News & Views | Published:


Pathways of parallel progression

Nature volume 540, pages 528529 (22 December 2016) | Download Citation


Two studies in mice identify mechanisms by which tumour cells disseminate in very early breast cancer. Both show that these cells colonize distant tissues more efficiently than their later counterparts. See Article p.552 & Letter p.588

The conventional linear model of cancer progression states that the cells of a developing tumour gradually pick up genetic mutations, with cells that accumulate optimal variants eventually acquiring the ability to migrate to and colonize other tissues in the body as metastases. This theory has a huge influence on current views of personalized medicine — for evidence, look no further than the US Cancer Moonshot initiative1, which recommends extensive sequencing of primary tumours to predict and understand treatment resistance. This represents a reasonable approach only if the disseminated cancer cells (DCCs) that form metastases are derived from cells that populate the primary tumour around the time of its detection. But several lines of evidence2,3,4,5,6,7,8,9 indicate that tumour cells can leave the primary site very early during tumour progression and evolve independently at the metastatic site. Online in Nature, two papers10,11 shed light on the mechanisms of early dissemination for the first time. Strikingly, they offer the first evidence that these early DCCs are more metastasis-competent than cells that leave the tumour at later stages.

The first compelling data to call the linear-progression model into question came from studies of human breast cancer. Genetic analysis of primary breast tumours and corresponding DCCs showed that, at the time of tumour detection, DCCs had fewer genetic alterations than primary cells, implying that DCCs seed the bone marrow early in disease progression, and evolve separately2. This theory of parallel progression5 was supported by the revelation that 20–30% of patients classified as having 'non-invasive' breast cancer have DCCs in their bone marrow7,9. Because up to 8% of 'non-invasive' breast cancers recur at distant sites12, it was assumed that at least some of these early DCCs had metastasis-initiating potential.

Animal models of breast cancer cast further doubt on the linearity of metastasis. In a mouse model in which breast cancer is driven by overexpression of the gene Her2, DCCs were detectable in the bone marrow by four weeks of age — just after Her2 expression begins4. Cancerous alterations in the mammary gland are detectable only by electron microscopy at this stage, and palpable primary tumours do not develop for another 14 weeks4.

Despite these data, the molecular mechanisms that underlie the early dissemination of tumour cells have not been identified. To address this, Hosseini et al.10 and Harper et al.11 returned to the Her2-driven breast-cancer mouse model.

Hosseini et al. (page 552) analysed gene expression in epithelial cells that line the ducts of the mammary gland, isolated before the mice reached nine weeks of age, when tumours are not yet palpable. Their analysis suggested that the hormone progesterone drives dissemination from these microscopic early tumours. The authors showed that progesterone triggers secretion of the proteins WNT4 and RANKL from cells expressing the progesterone receptor (PGR), and that these signals imbue epithelial cells that do not express PGR with increased migratory potential (Fig. 1a).

Figure 1: Mechanisms of early dissemination and metastasis in breast cancer.
Figure 1

Two papers outline molecular mechanisms by which cancerous epithelial cells in the mammary glands of mice, driven by overexpression of the gene Her2, disseminate from the gland to future sites of metastasis during very early stages of tumour development. a, Hosseini et al.10 find that WNT4 and RANKL proteins are secreted from cells in very early tumours that express the progesterone receptor (PGR) protein (PGR+ cells). Nearby cells that do not express PGR (PGR) become migratory owing to the secreted signals and invade the bone marrow as early disseminated cancer cells (DCCs). b, Harper et al.11 report that increases in the WNT signalling pathway lead to inhibition of the protein p38 and transition of early-tumour cells to an invasive DCC state. c, Cells from late-stage tumours also migrate to the bone marrow. Both studies provide evidence that cells that disseminate later in primary-tumour progression form metastases in the bone marrow less efficiently than early DCCs, implying that early DCCs might be a source of metastases.

The effect of progesterone becomes less apparent as tumour development progresses. The authors found that both higher cellular density and the increased HER2 protein levels driven by this change induce expression of microRNA molecules that inhibit the gene encoding PGR. In effect, a dissemination-to-proliferation switch occurs when a developing tumour region becomes crowded enough.

Harper et al.11 (page 588) identified an invasive cell population in microscopic tumours that is characterized by Her2-driven expression of WNT proteins. WNTs counteract the activity of the enzyme p38, which is typically expressed by non-proliferative (dormant) DCCs13. Downregulation of p38 leads to a decrease in the epithelial cell–cell junction molecule E-cadherin and to upregulation of several other genes involved in epithelial-to-mesenchymal transition — a change in cell state that facilitates invasion of mammary cells into the bloodstream, and ultimately into other tissues as early DCCs (Fig. 1b). Inhibition of p38 significantly increases the number of circulating tumour cells, as well as the number of DCCs in both the lung and bone marrow, confirming this protein's role in dissemination.

The authors found that DCCs regain some of their epithelial characteristics when they reach the lung and bone marrow. However, it is not clear whether dormancy-associated p38 signalling is restored when the cells enter these sites. Nor is it known whether these early DCCs are less responsive to quiescence-inducing cues from their new microenvironments14 than cells that leave the primary tumour at later stages.

Are early DCCs less likely to become — or to stay — dormant when they reach distant sites? Perhaps so. Adopting different experimental strategies, both groups demonstrated that early DCCs are inferior to late DCCs in their ability to form primary tumours if implanted in another mammary gland. However, early DCCs are substantially more metastasis-competent, forming metastases faster and more prolifically than their later counterparts (Fig. 1c).

The molecular mechanisms that drive dissemination from early mammary lesions identified in these studies might not apply across all subtypes of breast cancer or to other cancers. Nonetheless, the findings provide a general framework within which to study causality between early DCCs and metastasis — particularly for cancers in which early dissemination is a documented phenomenon, such as skin6 and pancreatic8 cancers.

These studies have major implications with regard to preventive therapies. First, given the microscopic stage of primary tumour formation at which dissemination of metastasis-competent cells occurs, developing means for early tumour detection may not be sufficient to prevent metastatic disease. And, because cells have probably disseminated by the time a primary tumour is detected, targeting the mechanisms of early dissemination identified in these studies may not be a viable therapeutic strategy either. Thus, we should aim to target the characteristic properties of early DCCs — their long-term survival and therapeutic resistance14.

As Hosseini et al. confirmed from their analysis of human tissue samples, early DCCs differ substantially from the primary tumour at the molecular level. Thus, to learn about early DCCs, we must increase the frequency with which DCCs are isolated from humans, profiled, and studied functionally in appropriate animal and culture models. This will aid the identification of elusive molecular targets for precision metastasis-prevention therapies based on demonstrable steps in cancer progression, instead of on an assumption of linearity.



  1. 1.

    Cancer Moonshot Blue Ribbon Panel Report 2016 (National Cancer Advisory Board, 2016); available at

  2. 2.

    et al. Proc. Natl Acad. Sci. USA 100, 7737–7742 (2003).

  3. 3.

    et al. N. Engl. J. Med. 353, 793–802 (2005).

  4. 4.

    et al. Cancer Cell 13, 58–68 (2008).

  5. 5.

    Nature Rev. Cancer 9, 302–312 (2009).

  6. 6.

    et al. J. Clin. Invest. 120, 2030–2039 (2010).

  7. 7.

    et al. Int. J. Cancer 129, 2522–2526 (2011).

  8. 8.

    et al. Cell 148, 349–361 (2012).

  9. 9.

    et al. Anticancer Res. 36, 2345–2351 (2016).

  10. 10.

    et al. Nature 540, 552–558 (2016).

  11. 11.

    et al. Nature 540, 588–592 (2016).

  12. 12.

    , , , & JAMA Oncol. 1, 888–896 (2015).

  13. 13.

    , , , & Mol. Biol. Cell 12, 863–879 (2001).

  14. 14.

    Nature Rev. Cancer 15, 238–247 (2015).

Download references

Author information


  1. Cyrus M. Ghajar is in the Public Health Sciences Division's Translational Research Program and the Human Biology Division, Fred Hutchinson Cancer Research Center, Seattle, Washington 98109, USA.

    • Cyrus M. Ghajar
  2. Mina J. Bissell is in the Biological Systems and Engineering Division, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA.

    • Mina J. Bissell


  1. Search for Cyrus M. Ghajar in:

  2. Search for Mina J. Bissell in:

Corresponding authors

Correspondence to Cyrus M. Ghajar or Mina J. Bissell.

About this article

Publication history



Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Newsletter Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing