Transition states that allow cancer to spread

Cancers of epithelial-cell origin often contain some tumour cells that have acquired traits of mesenchymal cells. How this leads to cancer spread has now been illuminated in mouse models.
Erik W. Thompson is in the Institute of Health and Biomedical Innovation, and the School of Biomedical Sciences, Queensland University of Technology, Brisbane QLD 4059, Australia; and is also at the Translational Research Institute, Woolloongabba, Australia.

Search for this author in:

Shivashankar H. Nagaraj is in the Institute of Health and Biomedical Innovation, and the School of Biomedical Sciences, Queensland University of Technology, Brisbane QLD 4059, Australia; and is also at the Translational Research Institute, Woolloongabba, Australia.

Search for this author in:

In cancers that arise from epithelial cells, some of the tumour cells adopt features of mesenchymal cells and lose their epithelial characteristics. This phenomenon is known as the epithelial-to-mesenchymal transition (EMT), and the emergence of cells with mesenchymal features is often associated with a poorer prognosis for patients. How this transition occurs and the implications of EMT for metastasis, the process by which cancer spreads, are not fully understood. In a paper in Nature, Pastushenko et al.1 report an analysis of mouse tumours undergoing EMT, revealing that distinct cell populations can be identified during this transition.

Cancers of epithelial tissues are called carcinomas and can give rise to cells with mesenchymal properties, such as an elongated shape. These mesenchymal cells have been implicated2 in tumour metastasis. The EMT process is also associated with cancer stem cells found in breast cancer and other cancer types, and is linked to the presence of the circulating tumour cells in the bloodstream that are a hallmark of tumour-relapse risk3. In numerous mouse models2, prevention of EMT or its reversal — a process known as MET, in which mesenchymal cells transition back to epithelial cells — can reduce the stem-cell-like properties of tumour cells and thereby reduce their ability to metastasize.

Although some studies4,5 indicate that not all tumour metastases require EMT, considerable amounts of data implicate6 the process of MET as a requirement for metastatic cells to successfully colonize locations beyond the initial tumour site. One way to address this controversy would be to gain a better understanding of how the EMT process occurs.

A hybrid EMT state, also known as a metastable state or partial EMT, in which individual cells express both epithelial and mesenchymal markers, has often been observed in carcinoma7, developmental processes and wound healing8. One model for how EMT might occur is that cells make a gradual transition from one state to the other along a continuous spectrum of change in which cells lose epithelial characteristics and concurrently gain mesenchymal ones. This is consistent with gene-expression analysis of cell lines from various types of carcinoma9. However, computational-modelling studies7 and in vitro observations8 indicate instead that distinctive, stable, long-lasting cell populations could represent discrete intermediate stages of EMT.

Pastushenko and colleagues investigated whether stable EMT states in tumours could be identified in vivo. The authors used genetically engineered mice that provide a model system in which to study skin carcinoma and have tumour cells10 that are known to exhibit EMT. Pastushenko et al. studied cells that had lost expression of the epithelial-cell marker EpCAM, and analysed the expression of more than 170 cell-surface proteins to try to identify useful markers of cell subpopulations. They then focused on three of these markers (CD61, CD51 and CD106) that are characteristic of a mesenchymal-cell state. This enabled the authors to identify distinct cell populations that had lost expression of the epithelial-cell marker EpCAM and expressed different combinations of these three mesenchymal markers.

Using the three markers as a tool to enable the isolation of cell populations present when EMT occurs, the authors isolated and characterized six distinctive cell subpopulations that lacked EpCAM expression. The authors used an impressive array of biomolecular analyses to assess the cellular characteristics of these populations, including their metastatic traits and their state of cellular differentiation. This work provides in vivo evidence of stable cell populations representing intermediate EMT states.

The authors also identified comparable cell subpopulations, characterized by the same markers, in both a mouse mammary-tumour model and human tumour samples transplanted into mice, suggesting that these subpopulations represent stable transitional cellular states found in different cancer types. This finding might have clinical relevance. The authors’ work therefore provides a platform for future investigations of this phenomenon in other cancer models, including studies of the regulatory networks that control these states. Such investigations might improve understanding of the molecular and cellular features that underlie the tumour-cell capabilities associated with these states.

Do these subpopulations exhibit behaviours that might offer therapeutic opportunities? One of the authors’ most striking findings, made in the mouse skin-carcinoma and mammary-tumour models, is that when comparing these cell subpopulations, two of them stood out as having a substantially higher metastatic potential (Fig. 1). This raises the question of whether these highly metastatic cells could be specifically targeted with drugs to arrest tumour progression. Interestingly, none of the stable subpopulations was superior to the others in terms of tumour-initiating capability or proliferation rate. Also, each of the subpopulations that had acquired mesenchymal characteristics exceeded the metastatic capabilities of the epithelial-cell population. These findings are relevant to current debates in this field, which include the question of whether a cycle of EMT followed by MET is required for successful metastatic colonization of a secondary location6,11.

Figure 1 | Populations of cancer cells in transition states. Cancers that arise from epithelial cells often contain some tumour cells that have acquired mesenchymal-cell characteristics, such as an elongated shape and the expression of genes associated with mesenchymal cells. This change is termed the epithelial-to-mesenchymal transition (EMT). Whether this transition occurs as a gradual, continuous change or by discrete stages is debated. Pastushenko et al.1 report mouse studies consistent with the latter, in which they identified cell populations in vivo that represent stable, distinct intermediates in this transition. The authors analysed cells that do not express the epithelial-cell receptor protein EpCAM, and isolated cell populations on the basis of their expression of the mesenchymal-cell receptor proteins CD51, CD61 and CD106. They monitored the level of proteins characteristic of the epithelial-cell state (such as keratin-14, yellow) or of the mesenchymal-cell state (such as vimentin, blue), and found cells that exhibited a hybrid of epithelial and mesenchymal characteristics. Cells that expressed EpCAM had low levels of vimentin and high levels of keratin-14. Once EpCAM expression was lost, the level of vimentin increased sharply and the level of keratin-14 declined incrementally as cell populations became more mesenchymal in character. The authors identified two cell populations that were the most likely to spread and form tumours elsewhere in the process called metastasis.

The authors’ work reveals the progressive acquisition of EMT features in EMT hybrid cells, supporting the previously proposed8 idea of stable transitional states. A process of stepwise change is supported by Pastushenko and colleagues’ data, indicating that distinct transcriptional and signalling processes govern intermediate aspects of EMT. The different cell subpopulations had characteristic transcriptional signatures regulated by distinct transcription-factor proteins, underscoring the reproducible and meticulously regulated nature of these hybrid cells.

Although these hybrid cell populations are a stable presence in tumours, the authors found that these cells also maintain a high degree of plasticity, given their ability to undergo MET and revert back to an epithelial-cell state. However, the authors found that the cell subpopulations that were best at undergoing MET during tumour metastasis to the lung in mice were not the most metastatic populations, which will fuel the debate6 about whether MET is a requirement for metastasis.

Elucidation of the circuitry and signalling feedback loops that might stabilize7,12 these distinct cellular states will be a worthy goal for future work. Studies of such cellular intermediates in patients’ tumours, circulating tumour cells and metastases would be an ideal way to extend this work in a clinical context. If EMT states can be characterized for a wide range of human tumours, this might offer a way to enhance personalized approaches for cancer treatment. The development of specific technologies to identify human EMT states might allow clinicians to predict a tumour’s metastatic potential and thereby plan the most effective treatment regimen.

Nature 556, 442-444 (2018)

doi: 10.1038/d41586-018-04403-x


  1. 1.

    Patushenko, I. et al. Nature 556, 463–468 (2018).

  2. 2.

    Chaffer, C. L., San Juan, B. P., Lim, E. & Weinberg, R. A. Cancer Metastasis Rev. 35, 645–654 (2016).

  3. 3.

    Francart, M. E. et al. Dev. Dynamics 247, 432–450 (2017).

  4. 4.

    Zheng, X. et al. Nature 527, 525–530 (2015).

  5. 5.

    Fischer, K. R. et al. Nature 527, 472–476 (2015).

  6. 6.

    Brabletz, T., Kalluri, R., Nieto, M. A. & Weinberg, R. A. Nature Rev. Cancer 18, 128–134 (2018).

  7. 7.

    Jolly, M. K. et al. Oncotarget 7, 27067–27084 (2016).

  8. 8.

    Nieto, M. A., Huang, R. Y.-J., Jackson, R. A. & Thiery, J. P. Cell 166, 21–45 (2016).

  9. 9.

    Tan, T. Z. et al. EMBO Mol. Med. 6, 1279–1293 (2014).

  10. 10.

    Latil, M. et al. Cell Stem Cell 20, 191–204 (2017).

  11. 11.

    van Denderen, B. J. W. & Thompson, E. W. Nature 493, 487–488 (2013).

  12. 12.

    Bracken, C. P. et al. Cancer Res. 68, 7846–7854 (2008).

Download references

Paid content

Nature Briefing

An essential round-up of science news, opinion and analysis, delivered to your inbox every weekday.