Targeting Cancer Cell Metastasis by Converting Cancer Cells into Fat

Breast cancer is a systemic disease that frequently results in metastatic dissemination. Tumor cell dissemination can be an early event in cancer progression, even in cancer types that are considered potentially curable (1). Upon systemic dissemination, metastatic cells need to survive the hazards of blood circulation, to evade immune defenses, and to respond to a constantly changing microenvironment, requiring a high ability to adapt their biological functions, a state that is referred to as high cellular plasticity (2–5).

Cancer cell plasticity plays a critical role in cancer survival, invasion and metastasis formation (2), as well as in tumor heterogeneity (6, 7) and the development of therapy resistance (8, 9). An epithelial-mesenchymal transition (EMT) and its reversal, a mesenchymal-to-epithelial transition (MET), are induced by external stimuli as a result of the tumor cell's interaction with the microenvironment at different sites and niches within the primary tumor (10–15). Lineage tracing models have revealed the role of immune cells' interactions with tumor cells as a driving force of EMT dynamics (14, 15). Specifically, the cytokine TGFβ seems to contribute to EMT progression by generating dedifferentiated, therapy-resistant cells that eventually lead to metastasis and disease recurrence (11, 14). Furthermore, activation of an EMT program correlates with increased metastatic potential and drug resistance also in high-grade tumors, such as in aggressive triple-negative breast cancer (16, 17). Thus, the therapeutic targeting of EMT-derived cells by specific cytotoxic compounds, such as salinomycin (9), or by inducing their redifferentiation via MET (18) has been proposed as an avenue to overcome malignant tumor progression, metastasis, and therapy resistance. Of note, such redifferentiation approaches require combination with chemotherapy or targeted therapies to achieve long-term clinical benefit (19).

For instance, mesenchymal invasive cells from mammary gland carcinomas were induced to undergo differentiation into epithelial cells by treatment with the histone deacetylase (HDAC) inhibitor SAHA. Treatment with SAHA reduced proliferation and induced differentiation in these cells (20). A different HDAC inhibitor was shown to induce differentiation and to increase sensitivity to chemotherapy in EMT-derived pancreatic cancer cells (21). Indeed, during the past years, numerous clinical trials with HDAC inhibitors have been performed for the treatment of various cancer types. However, HDAC inhibitors seem to exert varying effects on regulating cell state transitions, and the clinical results did not yet meet the expectations for this class of drugs (22).

Another redifferentiation approach has aimed to specifically induce a MET in breast cancer cells to overcome the cancer cell invasiveness and drug resistance correlating with an EMT. Screening for compounds that induce the upregulation of Cdh1 (the gene encoding for E-cadherin) revealed that the activation of adenylate cyclase (cAMP) could induce the acquisition of epithelial properties. The study further demonstrated a role for the cAMP-downstream effector protein kinase A in inducing a MET and maintaining an epithelial cell state (18).

However, while an EMT is an important mediator of primary tumor cell invasion, a MET has been shown to contribute to the proliferative outgrowth of disseminated cancer cells in distant organs (23–25). Hence, the therapeutic reversion of an EMT in cancer might be counterproductive (13). However, it has also been noted that cells undergoing an EMT and/or a MET are in a state of high cell plasticity and acquire stem cell–like characteristics (4, 5, 13, 26, 27). Notably, cells exhibiting an intermediate state of EMT/MET seem to carry the highest tumorigenic and metastatic capabilities (15, 28). It appears that the dynamic changes during an EMT and MET represent a state of high cell plasticity and are plausible in both well-differentiated cancer cells as well as in dedifferentiated mesenchymal-like cancer types, such as triple-negative breast cancer.

In our study (29), we have therapeutically exploited EMT-mediated cancer cell plasticity by inducing transdifferentiation of mesenchymal, malignant cancer cells into functional adipocytes. Adipocytes are specialized postmitotic cells. Given their inherent terminal differentiation and growth arrest, they are unlikely to further adapt or evolve and, thus, they lack cellular plasticity (Fig. 1). Moreover, we have hypothesized that due to the growth-arrested state of adipocytes cancer mutations may become irrelevant.

We initially asked whether breast cancer cells induced to undergo an EMT gain cellular plasticity and acquire the potential to differentiate into other cell types, such as adipocytes. To learn about the general process of cancer cell adipogenesis, preadipocytic 3T3-L1 fibroblasts were used as a “gold standard” control for inducing adipogenesis (30–32). Upon treatment with insulin, dexamethasone, and rosiglitazone, a synthetic agonist of peroxisome proliferator–activated receptor γ (PPARγ), 3T3-L1 cells upregulated adipogenesis transcription factors, such as PPARγ and C/EBPα, and gradually progressed into a substantial morphologic change with the accumulation of large lipid droplets bordered by specialized, adipocyte-specific proteins, such as perilipin (33–35). To test whether adipogenesis was also possible in breast cancer cells, we employed two mammary tumor–derived EMT models developed in our laboratory (36, 37). Epithelial MTflECad cells have been derived from a tumor of a murine MMTV-Neu transgenic mouse and carry conditional (floxed) alleles of the E-cadherin (Cdh1) gene. Upon genetic depletion of E-cadherin expression by Cre recombinase, these cells undergo an irreversible EMT (mesenchymal MTΔECad cells). Py2T cells have been isolated from a tumor of a MMTV-PyMT transgenic mouse and undergo an EMT upon treatment with TGFβ. Upon removal of TGFβ the cells undergo a MET, thus representing a reversible EMT model. Previously, it has been reported that BMP-2, a member of the TGFβ superfamily, can enhance the adipogenesis potential of mesenchymal stem cells (MSC; refs. 38, 39). Indeed, the addition of BMP-2 to the adipogenesis cocktail induced adipogenesis in EMT-induced Py2T (TGFβ-treated) and in MTΔECad cells, but not in their epithelial ancestors. This forced adipogenesis in EMT-derived breast cancer cells was most efficient when induced with the combination of BMP-2 and rosiglitazone. Of note, cancer-derived adipocytes did not revert back to the epithelial or mesenchymal state and maintained their adipocyte characteristics after removal of all differentiation inducers in vitro.

Next, we assessed whether cancer cell–derived adipocytes shared characteristics with normal 3T3-L1–differentiated adipocytes (40, 41). RNA sequencing analysis of 3T3-L1 and breast cancer cell adipogenesis kinetics demonstrated a high transcriptomic correlation between cancer cell–derived adipocytes and 3T3-L1–derived adipocytes (42). In both systems, upregulated expression of genes characteristic for white adipocytes was inversely correlated with the reduced expression of mesenchymal cell marker genes.

Molecular analysis revealed that, like mature adipocytes, both 3T3-L1–differentiated adipocytes and cancer-derived adipocytes secreted the adipokine adiponectin. Moreover, adipocytes responded to β-agonists by an increased release of fatty acids and glycerol and, indeed, 3T3-L1 and cancer cell–derived adipocytes released high levels of glycerol upon treatment with isoproterenol. Moreover, reliably recapitulating the physiologic function of bona fide adipocytes, the glucose transporter GLUT4 was expressed by cancer cell–derived adipocytes and translocated to the plasma membrane upon insulin stimulation. Finally, actin stress fibers, a typical mesenchymal cytoskeletal organization facilitating cancer cell motility, were changed during adipogenesis into a cortical actin organization, typical of sessile cells. Together, these and other experiments demonstrated the possibility of converting malignant breast cancer cells into bona fide adipocytes.

In this context, it is important to note that the interaction between adipose tissue and cancer cells can contribute to cancer progression (32). For example, tumor-associated adipocytes demonstrate increased secretion of adiponectin and FABP4, which was shown to directly promote cancer metastasis (43, 44). Interestingly, the adipocytes that have transdifferentiated from breast cancer cells express FABP4 and secrete adiponectin at comparable levels as 3T3-L1–derived adipocytes. Thus, a potential contribution of the cancer cell–derived adipocytes to tumor progression warrants further investigation.

Both transcriptional regulators of adipogenesis, PPARγ and C/EBPα, appear to be critical for the growth arrest that is required for adipocyte differentiation by mediating the expression of cyclin-dependent kinase inhibitors (45, 46). The simultaneous regulation of cell cycle and adipogenic regulators eventually results in cell-cycle exit and the generation of terminally differentiated, postmitotic adipocytes (47). Indeed, in our study, cancer cells gradually went into cell-cycle arrest during forced adipogenesis, and mature cancer cell–derived adipocytes ceased proliferation and thus overcame the initial oncogenic transformation. Interestingly, the adipogenesis progression of breast cancer cells directly correlated with the translocation of the Hippo signaling transcription factor YAP from the nucleus to the cytoplasm. This result is in line with published results describing YAP-regulatory effects in the differentiation of MSCs into osteocytes (nuclear YAP) or adipocytes (cytoplasmic YAP; ref. 48).

It has been established that TGFβ inhibits adipogenesis and impairs the development of adipose tissue. In fact, the inhibition of endogenous TGFβ signaling by either expression of a dominant-negative TGFβ receptor or by direct inhibition of SMAD3 leads to increased adipogenesis. SMAD3 has been found to bind to C/EBPs and to inhibit their transcriptional activity, including their ability to activate PPARγ2 expression (49). Thus, TGFβ-mediated inhibition of adipogenesis has been linked to its canonical signaling. However, the noncanonical TGFβ pathway involving MEK-ERK signaling has also been reported to interfere with PPARγ function, leading to obesity-linked insulin resistance (49). ERK directly phosphorylates PPARγ at serine 273 and induces diabetogenic gene expression in adipose tissues (50). Moreover, the phosphorylation of PPARγ by ERK1 inhibits the terminal phase of adipocyte differentiation (49).

However, as discussed above, TGFβ plays a major role in the induction of an EMT in tumor cells in vivo. Hence, to test our findings in proof-of-concept studies in preclinical models, we have faced the dilemma that TGFβ induces an EMT required for increasing cell plasticity, but at the same time represses adipogenesis of mesenchymal cells with high cell plasticity. Therefore, we tested whether MEK-ERK inhibition would facilitate adipogenesis, even in the presence of TGFβ. Indeed, the combination of MEK inhibitors with rosiglitazone was sufficient to efficiently induce adipogenesis of TGFβ-treated breast cancer cells in vitro, highlighting MEK-ERK activation as a major inhibitory axis of adipogenesis.

As we were motivated to test the forced adipogenesis of cancer cells as therapeutic approach, we tested the combination of MEK inhibitors with rosiglitazone in various transplantation mouse models of murine and human breast cancer and in patient-derived xenograft transplantation mice. Indeed, lineage tracing and immunofluorescence microscopy experiments demonstrated that the combination treatment with a MEK inhibitor and the diabetes drug rosiglitazone forced invasive cancer cells into adipogenesis. Of note, for the FDA-approved drugs rosiglitazone and trametinib dose selection was based on published preclinical in vivo studies (51, 52). Notably, although only a small fraction of the cancer cells transdifferentiated into postmitotic adipocytes (around 25 cancer-derived adipocytes/tumor section were quantified in the adipogenesis-treated preclinical models versus 0 for the control), the morphology of the primary tumor periphery had shifted from an invasive phenotype into a differentiated, smoothened tumor border upon combination treatment. As expected, primary tumor growth was already significantly repressed by the treatment with trametinib alone, yet strikingly, forced cancer cell adipogenesis by the combined treatment with trametinib and rosiglitazone efficiently inhibited the formation of lung metastasis.

Recent advances in EMT research suggest that cellular plasticity enhanced by EMT and MET is a major contributor to cancer metastasis. With our recent work, we have demonstrated that a moment of cellular plasticity can be exploited to specifically target cancer adaptation and dissemination mechanisms to force cells into a different cell type (29). In the context of breast cancer adipogenesis, EMT/MET regulate transcriptional changes that allow a forced transdifferentiation of cancer cells. TGFβ signaling acts as a central inducer of EMT, yet a repressor of adipogenesis. However, once the noncanonical TGFβ pathway executed by MEK-ERK signaling is inhibited, adipogenesis can occur (Fig. 2A). Our proof-of-concept study had its preclinical focus on modeling aggressive triple-negative breast cancer, yet EMT and cancer cell plasticity are inherent also to differentiated entities of breast cancer. Thus, adipogenesis treatment can potentially be proven beneficial when combined with state-of-the-art therapy for hormone receptor–positive or HER-2–positive breast cancer, or with conventional chemotherapy. Hence, further preclinical assessment will be required to test the effect of forced adipogenesis on different breast cancer subtypes and stages.

Cancer cell plasticity is considered a general characteristic of malignant solid tumors. Hence, it will be curious to investigate whether the same mechanisms of cancer adipogenesis would apply for other cancer types. Moreover, transdifferentiation into a postmitotic cell type other than adipocytes may also prevent malignant tumor progression and metastasis. Notably, in our report we have also shown that by utilizing adequate induction cocktails mesenchymal breast cancer cells can also be triggered to transdifferentiate into osteocytes or chondrocytes. The effect of the treatment on mesenchymal stem cells was not tested in this study and should be further assessed in the future, when testing the effect of adipogenesis treatment on the tumor microenvironment. Of note, mouse body weight was not increased as a result of the adipogenesis treatment when compared with controls.

Disseminated cancer cells often reside as dormant tumor cells in specialized niches resembling adult stem cell niches (2). Existing stem-cell niches are rich in extracellular signaling molecules, such as members of the TGFβ family, Hedgehog, Wnt, and the chemokine CXCL12 (2, 12, 53). These signals suppress redifferentiation and keep cells in a quiescent state, eventually leading to therapy resistance (5, 54). It seems plausible that due to the inherent cell-cycle arrest during adipogenesis, dormant cancer cells may also respond to adipogenesis treatment, thus preventing tumor relapse during or after therapy.

The use of MEK inhibitors in oncological therapeutics is currently restricted to mutational activation of the BRAF pathway in melanoma. Here, we show that delineation of the molecular mechanisms of cell plasticity can be used for transdifferentiation therapies regardless of oncogenic background and mutational load. Thus, adipogenesis therapy in breast cancer could potentially be beneficial in inhibiting cancer relapse and progression in any tumor subtype.

Past efforts to use differentiation therapies on solid tumors have focused on reverting cancer cells back into their original differentiated cell type. The limited benefit of such reversions (as in the case of MET of mesenchymal, invasive cancer cells) in preclinical and clinical settings may lie on the inherent plasticity of the reverted cells as well as the contribution of driver mutations and continuously activated oncogenic pathways initiating malignant progression. Our results demonstrate that transdifferentiation into a postmitotic cell type, such as adipocytes, can overcome oncogenic pathway activation as well as dynamic cellular plasticity. On the basis of our preclinical proof-of-concept studies, the specific drug combination of trametinib and rosiglitazone seems promising. However, in different cancer types other targets may prove beneficial in forcing such conversion. Further studies delineating the molecular mechanisms of cancer cell adipogenesis in a variety of cancer models may open new avenues for forcing such transdifferentiation therapeutic approaches. Moreover, one may speculate that a combination of the transdifferentiation therapy with conventional chemotherapy or targeted therapy may hit two birds with one stone: conventional therapy will kill highly proliferating cancer cells, while transdifferentiation therapy will incapacitate all cancer cells escaping chemotherapy by undergoing an EMT. This hypothesis clearly warrants experimental testing.

No potential conflicts of interest were disclosed.

This work was supported by a Swiss National Science Foundation project grant 310030B_16347 and a Swiss Cancer League/Oncosuisse project grant KFS-3479-08-2014 to G. Christofori. D. Ishay-Ronen received a MD-PhD student fellowship from the Botnar Foundation Basel.