, in DMEM supplemented with 2% charcoal-stripped of fetal calf serum (FCS). Migration assay were carried out in IBIDI chambers (Ramcon). MCF7 cells were seeded at confluence in DMEM supplemented with 2% or 10% of FCS into chambers fixed with collagenase to the lamella. After 48 hours, cells were released by gently removing the chambers for 15 hours. MCF10A stably expressing the ER-Src construct were kindly provided by Dr Kevin Struhl from Harvard Medical School, and were maintained in DMEM supplemented 5% of charcoal stripped horse serum (Invitrogen), with 20 ng/ml of EGF, 0.5 μ g/ml of hydrocortisone (Sigma-Aldrich), 100 ng/ml of cholera toxin (Sigma-Aldrich) and 10 μ g/ml of insulin (Sigma-Aldrich). When indicated, MCF10A ER-Src cells were treated with 1 μ M of 4-Hydroxytamoxifen (Sigma-Aldrich).

RNA extraction and qPCR

Total RNA extractions were performed using RNeasy kit (Qiagen) according to the manufacturer’s instructions. Retrotranscription was performed on 1 μ g of RNA with random primers using MMLV reverse transcriptase (Sigma). Quantitative RT-PCRs were performed using the iQ^™ SYBR Green supermix from BioRad (Bio-Rad, Hercules, CA, USA). The primer sequences used for qPCR are ki67 forward 5′ -ATAAACACCCCAACACACACAA-3′ and reverse 5′ -GCCACTTCTTCATCCAGTTAC- 3′ ; fhl2 forward 5′ -GGCAAGAAGTACATCCTGCG-3′ and reverse 5′ -CCACCAGTGAGTTTCTGCAC-3′ .

Immunofluorescence and antibodies

Cells were grown on 10-mm-diameter coverslips in 24-well plates in the presence of DMEM containing 2% charcoal-stripped FCS. Cells were fixed with PBS/4% PFA (paraformaldehyde) for 10 minutes and then permeabilized in PBS/0.3% Triton X-100 for 10 minutes. The antibodies used for immufluorescence analyses where used at 1∕1000 dilution and are indicated as follows: E-cadherin (ab15148; Abcam), MKL1 (ab113264; Abcam), H3K9ac (histone H3 acetylated at Lys 9; ab10812; Abcam), H3K9me3 (histone H3 trimethylated at Lys 9; ab8898, Abcam), H3 (ab1791, Abcam), PCNA (Dako, M0879), FHL2 (PA5-28643, Thermo Scientific), P27 (sc528, Santa Cruz), Cyclin D1 (Santa Cruz), H3pSer10 (Serine 10 Phosphorylated Histone H3, sc8656-K Santa Cruz) and FLAG (M2, Sigma-Aldrich). Images were quantified using ImageJ software.

Quantification and statistical analysis

Immunostaining signals were quantified from fluorescence microscopy images by averaging the intensity of the same area of 50 arbitrarily chosen cells or nuclei obtained from three independent experiments. Stars on histograms correspond to Mann–Whitney U tests. RT-PCR values are expressed as the mean of fold induction compared to controls of four independent experiments; ^∗ P < 0.05 with the Mann–Whitney U test.